基于分段分式的动态需求建模及多层全局优化的阶梯电价研究

Increasing block electricity pricing (BEP) is the key strategy devised to resolve the current electricity shortage, and other complex social and environmental problems brought by single electricity tariff system, such as unfair cross subsidies, social inequalities, energy scarcity, environment deterioration, etc. Due to BEP's unique non-linear pricing characteristics, it is difficult to identify the number of pricing variables and explore the patterns of household electricity consumption in the design. To tackle with these problems, this research aims to systematically explore the following three directions in BEP: 1) develop a novel piecewise and multi-stage dynamic demand model that is built upon multi-layer clustering analysis and customer behaviour analysis; 2) apply and improve the multi-level global optimisation model to minimise the electricity usage, whilst taking a wide variety of constraints into consideration; 3) build a comprehensive decision support system to offer assistance to BEP at its various design and implementation stages. From the theoretical point of view, this research would deepen and widen our theoretical insight in both BEP and its influence to electricity consumption behavior. From the practical side, this work would make the BEP more effective so as to achieve the goal of better energy conservation in the society. Moreover, the research outcomes could potentially be applied to other energy (e.g. water, gas, petrol) pricing reform in the near future.

实行阶梯电价已成为缓解我国能源供应紧缺、资源交叉补贴不合理等矛盾的重要举措。作为较独特的非线性定价，阶梯定价下的消费者行为经常难以甄别、存在价格变量难以确定等特征，这使得其设计机制的理论和实证研究都变得十分复杂。目前基于直观静态的设计方法在实行过程中已凸显出一些问题。为此，本课题将从以下三个模块对阶梯电价的设计进行系统科学的研究：1）提出一个全新的分段分式模型来刻画阶梯电价下的动态用电需求；并结合多层聚类分析，挖掘和解析不同特征的用电行为模式；2）通过改进多层优化算法，构建阶梯电价设计参数的全局最优模型；3）将多个理论研究成果集成到一个决策支持系统上，为阶梯电价设计和实施的不同任务提供系列支持。本课题将在理论层面上加深对能源消费行为与阶梯定价的内在平衡机制的科学认识，从实践层面为阶梯电价更好地发挥价格杠杆的作用、促进节能减排提供决策支持，还进一步可扩展到水、气、油等其他资源产品的价格改革。

相比传统的单一电价机制，阶梯电价机制的引入不仅会影响居民的电费支出，新机制所引起的用电行为差异将逐步凸显出来。本课题从现行递增阶梯电价动态需求的特征出发，挖掘和解析阶梯电价下不同特征的用电行为，并提出了一个全新的分类负载预测模型。通过实证数据的检验，新的负载预测模型不仅可以提高负载的预测精度，还提供了更为灵活的建模机制，可以针对不同的用电行为特征选择最为合适的负载预测模型并进行加总。研究结果为能源政策制定、定价机制设计、智能配电网供给侧和需求侧管理提供了理论和实践启示。本项目完成了课题申请书既定的研究目标并进一步开展新的相关研究。在课题的资助下，项目组成员累计参加国内外信息系统、计算智能、管理决策、能源环境等领域知名学术会议10次以上，项目组成员均到场做报告论文;在国内外学术杂志发表论文总共7篇，其中SCI期刊论文5篇，CSSCI论文1篇，包括Information & Management, Decision Support Systems, and IEEE Transactions on Smart Grid等，另外还有1篇在国际期刊 (Decision Support Systems)三审，其他若干篇论文正在审稿中;依托本课题培养了1位硕士研究生，该学生的毕业论文将在2017年5月进行答辩。