改进遗传K均值算法在负荷特性分类的应用

Application of Improved Genetic and K-means Algorithm on Load Characteristics Classification

负荷特性聚类是负荷建模的基础工作,精确的负荷特性分类对负荷建模十分重要。为适用于实际电网规划等对负荷分类精度要求更高的领域,针对负荷聚类时存在的聚类数目不确定、初始条件选择敏感的问题,提出了一种改进遗传算法的负荷特性分类新方法。该方法在遗传算法全局搜索能力的基础上,结合K均值聚类方法进行聚类分析,得到最优分类结果。同时,针对聚类中心敏感的问题,遗传算法中采用可变长编码方案进行优化,动态寻找最优聚类数目,确定最优聚类中心。实例分析结果表明,用改进遗传算法对负荷特性进行分类,能够有效避免初始条件对分类结果的影响,可以获得良好的分类效果。

Load characteristics classification is the basis of load modeling, precise classification is important for load modeling. To apply to actual power grid planning etc demands for high accuracy of classification. Directing at the clustering numbers are uncertain and the initial conditions are sensitive to choose, this paper proposes a new method based on improved genetic algorithm. Combined K-means clustering with genetic algorithm, this method utilizes the global search ability of genetic algorithm, and then gets the optimal classification results. On account of clustering centers are sensitive, optimizing this method by modified variable string length scheme, dynamic searching for the optimal number of cluster, this method can confirm optimal cluster centers. Experimental results demonstrate that utilizing this method to classify the characteristics of power loads can effectively avoid the influence of initial conditions and obtain good classification results.