基于遗传算法优化支持向量机的电能质量暂态扰动识别新方法

针对电能质量识别领域中,采用随机参数的支持向量机(SVM)分类器识别随机暂态扰动信号准确率低、优化耗时长等问题,提出一种基于遗传算法(GA)优化SVM识别电能质量暂态扰动(PQD)的新方法(GASVM)。首先,仿真生成具有随机噪声水平和扰动参数的9种PQD信号;接着,通过S变换,提取出6种信号特征构成输入特征向量,用于训练SVM分类器;再采用GA对SVM进行参数寻优,进而获得优化的GA-SVM分类器;最后,采用GA-SVM识别PQD信号。仿真对比试验表明,新方法能准确识别不同噪声环境下的9种PQD信号,分类准确率及优化所需时间均优于PSO优化SVM方法(PSO-SVM)。

New Method of Transient Power Quality Disturbances Recognition Using SVM Optimized by Genetic Algorithm

SVM classifier has low accuracy and long time consuming for identifying random disturbance signals in the field of power quality recognition. A new method of transient power quality identification is proposed based on the SVM optimized by genetic algorithm. Firstly, 9 type of PQD signals with random noise level and perturbation parameters are generated by simulation. Then, 6 type of signal features extracted through the S transform are constituted as the input vectors, which are used to train the SVM classifier. Furthermore, GA is applied to optimize the parameters of SVM and the optimized GA-SVM classifier is obtained. Finally, GA-SVM model is applied to identify PQD signals. The simulation experiments show that the new method is able to accurately identify 9 type of PQD signals in different noise environments. The classification accuracy and the optimization time are better than that of the PSO-SVM method.