基于改进FCM的超超临界机组过热器T-S神经网络模型辨识

构建了过热器多输入单输出的T-S神经网络模型,并针对输入变量空间划分问题提出了一种改进FCM算法。通过确定高斯型隶属函数参数,实现模型结构参数辨识,利用递推最小二乘法完成模型后件参数辨识。对华能海门电厂百万机组过热器模型辨识进行仿真,结果表明此方法具有较好的辨识效果,辨识出的过热器模型具有较好的精度和泛化能力。

T-S Neural Network Model Identification for Superheater of Ultra-supercritical Units Based on Improved FCM

The paper constructs of the T-S neural network model for the superheater with multiple inputs and single output,and presents an improved FCM algorithm aiming to solve the inputs’ space division problem.The function parameters of the Gaussian membership are obtained to identify the model structure and the recursive least squares method is adopted to identify model parameters.Simulations are implemented for the superheater of Haimen ultra-supercritical units.Results show that the improved method has good performance in model identification,and the identified models have preferable accuracy and generalization ability.