Optimal Charging Control Methods Of Large-scale Plug-in Electric Vehicles

With the increasing population of plug-in electric vehicles(PEVs),the interaction between PEVs and power grid imposes more significant impacts to each other.On the one hand,when tens of thousands of PEVs are connected to the power grid for charging,if no coordinate control or dispatch is exerted on the charging loads,the power grid will suffer from many negative influences,such as increase of peak load,voltage deviation on the feeder,overloading of distribution transformers,increase of power losses,degradation of operational efficiency,reduced stability of the power grid,etc.On the other hand,the batteries of PEVs can conveniently participate in the operation of power grid by using the technique that is known as vehicle-to-grid(V2G),in order to provide ancillary supporting services for the utility.On the abovementioned two aspects,this thesis conducted research work on the control and optimization methods of large-scale PEVs to participate in power system regulation.The main contents are comprised of the following four parts:(1)Due to the nonlinear constraints in PEVs coordinated charging optimization model,the traditional nonlinear algorithm becomes inefficient.To address this issue,a cutting plane algorithm was proposed to increase the computing speed.Furthermore,to relieve the impacts of the forecast errors of electricity prices,a robust optimization model was developed on the basis of the proposed charging optimization model,and three uncertainty sets with different structures were utilized.The numerical studies analyzed the impacts of different uncertainty sets on the optimality of solutions and compared the computing performance of the proposed method to the traditional interior point method.(2)When PEVs are integrated into the power grid for charging,they can also participate in the economic dispatch of power grid to provide spinning reserve.But the stochastic nature in the charging behaviors of PEV users will lead to uncertainty in day-ahead schedules and lower the reliability of real-time operation.Therefore,the driving patterns were analyzed based on the investigation data,and the three key parameters,i.e.,departure time,arrival time and daily driving distance,were used,in combination with the Copula model,to establish the probabilistic model of charging behaviors.Then we developed the robust unit commitment model,analyzed the impacts of budget of uncertainty on the operation cost of power grid and analyzed the impacts of uncertainty of electricity price on the optimization model.(3)For the problem of lower computation efficiency that exists when large-scale PEVs take part in economic dispatch of power grid,a hierarchical control strategy was proposed based on Alternating Direction Method of Multipliers.The PEVs aggregator is taken as the middleware between the grid operator and the PEVs,and the original optimal power flow(OPF)problem was decomposed into two parts.The grid operator is responsible for the OPF calculation and signal dispatch,and each PEV updates its charging profile by tracking the command signal.In comparison to the classic OPF approach,this proposed algorithm has the advantages of less runtime,less memory footprint,good scalability and easy parallelism.(4)The PEV battery has the capability of fast charging and discharging,which enables them to be suitable to provide primary frequency services.Hence,the dynamic model of PEVs was developed based on Thevenin equivalent circuits.Then the closed-loop state-space model of load frequency control is derived by taking PEVs into consideration.However,as mentioned above,the randomness in charging behaviors will cause the model parameters to be time-varying.Moreover,the transmission of the command signals will incur time delays and deteriorate the dynamic performance and stability of power grid.To address the two issues,a design approach of robust controller was proposed via utilizing the particle swarm optimization algorithm.At last,numerical experiments were conducted to validate that the proposed method was well immunized against the parameters variation and the time delays.