Research On IGBT Aging State Prediction Method Based On ISSA-BP

As the core power device in the power electronic converter system,Insulated Gate Bipolar Transistor(IGBT)has been widely used in smart grid,rail transit,new energy power generation and other fields due to its low switching loss and low driving power.In the entire power electronic system,as the service life of IGBT power modules increases,the risk of aging and failure of power devices in the later stages of use also increases,which can easily cause economic losses and safety hazards.Therefore,the aging state prediction of IGBT is of great significance to improve the reliability of power electronic systems.This paper establishes an IGBT aging state prediction model based on the combination of the improved sparrow search algorithm and Back Propagation(BP)neural network,and uses the IGBT collector-emitter turn-off transient peak voltage aging data set released by the NASA Research Center and the heat obtained from the self-built IGBT accelerated aging experimental platform.The anti-aging data set is used for model training and verification,and the experimental results show the effectiveness of the proposed model.This paper mainly conducts research from the following aspects:Firstly,the basic structure of IGBT,static characteristics and dynamic characteristics of IGBT work are summarized.On the basis of analyzing the packaging structure of IGBT modules,the failure types,failure points and failure modes of IGBT modules are further analyzed and summarized.The commonly used IGBT aging failure monitoring characteristic parameters are summarized,and the collector-emitter turn-off instantaneous peak voltage from the IGBT aging failure data set of NASA Research Center and the thermal resistance obtained from the self-built IGBT accelerated aging experimental platform are selected as the aging state monitoring characteristic quantities.Characterize the aging state of the IGBT.Secondly,the most commonly used accelerated aging strategies in IGBT accelerated aging tests are summarized,and the power cycle accelerated aging test method with constant shell temperature fluctuations is adopted to complete the construction of the IGBT accelerated aging test platform and obtain the corresponding IGBT accelerated aging data sets.The IGBT collector-emitter turn-off instantaneous peak voltage aging data set released by the NASA Research Center and the thermal resistance aging data set obtained from the self-built IGBT accelerated aging experiment platform were analyzed,and the normalization,exponential smoothing and random division methods were used to analyze the aging data.The above two different aging datasets are preprocessed.Thirdly,build an IGBT aging state prediction model based on BP neural network.On this basis,there is a long training time for the BP neural network prediction model training for the application of NASA open source IGBT aging data and the aging data obtained from the selfbuilt IGBT accelerated aging platform.In order to solve problems such as length,lack of global search ability and low prediction accuracy,the Sparrow Search Algorithm(SSA)is used to optimize the input weights and hidden layer thresholds of the BP neural network.Then build a simulation platform to complete the verification of the improved prediction model.Finally,aiming at the problems of sparrow population random initialization,poor population diversity and easy to fall into local optimum,the method of combining Tent chaotic map and firefly disturbance strategy is used to improve the sparrow search algorithm.The Improved Sparrow Search Algorithm(ISSA)is tested by using the benchmark function.The test results prove that ISSA is superior to other algorithms in terms of search speed and search accuracy.Set up a simulation environment to complete the modeling of the IGBT aging state prediction model based on the improved sparrow search algorithm optimized BP neural network(ISSA-BP),and use the open source IGBT collector-emitter turn-off instantaneous peak voltage aging data of the NASA Research Center and self-built The thermal resistance aging data obtained by the IGBT accelerated aging test platform are respectively verified for the proposed ISSA-BP-based IGBT aging state prediction model.By comparing with different prediction models,it is verified that the ISSA-BP prediction model has the advantages of predicting the IGBT aging state.Stronger applicability and higher prediction accuracy.