基于多主体复杂系统的电力智能交易演化理论及动态均衡特性研究

项目名称： 基于多主体复杂系统的电力智能交易演化理论及动态均衡特性研究

项目编号： No.51277059

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 电工技术

项目作者： 姚建刚

作者单位： 湖南大学

项目金额： 75万元

中文摘要： 自然科学基金规划指出："智能电网基础包括智能电网的电力市场机制与实现。"电力交易是电力系统节能减排、优化资源配置的主要手段。随着市场参与者数量的增加及交易范畴的扩大，电力交易面临更大的不确定性及市场风险，而不同的市场体制或交易模式的出台，其可行性与风险性需要大量的分析和论证。针对该类问题，传统方法主要通过跟踪少数参与者建立数学模型，进行特定情况下的博弈均衡分析及市场评估，所得到的结果一方面可能偏向于局部特性，另一方面均衡分析虽是古典经济的基本理论，但动态系统可能出现单个均衡、多个均衡或者混沌状态。本课题针对智能电网大规模电力交易面临的困境和难点，采用多主体复杂系统方法结合高效率的机器学习进行研究，旨在提出一种大规模电力智能交易自然属性模型，提升复杂条件下的交易主体学习效率，并探索复杂环境下的电力交易演化理论及动态均衡特性，为智能电网研究提供新的方法和思路，为电力智能交易提供理论支撑与保障。

中文关键词： 电力交易；过渡期电力市场；节能减排；报价策略；强化学习

英文摘要： The plan for Natural Science Foundation indicates: the basis for smart grid includes the market mechanism and its realization. Electricity trading is the primary mean to achieve the goals of energy conservation and emission reduction. With the increase of the number of market participants and the expansion of trading areas, the electricity trading is facing greater uncertainty and market risk. Extensive analysis and validation are needed before any market structures or trading patterns are introduced. In traditional way, the game equilibirim analysis and market assessment in special cases are carried out mainly with some mathematical models formulated by focusing on a small number of participants, which may tend to reflect some local properties. On the other hand, equilibrium analysis is the basic theory in classical economics, but dynamic systems may have one or several equilibrium states, or stay in chaos. So this project has fully investigated the difficulties involved in large-scale electricity trading in smart grid, and will apply multi-agent complex system analysis combined with efficient machine learning method to our research. The project aims to propose a large-scale electricity intelligent trading model with natural properties, enhance the agents' learning efficiency, explore the electricity trading

英文关键词： electric power transaction；power market in transition period；energy saving and emission reduction；quotation strategy；reinforcement learning