Research On Environmental Monitoring System And Ammonia Concentration Prediction Model Of Northern Closed Dairy Barn

The environmental quality of dairy cowshed is an essential factor which affects the growth of health,milk quality,and plays a crucial role in the breeding of cows.With the large-scale development of dairy farming in China,the temperature,humidity and harmful gas concentration in dairy cow shed have a severe impact on the growth and health of dairy cows.During the development process of large-scale dairy farming,scientific and refined breeding technology plays a vital role in improving the growth environment and milk production of cows.At present,because of the north cold enclosed dairy,there are nonlinear,time-varying,and hysteresis coupling among the environmental factors in cattle house,and the number of collection points is single.A single monitoring site cannot represent the whole environment of the cowshed,and there is a microclimate around the individual cow,which makes it difficult to predict the climate of the cowshed.Therefore,a set of wireless intelligent environment monitoring and acquisition device that can adapt to multiple points of dairy farms is designed.At the same time,the collected data were analysed and processed.The environmental ammonia concentration of dairy farms was predicted based on BP neural network.The predicted data can be used to regulate the environment in advance,to save the cost of production management,optimise the growth environment of cows,and improve the efficiency and quality of milk production.This project aims at the low temperature and high indoor humidity of closed dairy farms in the cold northern region in winter,and ammonia gas generated by the decomposition of nitrogen-containing organic matter such as straw feeding and cow feces.The microclimatic environment around individual cows as the research object,temperature,humidity and ammonia concentration in the growth environment of cows were selected as the leading environmental parameters for environmental prediction.Based on the BP neural network,a prediction model of ammonia concentration in multi-input and single-output dairy cow shed was established,and the fitting degree of the data was analysed by Matlab software.The main research work of this topic is as follows:(1)Several environmental factors affecting the growth of dairy cows were studied by referring to domestic and foreign literature,and the selection of ecological monitoring variableswas determined according to the environmental change characteristics of dairy cows in the cold northern region in winter.A set of MCU based multi-factor monitoring device for the environment of dairy cows was designed,which took temperature,humidity and ammonia concentration as the collection parameters and used wireless sensor technology to provide a basis for the subsequent prediction of environmental ammonia concentration based on the collected data.(2)The combination of Zigbee technology and wireless communication module is adopted to realise the design of multi-point monitoring system of the dairy cow shed environment.By storing the environmental parameters collected from the dairy farm by the upper computer and establishing a database,the centralised display of multi-parameter measurement data can be realised.(3)The received multi-point environmental parameters will be processed,and the ecological parameters collected by the same sensor at the same time will be prepared by the weighted average method.This data serves as the real-time data of the overall environment in the dairy cow shed at this time.(4)The establishment of the prediction model of environmental parameters of dairy cows was studied and the advantages and disadvantages of various optimized BP neural network algorithms were compared.Selected Lm-bp and br-bp optimization algorithms to predict the ammonia concentration in cow houses,and MATLAB was used to realize the establishment and simulation of early warning of cow houses.A BP neural network model with three layers was designed by normalising the real-time data of dairy cows.The input layer was temperature and humidity,and the output layer was ammonia concentration.The prediction effect was best when the number of hidden layer neurons was six.(5)The data predicted by the model and the collected data are analysed and verified.The expected value of the model was evaluated by comparing the root mean square error.At the same time,correlation coefficient and training steps were used to compare the prediction accuracy of the LM algorithm to optimise BP neural network and BR algorithm to maximise BP neural network.The results showed that the relative error between the predicted value of the br-bp model and the actual measured value was small,which could meet the demand of ammoniaconcentration prediction in dairy cows.