基于BAS-BPNN模型的季节性冻融期土壤含水率预测

Prediction of Soil Water Content during Seasonal Freeze-Thaw Period Based on BAS-BPNN Model

季节性冻融期土壤含水率对干旱半干旱区春耕下种及作物生长起决定性作用,土壤含水率预测对了解土壤墒情、实施春灌具有一定的指导意义。根据114组冻融期气象观测资料和土壤含水率实测数据,采用主成分分析法在影响土壤含水率的9个影响因子中提取出7个主要因子,建立了基于天牛须搜索(BAS)算法的优化BP神经网络(BPNN)模型,即BAS-BPNN模型,分析了模型的土壤含水率预测结果,并与同类的BP模型、PSO-BPNN模型和GA-BPNN模型的预测结果进行了对比。结果表明:BAS-BPNN模型训练集预测值与实测值的决定系数为0.9178,相对误差为8.65;测试集预测值与实测值的决定系数为0.9096,相对误差为9.08。BAS-BPNN模型比其他3种模型的决定系数高且相对误差小。与其他优化算法相比,BAS-BPNN模型寻优及收敛速度快,对冻融期土壤含水率预测精度更高,不失为一种较好的预测冻融期土壤含水率的方法。

Soil water content plays a decisive role in spring cultivation and crop growth in arid and semi-arid areas,and the prediction of soil water content has certain guiding significance in understanding soil moisture and implementing spring irrigation during the seasonal freeze-thaw period.In this paper,the seven major factors affecting soil water content were selected by the method of principal component analysis based on 114 groups of meteorological observation material and the measured data during freeze-thaw period,and it was set as the input layer of the BP neural network(BPNN)model optimized by the beetle antennae search(BAS)algorithm to simulate and analyze the model of prediction of soil moisture content.Its prediction results are compared with the predicted results of similar BPNN model,the PSO-BPNN model and GA-BPNN model.The results show that the determination coefficient between the simulated value and the measured value of the training set in the model is 0.9178,and the relative error is 8.65%.The determination coefficient between the predicted value and the actual value of the testing set is 0.9096,and the relative error is 9.08%.The BAS-BPNN model has a higher determination coefficient and less relative error than the other three models.Compared with other optimization algorithms,the BAS-BPNN model has a faster optimization and convergence speed,and has higher accuracy in the prediction of soil water content during freeze-thaw period.Therefore,it can be regarded as a better method for predicting soil moisture content during freeze-thaw period.