Multi-mode Coordinated Switching Control Method For Microgrid Based On Reinforcement Learning

With the increasing consumption of non-renewable energy and the aggravation of environmental pollution,countries begin to focus on low-carbon energy.Therefore,microgrid based on distributed energy resources emerges as the times require,but there are still many problems to be solved in the control strategy and switching stability between different working modes of microgrid system.Therefore,this paper studies the related problems in order to improve the effectiveness of the control strategy of microgrid system and improve the operation stability.Firstly,the control structures,operation modes,control strategies of microgrid and the mathematical models of common distributed generations are analyzed and introduced,which provides a theoretical basis for the establishment of microgrid model and the formulation of coordinated control strategy.Secondly,the single energy storage control strategy used in energy storage system has problems in the power fluctuation stabilization and the management of energy storage life.Therefore,the control strategy of coordinated switching of dual energy storage is proposed.The two groups of energy storage work separately or cooperatively according to different conditions,and the tasks of "peak clipping" and "valley filling" are undertaken by the two groups of energy storage respectively,which can not only improve the performance of suppressing the power fluctuation of distributed generation,but also effectively extend the life of energy storage.A power fluctuation suppression model based on the model prediction algorithm is established for the energy storage system to verify the fluctuation suppression effect of the dual energy storage strategy.Finally,under the dual energy storage strategy,different discharge lower limits of energy storage are simulated.Through fitting the life loss data,the discharge lower limit under the lowest life loss is obtained.Then,aiming at the problem of coordinated switching of each unit of the microgrid,a microgrid multi-modal switching control strategy based on reinforcement learning is proposed to optimize the coordinated control strategy of each unit and realize the coordinated work of each unit of the microgrid.Firstly,based on the behaviors of each unit of microgrid,the behavior models and state models which can describe all its behaviors and states are established.In the voltage hierarchical switching control unit,an improved voltage hierarchical switching control strategy is proposed to eliminate the problem that bus voltage fluctuation between layers may cause frequent triggering of control commands.According to the important loads and common loads in the load management unit,the partition management mechanism is proposed to ensure the stable operation of important loads.Finally,the system security is evaluated by voltage security index,and the reward and punishment strategy is designed based on this.