应用神经网络的变流器波形复合控制策略

Combined Control Strategy Application of Neural Network for Converter Waveform

针对港口现有岸电电源产品所采用控制策略的不足,研究分析了岸电电源PWM可逆变流器数学模型在DQ坐标系下的特点,提出了一种基于改进的重复控制和神经网络内模控制的变流器输出波形复合控制策略.采用BP神经网络结构作为内模控制器的预估模型和控制器,神经网络预估模型可在线学习建立与被控对象相匹配的精确模型,神经网络控制器动态响应快,输出无静差,扰抗性好.实验证明,应用该复合控制策略的系统整流功率因数接近于1;供电非线性混合负载输出波形失真率低于2%;动态响应快,在2个周期内恢复稳定输出.

To address the shortcoming of control strategy in shore power products used in harbor at present,this paper analyzed the characteristics of PWM Converter mathematic model based on DQ rotating frame.The author put forward the modified multiple control strategy with the blend between repetitive control theory and Neural Network internal model control theory.The BP neural network was adopted as the estimated model and control of internal control,which can establish the precise model of the controlled target on line.Meanwhile,fast dynamic response,with no static error in output and good interference rejection,can be achieved.The experiments have verified that the output waveform is stable and accurate,the load adaptability is powerful and the dynamic performance is good.