APPLICATION AND VIABILITY OF LONG DAY LIGHTING ON A LARGE MICHIGAN DAIRY FARM MILKING THREE TIMES EACH DAY By Benjamin VanZweden A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Biosystems Engineering—Master of Science 2019 ABSTRACT APPLICATION AND VIABILITY OF LONG DAY LIGHTING ON A LARGE MICHIGAN DAIRY FARM MILKING THREE TIMES EACH DAY By Benjamin VanZweden This study’s goal was to add to the knowledge base regarding the energy use, controls, possible galactopoietic effects, and economic feasibility of long day lighting (LDL) on a large commercial dairy farm. The implemented system maintained a minimum illumination level of 160 lux (15 foot-candles) for 16 hours and 40 lux (4 fc) maximum for 8 hours on a large dairy farm in Michigan that milks approximately 1,100 cows. The system used dimmable 150-Watt high-bay luminaires, dimmable 60-Watt canopy luminaires, and a dynamic control system. The control system used daylight harvesting to save energy, dimming the LED luminaires when natural light was available in the barns. The LDL system used 67% less energy than a theoretical system with metal halide HID luminaires, and 52% less energy than an identical system with LED luminaires but no dynamic controls. Due to unforeseen challenges, there were only 3 months of original milk production data available that could be compared with post implementation data. Analysis of the 300 highest producing cows showed an estimated 10.5% increase in milk production over July-September 2019 compared to July-September 2018. Using the estimated 10.5% increase in production, the implemented system would have a payback period of less than one year. If the estimated increase is maintained for one year, the production increase would be equivalent to adding 32 more cows to the highest producing groups. The estimated 32 cows that were matched in this calculated production increase would have produced about 13,000 pounds of CH4 annually, which is equivalent to approximately 331,000 pounds of CO2e. ACKNOWLEDGEMENTS This project could not have been completed without the help and support of many people and organizations. First, I would like to thank my committee members, Dr. Surbrook, Dr. Domecq, and Dr. Harrigan for their advice and guidance throughout this project. Dr. Surbrook was a fantastic advisor and his sense of humor in the face of difficulties really helped to lighten the mood. Al Go’s knowledge and previous experience with long day lighting were very helpful and his oversight kept this project on schedule. Mark VanEe’s electrical skills were indispensable when troubleshooting problems and installing equipment. Aryn Thomas wrote the grant application that provided the initial funds for this project, and her continual support throughout this research was greatly appreciated. This work would not have been possible without funding, so thank you to DTE Energy, Engineering Society of Detroit, and Michigan State University for providing the initial funds to start this work. A special thank you to the Michigan Agricultural Energy Council, which has been funding the upkeep and labor costs of this research. Their continued commitment to farm energy research is always appreciated. Finally, thank you to my family for their support, and above all to God, the giver of all knowledge and understanding. iii TABLE OF CONTENTS LIST OF TABLES ................................................................................................................................ vi LIST OF FIGURES ............................................................................................................................. vii KEY TO ABBREVIATIONS ................................................................................................................ viii 1. Introduction ................................................................................................................................ 1 1.1 Overview of Michigan Dairy Industry ................................................................................... 1 1.2 Increasing Profits by Increasing Production ......................................................................... 3 2. Literature Review ........................................................................................................................ 7 2.1 Fiscalini Dairy Farm, Modesto, CA ...................................................................................... 11 2.2 Large Commercial Dairy, NY ............................................................................................... 13 2.3 Wing Acres Dairy, Bellevue, MI ........................................................................................... 15 3. Research Objectives .................................................................................................................. 18 4. Materials and Methods ............................................................................................................. 19 4.1 Funding................................................................................................................................ 19 4.2 Research Site ....................................................................................................................... 20 4.2.1 Barn Architecture ......................................................................................................... 22 4.3 Equipment ........................................................................................................................... 23 4.3.1 Lighting System and Layout ......................................................................................... 23 4.3.2 Control system ............................................................................................................. 25 4.4 Installation .......................................................................................................................... 29 4.5 Data Collection .................................................................................................................... 30 4.5.1 Energy Data Collection ................................................................................................. 30 4.5.2 Milk Data Collection ..................................................................................................... 31 5. Results/Discussion .................................................................................................................... 34 5.1 Energy Data ......................................................................................................................... 34 5.2 Milk Data ............................................................................................................................. 37 5.3 Financial Analysis ................................................................................................................ 45 6. Conclusions/Future Work ......................................................................................................... 52 APPENDICES .................................................................................................................................. 61 Appendix A—DLC Requirements for Solid State Lighting and Lighting Controls ..................... 62 Appendix B—Project Specifications for LDL Lighting and Control Systems ............................. 64 Appendix C—Barn Architecture ................................................................................................ 67 iv Appendix D—Lighting layouts ................................................................................................... 70 BIBLIOGRAPHY .............................................................................................................................. 72 v LIST OF TABLES Table 1—Energy consumption for Michigan dairy farms (Van Zweden et al., 2018) ..................... 2 Table 2—Milk production of Michigan dairy farms (Van Zweden et al., 2018).............................. 2 Table 3—Annual cost per dairy cow and cost per cwt milk produced on MI dairy farms (Jones et al., 2018) .......................................................................................................................... 5 Table 4—Studies conducted in laboratories on the effects of LDL (Dahl et al., 2000) ................. 10 Table 5—Energy consumption of original, theoretical, and implemented systems ..................... 35 Table 6—Percent savings of the implemented LDL system compared to the theoretical HID layout and a featureless LED layout ....................................................................................... 35 Table 7—Cost savings of implemented system ............................................................................ 37 Table 8—Average production per parlor visit for Groups 1 and 3 with statistics ........................ 44 Table 9—Additional costs and capital investment associated with this project (west barn only, Groups 1 and 3) ............................................................................................................ 47 Table 10—Estimated additional milk, increased revenue, increased profit, simple payback, NPV, and IRR from implemented LDL (Groups 1 and 3) ................................................ 48 Table 11—Number of cows needed to match the production increase of LDL ............................ 49 Table 12—Greenhouse gas savings .............................................................................................. 50 vi LIST OF FIGURES Figure 1—Milk price per cwt in U.S. dollars 2016-2019 (Barchart Market Data Solutions, 2019) ............................................................................................................................................... 4 Figure 2—Satellite view of barns (LDL areas in white borders; North at top of image) ............... 21 Figure 3—Milk parlor cellar showing the individual metered lines from each stall ..................... 22 Figure 4—Left: dimmable high bay luminaire with weatherproof box, bird spikes, and antenna; Right: partially disassembled canopy luminaire showing the DIM10-250 controller (top) and dimmable driver (bottom) ............................................................................ 25 Figure 5—LDL schedule with milking schedules for Groups 1 and 3............................................. 29 Figure 6—West and east barn DAQ layout ................................................................................... 31 Figure 7—Power draw graph for the implemented dimmable LED layout, theoretical featureless LED layout, and theoretical HID layout ...................................................................... 36 Figure 8—Bulk tank monthly milk production from October 2018 to September 2019 compared to a 2-year average ...................................................................................................... 39 Figure 9—Monthly number of cows freshened from October 2018 to September 2019 compared to a 2-year average ...................................................................................................... 40 Figure 10—Monthly bulk tank data compared to monthly freshened cows ................................ 40 Figure 11—Groups 1 and 3 average production per parlor visit for the LDL period October 2018 through September 2019 ....................................................................................... 43 Figure 12—Average production per parlor visit for Groups 1 and 3 (June-Sept).......................... 43 Figure C-1—Barn cross sections illustrating the use of scissor trusses (top) and webbed coffer trusses (bottom) ................................................................................................................. 67 Figure C-2—West barn top view ................................................................................................... 68 Figure C-3—East barn top view .................................................................................................... 69 FIgure D-1—West barn lighting layout ......................................................................................... 70 Figure D-2—East barn lighting layout .......................................................................................... 71 vii KEY TO ABBREVIATIONS LDL—Long day lighting DLH—Daylight harvesting LED—Light emitting diode HID—High intensity discharge cwt—hundredweight kWh—kilowatt-hour fc—foot-candle viii 1. Introduction The state of Michigan has a large food and agricultural industry which contributes over $102B to the state’s economy and employs over 17% of the state’s workforce. The food and agricultural industry includes farms, food processors, storage, and transportation. Michigan produces over 300 different agricultural commodities and is one of the most agriculturally diverse states, second only to California. In addition, Michigan is also a large exporter of agricultural products. Michigan’s third largest agricultural export is dairy products, which contribute over $332M to the state’s economy (MDARD, 2016). Given the significance of Michigan’s agricultural industry to its economy, it is necessary to continually find and implement new and innovative ways to increase the efficiency of agricultural operations. This study will focus on the implementation of long day lighting in Michigan, which is used to increase the production efficiency of dairy cows. 1.1 Overview of Michigan Dairy Industry Michigan has almost 48,000 farms, 1,500 of which are dairy. Only 3% of those dairy farms are owned by corporations while the rest are family owned. Michigan also is the 5th largest producer of milk in the country, producing over 11 billion pounds annually (MDARD, 2019; UDIM, 2018). Dairy farming is very energy intensive, and 30% of the energy consumed by Michigan dairy farms is electricity. That percentage is even higher for large dairy farms (>500 cows), where electricity is 40% of total energy consumption. Electricity is used for lighting, vacuum pumps, milk pumps, refrigeration, compressors, ventilation, and manure pumps, among other things (Van Zweden, Go, & Surbrook, 2018). All of these areas can be opportunities for 1 improving energy efficiency on a dairy farm. Using electricity more efficiently can be attractive to Michigan dairy farmers since Michigan has the highest average electricity costs of any midwestern state (Hankey, Cassar, Liu, Wong, & Yildiz, 2017). In addition, although there is little dairy farm energy usage and production data available from other states, Michigan dairy farms generally consume more energy per hundredweight (cwt) of milk produced compared to other states with available data such as New York, Minnesota, and Wisconsin. However, the farms surveyed in Michigan by the Michigan Farm Energy Program were generally more productive than farms in the other states (Van Zweden et al., 2018). Table 1 shows energy and electricity consumed per milk cow and per hundredweight of milk produced on Michigan dairy farms, while Table 2 shows milk production figures for different sizes of Michigan dairy farms. Table 1—Energy consumption for Michigan dairy farms (Van Zweden et al., 2018) Herd size Energy used per cwt milk produced (kWh) Energy used per cow (kWh) Electricity used per cow (kWh) Electricity used per cwt milk produced (kWh) Electricity as % of total energy usage 20-99 100-249 250-449 450+ Totals: 1 22.29 12.27 9.71 8.07 9.97 2 4300 2854 2428 2213 2532 3 866 732 534 875 771 4 4.49 3.15 2.14 3.19 3.04 5 20% 26% 22% 40% 30% 6 Table 2—Milk production of Michigan dairy farms (Van Zweden et al., 2018) Number of Farms Average Annual Milk Production Per Cow (cwt) Herd Size 20-99 100-249 250-449 450+ Totals: 1 22 57 18 16 113 2 193 233 250 274 254 3 As can be seen from these tables, as farm size increases, individual cow milk production and overall energy efficiency increases. However, although the largest category of farms consumes less energy per cow and per cwt of milk produced, the amount of electricity 2 consumed per cow and per cwt of milk increases. This is because some of the methods used to increase cow milk production efficiency require an increase in electricity consumption. Some common energy intensive methods used today to increase milk production include milking three times per day and increasing ventilation to reduce heat stress. 1.2 Increasing Profits by Increasing Production As show in Tables 1 and 2, a common way for farms to increase profits is to increase milk production, preferably without increasing herd size. Increasing herd size adds more overhead and expenses, as well as risk since milk prices are volatile and sometimes dip below the farms’ production cost. This volatility can be seen in Figure 1, which shows the 4-year price trend of Class 3 milk traded on the NASDAQ. Increasing a dairy cow’s production efficiency increases the overall milk production efficiency, which increases profits and reduces the overall carbon footprint of dairy farms. This is achieved by using fewer cows to produce the same amount of milk, less farm infrastructure to produce the milk, and less required labor. 3 Figure 1—Milk price per cwt in U.S. dollars 2016-2019 (Barchart Market Data Solutions, 2019) Increasing individual cow milk production saves a lot of direct and indirect costs to the farm operation, since the only cost besides the method of production enhancement is the additional feed consumed per additional pound of milk produced each day. Table 3 shows the expenses associated with each dairy cow and cwt of milk produced on an average large Michigan dairy farm (>500 cows). The figures in Table 3 are a summary of information found in Michigan State University’s 2017 annual farm business survey. Individual farm management strategies can affect the actual overhead cost associated with producing milk. 4 Table 3—Annual cost per dairy cow and cost per cwt milk produced on MI dairy farms (Jones et al., 2018) Total direct expenses Total overhead expenses Total direct and overhead expenses $4,144 $15.17 2 $768 $2.81 $4,912 $17.98 3 4 Per cow Per cwt 1 Table 3 shows that if a farmer can avoid adding more cows to the herd and instead use a method to make the cows produce more milk, the yearly cost per cow may increase, but the cost per cwt of milk produced will decrease. Each additional pound of milk produced by a cow will require an additional 0.4 pounds of feed, but each additional cow added to the herd will require over 12 pounds of maintenance feed per day as well as 0.4 lb feed/lb milk produced (French, 2000; Herdt, 2018). According to a study of Michigan dairy operations in 2017, additional cost in feed for each additional cwt of milk produced by a cow is $6.32. This does not include the cost of the maintenance ration since an existing cow is becoming more efficient. The additional cost of feed per cwt for an additional cow would be higher. Since the average MI dairy cow on a large farm produces an average of 75 pounds of milk per day each year, a modest 5% increase in an existing cow’s production would only cost an additional $0.10 per day, decreasing the total cost per cwt of milk produced by over 4% to $17.25, while only increasing the total annual cost per cow by 0.7% (Jones et al., 2018). Until 2008, a common way to increase individual cow milk production was to treat the cows with recombinant bovine somatotropin (rbST), an artificial growth hormone. Bovine somatotropin is a natural hormone produced by female cows that regulates milk production. It does this by regulating the cow’s metabolism, allocating a greater portion of available nutrients toward milk production. Cows can be manipulated to produce more natural bST on their own 5 through selective breeding to improve genetics, but recombinant DNA technology can also be used to artificially produce rbST which is then injected into the cows (Bauman, 2009). Due to public demand in the mid-2000s, major retailers in Michigan started requiring that the milk supplied to them come from cows that were not treated with rbST. Due to these demands, Michigan’s largest milk cooperative, Michigan Milk Producers Association (MMPA), changed their pricing structure to discourage its members from producing and selling milk from rbST-treated cows. As a result, MMPA members stopped using rbST since it was no longer economically viable (Wolf, 2008). Since rbST increases milk production by 10-15%, the loss in production was noticeable (Flanders & Gillespie, 2015). A common method of increasing production and cattle wellbeing in the post-rbST market is to manipulate the surroundings and nutrition of the dairy cattle. Dairy farms now have access to animal nutritionists who derive the feed portions for the cattle. In addition, increased ventilation in free-stall barns during the summer months is commonly used to prevent heat stressing and subsequent production loss. Another method that has been researched for over 35 years but rarely implemented is called long day lighting (LDL). LDL manipulates the lighting in the free-stall barns to provide better distributed and more constant light that mimics the photoperiod of a day, and can increase production 5-9% (Thomas et al., 2017). 6 2. Literature Review Long day lighting (LDL) means exposing lactating dairy cows to 16-18 hours of equivalent daylight and 6-8 hours of equivalent moonlight. Studies have shown that the endocrine system of cows can be manipulated with LDL to produce more or less of certain hormones. These hormones then cause the mammary glands to secrete more milk. This increased milk secretion is called a galactopoietic effect (Collier, Dahl, & VanBaale, 2006). Although the mechanism behind the galactopoietic effects of LDL is not completely understood, scientists have identified a few different hormones that may be responsible. When light enters a cow’s eye, the photoreceptors in its retina are stimulated. The stimulated retina sends a signal to the pineal gland, part of the endocrine system. The pineal gland produces melatonin, a hormone responsible for regulating sleep and wake cycles, as well as photoperiodic responses. The signal from the stimulated retinal photoreceptor inhibits the pineal gland enzyme N-acetyltransferase, which is responsible for catalyzing the synthesis of melatonin. Therefore, the pineal gland produces less melatonin when the cow is exposed to daylight and more when light levels are at or below moonlight. Establishing a consistent melatonin production pattern in the cow is necessary to elicit a photoperiodic response to LDL. This is important because the cow’s melatonin pattern regulates the secretion of other hormones in the endocrine system. Therefore, continuous lighting is not beneficial to cows since they lose the circadian rhythm and no longer express the melatonin production pattern associated with increased milk production (Dahl, Buchanan, & Tucker, 2000). The first hormone that scientists noticed was expressed at increased levels in cows exposed to LDL was prolactin (PRL). PRL is a protein hormone secreted by the pituitary gland 7 that is responsible for a cow’s ability to produce milk. To determine whether it was responsible for the galactopoietic effects of LDL, scientists administered exogenous PRL to a control group of cows, which had no effect on milk yield. In addition, scientists noticed that the increased PRL in cows exposed to LDL was subdued during temperatures below freezing, but the positive galactopoietic effects of LDL still persisted. It was therefore concluded that PRL was not responsible for mediating the effects of LDL. Another hormone investigated was bST, which was discussed earlier. Scientists looked for increased concentrations of bST in cows exposed to LDL, but were unable to find increased secretion in those cows (Dahl et al., 2000). Scientists also noticed that cows exposed to LDL had an increased dry matter intake (DMI). This is because the increased milk production from exposure to LDL means more of a cow’s nutritional intake is being allocated to produce milk, so the cow consumes more food to make up for the lost nutrition secreted in its increased milk production. Therefore, increased DMI follows increased lactation, not the other way around (Collier et al., 2006). Through a series of experiments, scientists found that cows exposed to LDL had an increased amount of insulin-like growth factor (IGF-I), which they determined to be the cause of the galactopoietic effects of LDL. The concentration of IGF-I in a cow’s plasma fluctuates based on the photoperiod to which they are exposed, and cows exposed to long day photoperiods had increased concentrations of IGF-I. In addition, IGF-I concentration fluctuation is independent of bST levels. Although the exact mechanism that links increased IGF-I concentrations with increased lactation is still not completely understood, scientists have concluded that there is enough experimental evidence to link IGF-I concentration with lactation (Dahl et al., 2000). 8 Large dairy farms that milk at least three times per day may find it difficult to adjust their milking schedule to give the cows 6-8 hours of darkness. Therefore, scientists also decided to test whether feeding melatonin to cows at the appropriate levels would mimic the effects of the 6-8 hours darkness associated with LDL. Scientists did discover that feeding melatonin to mimic short days (16 hours darkness and 8 hours light) does decrease the mammary gland tissue growth of heifers, whereas the elevated concentrations of IGF-I associated with long days are associated with increased mammary gland tissue growth. Lactating cows exposed to constant light that were fed supplemental melatonin to mimic the melatonin they would have produced if exposed to 6-8 hours of darkness did not exhibit any increase in lactation. Therefore, it was concluded that the only way to get the galactopoietic effects associated with LDL was to actually expose the cows to LDL (Dahl et al., 2000). This can be a challenge for large producers who milk three times per day, since the cows are constantly being moved between the free-stall barn, holding pens, and milking parlor. Even smaller producers could experience challenges since a good lighting control system and consistent herd management are essential to experience galactopoietic effects from LDL. Scientists have already determined that LDL increases milk production in cattle. However, this was determined in dairy laboratories where every aspect of the cows’ life can be controlled by scientists. Table 4 shows a list of LDL studies with statistically significant (P<0.05) increase in milk production. These studies were conducted in laboratories around the Northern and Eastern USA, Canada, United Kingdom, and Europe. One important point to notice from the table is the type of lighting used in the studies. All lighting used was either fluorescent or metal halide. This means these lights cannot be dimmed, they are either ‘on’ or ‘off.’ In the case of 9 sodium vapor and metal halide lamps, there is a noticeable delay in reaching full brightness. This could be a problem for the sodium vapor and metal halide lights if there was a momentary interruption in power since they do not immediately relight when the power comes back on, which could inhibit the effects attributed to LDL. Table 4—Studies conducted in laboratories on the effects of LDL (Dahl et al., 2000) Location Light type Fluorescent Michigan Fluorescent Michigan Fluorescent Michigan Fluorescent Quebec Fluorescent Ontario Wales Fluorescent Maryland Metal halide 1 2 Cows' response to LDL Milk yield increase (kg/d) 2.0 1.4 2.2 2.0 2.8 3.3 2.2 3 Fat % DMI increase NA No change NA No change 6.1% -0.16 No change 4.0% No change No change -0.18 No change No change No change 4 5 The types of luminaires used in each Table 4 study are also outdated. New LED luminaires last much longer than fluorescent or metal halide luminaires and are inherently more efficient. New technology in the area of lighting controls has also made LED luminaires more efficient and capable of doing more with less extensive wiring. In addition to the studies highlighted in Table 4, which were conducted under very controlled conditions, other studies have been conducted on functioning dairy farms. These farms were willing to implement LDL and be the subjects of case studies. Below, three different cases will be described and evaluated. 10 2.1 Fiscalini Dairy Farm, Modesto, CA Fiscalini Dairy Farm is a large operation that has 540 acres of land and 2,800 dairy cows (Fiscalini Cheese Company, 2017). Scientists from the University of Arizona in Tuscon performed a case study on a small group of 158 milk cows from that farm, which effectively simulates a much smaller dairy operation. These cows were divided into two groups based on the number of calves they had birthed: 60 first-time mothers (primiparous) and 98 mothers who had already borne at least one calf (multiparous). Twenty days after giving birth, 30 primiparous and 49 multiparous cows were placed under LDL conditions, and the other 79 cows were placed under normal lighting conditions (VanBaale, Armstrong, Etchebarne, Mattingly, & Fiscalini, 2005). The LDL group of cows were exposed to 17 hours of daylight that was supplemented to keep the lighting levels above 161.4 lux or 15 foot-candles (fc). To convert lux to foot-candles, divide the lux value by 10.76. During the night, the LDL group was exposed to light that was less than 53.8 lux (5 fc) in intensity for seven hours (VanBaale et al., 2005). These values fall right in the center of the recommended exposure times and intensities for LDL systems (Thomas et al., 2017). The normal light (NL) cows were exposed to an average of 12 hours of light and 12 hours of darkness per day. To ensure consistency in the study, light intensities were measured biweekly in the LDL free-stall barn (VanBaale et al., 2005). During the 16-week study, there was no statistical difference in milk production between primiparous cows exposed to LDL and those exposed to NL (P=81). However, multiparous cows exposed to LDL did produce a statistically significant (P<0.01) 3.6 kg/day more milk than NL multiparous cows. DMI was only measured as a group as opposed to milk 11 which was measured per each individual cow. There was little difference in DMI between any of the pens, but actual DMI for each cow was not known (VanBaale et al., 2005). This case study is important since it found that LDL only worked in multiparous cows, and the galactopoietic effects were greater than any of the laboratory studies shown in Table 1. There was no explanation given for this result. The lack of increased DMI was puzzling, since increased production means there are more nutrients being allocated for milk secretion, but there could be several explanations for this. The most obvious possible explanation could be that due to exposure to LDL and subsequent endocrine system changes, the bodies of the LDL cows were simply utilizing their nutrient intake more efficiently. This could allow for more nutrient allocation to milk production without impacting the nutrition intake of other systems (Dahl et al., 2000). The other reason could have to do with the design of the study itself, since this study only lasted into 16 weeks of lactation. Therefore, DMI could have increased later on after the cows’ metabolism adjusted to the additional milk output. Another shortcoming of this case study is that it did not discuss the type of lighting used or the lighting control system. From the way it is worded, it seems that the lights in the LDL barn were simply turned on for 17 hours, then most were shut down for 7 hours of nearly complete darkness. There is no indication that sunrise or sunset was simulated with dimmable light sources. In addition, it does not seem to have permanent photo sensors in the barn to constantly measure light levels in real time, since the authors stated that they took measurements two times per week. Therefore, if there was an exceptionally cloudy day, the actual lighting levels in the barn could have been lower than the desired 161.4 lux (15 fc). Type 12 of lighting is also very important to know since lights vary in energy efficiency, dimmability, start-up time, and adaptability for use with control systems. 2.2 Large Commercial Dairy, NY Scientists from Cornell University’s Cooperative Extension conducted a LDL study on a 1000-cow commercial dairy farm in NY that milks the cows three times per day. They used two different test barns that each contained between 280-310 cows. Half of the cows in each test barn were multiparous and half were primiparous. The control barn only contained 155 multiparous cattle. The two test barns were either illuminated with T8 fluorescent lamps or all LED lamps, and were set to provide an 18-hour light period and a 6-hour dark period every day. The control barn was lit with T8 fluorescent lamps that operated for 12 hours per day. The lighting levels that were achieved ranged from 114-206.6 lux (10.6-19.2 fc) (Eiholzer & Capel, 2015). During the roughly 300-day study, there was no statistical difference in milk production between test and control cows. Per the previous study, the lack of production improvement in primiparous cattle is no surprise, but in this case, the multiparous cattle also showed no improvement. There could be a variety of reasons for this, one of which is improper implementation of LDL, which was mentioned by the authors in the conclusion. The key to getting galactopoietic effects from LDL is that the cows must be exposed to 6-8 hours of constant darkness. This is a challenge for large dairy farms that milk three times per day, since cows can be stuck in holding pens or the milking parlor during the times they should be in the darkened free-stall barn. In this study there was another issue: two of the cattle waterers froze 13 in one of the test barns which may have limited the water intake of the cows in that barn (Eiholzer & Capel, 2015). There are some possible improvements in the design of this experiment that could have made it more successful. Compared with the first study in California, the target lighting levels were lower for the LDL barns. The Fiscalini Dairy study kept light levels above the recommended 161.4 lux (15 fc), whereas this study only kept light levels above 114 lux (10.6 fc)—an older recommendation—which could have contributed to the lack of production gain. Another issue was that the control barn only contained multiparous cattle, which generally produce more milk than primiparous cattle under non-LDL conditions (Eiholzer & Capel, 2015). There is also no indication that there was a system in place to take light measurements in real time, and then adjust light levels through a control system based on those measurements. It only seems that light levels were measured at the beginning of the study to ensure proper lighting. Since the lights were on timers and constantly illuminated, this is as much of an energy savings issue as it is a galactopoietic issue, since during bright days the lights most likely could have been dimmed to still achieve the minimum required lighting levels for LDL. Another issue was that the LED lamps experienced water damage, condensation, and algae growth, which also could have affected their output. This highlights the importance of using light luminaire that are watertight and rated for use in wet conditions (Eiholzer & Capel, 2015). Although this study did not yield any quantitative galactopoietic effects, it did yield some qualitative effects that were highlighted by the dairy manager. He said that the quality and distribution of the light in the test barns was much better than the control, which improved the working conditions for the cow handlers. The better light distribution enabled the handlers 14 to work more efficiently and better see the condition of the cows, which led to quicker treatment of injuries and quicker identification of cows in heat (Eiholzer & Capel, 2015). 2.3 Wing Acres Dairy, Bellevue, MI Wing Acres Dairy is a small dairy farm that milks about 100 cows two times per day. Researchers from Michigan State University’s Department of Biosystems and Agricultural Engineering implemented an LDL project on Wing Acres Dairy with grant funding from the Michigan Energy Office in 2013. Since this was a small farm with minimal barn space, a separate control group could not be run simultaneously with the test group. Instead, energy use data and milk production data were collected for 2013, which would serve as a control. The manager of Wing Acres Dairy agreed to not make any herd or schedule changes during the entire period of the study. This project had a greater focus on the equipment used to provide the LDL to the cows compared to the previous two studies mentioned. This system used weather-proof LED luminaires, which avoided the problems encountered in the NY LDL study. In addition, these LED luminaires could be dimmed by varying the voltage supplying the luminaire, which means less energy is consumed when the lamps are dimmed. The system also had a dimming controller and light level sensors that measured light in real time at the cow’s eye-level, unlike the two previous studies. This allowed the lights to be dimmed or brightened based on the amount of sunlight that was lighting the barn, which can save energy and ensures that there is always 215.2 lux (20 fc) of light evenly distributed throughout the barn during the 16 hour light period. To account for short temporary periods of reduced light, such as a cloud briefly passing 15 over the sun, the control system was programed to only make light increases after at least 15 minutes of measured light levels below 215.2 lux (20 fc) (Thomas et al., 2017). Since the LEDs were dimmable, they were not simply turned ‘on’ or ‘off’—they were slowly transitioned during a half hour from low output to high output and vice versa to simulate dawn and dusk, another unique aspect of this project. This prevented the cows from becoming startled from suddenly turning the lights ‘on’ or ‘off.’ To ensure the barn never had more than 53.8 lux (5 fc) during the dark period, any light source external to the barn but in line of sight was shielded from the barn (Thomas et al., 2017). The control system was programmed to account for a variety of factors. Since no control system on the market was deemed adequate by the researchers, they used three different systems that were programmed to communicate with each other. To provide the best light distribution possible, each LED luminaire in the barn was individually controlled by the control system, which could independently adjust each luminaire’s light level based on the illumination sensors in the barn. The component that controlled the 16 hour light period had a battery back- up and a real-time clock, with daylight savings time turned off to avoid disturbing the circadian rhythm of the herd (Thomas et al., 2017). At two years, this study was much longer than the previous two. In the first year there was a 6.74% increase in milk production, and in the second year there was a 7.69% increase in milk production from the original yield. The average increase in milk production per cow during this study was 2.0 kg/day. However, there was no energy efficiency improvement from the system, due to the fact that the lights were lit for longer periods of time compared to the previous luminaires they replaced that were not used for LDL. 16 In addition to the increased milk production, the farm manager also noticed some qualitative changes to the herd, similar to the NY farm. Since the light was much more evenly distributed when the barn was equipped with LDL, the cows tended to spread out more, which reduced crowding and bullying between the cows. The cows were also more inclined to lay down, seemed calmer, and did not startle as easily. In addition, the cows were cleaner and had less ticks (Thomas et al., 2017). One area this study lacked was the separation of data for primiparous and multiparous cattle that was included in the first two studies. Since this was not done, there is no way of knowing which group contributed the most to milk production improvements, or if one group, such as the primiparous cattle, actually had no milk production increase. This information could be useful, since it would help farmers looking to implement LDL on their farms to prioritize which groups of cattle should be exposed to LDL for the greatest milk production increase. 17 3. Research Objectives Given the three projects previously reviewed and the positive and negative aspects of each, the goal of this project is to add to the knowledge base regarding the energy consumption, control schemes, possible galactopoietic effects, and economic feasibility of LDL on a large commercial dairy farm. To achieve this goal, the following objectives have been identified:  Design and implement an LDL system with dynamic controls on a large commercial dairy farm with three milkings per day and determine if an increase in milk production can be achieved.  Quantify the energy savings associated with implementing LDL with LED luminaires and dynamic controls compared to implementing LDL with metal halide HID luminaires or LED luminaires without dynamic controls  Perform a basic economic analysis comparing the cost of using LDL to increase production versus adding additional cows to the herd 18 4. Materials and Methods The following section will cover the details of the project. This includes funding, research site, equipment, installation, and data collection. 4.1 Funding Funding was provided through the E-Challenge Competition by DTE Energy and the Engineering Society of Detroit (ESD). The E-Challenge focused primarily on energy efficiency and not on additional benefits of incorporating energy efficiency measures. Applicant rules included that the project “integrates multiple existing technologies in new ways,” and must have a projected 8-year return on investment for energy reduction. Preference was also given to applicants who utilized “Michigan based technologies and businesses” (DTE Energy & Engineering Society of Detroit, 2017). This presented a challenge, since implementing long day lighting (LDL) usually means adding more luminaires to dairy barns, increasing energy consumption. Therefore, the team decided to present the project to the E-Challenge Competition in a different way. The case was made that since past attempts at implementing LDL in the literature usually involved using less efficient light sources such as metal halide (HID) or fluorescent luminaires, new research was needed to quantify the energy savings from using more efficient LED lighting combined with dynamic controls to implement LDL. The E-Challenge accepted this approach but stipulated that the lighting company chosen to provide the LED luminaires would also be required to develop lighting layouts for the barns using HID luminaires that would provide light levels and distribution similar to the LED lighting layout. They also required that all lights and control systems be listed on the Design Lights 19 Consortium (DLC) website. DLC requirements for lighting and control systems are listed in Appendix A. A large part of this study involved the quantification of the energy saving benefits of using LED luminaires with dynamic controls. This is because the funding source, DTE Energy, was primarily interested in the energy savings associated with lighting and controls since those are easy to quantify in the short period of time before the team had to present data to DTE Energy and Engineering Society of Detroit. DTE was also interested in using the comparative energy savings from this research as a guide in developing energy savings incentives for dairy operations in their service area. Quantification of the implemented system’s energy consumption is also important for the economic analysis of the project. 4.2 Research Site The research site was chosen based on its location and size. Other important considerations included the owner’s willingness to accommodate research and its associated disruptions and guests. One component of this was an agreement was that the farm owner makes no major changes to the herd for the course of the study. This includes major additions to the herd. The owner also had to be willing to cover the installation costs of the lighting system. The farm chose was Car Min Vu Farms in Webberville, Michigan. Located 21 miles from Michigan State University’s main campus, CMV Farms is a Grade A dairy owned by the Minnis family. All of CMV’s milk is sold to Michigan Milk Producers Association (MMPA), which tracks daily farm milk production quality and quantity. CMV milks approximately 1,100 Holstein cows at its Webberville location and houses the cows in two large free-stall barns. Figure 2 shows an aerial view of the two main barns and their connectors (highlighted in white) to the parlor 20 building. Each barn is segregated into 4 sections, each housing different ‘groups’ of cows based on milk production level, days in milk, and age after first freshening. Cows are moved through different groups throughout their lives. At the start of the project in 2017, the west barn housed the high producing cows (Groups 1-4), and the east barn housed the low producing cows (Groups 5-8). High producing cows are generally within the first 120 days of lactation, and/or are younger than the low producing cows. As cows’ milk production decreases, they are moved to Groups 5-8. Groups 2 and 4 are almost exclusively primiparous (first-time mothers), whereas Groups 1 and 3 are high producing multiparous (mothers who have had at least two calves). Groups 1 and 2 have 170 cows each, while Groups 3 and 4 have 130 and 90 cows, respectively. Figure 2—Satellite view of barns (LDL areas in white borders; North at top of image) The milk parlor uses equipment by DeLaval and milk data is recorded and managed through DelPro, a software offered by DeLaval. As cows enter the milk stalls in the parlor, their ear tags are scanned. Each stall in the parlor is equipped to measure individual milk production 21 for each cow during milking. However, the DelPro program has its limitations. Cows with missing ear tags do not have their milk weights recorded, and it randomly will stop recording a cow’s milk yield while the cow is in the parlor. This creates a discrepancy between the recorded yield data on the DelPro database and the actual milk yield data available from MMPA. The farm has slowly been repairing the failing individual milk meters. This discrepancy is addressed in a later section. Figure 3—Milk parlor cellar showing the individual metered lines from each stall In November 2018, the farm expanded despite the agreement between the owner and the team. There is now another location, called Car Min Vu West, which started housing the low producing cows, which were originally Groups 5-8, but are now Groups 9-12. Groups 5-8 in the east barn are now an entirely new set of purchased cows that are different in age and days after freshening. 4.2.1 Barn Architecture Appendices C-1 through C-3 illustrate the different architecture of the barns. The west barn exclusively uses laminated veneer lumber (LVL) high clearance scissor trusses. The smaller 22 east barn has LVL high clearance scissor trusses on its southern half, and low clearance webbed coffer trusses on its northern half. The east barn has four identically sized pens. The west barn has differently sized pens. The two northern pens are much larger than the two southern pens. 4.3 Equipment Based on the project needs and the stipulations laid out by the DTE energy, the team developed a list of requirements for lighting and control systems. These requirements are listed in Appendix B. All possible lighting and control systems were evaluated against these requirements 4.3.1 Lighting System and Layout The lighting system selected was provided by Everlast Lighting in Jackson, MI. Everlast assembles their luminaires in their Jackson facility, which fulfilled the requirement by DTE Energy that some components be made in Michigan. Everlast had the best price of the options evaluated and was also the most willing to work with the team to make modifications to their luminaires to accommodate the needs of the project. The research team worked with Everlast to develop a lighting layout for each barn that would provide the minimum amount of light required for LDL. The software used by the team and Everlast to develop the lighting layouts was Visual Lighting 2017. Based on their experience, designers from Everlast recommended four rows of luminaires in each barn and gave the team a starting point for the number of luminaires per row. The team then ran different layouts through the software to optimize light distribution while conserving cost. The final lighting layouts are in Figures D-1 and D-2 in Appendix D. These figures were designed to be an 23 installation guide for the electrician who would install the luminaires, and additional drawings were also supplied to the electrician that showed the luminaires in relation to the trusses. Due to the different trusses in the barns, two different types of luminaires were needed. Since the center rows of lights were all well over 12 feet from the ground, 150-Watt high bay LED luminaires were used. The outer light rows in the west barn and southern half of the east barn were also over 12 feet above the floor, so the same 150-Watt high bay luminaires were used in those locations as well. The outer rows of the east barn’s northern half presented a challenge, since the bottom of the webbed coffer trusses were 12 feet from the ground. The team originally specified using the 150-Watt high bay luminaires in these rows as well, but then discovered a problem—since the webbed coffer trusses are only on 4-ft centers as opposed the 12-ft centers of the scissor trusses, the bottom of the coffer trusses would block some of the light of the high bay luminaire. The lens of the high bay could have been lowered and mounted flush with the bottom of the coffer truss, but the models revealed that this would result in high intensity ‘hot spots’ directly under the luminaires, with low intensity ‘cold spots’ that were below the required light intensity in between the luminaires. After consulting with Everlast, the team decided to use 60-Watt canopy luminaires for the outer two rows in the northern half of the east barn. Canopy luminaires do not have the focused beams of light that are characteristic of high bay luminaires and are therefore ideal for use in lighting areas with low clearances. As can be seen in Appendix Figure D-2, more canopy luminaires were need per linear foot to achieve the required light levels compared to high bay luminaires. 24 The team had to work with Everlast engineers to incorporate the control modules from Synapse Wireless Inc. (discussed later) in each luminaire. The high bay luminaires required a custom designed weatherproof box to house the controller, whereas the canopy luminaires already had an open space above the light shield that could accommodate the Synapse controller using custom stamped brackets made by Everlast. The luminaires were delivered mostly assembled. The team only had to mount the wireless antenna and affix the MAC address label to the outside of each luminaire. Bird spikes were also added to the flat upper surface of the high bay luminaires. To make the MAC address labels easier to read from the ground, the team printed larger versions for the high bay luminaires’ MAC addresses. Figure 4 shows a completed high bay and canopy luminaire. Figure 4—Left: dimmable high bay luminaire with weatherproof box, bird spikes, and antenna; Right: partially disassembled canopy luminaire showing the DIM10-250 controller (top) and dimmable driver (bottom) 4.3.2 Control system The team evaluated several different control system options but found very few that could meet the requirements of the project and funding partners’ requirements. Some control systems met almost all of the requirements of the project and the project’s funding partners, but they had to be purchased together with a compatible lighting system from the same 25 supplier. This disqualified them since no components would be made in Michigan. There were a few options that could be purchased separately, but they were not DLC listed. The only system that met all of the requirements was from a company called Synapse Wireless in Huntsville, Alabama. The control system was compatible with the Michigan-made LED luminaires from Everlast, DLC listed, and met all the project requirements. One of the control system’s required capabilities was daylight harvesting (DLH). DLH is the ability of a control system to maintain consistent light levels in the barn as available natural light changes. Even during lighted day in central MI, there will be areas of the barns that will not have adequate light levels, so artificial light input is needed throughout the day. Instead of having all luminaires energized at their maximum power draw to ensure minimum light levels are met, they can be dynamically dimmed based on the levels of natural light available within the barns. There are two kinds of DLH control systems: closed loop and open loop. A closed loop system uses photocells that monitor the light levels inside the barn and dynamically dims and brightens the LED luminaires based on the actual light levels inside the barn. This method is the best for maintaining light levels and automatically adjusts for deteriorating light output from the luminaires as they age. An open loop system also uses photocells, but they face outside the lighted area and only monitor the natural light levels that are coming into the barn. The open loop method requires much more calibration and monitoring during implementation to ensure proper light levels in the barns. Synapse originally advertised their system as capable of closed loop DLH, which suited the needs of the project. However, before the luminaires were delivered, Synapse changed their DLH from closed loop to open loop. The team had already purchased photocells that were 26 compatible with Synapse’s advertised closed loop capability, but they would not work with open loop. Open loop DLH photocells can be exposed to direct sunlight and therefore require higher light input limits than closed loop photocells. Synapse supplied compatible photocells, and the team had to work with Synapse engineers to determine the placement of each photocell. Each photocell had to be connected to a Synapse AIM-121 sensor controller. AIM- 121s are capable of handling up to two different photocells. After consulting with Synapse engineers, it was determined that ten photocells were needed to control the luminaires in both barns: four in the west barn (west side: 1. east side: 2, peak: 1), five in the east barn (west side: 2, east side: 2, peak: 1), and one in the west barn milking parlor connector. The photocells on the barn peaks controlled the center two rows in each barn, and the number of photocells on the barn sides was determined by the architecture of the buildings. These architectural characteristics are illustrated in Figure 2. For example, the west side of the west barn has no attached structures, so only one photocell was needed for the entire west row of luminaires. The west and east sides of the east barn each have attached structures, so two photocells were needed for each side to accommodate shadowing throughout the day. All the luminaire and photocell controllers wirelessly communicate with a central gateway located in the west barn. This gateway connects to the farm’s internet router through an ethernet cable. To program light behavior, the gateway’s IP address must be entered into an internet browser on one of the farm’s computers. This brings up the login to the SimplySnap program. Once logged into SimplySnap, a census can be run, which automatically detects all energized luminaire controllers. The MAC addresses for each controller/luminaire combination 27 then appear in SimplySnap, and they can be renamed and grouped into different ‘Zones.’ ‘Scenes’ can then be created, which attribute a behavior to selected zones. The three different scene behaviors are: on, off, and dim. Dim levels can be set at one percent increments ranging from 0-100%. These scenes can then be called on the ‘Events’ calendar and set on repeat. Since Synapse did not offer a ramp up/ramp down feature for a simulated sunrise/sunset, fifteen separate scenes were created with varying dim levels. These scenes change light levels in one- minute increments for a fifteen-minute simulated sunrise/sunset each day. The ‘long day’ that was used was based on the sunrise/sunset of the longest day in central Michigan. At 5:59 a.m., only the center safety lights (non-dimmable 60-Watt canopy LED luminaires) are illuminated in the main barns. Three of the six high bay luminaires in the parlor connector are also energized at their lowest light output for safety. At 6:00 a.m., all of the dimmable luminaires in each barn are energized at their lowest output level, and sequentially get brighter in one-minute intervals until 6:14 a.m., when they reach full brightness. The center safety luminaires then turn off. When the sun rises and starts to provide natural light, the daylight harvesting system takes control of the luminaires and dims them as needed. On cloudy days, most of the luminaires are operating at 40-50% of their maximum output, whereas they only operate at 10% during a clear sunny day. At 9:15 p.m., the center safety luminaires turn on in the barns, giving the employees a five-minute warning that the main luminaires are about to shut down. At 9:20 p.m., all dimmable luminaires sequentially dim at one-minute intervals until they shut down at 9:35 p.m. Figure 5 shows the timeline of the LDL period, as well as the milking schedules of Groups 1 and 3. As can be seen, the actual light period is 15 hours, but each group 28 is milked once during the dark period in an illuminated parlor and holding area, so each group gets 16 hours of light during a 24-hour day. This is one of the challenges of implementing LDL on a farm that milks 3 times per day, since the cows do not always get the 16 hours of light during one period of time. Figure 5—LDL schedule with milking schedules for Groups 1 and 3 4.4 Installation Installation took place from May to September 2018. The farm owner covered the luminaire installation labor costs. There were some challenges with the installation which resulted in the five-month installation time. During the hottest weeks of the summer, almost no work was done since the ventilation systems in the barns could not be shut down. On cooler days, the electrician installed new circuit breaker panels for the light and control systems in each barn, so future work would not interrupt the ventilation systems. Running the wires and conduit for the luminaires required a construction lift, which was originally leased or borrowed. However, after some reliability issues occurred, the farm owner purchased his own 29 construction lift that the electrician used to finish the project. The team continued to use the lift on occasion when a faulty controller or luminaire needed to be replaced, since there was an additional issue with component failure in about 30 of the luminaire controllers. 4.5 Data Collection There are two different sets of data that needed to be collected: energy use and milk production. The energy data is the most important for this engineering project and is used to characterize the energy consumption of a dynamically controlled LDL system that utilizes LED luminaires with DLH. The milk production data is secondary since this is not an animal science study but will be necessary for any future animal science research that uses the implemented system to study the effects of LDL on cows. 4.5.1 Energy Data Collection Energy data was collected through a data acquisition (DAQ) system that monitored the power draw and energy consumption of the lighting circuits in each barn. The main DAQ server was an Acqusuite EMB from Obvius LLC and was placed in the west barn and was connected to the farm’s internet router through an ethernet cable. Figure 6 shows the DAQ system with its components in each barn. 30 Figure 6—West and east barn DAQ layout Three components were attached to the main DAQ server. The power meter measured power draw and energy consumption of the lighting circuits in the west barn, and the Modbus Flex input/output communicated with several photocells placed throughout the west barn to monitor actual light levels inside. The wireless Modbus communicated with an identical unit in the east barn that also had a power/energy meter and an array of photocells communicating with it. The light and energy data are collected in real time and recorded on a remote server hosted by Obvius LLC. This server can be remotely accessed, and the data can be downloaded as an Excel spreadsheet. Energy data is used to characterize the actual energy consumption of a dynamically controlled LED LDL system compared to an identical system without dynamic controls and a system utilizing HID luminaires instead of LED luminaires. This data is also used to calculate the annual energy cost of the system. 4.5.2 Milk Data Collection At the start of the project, the original plan for milk data collection was to use the overall farm milk production data provided by MMPA, the sole buyer of Car Min Vu Farms’ milk. 31 However, since the operation significantly changed the cow population of the east barn, contrary to the agreement at the start of the project, two methods have been used to monitor milk production. The first method is a comparison of MMPA data for the whole farm to a two-year average of milk production. This compared monthly and total milk production for a one-year period starting in October 2018, the first month of LDL implementation, to monthly average production for the previous two years. It also compares the total production between these two time periods. This data was collected, but the limitations of this method are discussed in greater detail in the results. Due to the changes in the herd, a second method was added later in the project to verify that Groups 1 and 3 in the west barn were experiencing an increase in milk production. Whole farm data may not tell the whole story of what is happening since so many new cows were added to the east barn. Attempting to quantify all of the influences on milk production associated with moving cows is outside the realm of this engineering study, but could be the subject of an animal science study. As mentioned earlier, there is a discrepancy in the data collected by DelPro and the data collected by MMPA. DelPro data is significantly less than MMAP data for the whole farm. Therefore, to individually track group data, average monthly production was used instead of total monthly production. Milk production data automatically deletes from the farms’ computer after it reaches six months. To ensure no data gets deleted, milk production data is downloaded monthly from the farm’s computer. There is currently no way to remotely download production data from DelPro, so a person has to visit the farm each month and download the data to a 32 portable memory drive. June through September 2018 data (before LDL) is available for Groups 1 and 3 and is compared to June through September 2019 average production (after LDL). 33 5. Results/Discussion This section contains three parts: energy data, milk production data, and a simple economic analysis combining the energy and milk data to quantify the actual implementation cost, the simple payback, and a comparison of LDL to adding more cows to the herd. 5.1 Energy Data Table 5 shows the energy consumption of several different lighting layouts for the west and east barns for a 315-day period of time from September 2018 through July 2019. Column 2 shows the energy consumption of the T-8 fluorescent luminaires that were originally in the barns. Since almost one third of all the 28-Watt T-8 lamps in both barns were not working, Column 2 represents the adjusted energy use of both barns as if all lamps were in working condition. This is important since after sunset, most areas of the barn had less than 1 footcandle (fc) (11 lux) of illumination, which can lead to dangerous working conditions. Columns 3 and 4 represent two theoretical values. The HID energy consumption in Column 3 represents the theoretical energy consumption of a comparable LDL system that used HID luminaires. Column 4 shows the energy consumption of the implemented LDL system without the dynamic controls, and Column 5 shows the actual energy consumption of the implemented system with daylight harvesting (DLH). Table 6 shows the percent energy savings of the implemented system compared to a theoretical HID layout and the implemented system without the dynamic controls (featureless LEDs). 34 Table 5—Energy consumption of original, theoretical, and implemented systems Barn West Barn East Barn Totals: 1 Original T-8 use Theoretical HID Featureless LED (kWh) 14,913 10,432 25,345 2 (kWh) 87,885 58,590 146,475 3 (kWh) 56,950 44,462 101,412 4 Dimmable LED with DLH (kWh) 27,565 21,040 48,605 5 Table 6—Percent savings of the implemented LDL system compared to the theoretical HID layout and a featureless LED layout Barn West East Totals: 1 % Diff HID % Diff Featureless 69% 64% 67% 2 52% 53% 52% 3 One of the objectives of this study was to evaluate the energy consumption of a long day lighting (LDL) system utilizing dynamically controlled LED luminaires compared to comparable systems that use HID luminaires or LED luminaires without dynamic controls. The theoretical HID layout developed by the team and the LED luminaire manufacturer, Everlast Lighting, on Visual 2017 utilized 72 250-Watt luminaires in the west barn and 48 250-Watt luminaires in the east barn. The theoretical energy consumption of this system is listed in Column 3 of Table 5. Column 2 of Table 6 shows that the implemented LED system with dynamic controls consumed 67% less energy than the theoretical HID system. This is because LED luminaires are inherently more efficient than HID luminaires, requiring less power for the same lumen output. The daylight harvesting (DLH) capability on the implemented system also added to the increased efficiency. The difference between Columns 4 and 5 of Table 5 shows the difficulty in predicting energy consumption of a dynamically controlled LDL system. Column 4 was calculated with the assumption that the luminaires would be illuminated at their maximum output for the entire 35 light period. This is a simple calculation, but it does not come close to the actual energy consumption of the implemented system. Column 2 of Table 6 shows that the implemented system with DLH consumes 52% less energy than the same system without DLH. The power draw graph in Figure 7 shows the actual west barn power draw of the implemented system, compared with a featureless system without dynamic controls and a theoretical HID system. As can be seen during sunny days, the lights are operating at less than 10% of their rated power draw, only ramping up to greater power draw levels when the amount of available natural light decreases, as can be seen on the morning of September 20. This causes the large energy savings over an identical system without dynamic controls. The featureless system and theoretical system would be drawing their full rated power during the ‘day’ period. Figure 7—Power draw graph for the implemented dimmable LED layout, theoretical featureless LED layout, and theoretical HID layout The addition of dynamic controls to implement DLH added about $5,000 to the total cost of the system. These extra components were discussed in the previous section dealing with the control system. DLH has saved an average of 168 kWh/day of electricity compared to a 36 featureless LED system and 311 kWh/day compared to a HID system. Car Min Vu Farms currently pays an average of $0.12/kWh of electricity, so this leads to an annual savings of over $7,000. Table 7 shows the energy and cost savings of the implemented system compared to a HID system and a featureless LED system. Compared to the existing fluorescent lighting system, the implemented LDL system consumes an additional 26,952 kWh per year, which is an annual cost increase of about $3,200 at $0.12/kWh. In total, this system costs the farm about $6,800 per year in energy costs. Table 7—Cost savings of implemented system Implemented System Compared to: HID System Featureless LED System * Calculated at $0.12/kWh 1 5.2 Milk Data Average Savings (kWh/day) Average Savings Annual Savings 310.7 $ 37.28 $ 13,609 167.6 $ 20.12 $ 7,343 ($/day) 2 3 4 During the beginning of this project, the intension was to track bulk tank production data for two years after implementation of LDL and compare this production to the previous two years of production before implementation. Bulk tank data from the Michigan Milk Producers Association (MMPA) not only shows total weight sold, but also offers an analysis on the quality of the milk. MMPA tank analysis shows somatic cell count (SSC), protein content, and butterfat content. Milk is more valuable if it has a low SSC and high protein and butterfat content. This would have added an additional aspect to this study to show if there was a change in the quality of the milk after implementing LDL. However, as mentioned before, the farm owner purchased over 400 additional cows and leased another location. He then replaced the cows in the east barn with the new cows he 37 purchased and moved the east barn cows to the new farm location. His new herd management strategy was to move cows that were confirmed pregnant and on the downward side of milk production to the new location, since confirmed pregnant cows do not require as much hands- on management. This changed the population of cows on Car Min Vu Farms’ original Webberville location, complicating the original plan to use bulk milk tank data as a comparison tool. Using overall tank data can also add variables that are hard to control, such as overall milking herd size and the overall age of the milking herd. Figure 8 shows the overall tank data for the Webberville location. It compares monthly tank weights from October 2018 through September 2019 to a two-year average over the same months. As can be seen, production dipped below the two-year average starting in April 2019. Since November 2018, the herd is typically ‘fresher’, meaning there are more cows than usual that are early in lactation. Cows reach peak lactation at 60 days from the freshening date, and then their production slowly tapers off. The farm has a larger total number of cows, and the cows closer to drying up and confirmed pregnant are now transferred to the new location, resulting in a ‘fresher’ overall herd at the Webberville location. Groups 5-8 have particularly been affected, since these groups used to house confirmed pregnant cows that passed peak production as well as cows nearing the end of their production cycle. Groups 5-8 now closely mirror Groups 1 and 3 in the west barn, and are now much ‘fresher’ than the cows that were housed in Groups 5-8 at the start of the project. There are also more primiparous cows in the east barn, which typically do not produce as much milk as multiparous cows. 38 k l i m f o s d n u o P 2900000 2700000 2500000 2300000 2100000 1900000 1700000 1500000 Oct Nov Dec Jan Feb Mar Apr May June July Aug Sept 2-yr Avg Oct to Sept Oct 2018 to Sept 2019 Figure 8—Bulk tank monthly milk production from October 2018 to September 2019 compared to a 2-year average There has been a large increase in fresh cows each month during the first year of LDL compared to the two-year average before LDL implementation, shown in Figure 9. Figure 10 combines monthly fresh cows with monthly milk production. Fresh cows are introduced into the herd 3 days after freshening. Approximately 60 days after freshening the cows reach peak production and maintain a relatively steady production for another 60 days. During the one- year LDL period, there was a 27% increase in number of fresh cows introduced to the herd compared to the previous two-year average over the same period of time. The increased numbers of fresh cows introduced to the herd causes an increase in herd production 60 days after freshening, but also causes a temporary dip in overall milk production for at least the first 30 days after freshening. It appears that the large number of fresh cows introduced each month negates the increased production of the cows that were freshened 60 days earlier. These are a few of the challenges of using whole farm milk tank data for this farm to track milk production. 39 250 200 150 100 50 0 Oct Nov Dec Jan Feb Mar Apr May June July Aug Sept Oct-2018 to Sept-2019 2-year avg Oct-Sept Figure 9—Monthly number of cows freshened from October 2018 to September 2019 compared to a 2-year average 2800000 2600000 2400000 2200000 2000000 1800000 1600000 1400000 1200000 1000000 2-yr Avg Oct to Sept (milk) Oct 2018 to Sept 2019 (milk) 2-yr Avg Oct to Sept (fresh cows) Oct Nov Dec Jan Feb Mar Apr May June July Aug Sept Figure 10—Monthly bulk tank data compared to monthly freshened cows 350 300 250 200 150 100 50 0 s w o c h s e r f f o r e b m u N ) s b l ( t h g e w k l i i M A better way to track milk production is to focus on smaller groups of cows that have similar characteristics over time. One of the difficulties of this project was the limited scope of time available to completion. Ideally, a control and test group would be monitored simultaneously, but in this case data from the previous year was used as the control. By the 40 time the team realized the farm management was changing the cows in the east barn from mid/late lactation to early/mid lactation cows, it was too late to recover individual group milk production data that was collected before June 2018, since the team was originally intending to use whole tank data during the LDL period and compare it to whole tank data for the previous two years. As mentioned above, whole tank data introduces more complications. For comparison purposes, individual group data is used for this project. Groups 1 and 3 in the west barn always contain 170 and 130 cows, respectively. These 300 cows are multiparous, high producing cows that are between 30 and 100 days in milk (DIM). Unlike other groups, these two groups have had the same type of cow in them throughout the course of the study. Therefore, comparing these groups’ average milk production before and after LDL implementation has the least amount of variability and should provide preliminary results that can be built upon in the future. In theory, each time a cow enters a milk stall its ear tag is automatically scanned and its milk yield is measured by the milking machine and recorded in pounds. This data is then stored for 6 months on a connected computer located on the farm. Milk reports can be generated that show each cow’s tag number, group number, and production weight every time it enters the parlor. Since cows are milked three times per day in Groups 1 and 3, there should be three entries in the milk report for each cow every day. Therefore, in a 30-day month, there should be 27,000 data entries. However, this is not the case. There are typically less than 20,000 data entries each month for the two groups. There are a few reasons for the discrepancy. One reason is that cows are being moved in and out of these groups every week, not just at the beginning and end of the month. Once a 41 cow reaches 100 DIM she is removed and sent to another group. A cow that has reached 30 DIM is moved into these groups on the closest herd sorting day, which is usually on a weekend. Another reason is that the ear tags that are scanned when the cows enter the parlor can fall off or be ripped off a cow’s ear, so the production data of that cow is not added to the group milk data. Cows in Groups 1 and 3 that become ill are moved to a hospital group in a different barn, so their milk production is not recorded during their hospital group visit. In addition, the milk measuring equipment on each stall in the milk parlor experiences frequent glitches. The machines stop recording milk data while the milk is still flowing, or simply do not record at all. This means a cow might only be recorded accurately once per day, and not recorded at all the other two times it was in the parlor. The herd managers on the farm have discovered a solution to this problem and are now implementing it when individual parlor milk meters fail. To handle these difficulties, a filtering program was written in Python which takes a month of cow data at a time from a milk report and separates out the cows that were recorded at least once per day for the entire month. This eliminates data from cows that were accidently put in Groups 1 and 3 and quickly removed, as well as cows that only spent part of the month in these groups. Figure 11 shows milk production data for Groups 1 and 3 for one year after LDL implementation. Each month shows an average weight that those groups produced each time they entered the milk parlor. Since they enter the parlor three times per day, average daily total production per cow is three times the weights listed for each month. However, as mentioned above, only individual group data from June-September 2018 and June-September 2019 are available for comparison. Figure 12 and Table 8 show data with comparisons for the last four months listed in Figure 11. 42 t i s i v r o l r a p / k l i m s b L 42 40 38 36 34 32 30 Oct Nov Dec Jan Feb Mar April May June July Aug Sept Figure 11—Groups 1 and 3 average production per parlor visit for the LDL period October 2018 through September 2019 LDL milk data June-September 2018 and 2019 t i s i v r o l r a p / k l i m s b L 41 39 37 35 33 31 29 27 25 June-Sept 2019 June-Sept 2018 June July Aug Sept Figure 12—Average production per parlor visit for Groups 1 and 3 (June-Sept) 43 Table 8—Average production per parlor visit for Groups 1 and 3 with statistics June July Aug Sept 1 Avg Yield (lbs) 27.9 27.9 31.5 34.4 2 2018 Std Dev 9.6 8.8 11.0 11.2 3 # Data Pts 7110 9483 4753 7065 4 Avg Yield (lbs) 39.1 32.2 33.4 38.1 5 2019 Std Dev 14.5 10.7 10.5 14.5 6 # Data t DF p Pts 13912 59.14 21020 < 0.00001 12557 31.88 22038 < 0.00001 19176 10.99 23927 < 0.00001 6107 16.46 13170 < 0.00001 7 10 8 9 As can be seen in Table 8, there is a statistically significant (p<0.00001) increase in average milk production using the t test for unpaired samples. The t test for unpaired samples was used since it is best suited for comparing change between members of a population during different periods of time. In addition, the unpaired t test is suitable for comparing populations with different numbers of data points, which is illustrated in Columns 4 and 7 in Table 8. ‘Data points’ are individual milk weight readings for a cow that has entered the parlor to be milked. Figure 12 shows a 17% increase in average monthly milk production for the four months listed. However, only half of June 2018 average parlor visit milk data was available, so it is impossible to tell if that number would be higher if all of June 2018 data was available. Since the exceptionally large difference between June 2018 and June 2019 could be more unreliable than subsequent months that had data available for every day, June data was dropped for the purposes of comparison. To calculate the percent increase in milk production over these three months, an average of the July-September 2018 production averages in Table 8 Column 2 was subtracted from the average of the July-September 2019 production averages in Column 5. This resulted in an estimated increase of about 3.3 pounds per parlor visit over those months for Groups 1 and 3. This translates into an estimated 10.5% increase in average monthly milk production when average production per parlor visit milk data for July through September 2018 44 and 2019 are compared. The calculated 10.5% increase is on par with the expected 5-9% increase in production that is suggested in the literature, although those studies were generally conducted on farms that milked their cows 2 times per day instead of 3 times. This estimated 10.5% increase in production was calculated based on data measured from July through September, which are months with relatively long lit days compared to the fall and winter months. If there were more data available, there should be a larger increase in milk production during the fall and winter months since the artificial lighting would be compensating for the naturally shorter days. This would also lead to an increase in energy consumption since the luminaires would be illuminated at their maximum output for a longer period. This calculated increase in production only reflects LDL’s possible effects on cows between 30 and 100 days in milk (DIM). Finally, since the data in Table 8 measures production per parlor visit, but does not necessarily capture all cows’ parlor visits each month, there could be additional variability in the data since cows may yield different production levels depending on what time of the day they are milked. 5.3 Financial Analysis The recorded energy consumption data of the implemented system is an additional cost to the farm, but the estimated 10.5% increase in milk production should increase revenue. Dairy science literature also projects that an increase in milk production will also cause the cows to increase their feed intake by 0.4 pounds of dry feed per pound of additional milk produced (French, 2000). The additional feed costs per hundredweight of milk produced that are outlined in Section 1.2 are combined with the increased annual energy costs of the implemented system and make up the total increased expenses. Whether or not there actually 45 is additional feed intake of 0.4 pounds per pound of milk produced is not entirely clear, which is shown in Table 4 of Section 2. Most of the studies listed in that table showed no increase in feed intake or were inconclusive. Part of the reason is that the effects of LDL on a cow’s metabolic mechanisms are not entirely known. The possibility exists that cows under LDL conditions use their available nutrients more efficiently and can therefore allocate more consumed nutrients for milk production. Previous studies on the effects of LDL on heifers showed that heifers exposed to LDL gained weight faster without eating more food, suggesting that they were utilizing their consumed nutrients more efficiently (Dahl et al., 2000). The studies in Table 4 may have also not been monitored for a long enough period of time to observe an increase in DMI, since previous animal science studies have found that any DMI increase possibly associated with LDL occurs after an observed increase in milk production, not before (Collier et al., 2006). Due to the variability in the literature and number of inconclusive studies in Table 4, the conventional dairy science feed numbers are used for additional milk production, since this represents the most conservative estimate for profitability. This is not an exhaustive list of expenses or potential benefits, since there may be other harder to quantify increased costs or benefits that are associated with an estimated 10.5% increase in milk production. However, the increased feed is well established in the literature, and the additional energy consumption of the new luminaires was directly measured. Since only Groups 1 and 3 were monitored, only the capital costs associated with the west barn are used in the financial analysis. In addition, the calculated 10.5% increase in production is only applied to the 300 cows in Groups 1 and 3. Table 9 shows the additional costs that are directly associated with this project in the west barn only. 46 Table 9—Additional costs and capital investment associated with this project (west barn only, Groups 1 and 3) Additional Feed (lb/yr) Additional Feed Cost ($/yr) Additional Energy Cost ($/yr) Total increased expenses ($/yr) Capital investment Installation cost 433,060 $ 68,424 $ 1,759 $ 70,183 $ 59,682 $ 12,600 1 3 4 2 5 6 Columns 1 and 2 of Table 9 quantify the additional feed intake and cost for the 300 cows in Groups 1 and 3 producing an estimated 10.5% more milk for an entire year. The average production per parlor visit in Table 8 Column 2 over July through September 2018 was used as a baseline average milk production that was then increased by 10.5% for this financial analysis. This yielded an average estimated additional 3.3 pounds of milk per parlor visit (9.9 lbs per day). Total increased expenses in Table 9 Column 4 are a combination of calculated additional feed costs for Groups 1 and 3 and measured energy consumption and costs for the west barn. Capital investment in Column 5 is the cost of all the equipment and materials needed to implement this system in the west barn. The equipment in west barn with the central control system and installation accounted for about 70% of the total project cost. The farm owner arranged the installation through his employees and associates, but the estimated installation cost for a system of this size in the west barn is about $12,600. Table 10 shows the increased revenue that comes from the estimated increased milk production. Milk prices have been very volatile over past few years. For most of 2018 through the first half of 2019, milk was selling for less than $16 per hundredweight (cwt). The last half of 2019 has seen a steep rise in prices, now approaching $19 per cwt. For the purposes of this study, a rate of $16 per cwt will be used. 47 Table 10—Estimated additional milk, increased revenue, increased profit, simple payback, NPV, and IRR from implemented LDL (Groups 1 and 3) Additional milk (cwt/yr) Total increased revenue @ $16/cwt Increased average Payback annual profit (yr) NPV Modified IRR Equivalent milk (cwt) 10,827 $ 173,224 $ 103,041 1 3 2 0.70 $409,895 4 5 49% 6 7,467 7 Average annual profit in Column 3 was calculated by subtracting the total increased revenue in Column 2 from the total increased expenses in Table 9 Column 4. This is the additional money the farm should be making from the increased milk production after subtracting the increased feed and energy costs. The simple payback period in Column 4 was calculated by dividing the capital investment cost in Table 9 Column 4 by the increased annual profit in Table 10 Column 3. Net present value (NPV) was calculated using a 5-year time period, since that is the length of the warranty on the Synapse luminaire controllers. The discount rate used was 2.25%, which is the federal loan discount rate and roughly the yield on a secured bond or time deposit. Modified internal rate of return (IRR) was used since the calculated IRR of 138% assumes an unrealistically high reinvestment rate of return. Modified IRR utilizes an estimated secured bond or time deposit interest rate of 2.25% and yields a more realistic 49% IRR. Column 7 shows the amount of milk that would need to be produced to pay for this system. This was estimated based on a milk price of $16/cwt and an estimated feed cost of $6.32/cwt. There is another cost savings that is not included in the financial analysis in Tables 9 and 10. This is the additional cows needed to match the estimated increased production in Groups 1 and 3. Table 11 shows the additional cows needed to match the estimated production increase from LDL. A cow located in Groups 1 and 3 is always between 30 and 100 days in milk (DIM) and 48 produces an average of 93.8 pounds of milk per day before implementation of LDL. After implementation of LDL, cows in Groups 1 and 3 produced an average of 103.7 pounds of milk per day. This daily 9.9 pounds in increased milk production per cow translates into over 1 million pounds of additional milk produced per year from the 300 cows in Groups 1 and 3. 32 additional cows producing milk at the original 93.8 pounds per day before LDL would be needed to match the additional 10,827 cwt of milk in Table 11 Column 1. Prices for dairy cows are volatile, but typically range from $1,000 to over $1,500 per cow. When Car Min Vu Farms recently purchased high producing cows, they paid about $1,100 per cow, so this is the number used to calculate the additional herd investment that was saved. Additional operational costs associated with 32 additional cows are hard to quantify for this study and are therefore not included in Table 11. Table 11—Number of cows needed to match the production increase of LDL Additional milk Additional cows (cwt/yr) saved Additional investment saved @ $1100/cow Additional feed costs saved ($/yr) 10,827 1 32 $ 34,779 $ 27,391 2 3 4 The addition of 32 cows also comes with increased feeding expenses. Table 9 outlined the increased feed expenses of existing cows that were becoming more productive. Adding additional cows also means that these cows must be feed their maintenance ration in addition to the 0.4 pounds of dry food per pound of milk produced. Typically, feed that supports milk production is 71% of a cow’s total ration while maintenance feed accounts for 29%. (Herdt, 2018; Jones et al., 2018). This is why making an existing herd more productive without increasing herd size is so beneficial, since the additional feed consumed directly relates to the additional milk produced, whereas adding cows also requires the additional basic maintenance 49 ration to keep the cow alive and healthy. Feed costs for a whole new cow are $8.85 per cwt of milk produced, including the maintenance ration. Column 4 of Table 11 illustrates the additional feed costs that were saved from using LDL instead of buying more cows. The annual feed costs for 32 additional cows would have been $95,815 annually at $8.85/cwt. This number was subtracted from the additional feed costs associated with LDL implementation and yielded an additional savings of over $27,000. There is also a positive environmental impact of using LDL to increase milk production instead of using additional cows to achieve the additional revenue. Dairy cows largely emit two greenhouse gasses, carbon dioxide (CO2) and methane (CH4). Since CH4 is over 25 times more harmful to the environment than CO2, CH4 emissions from cows are the focus of this comparison. There is a lot of variation in estimating the CH4 emissions from a cow, with some studies basing their emissions on the feed eaten by the cow while others take a total cow approach. Due to all of the variability between different studies, a whole cow emissions average from Penn State University is used for this comparison. The average annual CH4 emissions from one dairy cow every year is about 415 pounds (Hristov, Johnson, & Kebreab, 2014). The total CH4 emissions from an additional 32 cows is equivalent to 331,752 pounds of CO2, often referred to as carbon dioxide equivalent pounds (CO2e). Table 12 shows the environmental impact of 32 cows in terms of their CH4 and CO2e emissions. Table 12—Greenhouse gas savings Additional cows saved 32 1 Annual CH4 emissions saved @ 415 lbs/cow (lbs/yr) 13.121 2 Annual CO2e saved (lbs/yr) 330,752 3 Additional CO2e from LDL (lbs/yr) 22,855 4 Total CO2e saved (lbs/yr) 307,897 5 50 Since the implemented LDL system uses more energy than the original lighting, it is responsible for increased emissions. Using the Environmental Protection Agency’s emissions calculator, the additional annual 14,660 kWh of electricity translates into an additional 22,855 pounds per year of carbon dioxide equivalents (USEPA, 2018). Subtracting this number from the annual CO2e saved from not purchasing 32 additional cows, the total annual CO2e saved is 307,897 pounds. 51 6. Conclusions/Future Work The main goal of this project was to add to the knowledge base regarding the energy consumption, control schemes, possible galactopoietic effects, and economic feasibility of LDL on a large commercial dairy farm. The first objective to reach this goal was to design and implement an LDL system with dynamic controls on a large commercial dairy farm with three milkings per day and determine if an increase in milk production can be achieved. The farm used was a large commercial dairy farm like the one used in the New York case study. The system implemented was modeled after the one implemented on Wing Acres, but within the constraints set forth by the project’s funding partners. The LDL system on Car Min Vu Farms provided 15 hours of light and 9 hours of darkness. Generally, an LDL regime requires 16-18 hours of light, but each group of cows is milked and kept in a illuminated holding area and parlor for one hour during the dark period, which brings the total time of light for each group to 16 hours. The literature does not specify whether the 16 hours of light should be consecutive, and the preliminary milk production data suggests that there is a statistically significant increase in milk production for Groups 1 and 3 of 10.5% that was calculated over a three-month period. The second objective was to quantify the energy savings associated with implementing LDL with LED luminaires and dynamic controls compared to implementing LDL with metal halide HID luminaires or LED luminaires without dynamic controls. Similar LDL layouts were created with metal halide HID luminaires and featureless LED luminaires, and their calculated energy consumption was compared to the actual energy consumption of the implemented LDL system. 52 The energy consumption data for the implemented system was measured before and after implementation. 315 days of energy data were available from September 2018 through July 2018. Obvius LLC, the host company of the data, experienced a cyberattack on their servers in August 2019 and shut down remote access to their cloud data. The data is still held by Obvius LLC and will be sent as soon as possible. The presented and discussed energy data supports one of the purposes of this study— that empirical data is necessary to accurately characterize the energy consumption of a long day lighting system that utilizes dynamically controlled LED lights. The measured energy data from almost one year shows that the implemented system consumes significantly less energy than alternative systems utilizing HID or featureless LED luminaires. It does consume more energy than the existing luminaires, which is to be expected since dairy free-stall barns are typically underlit. Continued monitoring of the LDL system’s energy consumption will increase the accuracy of the energy consumption characterization for this system, which can then be used as a tool to predict energy consumption of other LDL systems using dynamic controls. In conclusion, this study shows that the additional control capabilities of LED lights make them an excellent option for LDL since they consume significantly less energy than other lighting alternatives. The third objective was to perform a basic economic analysis comparing the cost of using LDL to increase production versus adding additional cows to the herd. Since only Groups 1 and 3 were measured and compared to previous performance before LDL implementation, only expenses and revenues associated with the west barn and Groups 1 and 3 were used for the economic analysis. To perform an economic analysis, all major costs associated with the project 53 were quantified, and weighed against the perceived increase in revenue. Capital costs were easy to calculate from installation estimates and material and equipment quotes. The increase in energy consumption from the implemented system was based on almost one year of measured energy consumption. These two expenses will not be difficult to calculate for future projects’ economic analyses, since the capital costs come directly from the project budget, and energy consumption can be measured with readily available equipment. The other expense that was calculated was the cost of the increase in feed consumption associated with an increase in milk production. As mentioned in the discussion, the literature is not clear whether cows experiencing a positive galactopoietic effect from LDL in fact consumption the established 0.4 pounds of additional food for each additional pound of milk produced. The literature has established that LDL most likely influences and regulates hormone production in cows, which could cause cows to use their consumed nutrients more efficiently, rendering more consumed nutrients available for milk production. Literature has shown that heifers experience faster weight gain with no increase in food intake when exposed to LDL, so it is possible that lactating cows experience a similar effect (Dahl et al., 2000). Future work in this area can combine an LDL system similar to the one used in this project with an animal science study that precisely measures the food intake of the cows studied. This will provide a better estimate of the actual increase in feed costs associated with LDL. For the purpose of this project, the worst-case scenario for increased food consumption was used since the mechanisms by which LDL induces a galactopoietic effect of cows are not completely understood. Using 0.4 pounds of food per pound of milk produced represents an 8% increase in total dry mass intake (DMI) on Groups 1 and 3. The greatest increase in the 54 literature on Table 1 was 6.1%, but both studies that showed in increase in DMI did not show an obvious relation between increased milk production and increased DMI. Increased revenue was solely based on the increase in milk production. There is still uncertainty with the calculated 10.5% increase in milk production, since there was very little that could be used for comparison due to the major changes that were made to the herd. The economic analysis had to assume that the estimated 10.5% increase in milk production for Groups 1 and 3 over 3 months would be stable throughout the year. If the LDL is really having a positive galactopoietic effect on the cows, a larger increase in milk production should be seen during the fall and winter when the days are naturally short. The supplemental illumination provided by LDL should cause an increase in production over cows that live in free-stall barns that have little to no supplemental illumination. The uncertainty in milk production data also could have an effect on the increased DMI calculations discussed in the expenses, since the increased feed costs are directly tied to an increase in milk production. Although this is an engineering research project and not an animal science research project, milk production data is still necessary to illustrate the potential revenue gains that this technology could offer. The few months of monthly group data that are available from DelPro to be used as a comparison are not are not enough to draw a solid conclusion about the galactopoietic effects of LDL over the entire milk cycle of a cow. The 3 months of comparisons showed an estimated 10.5% increase in milk production over a July-September time period for cows that were between 30 and 100 days in milk (DIM). An entire year of group production data from DelPro should be compared before and after implementation of LDL so see the effects of LDL during the winter months when the days are naturally their shortest. 55 Since there are only three months of reliable group data from July-September 2018 that can be used for comparison, the future of this study will require that the west and east barns be exposed to different lighting protocols and groups from each barn studied simultaneously. Management made major changes to the herd makeup of the east barn, making the type of cows that occupy it similar in days in milk to the west barn cows. In the future, the west barn could stay on an LDL protocol, whereas the lighting schedule of the east barn could be changed to 12 hours on and 12 hours off, or constantly energized at the luminaires’ lowest light output. Two groups from the east barn could then be selected as a control groups, and Groups 1 and 3 from the west barn would be the test groups. Since all cows in both barns are currently under LDL, a future study could make these adjustments to the east barn lighting and track milk production to establish the period of time it takes for cows under LDL to lose any potential galactopoietic effects of LDL. This could be very important research since not all farms have backup generators that can handle the entire farm, so if electrical service is lost, the cows could be without LDL for hours or days. Also, if it was found that it takes hours or days of lost LDL to decrease milk production, capital investment and energy could be saved because less lighting would be needed in walking corridors leading to and from the parlor since cows rarely spend more than an hour in them at a time. The financial analysis indicated a simple payback period of only 0.7 years, which is well below the typical maximum investment payback period of 5 years that the Michigan Farm Energy Program uses for energy investment upgrades. If continued study on this project shows a production increase lower than the estimated 10.5%, the calculated payback period will increase. The other contingency is the increased DMI, which may be lower than the assumed 56 values used from the study of large Michigan dairy farms. If the DMI is less than the amount used for the calculations, the payback period will increase. 32 additional cows producing milk at the pre-LDL average production levels of Groups 1 and 3 would be needed to produce the calculated increase in milk production due to LDL in those two groups, which adds additional opportunity cost savings. These savings were not included in the economic analysis but can be used as an additional positive quality when proposing the implementation of LDL to farmers. As Michigan loses dairy farms, the remaining farms are expanding their herds to increase production, which can require an enormous capital investment in cows and barn space. Simply increasing the productivity of their existing herd with an automated system will not significantly increase the time needed to manage the herd compared to adding hundreds of more cows. In addition, the savings from utilizing dynamically controlled luminaires compared to featureless luminaires more than made up for the additional cost of the dynamic controls. As mentioned in the discussion, not all costs could be captured, but not all benefits could be quantified either. Literature suggests that heifers similar to the cows in Group 2 gain weight more efficiently under LDL which is a potential added benefit that could not be measured for this study. The additional cow savings also reflect other hard to quantify savings, such as reduced veterinary expenses, less infrastructure such as barns, decreased herd management time, and less effort in manure handling. In conclusion, although more data is needed, this project has a lot of potential and should reveal that investing in dynamically controlled LDL will help dairy farms become more efficient and increase their production without utilizing more cows. 57 Like the Wing Acres Dairy case study of Section 2, the herd manager at Car Min Vu Farms mentioned that he also saw qualitative improvements in his cows and their environment. He noticed that there tended to be less crowding and the cows looked much cleaner. He also commented that the better lighting conditions made the work environment much safer for the employees who routinely walk among the cows to check them or move them to and from the parlor. He said that the simulated sunrise and sunset each day also made it easier for the employees’ eyes to adjust to the changing light levels at the beginning and end of the lit day. Maintaining more consistent light levels in the free-stall barns was also shown to reduce skin parasite infestations on Wing Acres, but that was not qualitatively observed on this project. Although these findings were the subjective opinions of a third party and not the area of focus for this engineering study, future studies could attempt to quantify some of the other possible positive effects of LDL. If this project were to be conducted again, there are a few things that should be done differently with the type of equipment used. From the beginning, this project had limitations from the funding partners that limited the type of equipment that could be used. The main constraint that would be changed is requirement that both the lighting and control system must be DLC listed. While searching for control systems that met the needs of the study, the team encountered other control systems that offered more granular control of the lights. Although they had features similar to systems listed on DLC, those particular companies had not paid DLC to review their system and therefore were not listed on the DLC website. In future replications of this study, the team would evaluate control systems solely based on their features to determine which system would be best suited for this unique application. Synapse 58 acknowledged that this project was very complicated for them, but the team had to work with this company since they were the only usable option that met all of the project requirements. Future replications of this project should use control systems that handle daylight harvesting with closed loop sensors instead of open loop sensors. A closed loop DLH system could save more energy since luminaires that are in darker, shaded areas of the barn could be grouped together and controlled based on actual light levels inside the barn. Currently, these luminaires are grouped together but are controlled by light levels outside of the barn, which requires more trial and error to set up the system and requires a bigger safety factor to ensure that interior light levels do not fall below the targeted 15 foot-candles. A system similar to the one implemented on Car Min Vu Farms could be used by animal scientists to study the mechanism by which LDL produces a galactopoietic effect in cows. If conducted inside a closed building with no natural light, DLH capability would not be necessary, but the automated and individual controls for each luminaire would allow the system to be customized to the test barn. Future animal science studies can expand on the work of this engineering research project, controlling more variables, testing inputs such as DMI, and outputs such as milk quality and hormone levels in cows in the control and test groups. Smaller studies with more control over variables should still milk the cows three times per day, since dairy farms continually get larger and practice three times per day milking. Finally, any future LDL project implemented on a privately-owned farm should require a written contract between the owner and the research team to avoid the major herd changes that happened during this project. Any farm selected for such a project should have a reliable means of tracking individual cow milk production, and this data should be collected from the 59 beginning of the project in conjunction with bulk tank data. The overall monthly production weights from the bulk tank data may have too many variables to be useful in an animal science study, but the overall milk quality analysis could still be useful in tracking changes in butterfat, protein, and somatic cell count, since each quality marker is given as a percentage or count per pound. Although this engineering study had limitations, it provides a solid foundation for future research and the data collected so far is promising. The implemented system can be customized through a computer interface for a variety of animal science research projects that can further explore the benefits of LDL on milk production. The current and future additions to the knowledge base of LDL will continue to assist dairy farmers in making capital investment decisions in an increasingly competitive market. 60 APPENDICES 61 Appendix A—DLC Requirements for Solid State Lighting and Lighting Controls DLC capability requirement for solid state high-bay and low-bay luminaires (DesignLights Consortium, 2018b):  Minimum Light Output—5,000 Lumens (lm)  Minimum Efficacy—105 lm/W  Minimum Warranty—5 years  Maximum Color Temperature (CCT)—5,700 K  Minimum Color Renditioning Index (CRI)—70  L70 (30% reduction in light output)—50,000 hours DLC capability requirements for interior networked lighting control systems (DesignLights Consortium, 2018a):  Networking of Luminaires and Devices—each luminaire and controller must be able to communicate with other linked luminaires and controllers in the same system. This capability is not required at the whole building level but is required for all lighting system components (luminaires and controllers) within a lighted space. It does not apply to non-lighting systems  Occupancy Sensing—the ability to control the lighting levels in a room based on its occupancy of moving objects. The ability to communicate with external occupancy sensing systems can also fulfill this requirement 62  Daylight Harvesting/Photocell Control—the system is able to change the output of artificial light based on the amount of natural daylight that is present in the lighted space.  High-End Trim—at any time the maximum light output from a luminaire can be set to a value below the listed maximum light output value  Zoning—groups of luminaires can be uniquely controlled together through software. This means different control zones can be made that each have a unique control protocol (i.e. scheduling, brightness, etc.)  Luminaire and Device Addressability—each luminaire, controller, user interface device (wireless switch), and sensor must be uniquely addressable to allow configuration that is independent of the actual electrical circuit wiring.  Continuous Dimming—the control system must have at least 100 dimming levels to have ‘smooth’ light level changes 63 Appendix B—Project Specifications for LDL Lighting and Control Systems Lighting requirements:  Intensity—lights must have adjustable intensity to simulate day and night in main cow housing areas of the barn o Daytime—minimum of 15 fc, 36 inches from ground and evenly distributed, measured horizontally o Nighttime—1-4 fc, 36 inches from ground, measured horizontally o Lights for the main isles should have the option of being always on and emitting 3-5 fc. This lighting must be directional, and not bleed into the main cow areas at an intensity greater than 4 fc  Lighting must be directional  Operating temperature: -25oF to 125oF  Color temperature: 5100K-5500K  Color Rendition Index (CRI)—as close to 100 as possible  Durability: o Dust-proof o Water-resistant o Can withstand physical jolts from when cows or equipment bump into any support beams to which the lights may be mounted  Assume zero reflectivity from floors, ceiling and walls for your analysis 64  The potential energy usage of the dimmable LED units recommended should be less than non-dimmable metal halide lights with similar light output Control System Requirements:  Hardware/software requirements o Main control box:  Throughout the day the system must dynamically fade or brighten luminaire output to maintain the required foot-candle level while allowing natural light to contribute as much as possible.  Ability to gradually ramp up or ramp down (not step) the light intensity when transitioning from the ‘day’ to the ‘night’ photoperiod  ~ 5-year warranty o Sensors, communication/junction devices, and any luminaire mounted controls:  Outdoor rated  Operating temperature range: close to -25oF to 125oF  ~5-year warranty  Control Pattern requirements: o ‘Day’ photoperiod:  Lights gradually ramp up in intensity (~15 min period, simulating sunrise) until the sensors tell the control box that the minimum required light intensity for ‘daytime’ has been achieved  Light intensity sensors monitor the light levels in the barn throughout the day in real time and make adjustments as needed: 65  If there is plenty of sunlight, the sensors will tell the control box that there is more than enough light to meet the minimum LDL requirements. The control system will then dim the lights until the sensors tell it that an intensity value that is close to the minimum requirement has been achieved  If there is not enough light due to storms, overcast conditions, or early sunset, the sensors will tell the control system that there is not enough light to meet the minimum LDL requirements. The control system will then ramp up the light intensity until the sensors tell it that the barns are adequately lit o ‘Night’ photoperiod:  Lights gradually ramp down in intensity (~15 min period, simulating sunset) until they are completely off 66 Appendix C—Barn Architecture Figure C-1—Barn cross sections illustrating the use of scissor trusses (top) and webbed coffer trusses (bottom) 67 Figure C-2—West barn top view 68 Figure C-3—East barn top view 69 Appendix D—Lighting layouts FIgure D-1—West barn lighting layout 70 Figure D-2—East barn lighting layout 71 BIBLIOGRAPHY 72 BIBLIOGRAPHY Barchart Market Data Solutions. (2019). Milk. Retrieved from http://www.nasdaq.com/markets/milk.aspx?timeframe=4y Bauman, D. E. (2009). Facts about recombinant bovine somatotropin (rbST). Cornell University Department of Animal Science. Collier, R. J., Dahl, G. E., & VanBaale, M. J. (2006). Major advances associated with environmental effects on dairy cattle. Journal of Dairy Science, 89, 1244-1253. Dahl, G. E., Buchanan, B. A., & Tucker, H. A. (2000). Photoperiodic effects on dairy cattle: a review. Journal of Dairy Science, 83, 885-893. DesignLights Consortium. (2018a). Networked lighting control system technical requirements V3.0. Retrieved from https://www.designlights.org/lighting-controls/qualify-a- system/technical-requirements/ DesignLights Consortium. (2018b). Technical requirements V4.4. Retrieved from https://www.designlights.org/solid-state-lighting/qualification-requirements/technical- requirements/?scrollTo=luminaires DTE Energy, & Engineering Society of Detroit. (2017). E-Challenge Competition for Colleges and Universities. In DTE Energy (Ed.). Eiholzer, L., & Capel, M. (2015). Effects of long-day energy-efficient lighting on milk production in dairy operations. New York State Energy Research and Development Authority. Fiscalini Cheese Company. (2017). Our farm. Retrieved from https://www.fiscalinicheese.com/pages/our-farm Flanders, F. B., & Gillespie, J. R. (2015). Modern livestock & poultry production (9 ed.): Cengage Learning. French, P. D. (2000). Applied animal nutrition ration formulation. Retrieved from https://oregonstate.edu/instruct/ans310/ Hankey, R., Cassar, C., Liu, J., Wong, P., & Yildiz, O. (2017). Electric power montly. United States Department of Energy Energy Infromation Adminstration. Herdt, T. H. (2018). Nutritional requirements of dairy cattle. In: Merck Sharp & Dohme Corp. 73 Hristov, A. N., Johnson, K., & Kebreab, E. (2014). Livestock methane emissions in the United States. Retrieved from https://extension.psu.edu/livestock-methane-emissions-in-the- united-states Jones, J., Betz, R., Stein, D., LaPorte, J., Wardynski, F., Lee, K., & Moore, S. (2018). 2017 Michigan farm bussiness summary. In: Michigan State Univeristy. MDARD. (2016). Michigan's food and agricultural industry. In Michigan Department of Agriculture & Rural Development (MDARD) (Ed.). Michigan: State of Michigan. MDARD. (2019). Facts about Michigan agriculture. Retrieved from https://www.michigan.gov/mdard/0,4610,7-125-1572-7775--,00.html Thomas, A., Thelen, J., Go, A., Surbrook, T., Vanee, M., & Althouse, J. (2017). A dairy long day lighting success story: MI dairy increases production and cuts costs. Paper presented at the 2017 ASABE Annual International Meeting, Spokane, Washington. UDIM. (2018). Michigan dairy facts. Retrieved from https://www.milkmeansmore.org/milk- local/michigan-dairy-facts/ USEPA. (2018). Greenhouse gas equivalencies calculator. Retrieved from https://www.epa.gov/energy/greenhouse-gas-equivalencies-calculator Van Zweden, B., Go, A., & Surbrook, T. (2018). Developing energy use and savings indices for Michigan dairy operations. Paper presented at the 2018 ASABE Annual Internation Meeting, Detroit, Michigan. VanBaale, M. J., Armstrong, D. V., Etchebarne, M. A., Mattingly, R. M., & Fiscalini, J. B. (2005). Case study: supplemental light for lactating dairy cows. University of Arizona. Wolf, C. (2008). Farm decisions related to rbST use in Michigan. Retrieved from eXtension website: http://articles.extension.org/pages/9833/farm-decisions-related-to-rbst-use- in-michigan 74