Alapadatok
Év, oldalszám:2009, 5 oldal
Nyelv:angol
Letöltések száma:4
Feltöltve:2021. május 31.
Méret:731 KB
Intézmény:
-
Megjegyzés:
Csatolmány:-
Letöltés PDF-ben:Kérlek jelentkezz be!
Értékelések
Nincs még értékelés. Legyél Te az első!Legnépszerűbb doksik ebben a kategóriában
Tartalmi kivonat
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 Implementation of an application for daily individual concentrate feeding in commercial software for use on dairy farms E.JB Bleumer*, G. André, G van Duinkerken Wageningen University and Research Centre, Animal Sciences Group, Business unit Animal Production, P.O Box 65, 8200 AB Lelystad, the Netherlands *edwin.bleumer@wurnl Abstract Daily concentrate allowances for individual dairy cows are usually based on empiric models. These models are generally based on regression equations derived from population data and do not take into account individual and temporal variation. An application was implemented in common practice which consists of an adaptive model for estimating the actual individual response in milk yield on concentrate intake using individual real time process data. Before the application was implemented, a prototype was developed by a
team consisting of biometricians, animal nutritionists and ICT application specialists. It was tested in an animal experiment and further developed into a proof of principal, which was implemented for testing in a common practical setting on a research farm. Because the results were very promising, a workshop was organised to introduce the concept to software, hardware and feed industries where they were challenged to participate. In the next collaborative phase with industry involvement the further implementation into a management system was stepwise: 1) technical documentation of algorithms, 2) programming, 3) verification of algorithms, 4) onfarm implementation of the integrated software and, 5) on-farm evaluation. During the implementation it became clear that steps 1 to 3 were not difficult to perform and did not take much time. Steps 4 and 5 were more complicated because: 1) correct data must be generated from the management system as an input for the model and, 2) the output of
the model has to be interpreted correctly for calculating concentrate allowances in the management system. However, not only technical aspects of an implementation process are important, also the communication with end users and stakeholders requires particular attention, for successful implementation of a new concept. While testing and implementing the application it became clear that end users and stake holders were willing to accept and use the innovative concept but interpreted the outcome based on traditional population knowledge and paradigms. Keywords: management system, operational model, economical profit, concentrate feeding Introduction A lot of information of an individual animal is automatically collected, integrated an saved in management systems on a dairy farm (Frost et al., 1997) Dairy farmers are using this information for their management of their animals and they are challenged to use this information to make the economic profit as high as possible (Doluschitz,
2003). The economic profitability in dairy farming depends to a large extent on milk benefits and concentrate costs. Common calculations of individual concentrate allowances are based on the assumed energy balance of the animals and are based on empirical knowledge of the population. This means 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 that for similar cows, with a similar energy requirement, the same amount of concentrate is advised (Broster and Thomas, 1981). However, it is known that there is a lot of variation in feed efficiency between individual cows and also within cows over time. Dairy cows could be considered as complex, individual and time-variant (CIT-systems), Berckmans, 2004), so the basic principles of precision livestock farming (PLF) can be applied (Wathes et al., 2008) A model, which calculates concentrate allowances based on the individual daily milk yield
response on concentrate intake and milking frequency was developed by André et al. (2007) Individual milk prices and concentrate price are taken into account to maximize the gross margin milk returns minus concentrate costs. The model was tested on an research dairy farm and developed into a proof of principle (step 0). Because the test results were very promising, it was attempted to find partners to implement this approach into common dairy farming practice. The following steps were performed in partnership with industry: 1) technical documentation of algorithms, 2) programming, 3) verification of algorithms, 4) on-farm implementation of the integrated software and, 5) on-farm evaluation. In this paper we describe and discuss the process steps and their results. Materials and methods Brief outline of the model Before describing the stepwise implementation of the model a brief outline of the model is given. The implemented concept is based on the model described by André et al
(2007), but restricted to the effect of the daily individual concentrate intake on milk yield response. The model consists of two parts, an adaptive model and a control algorithm. The adaptive model estimates the response parameters from real time process data stored in the database of the farm management system. The control algorithm calculates the optimal concentrate allowances based on prices and targets. These optimal concentrate advises are put back into the management system to control the automated concentrate feeders. For determination of the optimal concentrate allowances various input variables are necessary: individual milk yield accumulated per day, the daily concentrate intake, concentrate price and milk price. The individual milk price is calculated from the milk constitution, because fat and protein are not valued equally. The economic optimum for concentrate allowances is determined by maximizing the gross margin milk returns minus concentrate costs. The optimal setting
for concentrate allowance is transformed into a practical setting for concentrate allowance, to avoid to big changes in concentrate intake. Prototyping and testing on HTB Based on the model, a prototype of an application for optimizing the concentrate allowance was developed by a team of biometricians, animal nutritionists and an ICT specialist. It was tested on a research dairy farm in common practical circumstances. On average, there were 66 Holstein Frisian cows in milk with an average production of approximately 30 kg per cow per day. The farm was equipped with a single unit automatic milking system The cows were fed in a robotic feeding system which provided roughage-concentrate mixtures on an individual basis. At first, advises were calculated with support of different software programs. MS Access 2003 was used as a database for collecting the data as input for the model. The model itself was built in GenStat 8th edition, in which also data analyses and calculation took place.
The results of the calculations were imported in MS Excel 2003 where the allowances per cow 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 were presented and visually checked. Next the advises were put in the management software manually. Although input was gathered daily, concentrate allowances were calculated daily Before concentrate allowances were put in the management software, parameters of the model and advises were checked and, if necessary, adapted manually by the farm management. In time, the application was corrected and improvements were made based on practical settings and knowledge of the farm management, and further developed into a proof of principal. Marketing: participation of stakeholders for application building and introduction in practice. To introduce this new concept, a workshop was organized and the
results of the proof of principal was presented. Approximately 30 companies from software, hardware and feed industry were invited and asked to participate in developing this concept into a commercially product, which resulted in further collaboration with a software and two companies from feed industry. Results and discussion I. Application building, implementation and testing Implementation 1) technical documentation of algorithms In the first step, technical documentation of the model was written for the programmers. This documentation consisted mainly of mathematical equations, tables and examples to clarify the parameters and variables used in the model. The documentation was updated during the application phase based on remarks and questions of programmers and end-users. Statistical equations in the technical documentation appeared to be rather complicated. However, the logic structure of these equations is very similar to the logic structure of the source code of software.
Despite the logic structure, interpretation of the model was sometimes difficult Implementation 2) programming In this second step the equations were programmed by the software manufacturer, not directly into their main software but in a separate module so it could be tested without the use of the main software program. Although the module was not integrated immediately, conditions for integration in a later stage were already developed. Not only was integration within the main management software necessary, but also connections with the surrounding software architecture. This third party software is used for downloading milk price, concentrate price and for communication with milking and feeding equipment. For an overview of the model and it’s integrations see figure I. Implementation 3) verification of algorithms Algorithms were verified using an backup of the algorithms a backup of a database with the required input. It was enough to use the input data of a few days from only one
cow to see if the parameters and variables were calculated correctly. By using the same input data in our own prototype the algorithms were validated. Bugs were reported to the software manufacturer and fixed. One on one contact with the programmers of the software company proved to be very successful in solving bugs. The data flow connections with third party software was tested by the company itself. 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 Real time individual process information Management software Third party software Adaptive model Control algorithm Calculated settings Figure I. Overview of dataflow of the implemented model and it’s integration with other software components Implementation 4) on-farm implementation of the integrated software Important in this phase was the exchange of data between module, main
program and the process software. Only one concentrate type was optimized Correct data exchange between the management system and the process software was inevitable. Implementation 5) on-farm evaluation. On farm testing was performed in step five, starting with the implementation on one research dairy farm. The most important goal was to find out if the software would run correctly under farm circumstances. Not only should input data for the model be generated correctly, but also the output of the model had to be interpreted correctly. Furthermore disturbances in the data exchange process, which can lead to missing data or other errors and malfunctioning, had to be monitored thoroughly. After successful implementation on this farm, the module was installed on four common dairy farms. This was the first step in introducing the model at common dairy farms II. Commercial distribution Not only farmers had to accept this new way of optimizing and calculating concentrate allowances, also
commercial agents of the feed industry had to become accustomed to it. To get commercial agents and dairy farmers acquainted with the model a few methods were used. At the same time the software became commercially available, the software company launched a website, where information on the concept could be found. Meetings, where an animal nutritionist and a biometrician explained the way the model works, were organized for commercial agents of the feed industry. Articles in popular farm magazines contributed to expand the knowledge of this new approach, not only to the commercial agents but also to dairy farmers as well. Dairy farmers who had used the model for a few months were interviewed to share their experiences. III. Future developments The described model is only useful for daily operational use; long term tactic and strategic decisions can not be made. Long term effects on animal health, roughage stock, milk quota, nitrogen excretion, economic characteristics, and so on, have
not been described yet. Currently, there is not enough information available to evaluate these long term effects. It is clear that these effects can not be ignored an may lead to a different approach towards farm management. In future it’s not only possible to optimize on economical results but, for example, also nitrogen utilization. 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 Conclusion Close cooperation between developers, stake holders and end-users is crucial for successful implementation of new software concepts in common practice. Literature André, G., Ouweltjes, W, Zom, RLG, Bleumer, EJB 2007 Increasing economic profit of dairy production utilizing individual real time process data. In: Cox, S (ed) Precision livestock farming ’07. Wageningen Academic Publishers, The Netherlands pp 179-186 Berckmans, D. 2004 Automatic on-line monitoring of animals by precision livestock farming. In: F Madec (ed),
Animal production in Europe: The way forward in a changing world, Saint Malo. pp 27-30 Broster, W.H, Thomas, C 1981 The influence of level and pattern of concentrate input on milk output. In: W Haresign (ed), Recent advantages in animal nutrition, Butterworth, London. pp 76-96 Doluschitz, R. 2003 Precision Agriculture – Applications, Economic Considerations, Experiences and Perspectives. Proceedings of the EFITA 2003 conference July 5-9, 2003 Debrecen, Hungary, pp. 541-546 Frost, A.R, Schofield CP, Beaulah SA, Mottram TT, Lines JA, Wathes CM 1997 A review of livestock monitoring and the need for integrated systems. Computers and Electronics in Agriculture 17: 139-159 Wathes, C.M, Kristensen, HH, Aerts, J –M, Berckmans, D 2008 Is precision livestock farming an engineer’s daydream or nightmare, an animals friend or foe, and a farmer’s panacea or pitfall? Computers and Electronics in Agriculture 64: 2-10
team consisting of biometricians, animal nutritionists and ICT application specialists. It was tested in an animal experiment and further developed into a proof of principal, which was implemented for testing in a common practical setting on a research farm. Because the results were very promising, a workshop was organised to introduce the concept to software, hardware and feed industries where they were challenged to participate. In the next collaborative phase with industry involvement the further implementation into a management system was stepwise: 1) technical documentation of algorithms, 2) programming, 3) verification of algorithms, 4) onfarm implementation of the integrated software and, 5) on-farm evaluation. During the implementation it became clear that steps 1 to 3 were not difficult to perform and did not take much time. Steps 4 and 5 were more complicated because: 1) correct data must be generated from the management system as an input for the model and, 2) the output of
the model has to be interpreted correctly for calculating concentrate allowances in the management system. However, not only technical aspects of an implementation process are important, also the communication with end users and stakeholders requires particular attention, for successful implementation of a new concept. While testing and implementing the application it became clear that end users and stake holders were willing to accept and use the innovative concept but interpreted the outcome based on traditional population knowledge and paradigms. Keywords: management system, operational model, economical profit, concentrate feeding Introduction A lot of information of an individual animal is automatically collected, integrated an saved in management systems on a dairy farm (Frost et al., 1997) Dairy farmers are using this information for their management of their animals and they are challenged to use this information to make the economic profit as high as possible (Doluschitz,
2003). The economic profitability in dairy farming depends to a large extent on milk benefits and concentrate costs. Common calculations of individual concentrate allowances are based on the assumed energy balance of the animals and are based on empirical knowledge of the population. This means 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 that for similar cows, with a similar energy requirement, the same amount of concentrate is advised (Broster and Thomas, 1981). However, it is known that there is a lot of variation in feed efficiency between individual cows and also within cows over time. Dairy cows could be considered as complex, individual and time-variant (CIT-systems), Berckmans, 2004), so the basic principles of precision livestock farming (PLF) can be applied (Wathes et al., 2008) A model, which calculates concentrate allowances based on the individual daily milk yield
response on concentrate intake and milking frequency was developed by André et al. (2007) Individual milk prices and concentrate price are taken into account to maximize the gross margin milk returns minus concentrate costs. The model was tested on an research dairy farm and developed into a proof of principle (step 0). Because the test results were very promising, it was attempted to find partners to implement this approach into common dairy farming practice. The following steps were performed in partnership with industry: 1) technical documentation of algorithms, 2) programming, 3) verification of algorithms, 4) on-farm implementation of the integrated software and, 5) on-farm evaluation. In this paper we describe and discuss the process steps and their results. Materials and methods Brief outline of the model Before describing the stepwise implementation of the model a brief outline of the model is given. The implemented concept is based on the model described by André et al
(2007), but restricted to the effect of the daily individual concentrate intake on milk yield response. The model consists of two parts, an adaptive model and a control algorithm. The adaptive model estimates the response parameters from real time process data stored in the database of the farm management system. The control algorithm calculates the optimal concentrate allowances based on prices and targets. These optimal concentrate advises are put back into the management system to control the automated concentrate feeders. For determination of the optimal concentrate allowances various input variables are necessary: individual milk yield accumulated per day, the daily concentrate intake, concentrate price and milk price. The individual milk price is calculated from the milk constitution, because fat and protein are not valued equally. The economic optimum for concentrate allowances is determined by maximizing the gross margin milk returns minus concentrate costs. The optimal setting
for concentrate allowance is transformed into a practical setting for concentrate allowance, to avoid to big changes in concentrate intake. Prototyping and testing on HTB Based on the model, a prototype of an application for optimizing the concentrate allowance was developed by a team of biometricians, animal nutritionists and an ICT specialist. It was tested on a research dairy farm in common practical circumstances. On average, there were 66 Holstein Frisian cows in milk with an average production of approximately 30 kg per cow per day. The farm was equipped with a single unit automatic milking system The cows were fed in a robotic feeding system which provided roughage-concentrate mixtures on an individual basis. At first, advises were calculated with support of different software programs. MS Access 2003 was used as a database for collecting the data as input for the model. The model itself was built in GenStat 8th edition, in which also data analyses and calculation took place.
The results of the calculations were imported in MS Excel 2003 where the allowances per cow 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 were presented and visually checked. Next the advises were put in the management software manually. Although input was gathered daily, concentrate allowances were calculated daily Before concentrate allowances were put in the management software, parameters of the model and advises were checked and, if necessary, adapted manually by the farm management. In time, the application was corrected and improvements were made based on practical settings and knowledge of the farm management, and further developed into a proof of principal. Marketing: participation of stakeholders for application building and introduction in practice. To introduce this new concept, a workshop was organized and the
results of the proof of principal was presented. Approximately 30 companies from software, hardware and feed industry were invited and asked to participate in developing this concept into a commercially product, which resulted in further collaboration with a software and two companies from feed industry. Results and discussion I. Application building, implementation and testing Implementation 1) technical documentation of algorithms In the first step, technical documentation of the model was written for the programmers. This documentation consisted mainly of mathematical equations, tables and examples to clarify the parameters and variables used in the model. The documentation was updated during the application phase based on remarks and questions of programmers and end-users. Statistical equations in the technical documentation appeared to be rather complicated. However, the logic structure of these equations is very similar to the logic structure of the source code of software.
Despite the logic structure, interpretation of the model was sometimes difficult Implementation 2) programming In this second step the equations were programmed by the software manufacturer, not directly into their main software but in a separate module so it could be tested without the use of the main software program. Although the module was not integrated immediately, conditions for integration in a later stage were already developed. Not only was integration within the main management software necessary, but also connections with the surrounding software architecture. This third party software is used for downloading milk price, concentrate price and for communication with milking and feeding equipment. For an overview of the model and it’s integrations see figure I. Implementation 3) verification of algorithms Algorithms were verified using an backup of the algorithms a backup of a database with the required input. It was enough to use the input data of a few days from only one
cow to see if the parameters and variables were calculated correctly. By using the same input data in our own prototype the algorithms were validated. Bugs were reported to the software manufacturer and fixed. One on one contact with the programmers of the software company proved to be very successful in solving bugs. The data flow connections with third party software was tested by the company itself. 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 Real time individual process information Management software Third party software Adaptive model Control algorithm Calculated settings Figure I. Overview of dataflow of the implemented model and it’s integration with other software components Implementation 4) on-farm implementation of the integrated software Important in this phase was the exchange of data between module, main
program and the process software. Only one concentrate type was optimized Correct data exchange between the management system and the process software was inevitable. Implementation 5) on-farm evaluation. On farm testing was performed in step five, starting with the implementation on one research dairy farm. The most important goal was to find out if the software would run correctly under farm circumstances. Not only should input data for the model be generated correctly, but also the output of the model had to be interpreted correctly. Furthermore disturbances in the data exchange process, which can lead to missing data or other errors and malfunctioning, had to be monitored thoroughly. After successful implementation on this farm, the module was installed on four common dairy farms. This was the first step in introducing the model at common dairy farms II. Commercial distribution Not only farmers had to accept this new way of optimizing and calculating concentrate allowances, also
commercial agents of the feed industry had to become accustomed to it. To get commercial agents and dairy farmers acquainted with the model a few methods were used. At the same time the software became commercially available, the software company launched a website, where information on the concept could be found. Meetings, where an animal nutritionist and a biometrician explained the way the model works, were organized for commercial agents of the feed industry. Articles in popular farm magazines contributed to expand the knowledge of this new approach, not only to the commercial agents but also to dairy farmers as well. Dairy farmers who had used the model for a few months were interviewed to share their experiences. III. Future developments The described model is only useful for daily operational use; long term tactic and strategic decisions can not be made. Long term effects on animal health, roughage stock, milk quota, nitrogen excretion, economic characteristics, and so on, have
not been described yet. Currently, there is not enough information available to evaluate these long term effects. It is clear that these effects can not be ignored an may lead to a different approach towards farm management. In future it’s not only possible to optimize on economical results but, for example, also nitrogen utilization. 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 Conclusion Close cooperation between developers, stake holders and end-users is crucial for successful implementation of new software concepts in common practice. Literature André, G., Ouweltjes, W, Zom, RLG, Bleumer, EJB 2007 Increasing economic profit of dairy production utilizing individual real time process data. In: Cox, S (ed) Precision livestock farming ’07. Wageningen Academic Publishers, The Netherlands pp 179-186 Berckmans, D. 2004 Automatic on-line monitoring of animals by precision livestock farming. In: F Madec (ed),
Animal production in Europe: The way forward in a changing world, Saint Malo. pp 27-30 Broster, W.H, Thomas, C 1981 The influence of level and pattern of concentrate input on milk output. In: W Haresign (ed), Recent advantages in animal nutrition, Butterworth, London. pp 76-96 Doluschitz, R. 2003 Precision Agriculture – Applications, Economic Considerations, Experiences and Perspectives. Proceedings of the EFITA 2003 conference July 5-9, 2003 Debrecen, Hungary, pp. 541-546 Frost, A.R, Schofield CP, Beaulah SA, Mottram TT, Lines JA, Wathes CM 1997 A review of livestock monitoring and the need for integrated systems. Computers and Electronics in Agriculture 17: 139-159 Wathes, C.M, Kristensen, HH, Aerts, J –M, Berckmans, D 2008 Is precision livestock farming an engineer’s daydream or nightmare, an animals friend or foe, and a farmer’s panacea or pitfall? Computers and Electronics in Agriculture 64: 2-10
