基于改进共轭梯度优化BP神经网络的风电机组变桨距控制

Wind turbine variable pitch control based on BP neural network with improved conjugate gradient optimization

根据共轭梯度算法和传统BP神经网络的变桨距控制器的原理,针对兆瓦级风电机组变桨距控制设计了一种改进共轭梯度优化BP神经网络的变桨距PID参数自整定控制器,此控制器采用改进共轭梯度法修正BP神经网络的权值和阈值,实现BP神经网络变桨距PID控制器的在线整定。在Matlab/Simulink中仿真,仿真结果表明,采用此变桨距控制器可以在额定风速之上快速响应,在相同风速状况下使发电机桨距角调节命令更加准确,风轮转速更加平稳,输出功率维持在额定功率附近,取得了很好的变桨距控制效果。

According to the principle of variable pitch conjugate gradient algorithm and the traditional BP neural network controller, an variable pitch PID parameters self-tuning controller with improved conjugate gradient optimization BP neural network is proposed for mega-watt graded wind turbine pitch control. This controller uses an improved conjugate gradient method to correct BP neural network weight and threshold for online tuning of pitch BP neural network PID controller. The Matlab/Simulink simulation shows that use of this variable pitch controller could quickly response to the wind speed above rated value; at the same wind speed condition, the generator pitch modulation command is more accurate ,rotating speed of the wind wheel is more stable, the output power is maintained at near rated power, which has a good pitch control effect