基于改进径向基神经网络的风电叶片模温串级PID控制算法

CASCADE PID CONTROL ALGORITHM FOR WIND TURBINE BLADE MOLD TEMPERATURE BASED ON IMPROVED RBF NEURAL NETWORK

针对风电叶片模具电加热系统中被控对象存在的大惯性、非线性、干扰多等问题,提出一种基于改进径向基(RBF)神经网络的串级PID温度控制方法。首先,采用RBF神经网络结构对常规PID串级控制主回路结构进行优化,在此基础上,引入双动量因子,对主控制回路的输出Jacobian信息进行系统辨识,进而实现对控制器参数的自适应整定;其次,采用Kalman滤波器对主回路的输出噪声进行滤波,以消除外部扰动对系统辨识效果的影响;最后,搭建电加热试验平台,通过现场试验对上述算法的控制效果进行分析。仿真及现场试验结果表明:改进的径向基神经网络串级PID温度控制系统相较于常规串级控制具有响应快、超调低、抗干扰能力强等优点,且在主控制回路中的Kalman滤波算法能有效消减系统的输出噪声,可在很大程度上提高控制性能。

To solve the problems of large inertia,non-linearity,and many disturbances of the controlled object in the electric heating control system of wind turbine blade mold,a cascade temperature control system based on improved radial basis(Radial-Basis Function,RBF)neural network was proposed.Firstly,the RBF neural network structure was used to optimize the main loop of the conventional PID cascade control.On this basis,the dual momentum factor was introduced to identify the output Jacobian information of the main control loop,and then achieve the adaptive tuning of the controller parameters.Secondly,the Kalman filter was adopted to perform algorithm processing on the output noise concentration of the main loop to eliminate the influence of external disturbance on the system identification effect.Finally,an electric heating test platform was built and the control effect of the algorithm was analyzed through field tests.Simulation analysis and field test results show that the improved RBF neural network cascade PID temperature control system was compared with the conventional cascade control has quick response and lower,and the advantages of strong anti-interference ability.In the main control circuit,Kalman filter algorithm can reduce the system output noise effectively and improve the control performance greatly.