Fault Reconstruction Of All-electric Propulsion Ship Based On Multi-agent Game Theory

As the industry develops,modern ships are gradually moving towards full electric propulsion,resulting in a complex structure of the ship’s electrical network and an increasing proportion of dynamic loads.Therefore,the dynamic characteristics of the network after a fault occurs become more prominent,which poses a significant threat to the stable operation of the ship’s electrical network.Shipboard electrical network fault reconfiguration aims to restore power to the load as much as possible in a timely manner after a fault occurs to ensure normal ship operation.This paper designs shipboard electrical network steady-state and dynamic evaluation indicators based on steady-state analysis and dynamic simulation results.A multiagent game model is constructed based on game strategy,and the fuzzy genetic algorithm is used to solve the fault reconfiguration problem,ensuring the stable operation of the shipboard electrical network.The thesis uses PSCAD transient simulation software to build a comprehensive simulation model of the all-electric propulsion ship power system and completes the model simulation verification under normal operation and typical fault scenarios.Considering that existing intelligent optimization algorithms are prone to getting stuck in local optima,the thesis proposes an improved genetic algorithm,the fuzzy reasoning genetic algorithm.By designing adaptive mutation operators and cross operators to ensure population diversity and introducing fuzzy reasoning to nonlinearly normalize multiple objective functions to obtain the fitness function,the algorithm’s tendency to premature convergence is improved,and the convergence speed is increased.Considering the high proportion of dynamic loads in the electric power propulsion system of ships,the dynamic characteristics of the system are more significant when faults occur.Therefore,this thesis proposes to include the evaluation of dynamic characteristics as a selection criterion in the fault reconstruction scheme and constructs an evaluation index that reflects the dynamic characteristics of the ship’s power grid.Considering the risk of “Single point failure” faced by centralized reconstruction algorithm,multi-agent genetic algorithm is introduced to ship power grid fault reconstruction.Through practical examples,the dynamic index is improved by an average of about 20%,and the dynamic performance of the power grid is significantly improved.The average convergence time of the algorithm is 8 generations,which demonstrates the feasibility of the algorithm.As the scale of the electric power propulsion system in ships continues to expand,distributed multi-agent reconstruction algorithms need to explore the competition and cooperation between different areas of the ship’s power grid.To address this issue,flexible game strategies are introduced,and a multi-agent game model is established to study the game between agents in both cooperative and non-cooperative modes.A multi-agent game algorithm for ship power grid fault reconstruction is proposed.Finally,the algorithm is validated in typical fault scenarios.The average convergence time of the algorithm is 7 generations,and the dynamic index is improved by up to 30%,demonstrating the advantages of the multi-agent game algorithm.This thesis constructs a power flow calculation model and a transient simulation model to comprehensively consider the overall operational status of the ship’s power grid.By evaluating the fault reconstruction scheme using both steady-state analysis and dynamic simulation results,this thesis provides a new perspective on studying the problem of ship power grid fault reconstruction.To avoid premature convergence of the algorithm,fuzzy logic is introduced into the genetic algorithm and adaptive operators are added to effectively improve the diversity of the population and global search capabilities.Through experimentation,the convergence accuracy and speed of the algorithm are significantly improved.Finally,the multi-agent game model and algorithm are presented,and fault reconstruction is carried out in cooperation and non-cooperation modes.The algorithm shows good performance in both modes in terms of convergence and search ability,and the advantages of the algorithm are demonstrated in the cooperative mode.