Adaptive Cooperative Control And Capacity Configuration Optimization Of Supercapacitor Energy Storage Systems In Urban Rail Transit

With the rapid development of urban rail transit in China,the problem of energy consumption has become increasingly severe.With the aim of reducing energy consumptions of urban rail transit systems,recently,more and more energy recovery devices are put into operation on metro lines in China.Super capacitors(SC)have the advantages of high power density,fast charge and discharge speed,and long cycle life,thus possess great potential for regenerative braking energy recovery.In order to improve the energy-saving and voltage-stabilizing effect,as well as economic benefit of stationary supercapacitorbased energy storage systems(SCESS),this dissertation focuses on the optimum design and optimum control of SCESS,conducts theoretical and experimental researches on energy management strategy(EMS)for single SCESS,cooperative control algorithm between multiple SCESSs,and synthetic optimization of ESS capacity configuration,train operation and power system parameters.In the complex environment where multiple trains operate simultaneously and system energy interacts in real-time,the energy-saving and voltage-stabilizing effects of SCESS are affected by multiple factors,such as the substation no-load and the train operating state.The EMS determines the charging and discharging modes of ESS at different system states,thereby regulating the regenerative braking energy distribution of the system in steady-state level,and is essential to improvement of energy-saving and voltage stabilizing effects of ESS.Therefore,based on the line-voltage-based hierarchical control scheme,an EMS based on deep reinforcement learning(DRL)is proposed in this dissertation,and the off-line training procedure as well as on-line decision method for the proposed EMS are designed.The policy of the ESS agent is trained and improved based on the mechanism of ”trial” and ”feedback” in the learning process,and the control parameters of ESS are adjusted dymanically according to the status of trains,substations and ESS during real-time operation.Since the no-load voltage fluctuation of the substation will cause the ESS to operate in an unreasonable working state such as ”storage without discharge”,this dissertation incorporates the substation-characteristic-fitting-based no-load voltage identification module into the EMS,thus the online adaptability of the EMS to environmental changes is improved.The optimality and adaptability of the proposed EMS are verified through simulation based on actual metro line conditions.Gnearaly,multiple SCESSs are installed in different substations along the metro line,and the energy transmission efficiency between the SCESSs,the rectifier units and the trains is comprehensively affected by the control parameters of the multiple SCESSs.In this dissertation,the influences of the control parameters on the overall recovery efficiency and the regenerative energy distribution among different SCESSs are analyzed under different train operating conditions.Considering the ”dynamic” and ”cooperative” characteristics of the decision-making process for multi-SCESSs,this dissertation formulates the decision-making process as a cooperative Markov Game,and proposes a cooperative control algorithm based on multi-agent deep reinforcement learning,where the ”centralized learning,decentralized control” framework stabilizes the multi-agent learning process,and overall energy-saving effect of multi-SCESSs are improved through coorperation between SCESS agents.The proposed cooperative control algorithm is verified through simulation,and power distribution in several train operation scenarios are presented to illustrate the power dispatch mechanism of proposed algorithm.Current reaearches on capacity configuration optimization of ESSs rarely considers the impact of train operating characteristics and parameters of the traction power system.However,for the entire traction power system with close interaction between compoments,the energy-saving effects of SCESSs should be analyzed and optimized from the systematic perspective.In this dissertation,the capacity configuration of SCESSs,the output characteristics of the substation,the control schemes of braking resistors and the train operation diagrams are considered comprehensively.Based on the equivalent circuit model,the influences of traction power system parameters on the energy transmission between powering/braking trains,substations and SCESSs are analyzed,and the relationships between the train operation parameters and system energy consumptions are revealed.A multi-variable synthetic optimization method is proposed in this dissertation to optimize the capacity configuration of ESSs,train operation diagrams and power supply system parameters collaboratively.In order to reduce the search space of the optimization algorithm and improve the solution efficiency,this dissertation establishes a hierarchical optimization model,based on which the design variables and control variables are improved iteratively.Since the traffic density has a great impact on the regenerative energy distribution of the system,the optimization objective in this dissertation considers the frequency distribution characteristics of the train departure interval throughout the day.The flowchart of the two-stage collaborative optimization algorithm is desgined by combining the elitist non-dominated sorting genetic algorithm(NSGA-II)with the traction power simulation platform.Based on simulation study of Batong Line,the Pareto sets of the algorithm are obtained,and reductions on system energy consumption and configuration costs are validated.In order to verify the proposed control strategies,a power hardware-in-the-loop experimental platform for traction power system with SCESSs is developed in this dissertation.It realizes co-operation of the RT-LAB real-time simulator and the physical SCESS through the power amplifier,and experimentally emulates the traction power system with multi-train operation.Besides,A hierarchical control system,in which the personal computer(PC)executes the energy management strategy,and the DSP controller implements DC/DC converter control is designed.Based on the experimental platform,the EMS based on DRL,the no-load voltage identification method and the cooperative control algorithm of multi-SCESSs are implemented experimentally,and the feasibility and effectiveness of the strategies are demonstrated.