基于GA-BP神经网络的水轮机非线性建模方法研究

Research on Nonlinear Modeling Method of Hydraulic Turbine Based on GA-BP Neural Networks

建立精确的水轮机模型是水轮机调节系统有效建模仿真的关键。运用基于遗传算法改进反向传播神经网络的GA-BP神经网络对水轮机工作特性进行非线性拟合建模。详细介绍了利用水轮机模型综合特性曲线与飞逸特性曲线获取水轮机流量特性与力矩特性样本数据的方法,并对基本样本数据进行补充延伸。结合BP神经网络与遗传算法两者优点构建了GA-BP神经网络,利用所得样本数据进行训练,获得了基于GA-BP神经网络的水轮机非线性模型,并与传统BP神经网络在水轮机流量特性和力矩特性拟合效果上进行对比试验。仿真结果验证了论文提出方法的可行性和优越性。

The establishment of accurate hydraulic turbine model is a key to the effective modeling and simulation of hydraulic turbine gover- ning system. GA-BP neural networks based on back propagation neural networks with genetic algorithm is used to model the nonlinear charac- teristics of the hydraulic turbine. This paper presents the method of obtaining the sample data of hydraulic flow characteristics and torque characteristics of turbine by using the comprehensive characteristic curve and the runaway characteristic curve of hydraulic turbine model. And the basic sample data has been extended as much as possible. GA-BP neural network is developed with BP neural networks and genetic algorithm. The nonlinear model of hydraulic turbine based on GA-BP neural network is obtained by using the sample data for training. Com- pared with the BP neural networks in the simulation of the flow characteristics and torque characteristics of hydraulic turbines. The simulation results show that the proposed method is feasible and superior.