基于用户用电行为建模和参数辨识的需求响应应用研究

发布时间：2018-12-27 19:46

【摘要】：需求响应能够实现供电侧与用电侧的有效资源互动,提高系统运行效率和可靠性。有效利用需求侧资源与传统的单一增加调峰电源或加强电网建设的做法相比,可以极大减轻基础设施的投资压力。目前基于激励的需求响应项目大多针对工业用户实施,而居民、商业和办公等用户虽然单体容量小但数量庞大、分布广泛,具有很大的需求响应潜力,且更适合开展基于电价的需求响应项目。因此在智能用电环境下,开展用户用电行为分析,是需求响应应用的基础性工作。本文从单个用户的用电情况监测出发,搭建多维度负荷分类体系,根据典型用电负荷分类,对单个用户的用电负荷进行分解,并对多个用户进行用电行为分析,形成了具备不同价格敏感度的用户聚类。针对已得到的聚合用户,构建实时市场环境,实现对聚合用户的电价调度。首先,分析了用户参与需求响应的必要性和可行性,概述了目前需求响应在国内外的发展应用现状,并从电价调度、用电行为分析和模型参数辨识三个方面总结了学术界的研究现状,为本文的后续研究奠定了基础。其次,对于单个用户,针对目前常用负荷分解方法存在的不足,综合考虑用户常用设备的用途和特性,提出了多维度负荷分类体系,分别从功能维度、时间维度和功率维度进行设备分类,并基于模糊隶属度函数构建用户负荷分解模型,以模糊隶属度表征负荷分解结果的可信度,实现用户用电负荷的分解,为需求响应项目的潜力分析提供基础。然后,对于多个用户,基于改进K-Means聚类算法,在峰谷电价环境下,选取峰谷平各时段的用电量占比和负荷率作为用户的用电特征量,构建用电行为聚类模型,对用户用电行为进行聚类分析,并针对价格转换点前后的用户用电量变化情况,提出了用户价格敏感度的计算方法,构建用户筛选模型,从而确定适合于价格需求响应的敏感对象。最后,对于聚合用户,在基于消费者心理学构建的实时市场环境下,应用支持向量机进行模型参数辨识,构建需求响应电价的计算模型,并对电价调度误差和应用场景进行分析。对比通过人工神经网络和回归分析方法得到的误差,提出将这三种参数辨识法进行加权的组合分析法,构建综合调度模型,提高电价调度的准确度和有效性。
[Abstract]:The demand response can realize the effective resource interaction between the power supply side and the power side, and improve the efficiency and reliability of the system. Compared with the traditional method of increasing peak-shaving power or strengthening power grid construction, the effective use of demand-side resources can greatly reduce the investment pressure of infrastructure. At present, most of the demand response projects based on incentives are aimed at industrial users. Although the individual capacity of residents, businesses and offices is small, the number of users is large, and they are widely distributed, so they have a great potential for demand response. And more suitable for the development of electricity-based demand response project. Therefore, it is the basic work for the application of demand response to carry out the analysis of the user's power consumption behavior in the intelligent power environment. In this paper, a multi-dimensional load classification system is built based on the power consumption monitoring of a single user. According to the typical power load classification, the power load of a single user is decomposed, and the power consumption behavior of multiple users is analyzed. A user cluster with different price sensitivity is formed. A real-time market environment is constructed for the aggregate users, and the electricity price scheduling for the aggregate users is realized. First of all, the necessity and feasibility of user participation in demand response are analyzed, and the current development and application of demand response at home and abroad are summarized. The analysis of electrical behavior and the identification of model parameters summarize the current research situation in academic circles and lay a foundation for the further study of this paper. Secondly, for a single user, considering the shortcomings of current load decomposition methods, a multi-dimensional load classification system is proposed, which is based on the functional dimension, considering the usage and characteristics of the equipment commonly used by users. Time dimension and power dimension are used to classify equipment, and user load decomposition model is constructed based on fuzzy membership function. The reliability of load decomposition result is represented by fuzzy membership degree, and user power load decomposition is realized. Provides the basis for potential analysis of demand response projects. Then, for multiple users, based on the improved K-Means clustering algorithm, under the peak-valley electricity price environment, the power consumption ratio and load rate of each period of peak and valley level are selected as the characteristics of the user, and the electricity behavior clustering model is constructed. Based on the clustering analysis of the user's electricity consumption behavior and the change of the user's electricity consumption before and after the price conversion point, the calculation method of the user's price sensitivity is put forward, and the user screening model is constructed. In order to determine the appropriate price response to the sensitive object. Finally, for aggregate users, in the real-time market environment based on consumer psychology, support vector machine is used to identify the model parameters, and the calculation model of demand response price is constructed. And the electricity price scheduling error and application scenario are analyzed. By comparing the errors obtained by artificial neural network and regression analysis, a combined analysis method weighted by these three parameter identification methods is proposed to construct a comprehensive scheduling model to improve the accuracy and effectiveness of electricity price scheduling.
【学位授予单位】：东南大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TM73

【参考文献】