基于高斯混合模型的居民聚合响应潜力多重置信评估

针对居民用电负荷与源端出力多变背景下传统电力系统运行灵活性不足的现实问题，需求响应可有效提高系统运行灵活性与安全经济效益，价值尤为凸显，而响应潜力的精细化评估是其重要基础支撑。文中提出一种在缺少历史响应数据支撑时基于高斯混合模型的聚合响应潜力评估方法。首先，通过家庭及相似日的两次聚类分析选取典型样本数据，强化数据的代表性；然后，引入高斯混合模型精准挖掘家庭用电行为的概率分布，形成单个家庭的响应潜力；最后，自下而上加权汇总，实现多重置信情景下聚合需求响应潜力的评估。实验分析表明该方法能够仅从历史用电数据中挖掘出小时级的居民需求响应潜力信息，充分反映用电负荷分布及响应潜力分布特征，并通过对比分析验证了两次聚类选取典型样本数据的有效性。

Evaluation of residential demand response potential under multiple confidence scenarios based on Gaussian mixture model

Given the practical issues of the traditional power system''s insufficient operational flexibility in the face of changing residential power load and source-side output, demand response (DR) can effectively improve the flexibility, safety and economic benefits of system operation. Its value is especially noticeable , among which the refined assessment of DR potential is an important basic support. This paper proposes a method for evaluating aggregate DR potential in the absence of historical DR data based on Gaussian mixture model. Firstly, the typical data are selected through two-stage clustering of households and similar days to improve data representativeness. Then the Gaussian mixture model is introduced to accurately explore the probability distribution of household electricity consumption behavior and calculate individual households'' DR potential. Finally, the bottom-up weighted aggregation is implemented to evaluate the aggregate DR potential under multiple confidence scenarios. According to empirical analysis, this method can mine hourly information of residential DR potential from historical electricity consumption data, which can reflect the distribution of power load and DR potential. Comparative analysis is used to validate the validity of typical data selection by two-stage clustering.