DC-DC Circuit Fault Prediction Method Based On Adaptive Data Window

With the rapid development of power electronics technology,a variety of electronic products are increasingly appearing and widely used.The main components of the power system and the driver we are familiar with are DC-DC circuits.Therefore,the DC-DC circuit malfunctions,doing harm to electrical products,and causing casualties and threatening social harmony and stability.In order to reduce the impact of the DC-DC circuit on the system due to random faults,This paper has significant value in the research of fault prediction methods based on adaptive data windows.In order to improve the fault prediction accuracy of DC-DC circuits,the following work has been done:Firstly,analyze the key components that make up the DC-DC circuit.It mainly analyzes its equivalent model,failure mechanism and degree of degradation.The characteristic parameters of each device performance degradation and its failure threshold are identified.Secondly,the establishment of the simulation circuit of the DC-DC circuit switching power converter and the design and performance indicators of its peripheral components is analyzed.The working principle of the DC-DC circuit is discussed.Introduced the performance degradation of each key component in the DC-DC circuit.The circuit-level characteristic parameters and component-level characteristic parameters used throughout this paper are analyzed.Thirdly,the DC-DC circuit fault prediction method is deeply studied.The ELM algorithm,LSSVM algorithm and LSTM algorithm are introduced respectively.Analyzed the settings of the data window,the operation of multiple data windows can solve the current problem of inaccurate modeling of the overall operating conditions.The ELM?LSSVM and LSTM algorithms are combined with the adaptive data window to predict the component parameters of the DC-DC circuit,and the prediction accuracy is enhanced.The reason why the ripple voltage is invoked as the circuit-level characteristic parameter of the DC-DC circuit in this paper is analyzed in the four parameters of output voltage average,output voltage ripple value,ripple ratio and loss function.Among them,the ripple voltage is obtained by denoting the output voltage.Finally,prediction models based on WDD-ELM?WDD-LSSVM and WDD-LSTM algorithms are used to adaptively predict the circuit-level feature parameters based on the data window,and simulation experiments are carried out to prove the feasibility of this method.In this paper,the experimental results verify the advantages of adaptive data window and predictive model for fault prediction of DC-DC circuits.Not only can save system storage space,Making prediction accuracy better,you can also use the least amount of data to achieve more accurate predictions.Shortening the amount of computation for each iterative process of the prediction algorithm,this reduces the prediction time.Figure [71] table [17] references [68]...