【摘要】 通过分析控制器参数学习率和控制器性能之间的关系,设计一种基于可变学习速率反向传播算法VLRBP和模糊神经元网络的变频空调控制系统。该系统不仅可以通过反传误差信号训练控制器参数,而且可以根据网络的当前状态朝最优化方向调整控制器参数的学习率。实验结果表明,该控制系统不仅比传统的空调PID控制器和模糊控制器具有更好的控制性能,而且相比基于标准BP算法和动量BP算法的模糊神经网络控制系统,也具有更快的收敛速度和更好的控制精确度。更多还原

【Abstract】 After analyzing relationship between learning of rate of controller parameters and the controller performance, a control system for inverter air-conditioner is designed, which is based on variable learning rate back-propagation arithmetic and fuzzy neural network. In this system, the controller parameters are trained by back-propagation error signal and the learning rate of parameters are adjusted by optimization direction of network’s current state. The experiment shows that the new control system not only has better control performance than conventional PID controller and fuzzy controller, but also has a faster constringency rate and better control definition than the fuzzy neural network control system based on normal BP arithmetic or momentum BP arithmetic.更多还原