| 引用本文: | 余洋,孙梓朔,王中晶,李君卫,庞淇文,樊蕊.基于BKA-GMM算法的需求侧高价值用户划分及筛选方法[J].电力自动化设备,2025,45(9):165-173. |
| YU Yang,SUN Zishuo,WANG Zhongjing,LI Junwei,PANG Qiwen,FAN Rui.Demand-side high-value user classification and screening method based on BKA-GMM algorithm[J].Electric Power Automation Equipment,2025,45(9):165-173. |
|
| 本文已被：浏览 1145次 下载 804次 |
码上扫一扫！ |
|
|
| 基于BKA-GMM算法的需求侧高价值用户划分及筛选方法 |
|
余洋1,2, 孙梓朔1,2, 王中晶1,2, 李君卫1,2, 庞淇文1,2, 樊蕊1,2
|
|
1.华北电力大学(保定)新能源电力系统全国重点实验室，河北 保定 071003;2.华北电力大学(保定)河北省分布式储能与微网重点实验室，河北 保定 071003
|
|
| 摘要: |
| 针对调峰场景中传统用户分类方法评价标准单一导致的聚类效果不佳、大数据处理效率与精度失衡、划分结果可解释性较低等问题，提出基于黑鸢算法(BKA)优化高斯混合模型(GMM)算法的需求侧高价值用户划分与筛选方法。为解决用户分类不明确和系统匹配不良的问题，结合用户心理学与峰、谷分时电价的负荷转移率模型，引入改进S形函数的用户负荷削减率，构建用户激励-潜力模型；针对传统聚类算法在大规模用户场景下精度与效率不足的问题，提出BKA-GMM算法，预先确定最佳聚类数和正则化参数，结合基于峰、谷时段调度潜力的划分依据，实现用户的初步分类；为提高分类结果的解释性，根据潜力系数及3个用户群特征制定筛选条件，精确识别出削减型、灵活型和复合型的高价值用户。仿真结果表明，将所建模型作为划分条件，显著提升了用户划分效率和参与实际响应的概率，所提筛选方法不仅加快了划分速度，提高了准确性，而且清晰地筛选出了不同类型的高价值用户。 |
| 关键词: 调度潜力 高价值用户 BKA优化算法 GMM聚类分析 用户划分 需求响应 |
| DOI：10.16081/j.epae.202506012 |
| 分类号: |
| 基金项目:国家重点研发计划项目(2018YFE0122200)；河北省在读研究生创新能力培养资助项目(CXZZBS2025192) |
|
| Demand-side high-value user classification and screening method based on BKA-GMM algorithm |
|
YU Yang1,2, SUN Zishuo1,2, WANG Zhongjing1,2, LI Junwei1,2, PANG Qiwen1,2, FAN Rui1,2
|
|
1.National Key Laboratory of Renewable Energy Power System, North China Electric Power University(Baoding),Baoding 071003, China;2.Hebei Provincial Key Laboratory of Distributed Energy Storage and Microgrid, North China Electric Power University(Baoding),Baoding 071003, China
|
| Abstract: |
| Aiming at the problems of poor clustering effect, imbalanced efficiency and accuracy of big data processing, and low interpretability of classification results caused by the single evaluation criterion of the traditional user classification methods in peak-shaving scenario, a demand-side high-value user classification and screening method based on Gaussian mixture model(GMM) algorithm optimized by black kite algorithm(BKA) is proposed. In order to solve the problems of unclear user classification and poor system matching, a user incentive potential model is constructed by combining the user psychology and the load transfer rate model of peak and valley time of use electricity price and introducing the user load reduction rate of improved S-shaped function. Aiming at the problem of insufficient accuracy and efficiency of the traditional clustering algorithms under large-scale user scenario, the BKA-GMM algorithm is proposed, which pre-determines the optimal clustering number and regularization parameters, and combines the classification criteria based on peak and valley scheduling potential to realize preliminary user classification. In order to improve the interpretability of classification results, the screening criteria are developed according to the potential coefficient and three user group characteristics, which accurately identifies reducing, flexible, and composite types of high-value users. The simulative results show that the user classification efficiency and the probability of participating in actual response are significantly improved by taking the constructed model as a partitioning condition, and the proposed screening method not only enhances the classification speed and accuracy, but also clearly screens different types of high-value users. |
| Key words: dispatching potential high value user BKA optimization algorithm GMM clustering analysis user classification demand response |
