基于BP神经网络PID控制的PMLSM调速系统设计

Design of Speed-regulation System of PMLSM Based on BP Neural Network PID Control

针对永磁直线同步电动机提升系统的非线性、时变性、易受扰动等特性,在所建立的永磁直线同步电动机d-q轴动态数学模型的基础上,设计了一种改进型BP神经网络PID控制的PMLSM调速系统。该系统将BP神经网络算法中固定的学习速率改为自适应可调,同时添加动量项以减小学习过程中的振荡趋势,极大地改善了算法的收敛速度,避免了网络落入局部最小值的结果。仿真结果表明,使用改进的BP神经网络PID控制器可使PMLSM调速系统的调节时间和超调量大幅减小,响应速度加快,使提升系统具备较好的动态性能和较强的鲁棒性。

For characteristics of nonlinearity, time-varying volatility and nonstationarity to disturbance of hoisting system of PMLSM, an improved PMLSM speed-regulation system was designed on basis of established PMLSM d-q axis dynamic mathematical model and t3P neural network PID control. The system modifies the fixed learning rate in BP neural network to a self-adaptive one, and adds momentum to reduce oscillation tendency in learning process, so it greatly improves convergence speed and avoid that network falls into a local minimum. The simulation result showed that using the improved BP neural network PID controller can make PMLSM speed-regulation system reduce adjusting time and overshoots greatly and accelerat response speed, and make the hoist system have better dynamic performance and robustness.