基于NNARX水轮机调节系统辨识的预测控制与故障诊断

Researchon the NNARXidentification and prediction control of the hydraulic turbine system

针对水轮机系统参数不确定性、非线性及非最小相位特点,基于非线性动态神经网络无限逼近能力对水轮机系统采用L-M算法整体辨识,且该算法对两层小网络结构训练学习是最有效的方法。利用辨识的模型计算优化未来的输出并与传统PID控制进行对比和已训练好的网络函数对水轮机故障进行判断。仿真结果表明,该控制器在动态状况下具有很好的鲁棒性、快速性,满足系统内部参数的变化规律,同时此网络函数可以有效的诊断故障。

Because of the uncertainty, nonlinear and non-minimum-phase characteristics of hydraulic turbine, this paper based on theinfinite ability of nonlinear dynamic neural networkusing the algorithm of L-M overall recognition hydraulic turbine system, and the al- gorithm of two layers of small structure training is the most effective way to learn. The identification model is used to calculate the future output and compare with traditional PID control and have trained network functions to judge hydraulic turbine faults. Simulation results show that the controller has good robustness and quickness under dynamic condition, so that satisfy the change rule of interior parameters of the system, as well as the network function can effective fault diagnosis.