区域尺度住宅建筑日用电负荷模型构建方法研究

面对新时代能源发展形势，有效整合电力用户、电网企业及供应商信息，全面深度分析“源网荷储”特性是推进能源互联网发展的重要举措，模拟建筑用电这一典型负荷已尤为重要。关注区域规模居民用电负荷，提出用电负荷数据预处理分析方法和区域尺度住宅建筑日用电负荷模拟方法。区域尺度的用电数据在空间规模和时间跨度上均具有高维特征，分析大量数据并再现区域规模用电负荷是其主要难点。为此，提出结合自编码和k近邻算法的数据异常值剔除方法。在数据预处理的基础上，提出基于聚类分析的区域尺度住宅日用电负荷模型和模型检验方法，以单个住户的日均用电和全年最大日负荷为指标聚类，基于聚类分析结果提出随机用电模型，模拟区域尺度住宅建筑逐户逐日用电负荷。应用华东某重点城市智能电表采集的整年居民用电数据开展研究分析案例，实现区域住宅建筑居民日用电负荷的模拟再现。提出“数据异常值预处理、聚类分析、模型构建和检验”系列研究分析方法，可有效满足能源互联网建设对大量末端用户用电负荷的模拟分析需求。

Residential Electricity Load Model Construction in District Scale

Faced with the energy development situation in the new epoch, it is key measures for promoting the development of power energy Internet to effectively integrate the information of power users, grid companies and suppliers, and to comprehensively analyze the source network load and storage characteristics. Therefore, it is crucial to simulate the typical electricity load in buildings. This study focuses on the electricity consumption of different households in district scale and proposes a preprocessing method of electricity load of district-scale users, and a residential daily electricity load simulation method. The electricity consumption data of urban scale has high-dimensional characteristics in both spatial scale and temporal span. Characterizing the district-scale electricity consumption is the principal difficulty. To this end, this study proposes a method integrating auto-encoding and neighbor clustering algorithms for data outlier detection for data with high-dimensional characteristics at the spatial and temporal scales. Based on the results of preprocessing, the study proposes a district-scale residential electricity load simulation model based on clustering analysis, and the evaluation method. Clustering analysis was performed on the average daily electricity consumption and the daily maximum load of the whole year of single household. Based on the clustering analysis, a random electricity model is proposed to simulate the daily electricity consumption of residential buildings on the district scale. In this study, part of the household in one key city in East China is used as a case study to demonstrate the proposed method. The raw data comes from the daily electricity consumption measurement of multi-family smart meters, lasting for one year. To conclude, this study builds a complete research method of data outlier preprocessing, clustering analysis, model construction and model verification, it can effectively solve the needs of power energy Internet construction for the end use energy consumption analysis of a large amount.