利用居民用电量的住房面积预测算法设计

Design of a Housing Area Prediction Algorithm Based on Residential Electricity Consumption

从供电企业的角度出发,为了向用户提供相似邻里用电比较服务,引导居民节约用电,提出了利用居民用电量的住房面积预测算法,在对居民用电量和住房面积进行数据清洗的基础上,分别以年制冷与取暖电量、年基本生活电量以及最近12个月有效月电量为建模特征量,借助支持向量回归、神经网络、K-means聚类等算法工具,构建了 4种模型,比较并验证了模型的效果,其中结合K-means聚类的神经网络模型预测效果最好,平均预测偏差为19.755%。结果表明,该算法能通过居民用电量对住房面积进行较准确的预测,为相似邻里用电比较服务提供重要支持。

From the perspective of power supply companies, in order to provide users with similar neighborhood electricity usage comparison services for guiding residents to save electricity, a housing area prediction algorithm based on residential electricity consumption is proposed. On the basis of data cleaning of residential electricity consumption and housing area, using the annual cooling and heating power, annual basic living power and the effective monthly electricity consumption in the last 12 months as modeling features, four models are built with support vector regression, neural network,K-means clustering and other algorithm tools and the effect of models are compared and verified. The neural network model combined with K-means clustering has the best prediction effect, and the average prediction deviation is 19.755%. The results show that this algorithm can accurately predict the housing area through residential electricity consumption, and provide important support for similar neighborhood electricity usage comparison services.