关键词: 温室, 温度, 时间序列, 长短期记忆网络, 灰狼优化算法

Abstract: Temperature is one of the main limiting factors of crop growth in facility production. It is of great guiding significance to predict air temperature in the greenhouse in advance for managing and controlling the environment in the greenhouse accurately. Long ShortTerm Memory network (LSTM) based on Grey Wolf Optimization (GWO) model was proposed to predict air temperature in the greenhouse in this paper. This model used GWO to adjust and optimize the parameters of LSTM, which could avoid manual adjustment of parameters and improve the efficiency of model parameter adjustment. The experimental greenhouse was located in Jiangsu Academy of Agricultural Sciences. The data of environment and control device operation status were collected from September 23rd, 2020 to December 21st, 2020 during the experiment. The results showed that when the predicted time step was 30 min, the root mean square error, mean absolute error, mean absolute percentage error and determination coefficient of GMO-LSTM prediction were 0.677 6、0411 4、0.168 7 and 0.960 4, respectively. In the prediction time step of 60 min, the prediction accuracy of GMO-LSTM was higher than that of standard LSTM and Back Propagation Artificial Neural Network (BP-ANN). In summary, GWO-LSTM model proposed in this paper could accurately predict the future temperature change in the greenhouse, which could also provide the effective data support for developing intelligent control strategy of the environment in the greenhouse.

Key words: greenhouse, temperature, time series, Long ShortTerm Memory network, Grey Wolf Optimization algorithm