基于K-means聚类的短期负荷预测研究

本文的中心思想是采用多种方法组合进行预测,对30个种类的用户全年8760个小时的用电量做全年最后一天的日负荷预测。本文根据聚类算法对数据进行归纳整理再拆分的特点先将数据聚成5类,根据5类用户曲线的特点分别采取不同的适用于各曲线的方法进行预测。本文在k-means聚类的基础上又采取了三种方法进行预测。预测出结果的MAPE值不高,证明预测的精度的准确定。聚类分析的归纳整理功能为预测节省了大量时间,提高了预测速度,而以此为基础采用不同适合的方法进行预测又提高了预测精度,这在一定程度上解决了当前预测快速与准确无法兼具的问题。

Short-term Load Forecasting Based on K-means Clustering

The central idea of this paper is to use a combination of methods to predict the daily load forecast for the last day of the year for a total of 8760 hours of electricity consumption for 30 types of users.According to the clustering algorithm,the data is summarized and re-segmented.The data is first clustered into five categories.According to the characteristics of the five types of user curves,different methods suitable for each curve are used for prediction.Based on the k-means clustering,this paper adopts three methods to predict.The MAPE value of the predicted result is not high,which proves the quasi-determination of the accuracy of the prediction.The inductive sorting function of clustering analysis saves a lot of time for forecasting and improves the forecasting speed.Based on this,using different suitable methods for forecasting and improving forecasting accuracy,this solves the current forecast quickly and accurately.Both problems.