中文摘要：

基于逆动力学控制的思想，提出一种带前馈的RBFN逆模型控制策略，并将该控制策略对多变量非线性系统进行了在线解耦与控制；采用3个RBF神经网络分别对被控对象的3个输入进行逆辨识，每相神经网络逆辨识模型反向作为逆控制器模型与每相串联，从而构成3个已解耦的独立的伪线性对象，进而针对3个独立的伪线性对象进行线性控制，从而实现了对三相耦合系统的精确控制；经过实验室模拟调试的实测三相电流波形表明，在电流设定值为2100A时，升温速率为5．3℃／min，电流波动范围≤±10％。

英文摘要：

An inverse control strategy based on RBFN with feed forward was presented in this paper. The non--linear multi--variable system was controlled by an decoupled control method. Three inputs of the plant was identified by three RBFNS. Three dynamic pseudo line- ar systems were formed through The models of controllers and the plants are in series to be decoupled into three independent pseudo--linear plant. So we can use linear method to control those plants. The actual measurement current waveforms show that the heating rate is 5.3℃/ min and current fluctuation range is≤±10% when the current set at 2100A.