【作者】 刘晓峰；

【导师】 高丙团；

【作者基本信息】 东南大学 ， 电气工程， 2019， 博士

【摘要】 随着高级量测、通信技术的发展以及电力体制改革的不断深入,电力需求侧管理(demand side management,DSM)能够从技术层面以需求响应(demand response,DR)的形式参与电网调度。居民用户作为需求侧的重要组成部分,具有丰富的柔性负荷资源,但其用电行为的随机性、决策主体的多样性以及市场信息的不完全性,使得居民最优决策变得极具挑战性,同时也限制了居民侧DR的进一步发展。因此,居民负荷用电行为以及多主体决策优化的研究对于深入挖掘居民DR潜力、缓解电网供需矛盾具有重要意义。鉴于此,本文在研究居民用电行为特征的基础上,利用多主体决策优化工具博弈论,从市场博弈信息的完整性层面,对完全信息下居民柔性负荷博弈优化、不完全信息下居民分布式能源双向交易及DR资源日前市场投标博弈优化开展了系统性研究。本文主要研究工作如下所示:首先,针对居民负荷需求以及柔性负荷DR调度时段确定问题,提出了基于自下而上的居民用电行为及负荷需求预测方法,从而为后续居民侧DR博弈优化调度提供基础。居民用电行为及负荷需求预测主要包括历史相似日提取、用户用电行为特征分析和负荷需求预测。其中,历史相似日提取部分通过构建相似度特征向量来筛选与预测日相似度最高的历史相似日;用户用电行为特征分析部分通过统计相似日家庭负荷用电行为概率来预测预测日负荷用电行为,进而可确定柔性负荷DR调度时段;负荷需求预测部分则通过构建居民用电行为特征模型和居民负荷电器特征模型来预测居民负荷。在此基础上,针对完全信息下居民参与DR时柔性负荷优化管理问题,分别提出了居民社区非合作博弈与合作博弈优化方法,以对居民储能容量配置和负荷调度进行优化。在非合作博弈建模部分,建立了以各社区日总费用最小为目标的博弈模型,给出了纳什均衡及最优容量存在性证明,提出了粒子群算法和内点法相结合的分布式算法。在合作博弈建模部分,建立了以所有社区日总费用最小为目标的博弈模型,从概率角度提出了合作博弈分配方法,给出了合作博弈分配个体理性、等价对待性证明。其次,针对不完全信息下居民分布式能源参与电网双向交易过程中的优化调度问题,以电动汽车为媒介提出了贝叶斯博弈优化方法。对于居民社区电动汽车双向交易不完全信息博弈场景,给出了不完全信息的基本假设,建立了电动汽车购电费用模型和售电收益模型。进而,以各社区期望收益最大为优化目标,以电动汽车充放电策略为优化策略,构建了不完全信息下的居民社区贝叶斯博弈模型,给出了贝叶斯纳什均衡的存在性与唯一性证明。进一步,针对DR日前市场信息不完全以及居民用户违约可能下DR资源投标决策问题,提出了居民DR资源日前投标贝叶斯博弈优化方法。对于居民用户参与DR违约问题,提出利用燃气锅炉和储能系统等辅助设备来降低资源违约率,建立了日前投标价格、辅助设备、DR资源违约模型,并以此建立了社区运营商售卖DR资源以及电热能收益模型。进而,构建了社区运营商贝叶斯博弈投标决策模型,设计了分布式算法求解投标决策博弈均衡。最后,针对居民需求侧博弈优化技术在实际系统中可行性问题,设计了居民负荷博弈优化管理实验研究。结合居民负荷能量管理系统,建立了实验系统完全信息博弈模型和不完全信息博弈模型,将居民博弈决策量离散为各类负荷开关量。进而,利用实验系统对所建立模型进行了验证,实验结果表明所提博弈方法能为系统负荷的运行方式做出最优决策,具有良好的可行性。本文系统的给出了居民用电行为预测、柔性负荷完全信息博弈优化、分布式能源双向交易以及DR资源日前投标不完全信息博弈优化技术,研究成果可为居民用户参与DR方式及机制的制定提供理论支撑。更多还原

【Abstract】 With the development of advanced measurement technology,communication technology,and electricity market reform,demand side management(DSM)can take part in power grid dispatch in the form of demand response(DR)project.As an important part of demand side,residential users have rich flexible load resources.However,the optimal decision of residents becomes extremely challenging due to the randomness of energy consumption,the diversity of decision-making subjects,and the incompleteness of market information.At the same time,such factors also limit the further development of residential DR.Therefore,the research on energy consumption behavior and multi-subject decision-making optimization will contribute the in-depth excavation of residential DR and alleviate the contradiction between energy supply and demand.Under such background,based on the study of residential consumption behavior,this dissertation focuses on the DR decision problem in residential flexible load,distributed energy equipment,and DR day-ahead bidding from the aspect of game information’s integrity.The main research work of this dissertation is as follows:Firstly,aiming at the problem how to obtain residential load demand and flexible load dispatching interval,a bottom-up prediction model of consumption behavior is proposed in the dissertation.Prediction model of consumption behavior mainly includes extraction modular of historical similar days,analysis modular of consumption behavior,and prediction modular of energy consumption.In which,extraction modular of historical similar days is proposed to select similar days from historical days by formulating similarity eigenvector;analysis modular of consumption behavior is proposed to forecast appliances’ consumption behavior by analyzing appliances’ behavior in historical similar days;prediction modular of energy consumption is proposed to forecast residential energy demand by formulating consumption behavior model and appliance’s electrical model.Secondly,aiming at the optimal problem of flexible load in DR process under complete information,non-cooperative game approach and cooperative game approach are proposed,to optimize storage capacity and load consumption.In the non-cooperative game,game model is founded for the minimal daily cost of each community;then,Nash equilibrium and optimal storage capacity are proved;last,distribution algorithm is proposed combining particle swarm optimization and interior point method.In the cooperative game,game model is founded for the minimal daily cost of all communities;then,cost allocation method is promoted from the perspective of probability;last,individual rationality and equivalence of allocation method is proved.Thirdly,aiming at the optimal problem in two-way energy trading with grid under incomplete information,economic dispatch problem of EV is proposed with Bayesian game approach.For the two-way energy trading scenario with incomplete information,basic assumptions are stated at the beginning.And then,cost model and profit model of EVs are founded.Furthermore,Bayesian game model is founded for the maximal expected profit of each community by optimizing EVs’ charging and discharging strategy.And then,the existence and uniqueness of Bayesian equilibrium is proved.Fourthly,aiming at the flexible load resource breach and incomplete information problem in DR day-ahead bidding market,game-theoretic bidding strategy is proposed under the electricity market.For the breach problem of DR resource,auxiliary equipment,such as gas boiler and storage,are used to reduce breach ratio of DR resource.And then,day-ahead bidding price model,auxiliary equipment model,and DR resource breach model are founded,respectively.Basically,profit model is founded by selling DR resource,electricity and heat.Furthermore,Bayesian game model is founded for community operator’s bidding strategy,and then distribution algorithm is designed to search the equilibrium.Finally,aiming at the practicability problem of game-theoretical technology in the real system,experimental study is designed with residential load management system.Combining the experimental condition,complete information game and incomplete information game are founded for experimental system,in which residential game decision strategy is discretized into load switching variables.Furthermore,the founded game models are verified with experiment.Experimental result shows that the proposed game approach can make the optimal decision for the system operation and has a good practicability.This dissertation has given the modeling techniques on residential consumption behavior prediction,complete information game-based optimization for flexible load,Bayesian game game-based optimization for distributed energy equipment and day-ahead bidding of DR resource.The result of the dissertation can provide theoretical support for the design of operation mechanism in residential DR.更多还原