基于自适应径向基函数神经网络的无刷直流电机直接电流控制

提出了基于自适应径向基函数（Radial Basis Function)神经网络的无刷直流电机直接电流控制新方法。该方法构造了一个隐层节点初始个数为零的RBF网络，通过在训练过程中不断地按照自适应算法添加和删除隐层单元，形成一个结构简单、紧凑的RBF网络来实现电机电压、电流与功率开关导通信号之间的非线性映射，直接控制功率开关的通断，实现无位置传感器的直接电流控制。网络训练采用离线训练和在线训练相结合的方法。首先利用来自实验数据的训练样本按给出的自适应算法对网络进行离线训练，确定RBF网络隐层节点的个数及位置；再按递推最小二乘法(RLS)在线修正隐层与输出层之间的连接权：最后，用数字处理器(DSP)实现在线控制算法。实验结果表明，该控制方法具有较高的鲁棒性和控制精度。

DIRECT CONTROL OF CURRENTS BASED ON ADAPTIVE RBF NEURAL NETWORK FOR BRUSHLESS DC MOTORS

In this paper, a novel approach based on an adaptive Radial Basis Function (RBF) neural network is proposed, and through this approach, the direct control of the currents for a brushless DC motor without position sensors is realized. In the proposed RBF network, there is no hidden units at the beginning, and during the process of learning, they are increased or decreased according to an adaptive algorithm so that a RBF network is built with a much simpler and tighter structure to form an efficient nonlinear map from terminal voltages and phase currents of the motor to the states of the power switches, facilitating the elimination of the position sensors. The RBF network is trained both offline and online. In the offline methods with the training data collected from experiments, the number and locations of the hidden units of the RBFNN are obtained; while online learning, the weights between the hidden layer and the output layer are updated according to the recursive least squares (RLS) algorithm. The RBF network is used to realize the direct control of currents for the BLDC motor in experiments based on DSP, and the results show that with the proposed method, and higher control precision and robustness is obtained.