Prediction of soil moisture based on BP neural network optimized search algorithm

Abstract

Abstract The accuracy of traditional soil moisture prediction method is low and the training period is long. In this paper, the BP neural network prediction model is studied, and a longicorn beetle search algorithm (BAS) optimized BP neural network prediction method is proposed. Spelman rank correlation coefficient method is used to analyze the correlation between the change of soil moisture and each variable. In this paper, evaporation, ground temperature, precipitation, air pressure, sunshine hours, air temperature and wind speed are taken as independent variables to analyze the Spelman correlation with soil moisture, and the correlation between each variable and soil moisture change is obtained. The longicorn beetle search algorithm is used to optimize the initial weight and threshold of BP neural network, and the prediction model of BAS-BP neural network is established. The soil moisture prediction model of BAS-BP is compared with different prediction models of GA-BP and BP. The results show that the average absolute error and average relative error of BAS-BP are 9.1936 and 0.1333 respectively, which is lower than that of GA-BP and BP model. The shortcomings of long training time and slow convergence speed are overcome by BAS-BP neural network, and the accuracy of prediction is improved.