Research And Application Of Environmental Monitoring And Ammonia Prediction Model For Pig Breeding Based On Internet Of Things

The breeding environment is an important factor affecting the quality and production efficiency of pigs.With the collectivization and intensification development of pig industry,it has become the primary task of pig industry to strengthen the monitoring of breeding environment,improve the ability of breeding prevention,and ensure the healthy production of pig breeding.In pig breeding,ammonia is an important evaluation indicator of piggery environmental.Excessive ammonia concentration will adversely affect the health of pigs.Real-time monitoring of piggery environment and the establishment of an efficient and suitable ammonia prediction model are of great significance to pig industry.This article takes ammonia as the research object,builds a multi-parameter environmental monitoring system for pig breeding,and establishes a fast and accurate online ammonia prediction model.The main research contents are as follows:(1)Realize the hardware design and software design of the multi-parameter environmental monitoring system for pig breeding,including three modules: on-site information perception,wireless remote communication and breeding monitoring application software.At the same time,the online prediction function of ammonia on the Web and Android terminal is realized.The system can monitor the piggery environmental in real time,and provide a comprehensive and accurate data source for the ammonia prediction in the follow-up research.(2)According to the characteristics of nonlinearity and nonstationarity of ammonia concentration time series in pigsty,a single-factor ammonia concentration time series prediction model is proposed.The model first decomposes the original ammonia time series into a series of different time scales intrinsic mode function and residuals through ensemble empirical mode decomposition;then reconstructs the decomposed results through phase space reconstruction;then uses the sparrow search algorithm to optimize the Elman neural network to predict each sub-sequence;finally,the ammonia prediction results of each sub-sequence are added and reconstructed,so that the final prediction results of ammonia gas can be obtained.The experimental results show that this model has high accuracy in RMSE,MAE and MAPE,and is effective and suitable for short-term ammonia prediction.(3)Aiming at the influence of various environmental factors on the ammonia concentration in pigsty,a multi-factor ammonia concentration prediction model is proposed.The model first uses gray correlation analysis to screen the key factors affecting the ammonia concentration in pigsty;then the similarity degree is defined to count the samples of similar days,and the historical samples are divided into several categories by using the method of ISODATA clustering,and the category with the greatest similarity with the predicted day is identified;finally,the most similar type of sample and the predicted day sample are used as input samples to establish the Elman neural network ammonia prediction model.After experimental verification,compared with the model without ISODATA clustering,the proposed model has higher prediction accuracy and stronger robustness,and is more suitable for long-term ammonia prediction.The experimental results of the system in a pig breeding base show that: the system data collection is accurate,information transmission is reliable,the control effect is good,the prediction accuracy is high,the overall performance is stable,and it can meet the requirements of fine pig breeding with the Internet of Things.