BP神经网络与GA-BP农作物需水量预测模型对比

Comparison between BP neural network and GA-BP crop water demand forecasting model

农作物需水量预测是制定合理灌溉制度的重要依据.针对BP神经网络的不足,利用遗传算法(GA)具有全局搜索能力强的特点,建立基于GA-BP神经网络的农作物需水量预测模型.以广州辣木农庄试验田农作物作为研究对象,结果表明:基于BP神经网络农作物需水量预测模型测试集均方误差和确定性系数分别为0.037和0.648;GA-BP神经网络农作物需水量预测模型测试集均方误差和确定性系数分别为0.013和0.882,GA-BP农作物需水量预测模型收敛速度、确定性系数和性能均优于BP农作物需水量预测模型.

Forecasting crop water requirement is an important basis for formulating a reasonable irrigation system. In view of the deficiencies of BP neural network, genetic algorithm（GA）has the characteristics of strong global search capability, and a prediction model of crop water requirement based on GA-BP neural network was established. Taking the experimental crops of the Lamu Farm in Guangzhou as the object of research, the results show that the mean square error and certainty coefficient of the test set of crop water requirement prediction model based on BP neural network are 0.037 and 0.648 respectively. The square error and the certainty coefficient of the rest set of crop water demand forecasting model based on the GA-BP neural network are 0.013 and 0.882 resiectively. The GA-BP crop water demand forecasting model has a convergence rate, certainty coefficient and performance better than the BP crop water demand forecasting model.