摘要
针对水轮机调节系统的高阶、非线性及非最小相位的特点,设计了基于非线性自回归动态神经网络(NNARX)的水电机组预测控制系统。为了更好地得到过程参数及对象模型,先利用NNARX神经网络对水电机组整体进行辨识,再利用此网络对水电机组进行预测控制,并给出了模型算法及处理过程。由于NNARX动态网络误差曲面比较复杂,利用L-M算法对其进行训练。仿真结果表明,基于NNARX的动态神经网络模型具有很好的收敛性,辨识精度高,预测控制效果良好。
This paper designed a neural network predictive controller （NNPC） of hydraulic generator in order to solve the difficulties caused By its own characteristics which include high order, nonlinear and non-minimum phase. For obtai ning the mathematical model and process parameters, neural network autoregressive with exogenous input （NNARX） is used in this paper. The predictive controller can work well after the neural network was trained. All the processing procedure is elaborated. Due to the complex error surface produced by NNARX, L-M algorithm is used in the training process. The simulation results indicate that NNARX has good convergence, high accuracy and good predictive control effect.
出处
《水电能源科学》
北大核心
2014年第3期192-195,191,共5页
Water Resources and Power
基金
国家自然科学基金项目(51379080)
湖北省自然科学基金项目(2010CDB02503)
湖北工业大学科研基金项目(BSQD12107)
关键词
神经网络
水电机组
辨识
预测控制
neural network
hydroelectric generator
identification
predictive control
