Design of Irrigation Control System for Vineyard Based on LoRa Wireless Communication and Dynamic Neural Network

Abstract

In grape cultivation, precise irrigation control systems can improve the yield and taste of grapes and enhance the efficiency of water use. A drip irrigation control system, based on LoRa remote wireless technology, and a dynamic neural network, is designed in this paper. Lora remote wireless communication terminal, which can connect sensors or control devices, such as soil temperature, digital flow meter, and so on, at the same time, can be remote control such as pumps, water valves, and other equipment operation. A prediction model was established using the Long Short-Term Memory Network (LSTM) which is a class of the Recurrent Neural Network (RNN). The trained model uses past soil moisture, precipitation and climate measurements to predict the moisture content of vineyard soils based on a pre-set amount of time, thereby calculating the amount of irrigation and determining the timing of irrigation. According to the training results of the model, the root mean square error (RMSE) is $0.116 \sim0.171$, the range of R2 is $0.941 \sim0.986$, and the range of mean absolute error (MAE) is $0.071 \sim0.081$. For predicting water content at different soil depths, R2>0.9 shows good model performance and prediction effect. The system has good working stability, high prediction accuracy, and slight deviation. The planter can apply it to the irrigation of large, medium, and small-scale grape plantations in the actual planting environment. Compared with the traditional irrigation control system, advanced irrigation methods can be realized, improving the irrigation rate and saving irrigation water.