Research On Power Management System Of Ship Power Station

Ship power station refers to the electric power device set up to meet the needs of ship navigation and life,and is the "power center" of the ship.With the increase of global energy demand and the improvement of environmental protection awareness,the development of ship power station also gradually develops in the direction of high efficiency,energy saving,environmental protection and intelligence,coupled with the rapid improvement of computer technology,ship power station also gradually introduces intelligent technology,such as Internet of Things,cloud computing,artificial intelligence,etc.Therefore,it is of great significance to study the application of artificial intelligence algorithm to the power management system of ship power station to improve the operation efficiency,safety and intelligence level of ship power station.This thesis introduces the time series-based LSTM neural network to analyze the routine fault diagnosis of ship power system and realize the real-time tracking and prediction of power grid fault data by building the simulation model of ship power station in Simulink as the research basis.The details are:(1)by analyzing the physical model of ship power station,inductively deriving the mathematical model of each module,completing the simulation model of ship power station in Simulink according to the established mathematical model;using the model to conduct a series of simulation experiments such as single-unit no-load start,sudden addition and removal of load,double-unit parallel,single-phase grounded short-circuit fault,phase-to-phase short-circuit fault,three-phase short-circuit fault,etc.to verify the The accuracy and reliability of the power station model were verified.(2)Then the simple monitoring system of this power station was built in King SCADA,and the data intercommunication between this monitoring system and the ship power station model was realized through SQL Server database,and the fault simulation and load control operations were realized in this monitoring interface.(3)Based on the simulation model of ship power station constructed in this paper and the data obtained from the fault simulation,the application of deep learning in the fault diagnosis of ship power station is studied.The BP neural network,which is the most representative feedforward neural network in deep learning,and the LSTM neural network,which is the most representative feedback neural network,are analyzed,and the performance of the BP network is optimized by genetic algorithm;then three fault diagnosis models are built in Matlab: BP,GA-BP and LSTM models.The data obtained from the fault simulation were input into these three models for training and testing,and the results showed that the BP model had the lowest fault diagnosis accuracy,the optimized GA-BP had the best effect,and the LSTM model had a diagnosis effect close to and much better than the GA-BP model.(4)Finally,the concept of time series is introduced,as LSTM is able to capture long-term dependencies in time series and thus can be used to predict future trends and changes.This property is applied to ship power plant fault prediction,and it is verified by simulation that this time-series based LSTM neural network model can monitor the power grid well in real time and make real-time fault prediction by the operational status of the power grid.