基于自适应学习速率的模糊神经网络控制器

Fuzzy Neural Network Controller Based on Adaptive Learning Rate

针对模糊神经网络控制器中很难确定一个最佳学习速率的问题,将带有动量因子的自适应学习速率BP算法引入模糊神经网络控制器中。采用模糊推理自适应调节学习速率,同时引入动量因子,提高系统的收敛速度,并基于Lyapunov定理给出了系统稳定的证明过程。针对同一数学模型,用Matlab编程仿真3种方法的实验结果表明:优化后的模糊神经网络控制器较普通模糊神经网络控制器和模糊控制器具有更优越的控制性能。

Aiming at the difficulty in determining optimal learning rate in the fuzzy neural network controller, the adaptive learning rate’s BP algorithm with momentum factor was introduced to the fuzzy neural network controller.The method has the fuzzy inference adopted to adjust adaptive learning rate and the momentum fac-tor introduced to improve convergence speed of the system,as well as the Lyapunov principle based to provide certification process for the system stability.Regarding the same mathematical model,simulation with Matlab shows that the optimized fuzzy neural network controller outperforms both ordinary fuzzy neural network control-ler and fuzzy controller in the control performance.