基于GA-SVM的电能质量扰动分类方法研究

Research on Power Quality Disturbance Classification Based on GA-SVM

电能质量扰动的准确分类,是电能品质改善和治理的重要决策依据.为解决支持向量机(SVM)分类器在多分类问题中的不足,采用模式识别领域中聚类分析的思想,提出了一种基于遗传算法(GA)的SVM决策树多分类电能质量扰动识别方法.该方法首先对参数进行初步最优值筛选,将得到的初步最优值作为遗传算法初始值进行编码,根据设立的适应度函数完成GA中的选择、交叉、变异等操作,进一步搜索最优值,再以最优决策树构建SVM分类器,最终实现SVM的多分类.仿真结果表明,相比未经优化的SVM模型,基于GA算法优化的SVM具有较高的识别精度和抗噪能力.

The accurate classification of power quality is an important basis for the improvement and management. In or- der to solve the lack of support vector machine （SVM） classifier in multi classification problem, this paper proposes a new power quality disturbance classification method based on genetic algorithm （GA） and SVM with the cluster analysis in pat- tern recognition. Firstly, the method is used to select the optimal parameters, and then the results are encoded as the initial values of the genetic algorithm. According to the fitness function, the selection, crossover and mutation operations of GA are completed, and the optimal values are searched further. Finally the optimal decision tree is used to construct the SVM multi classifier. The simulation results show that the optimized SVM based on GA has higher recognition accuracy and anti noise ability than the non-optimized SVM model.