Energy Management And Control Of Microgrid Based On Solid Oxide Fuel Cells

The promotion and application of microgrids,including renewable energy(photovoltaic,wind turbine,etc.)and high-efficiency and low-pollution solid oxide fuel cells(SOFC),is an effective way to promote "carbon peaking,carbon neutrality" and limit global warming.However,reducing investment costs,operating costs,and SOFC degradation of microgrids is still a challenge under the intermittency and randomness of renewable energy,as well as the uncertainty of load power.In addition,due to fuel transfer delays,thermal constraints,degradation,etc.,SOFC is difficult to achieve fast and accurate power tracking,resulting in power imbalance and DC bus voltage fluctuations in the microgrid.Therefore,the key issue in SOFC-based microgrids is the economical and reliable size configuration,beneficial operation,and control strategies.Based on the above needs,this thesis conducts research on size optimization,energy management,and control strategies for SOFC-based microgrids.The main contents are as follows:(1)This thesis first establishes two types of models,namely mechanism models and simplified power models,for renewable energy power generation systems(photovoltaic,wind turbine),auxiliary power generation systems(SOFC),and energy storage systems(lithium batteries,electrolyzer)in isolated DC microgrid systems,respectively,for the optimization of microgrid size configuration,verification of energy management and control algorithms.After model verifing,the Genetic Algorithm is used to optimize the microgrid size configuration with lowest cost based on real historical weather and residential electricity consumption data from multiple regions in China.(2)This thesis proposes a multi-time scale energy management strategy(EMS)to reduce the operating costs and SOFC degradation of microgrids under the influence of intermittency and randomness of renewable energy,as well as the uncertainty of load power.The multi-time scale EMS is divided into two stages: in the first stage,the Deep Deterministic Gradient Policy reinforcement learning algorithm is used to plan the SOFC generation in day-ahead based on the day-head net power predicted by the Long ShortTerm Memory neural network;In the second stage,real-time power scheduling of microgrids is carried out through the Generalized Model Predictive Controller to balance real-time power demand and eliminate prediction errors caused by randomness and uncertainty.The final results indicate that compared to the real-time optimization-base EMS,the multi-time scale EMS proposed in this thesis effectively reduces the power fluctuation of SOFC,delays the degradation of SOFC,and reduces the cost of power generation.(3)This thesis establishes a distributed collaborative control strategy for microgrids based on the distributed consensus algorithm to compensate for the power balance error mainly caused by the power tracking delay of SOFC in real time.The results show that compared to the decentralized power tracking control strategy,the control strategy established in this thesis can effectively reduce power balance errors and improve the voltage stability of microgrids.