基于支持向量机的海量电力数据智能分类方法

Intelligent classification method of massive power data based on support vector machine

针对电力公司海量数据分类问题,提出一种改进的k-means数据分类方法。在k-means算法的基础上,应用PCA对k-means算法进行降维处理,用canopy算法优化最佳簇集数、初始聚类中心。然后,应用改进的k-means算法对居民用户用电进行聚类;最后以该聚类结果为基础,建立LSTM预测模型。通过LSTM预测模型对某小区90户居民用电数据进行仿真实验,并对比分析了传统聚类、改进聚类和不适用聚类下LSTM三种模型的预测结果。结果表明,未使用任何聚类算法构建的LSTM模型在进行电力负荷预测时,预测结果的精度最低;应用改进的k-means算法构建的LSTM模型预测结果精度最佳。

Aiming at the problem of massive data classification in power companies,an improved k-means data classification method is proposed.On the basis of K-means algorithm,PCA is used to reduce the dimension of K-means algorithm,and canopy algorithm is used to optimize the optimal number of clusters and the initial cluster center.Then,the improved k-means algorithm is applied to cluster the power consumption of residential users;finally,based on the clustering results,the LSTM prediction model is established.Through the application of LSTM prediction model to 90 households’electricity consumption data of a residential area,the prediction results of three kinds of LSTM models under traelitional clustering,improved clustering and inapplicable clustering are compared and analyzed.The results show that the LSTM model without any clustering algorithm has the lowest accuracy in power load forecasting,and the LSTM model constructed by the improved k-means algorithm has the best prediction accuracy.