基于遗传算法的BP神经网络在水利定额编制中的应用

Application of BP Neural Network Based on Genetic Algorithm in Water Resources Quota

针对BP神经网络易陷入局部最优和遗传算法全局搜索速度过慢的缺点及水利定额编制中存在非线性和复杂性的实际状况,提出采用遗传算法(GA)优化BP神经网络在水利定额编制中的问题。实例分析表明,优化后模型(GA-BP神经网络)结合了BP神经网络的非线性逼近、局部寻优能力和遗传算法的全局搜索特性,在稳定性、预测精度、收敛速度上均优于BP神经网络,可运用于水利定额编制。

Aiming at the shortcomings of BP neural network being easy to fall into local optimum and the global search of genetic algorithm being too slow, the actual situation of non-linearity and complexity in water conservancy quota system, BP neural network optimized by genetic algorithm （GA） was proposed to solve water quota system. The case study shows that the optimized model （GA-BP neural network） combines the non-linear approximation of BP neural net- work and the local search ability and the global search of GA. It is concluded that the GA-BP model is superior to BP net- work in stability and prediction accuracy and convergence rate, and it is better to use in water quota preparation.