基于ARIMA和BP神经网络的猪舍氨气浓度预测

为了从源头减少生猪养殖过程中的氨气排放，降低猪舍氨气浓度，提出了基于ARIMA-BP神经网络的猪舍氨气浓度组合预测方法，分别从最优权重和残差优化角度对基于ARIMA-BP神经网络的组合预测方法进行了对比研究.将该预测方法应用于江苏省宜兴市某养猪场的氨气浓度预测中，预测结果表明：基于ARIMA-BP神经网络残差优化组合预测方法的预测精度最高，与BP神经网络、ARIMA预测方法和基于ARIMA-BP神经网络最优权重组合预测方法对比分析，评价指标MAE、MAPE和RMSE分别为0.0319、0.1580%和0.0365.本文提出的氨气预测方法可以为猪舍环境精准化调控管理提供科学依据以减小猪舍氨气排放对生态环境的污染.

Prediction of ammonia concentration in piggery based on ARIMA and BP neural network

In order to reduce ammonia emissions from the source during pig breeding and reduce the ammonia concentration in piggery, this paper proposed a combination prediction method based on ARIMA-BP neural network for the concentration of ammonia in piggery, and compared with the combined prediction method based on ARIMA-BP neural network, from the perspective of optimal weight and residual optimization. The proposed prediction method was applied to the prediction of ammonia concentration in a piggery in Yixing, Jiangsu province. The results of the prediction experiments showed that the prediction accuracy of the combination prediction method based on ARIMA-BP neural network residual optimization was the highest. Compared with the BP neural network, ARIMA prediction method and the optimal weight combination prediction method based on the ARIMA-BP neural network, the evaluation indexes MAE, MAPE and RMSE were 0.0319, 0.1580% and 0.0365respectively.The ammonia prediction method proposed in this paper can be used as a scientific basis for the precise control and management of piggery environment in order to reduce the ecological environmental pollution caused by ammonia emission from piggery.