关键词: 神经网络, 需水量, 物联网, 智能灌溉系统, 灰色关联分析

Abstract: In view of the low efficiency of agricultural production and the serious waste of agricultural water in our country, an intelligent irrigation system based on LSTM neural network was designed. The system used raspberry pie as the lower machine controller and Ali cloud server as the upper computer. The grey correlation analysis method was used to determine that the correlation coefficients between average air temperature, solar radiation, sunshine hours, average wind speed, relative humidity, pressure and crop water demand, which were 0.636 52, 0.510 42, 0.444 56, 0.440 29, 0.343 50 and 0.287 87 respectively. The meteorological factors with large correlation coefficient were selected as the characteristic input vector of the LSTM prediction model, and the crop water demand was used as the output vector, LSTM neural network prediction model was constructed. Meanwhile, the corresponding Internet of Things platform and mobile platform for the system was developed, which achieved data visualization and switch remote control functions. The results showed that the fitting coefficient R2 of the prediction model was equal to 0.987 2, the residual was relatively stable and the prediction accuracy was high. In the actual irrigation experiment, the average absolute error of the model prediction was 3.02%. The system runs stably and has a high degree of intelligence, which provides a certain reference for the development of agricultural modernization.

Key words: neural networks, waterdemand, Internet of Things, intelligent irrigation system;gray correlation analysis