基于支持向量回归机的同步发电机励磁电流预测方法

为了能更准确、容易地在线诊断出同步发电机转子绕组匝间短路故障,提出了一种基于支持向量回归机的励磁电流预测方法.利用同步发电机正常运行时不同工况下的机端电压、有功功率、无功功率和励磁电流来建立发电机励磁电流的支持向量回归机预测方法.利用该方法预测正常运行时所需励磁电流,并与在线实测的励磁电流进行比较,误差(相对误差)超过阈值就诊断为发生匝间短路故障.通过微型同步发电机动态模拟实验表明,该方法的精度优于BP神经网络法和遗传规划法.

Prediction Method of Synchronous Generator Field Current Based on Support Vector Regression Machine

For more accurate and easy online diagnosis of synchronous generator rotor winding inter-turn short-circuit fault,this paper puts forward a novel field current prediction method based on support vector regression（SVR） machine. We use terminal parameters （voltage, active power, reactive power） and rotor field current under different fault-free operating conditions of synchronous generator to create this SVR pre- diction method. If online measured terminal parameters are input to this method, a predicted field current will be obtained. Then, by comparing the predicted field current with the corresponding online measured field current, if error（relative error） exceeds a specific threshold, a synchronous generator rotor winding in- ter-turn short-circuit fault is diagnosed to occur. The micro-synchronous generator dynamic simulation re- suits show that this method has better accuracy than BP neural network method and genetic programming （GP） method.