Precision fertilization method of field crops based on the Wavelet-BP neural network in China

Abstract

Abstract In the process of agricultural production, the yield of field crops is affected by the interaction of various factors. This relationship between soil nutrients, fertilization, and yield is very complicated. It has strongly non-linear and black box characteristics and is difficult to quantify with traditional analytic methods. This paper proposes a method for the precision fertilization of maize based on the wavelet-BP neural network. Firstly, the existing “3414″ experimental data was chosen and interpolated as the performing data for the modeling. Then, the general condition of low frequency and details of high-frequency of the yield were calculated via the wavelet decomposition and reconstruction method. After that, the wavelet-BP neural network was established. Three different BP neural sub-networks for each component of the yield after wavelet analysis were established, and the output of each sub-network was summed to obtain the predicted corn yield. The results show that the model that combines wavelet analysis with the BP neural network achieves better performance than the traditional BP neural network, Support Vector Machine (SVM), and Random Forest regarding accuracy and stability, which indicates the feasibility of this method. Finally, the best fertilizer amount that achieves the maximum yield or maximum profit is calculated based on nonlinear programming. The proposed method of the wavelet-BP neural network provides a new technique for precision fertilization research and enriches the existing fertilization system. The application of precision fertilization based on the wavelet-BP neural network has important practical significance that guides and increases maize production, while reducing production cost and agricultural pollution. At the same time, it also provides a way to solve similar problems in the entire field of cleaner agricultural production.