牛舍甲烷浓度时序预测模型研究

打开文本图片集

中图分类号：TP391 文献标识码：A 文章编号：1673-9868（2025）10-0234-11

Abstract：Methane concentration has a large impact on catle growth and development，and the prediction of methane concentration in cattle barns can provide a scientific reference for the precise control of cattle barn environment. A methane prediction model based on the combination of gated recurrent neural network，improved sparrow search algorithm （ISSA）and back propagation algorithm （BP） was constructed for cattle barn. Firstly，the Gated Recurrent Unit （GRU） model was used to extract the nonlinear features from environmental data of the barn，and then the hyperparameters of the GRU model were optimized by ISSA to obtain the optimal GRU model，soas to improve the models ability to fit the nonlinear features. Finally，the BP model was used to further fit the predicted residual features after optimization of the ISSAGRU model，so as to improve the model prediction accuracy. Using the data collected in the experimental barn from April 29 to June 30，2024 for training and testing，the results showed that the model was able to effectively fit the multi-environmental parameters of the barn. Compared with BP，GRU，GRU-BP，SSAGRU，ISSA-GRU and ISSA-GRU-ARIMA，the ISSA-GRU-BP model proposed in this paper has a high prediction accuracy，with R2 ， RMSE ，and MAPE of 0. 934， 0.899×10-6 ，and 9.638% ，respectively.

Key words： cattle barn; methane concentration prediction; gated recurrent neural network;improved sparrow search algorithm;back propagation algorithm;residuals

在全球范围内，甲烷（ CH4 ）是影响全球温室效应的一种重要温室气体，其排放带来的增温潜力是二氧化碳的28倍，而反刍动物是甲烷排放的重要贡献者[1]。（剩余12463字）