BP神经网络在飞轮电池控制系统中的研究  

Study of BP Neural Network on Flywheel Battery's Control

作　　者：蒋宇[1] 李志雄[2] 汤双清[2] 

机构地区：[1]黄山学院,安徽黄山245013 [2]三峡大学,湖北宜昌443002

出　　处：《微特电机》2009年第6期29-32,共4页Small & Special Electrical Machines

基　　金：湖北省自然科学基金(2005ABA294);湖北省教育厅自然科学计划资助项目(2003A001)

摘　　要：飞轮储能用电机的非线性和变参数特性使得传统的PID控制很难取得较好的效果,而人工神经网络在一定的条件下可以逼近任意非线性函数,且具有较强的自学习、自适应、自组织能力,故将其与传统控制相结合构成神经网络自适应控制策略,应用于飞轮储能电机以实现高性能控制。同时采用变学习速率的神经网络学习算法,学习速率随收敛过程误差的大小而自适应地进行调整,可大大加快神经网络学习训练的收敛速度,进一步提高系统动态响应速度。仿真结果表明,在充放电两端,系统的动态响应快,超调小,稳态精度高,鲁棒性强,抗扰动能力强,控制效果较好。The inherent nonlinearity of flywheel battery＇s motor makes it hard to get a good performance by using the conventional PID controller. An artificial neural network （ANN） based adaptive PID controller for flywheel battery was introduced. ANN can approximate to any nonlinear function with arbitrary precision under certain condition, and it also has a strong ability of adaptiveness, self-learning and self-organization. Thus, it is possible to combine the two together to construet an ANN based adaptive PID. Applying it to flywheel battery control, a good performance could be gotten. Meanwhile an adaptive learning algorithm was adapted to adjust the learning rate according to the en＇or. This could increase the convergence speed of ANN and made the system respond quickly. The simulation results demonstrated that a high control performance was achieved in both charging and discharging. The system responded quickly with little overshoot and zero steadystate error, and it was robust to load torque and speed disturbance.

关 键 词：飞轮电池 BP神经网络 PID控制 SVPWM 

分 类 号：TM351[电气工程—电机]