基于神经网络的水轮机调节系统非线性仿真

针对水轮机调节系统非线性特性难以准确描述及非线性仿真的复杂性,利用神经网络强大的非线性逼近能力,对水轮机调节系统进行在线非线性实时仿真。对导叶开度、单位流量及单位转速之间的非线性关系建立神经网络结构,用Matlab中的trainlm函数和训练样本对网络进行学习训练。非线性关系的插值拟合过程在PC机中进行,仅将得到的非线性关系的权值和阈值传输给下位机控制核心DSC系统,有效地提高了非线性实时仿真速度。针对HL220型实际机组探讨了基于神经网络的非线性仿真的实现过程,并对结果进行了解析。

Research on Nonlinear Simulation Algorithms of Water Turbine Regulation System Based on Neural Network

Online nonlinear real time simulation is carried out to solve the difficulties of describing nonlinear properties of water turbine regulation system and the complexity of nonlinear simulation by using neural network which has strong nonlinear approximation ability. Firstly the neural network is used to set up the network structures of the nonlinear relationship between the three main parameters that is the guide vane opening, unit flow and unit speed. Then the trainlm function in MATLAB and samples are applied to train the networks. The interpolation fitting process of the nonlinear relation is performed on the PC and only the right weight value and threshold of the nonlinear relationship are transferred to the control core DSC of the local computer, which effectively improved the nonlinear real time simulation. The process of realizing nonlinear simulation based on neural network is applied to HL220 water turbine and it analyzes the results.