Multi-Objective Lower Irrigation Limit Simulation and Optimization Model for Lycium Barbarum Based on NSGA-III and ANN

Lycium barbarum has rich medicinal value and is an important medicinal and economic tree species in China, with an annual output value of 21 billion RMB. The yield and the quality of Lycium barbarum dry fruit are the crucial issues that affect the cultivation of Lycium barbarum and the income of farmers in the Ningxia water shortage area. According to the local acquisition standard of Lycium barbarum, the amount of dry fruit per 50 g (ADF-50) is the key factor in evaluating the quality and determining the purchase price. In order to optimize the irrigation lower limit of automatic drip irrigation system with multiple objectives, the yield and ADF-50 are selected to be optimal objectives. The lower irrigation limits of the automatic drip irrigation system in the full flowering stage, the summer fruiting stage, and the early autumn fruiting stage are optimized by the third generation of non-dominated sorting genetic algorithm (NSGA-III) in this paper. The mathematical relationships between irrigation lower limit and irrigation quantity, irrigation amount, yield, and ADF-50 were established by the water balance model, water production function (WPF), and artificial neural network model (ANN), respectively. The accuracy of the water balance model and ANN were verified by experiments. The experiments and optimization results show that: (1) irrigation quantity and ADF-50 calculated by the water balance model and ANN are accurate, and their Nash-Sutcliffe coefficient are 0.83 and 0.66; (2) In a certain range of irrigation quantity, ADF-50 and Lycium barbarum yield show competitive relation. By solving the NSGA-III optimization model, the lower irrigation limits schemes, which tend to different objectives, and a compromise scheme can be obtained; (3) Compared with the original lower limit of irrigation water, the compromise scheme's yield and quality of Lycium barbarum are improved 10.7% and 8.8% respectively. The results show that the automatic drip irrigation system's lower irrigation limit scheme optimized by the model can improve not only the yield but also the quality of Lycium barbarum. This provides a new idea for establishing the irrigation lower limit of the automatic drip irrigation system in the Lycium barbarum planting area.