【作者】 喻磊；

【导师】 陈民铀；

【作者基本信息】 重庆大学 ， 电气工程， 2014， 博士

【摘要】 微电网作为分布式发电的有效管理形式，对于推进清洁能源的发展，增加新能源在配电网中的渗透率，提高配电网的稳定性、可靠性具有重要的意义。但是由于微电网内部分布式电源的间歇性和波动性以及微电网分布式接入配电网的方式，导致微电网本身以及微电网与配电网之间的优化控制问题难以解决。研究微电网的优化控制方法，多微电网之间的协调控制策略对于保证微电网的可靠经济运行具有重要的意义，为推进智能配电网的建设提供理论指导。由于微电网运行模式多样、可控程度不同，优化问题的局部约束条件各异，全局运行信息难以收集等问题，造成传统的集中式优化控制方法无论是从模型构建还是计算成本方面都难以胜任含多微电网复杂系统的优化控制问题求解，而基于多代理的分布式控制方法对于求解这类分布式问题具有明显优势。因此，本文采用多代理控制理论构建了微电网混合优化控制框架结构，依据微电网的不同运行环境，从系统到个体设计了相应的优化控制算法和策略。首先针对含多微电网的配电网系统，提出了分布式约束优化算法对微电网进行分布式优化控制，解决集中控制难以求解难题，其次针对配电网中个体微电网，建立了集中式优化算法对微电网进行集中式控制，保持集中控制对小规模系统快速求解优势，最后结合微电网内不同的DG给出了各自的执行控制方法。本文的主要研究内容和创新点包括：（1）结合多Agent理论，建立了完整的微电网混合式优化控制结构，研究基于多Agent分布式理论的分布式优化控制算法，将分布式优化控制问题视为分布式约束优化问题（DCOP），以势博弈理论为基础，提出了符合多代理（MAS）分布式理论的DCOP求解算法Weighted Regret Monitoring Distributed SimulatedAnnealing（WRM-DSAN）。算法以局部优化为基础，通过Agent之间的通信交流以及优化控制规则进行全局优化。测试结果表明所提出的分布式优化算法结构上满足分布式设计要求，并且能够保证算法几乎处处收敛到全局最优纳什均衡，较集中式有更好的复杂网络分布式求解能力。（2）以IEEE39节点系统为基础，通过微电网优化配置，建立了含多微电网的配电网系统。按照DCOP理论将微电网优化控制模型转化成为适合分布式求解的分布式约束优化问题，设定微电网中每一个Agent的优化目标函数，同时考虑局部优化约束和全局优化约束的关系，采用所提出的势博弈分布式优化算法WRM-DSAN进行求解，考虑分析Agent环状和网状控制结构对优化控制结果的影响，证实了微电网分布式优化控制的可行性，通过结果对比表明，系统在优化后获得了更好的整体效益，且网状控制结构具有更好的故障容忍能力。（3）针对小规模微电网的动态集中优化控制方法，提出了基于粒子群算法理论的Information Exchange Particle Swarm Optimization（IEPSO）集中式优化控制算法，建立了微电网动态集中优化控制模型，包括DG优化配置，风电以及负载的波动模型，集中优化控制的数学模型。考虑DG以及负载变化的情况下，制定微电网储能的控制策略，分析了不同储能贡献率对优化运行的影响，结果表明动态储能贡献率具有更好的协调效果。（4）建立了包括风电，小水电和储能底层执行DG的控制模型，并按照CERTS规范建立了微电网仿真平台，对组网后微电网进行仿真，同时研究并搭建了微电网在Matlab/simulink环境下微电网的数字仿真平台，为后续微电网的研究提供实验环境。更多还原

【Abstract】 As an effective manager for distributed generation, microgrid plays an importantrole in clean energy development, stability and reliability of future distribution network,and can also improve the penetration of renewable energy. However, DGs in microgridare characterized by fluctuation and intermittent and microgrids are distributedembedded into distribution network, which lead to the microgrid and microgrids indistribution network is difficult to control. The research of optimization control methodfor microgrid and coordination control strategy between microgrids could providevaluable references for reliable and economic operation of microgrid and developmentof smart distribution network.Generally, different controllable level of microgrid has different operation mode,where the local constraints for optimization control are varied and sometime the globalinformation is not available. For these complex distribution networks withmulti-microgrid, the centralized optimization control method is not suitable for dealingwith the control problem because of model building and computing cost, but it issuggested that multi-agent based distributed control method should be good at solvingsuch problems. Therefore, a hybrid microgrid optimization control framework is buildin this thesis based on multi-agent control theory, and the corresponding optimizationcontrol algorithms and strategies from top to bottom structure are designed for differentoperating environment. First, a distributed constrained optimization algorithm which isused to distributed optimization control for microgrid is proposed on the top controllevel of multi-microgrid; Second, in order to maintain a rapid control advantage ofcentralized method for small system, a centralized optimization algorithm is presentedin the middle control level for single microgrid; Finally, different execution controlapproaches for different DG at the bottom control level in microgrid are build. The mainresearch contents of this thesis are shown as follows:(1) Based on multi-agent theory, the optimization control system of microgrid isestablished first in this thesis, where the distributed control problem is regarded as adistributed constraint optimization problem (DCOP). To meet the requirement ofpotential game theory, the distributed optimization control algorithm WRM-DSAN isproposed. In this algorithm, global optimization that is based on local searching can beexecuted due to the communication between Agents and optimization control rules. Test results show that the proposed distributed optimization algorithm can satisfy therequirement of distributed design structure and the algorithm can converge to globaloptimal Nash equilibrium almost everywhere. It is proved that distributed optimizationalgorithm is more suitable for solving complex system problems in contrast withcentralized algorithms.(2) Distribution network with microgrids is build based on the IEEE39system.According to DCOP theory, the optimization control model of microgrid is consideredas an distributed constraint optimization problem. Utility function for each Agent inmicrogrid are formulated with the local optimization constraints. And potential gamedistributed optimization algorithm, WRM-DSAN, is applied to solve this distributedconstraint optimization problems. Under the ring and mesh control structure, theoptimal control results are analyzed, and it is indicated that distributed optimizationcontrol method is viable for multi-microgrid. After optimization, it is easy to find thatthe mesh control structure has better fault tolerance ability.(3) Firstly, a centralized optimization control algorithm IEPSO is designed fordynamic centralized control of small microgrid system. Then a dynamic centralizedoptimization control model of microgrid is established, including DG placement model,wind power and load fluctuation model and centralized optimization controlmathematical model. Considering the change of load and DG, energy storage controlstrategy of microgrid is formulated, and the effects of different DG contribution rate onthe optimal operation are analyzed. Results indicate that the dynamic DG contributionrate can provided more effective coordination control.(4) Microgrid simulation platform is built under the standard of CERTS with windpower, small hydro and energy storage model. Control performance of microgrid issimulated under island operation mode. Finally, the digital simulation platform ofmicrogrid is built in Matlab/simulink for future research of microgrid control.更多还原