| As an effective tool to improve the reliability of power electronics,condition monitoring technology has been widely used in the industrial field where safe production and intelligent operation and maintenance are advocated.It can avoid catastrophic failure by real-time evaluating the health status of the system and timely warning the aging of devices,which greatly reduce the maintenance cost and enhance the operation safety.However,there are not many successful applications of condition monitoring system in the field of power electronic converter,and there is still a certain distance between theoretical research and industrial application.Several reasons that hinder the popularization of existing research results to engineering include: 1)the dependence on extrafrontal sensory information reduces the feasibility of application,2)the dependence on specific topology reduces the generalization universality,3)the calculation accuracy is insufficient due to the neglect of nonideal factors in the modeling of parameter identification,and 4)the performance of the pre-defined evaluation model is questionable when applied to the target scenario.In order to solve the above problems,this paper takes the engineering application of the condition monitoring strategy in converter power module unit as the primary goal,and treats the two key components of the converter module and the three relevant links in the multi-level condition monitoring strategy as the main research objects.The main research contents include the following points:(1)Aiming at the problems of serious parameter coupling and low aging sensitivity in the physical layer evaluation model of IGBT(insulated gate bipolar transformer)module,an IGBT aging evaluation method based on the zero state opening process is proposed.This method does not need to use other sensor information to decouple the aging state,and can achieve relatively high aging evaluation accuracy.Firstly,the variation law of parasitic parameters during the package aging process of IGBT module is analyzed,and the formation mechanism of collector-emitter voltage of IGBT module during the opening process under zero dc voltage and zero load current is revealed.Secondly,the functional relationship between collector emitter voltage and parasitic parameters is established,and three new health indicators are proposed to evaluate the package aging status of the IGBT module.Finally,a temperature compensation method using off-time collector emitter coltage is proposed to realize full parameter decoupling and high-precision aging evaluation of IGBT module packaging.(2)Aiming at the problems of strong hardware dependence and poor application universality of the physical layer evaluation model of dc-link capacitors in ac-dc-ac converter system,a real-time equivalent capacitance calculation method enabled by unilateral information is proposed.This method does not need additional hardware and does not interfere with the normal working process of the system,and realizes the aging condition evaluation of capacitors with high response speed and high calculation accuracy.Firstly,the functional relationship between the injected signal and current harmonic is established,and a current harmonic suppression method based on signal frequency selection and feedforward beat frequency control strategy is then proposed.Secondly,a capacitive current calculation method enabled by unilateral information is proposed,and the capacitive current signal reconstruction without dc sensor is realized with the help of modulation information and power analysis.Finally,a framework of equivalent capacitance calculation based on sliding Fourier algorithm is proposed,and the real-time calculation of capacitance under variable frequency and variable load operating conditions is realized.(3)Aiming at the problems of low data transmission efficiency and high storage unit cost in the multi-level state monitoring system,a signal processing framework based on compressed sensing algorithm is proposed,which improves the signal transmission speed,storage space utilization efficiency and algorithm execution efficiency,and contributes to the construction of multi-level state monitoring strategy.Firstly,the basic theory of compressed sensing algorithm and the construction of sparse transformation matrix are studied,and a signal processing framework of multi-level state monitoring system based on compressed sensing algorithm is proposed.Secondly,a new performance evaluation index of signal reconstruction is defined,which provides guidance for the adaptive selection of sparse matrix and measurement matrix.Finally,a measurement matrix design method based on verification and training process is proposed to realize dimension reduction of matrix and improve the algorithm efficiency,which lays a foundation for the digital hardware implementation of high-performance signal transmission system of multi-level condition monitoring strategy.(4)Aiming at the problem of doubtful prediction accuracy of the basic evaluation model under nonideal and differential systems,a correction method for regression model and classification model based on genetic programming system is proposed.This method outputs simple and intuitive explicit function expressions,which provide an important technical means to solve the prediction accuracy of the basic evaluation model in the real environment.Firstly,a regression correction method based on adaptive feature selection is proposed,which can improve the calculation accuracy of the equivalent capacitance value with low complexity and strong interpretability,and reduce the final output error to 1%.Secondly,a multivariate classification model based on feature mapping function and support vector machine is proposed,which realizes the nonlinear fusion of multiple health indicators of IGBT module,and improves the accuracy and efficiency of aging status evaluation.Finally,a transfer learning method enabled by repeated experiments is proposed,which improves the prediction accuracy of the basic evaluation model in the different application.The assessment of IGBT module aging degree can be achieved with a success rate of more than 80% and a correctness rate of more than 95% can be obtained.In this paper,two physical evaluation models for IGBT module and dc-link capacitor are proposed,the relevant algorithms are introduced to improve the output accuracy of the model,which is of great significance to the promotion of device level condition monitoring strategy in engineering application... |
PDF Full Text Request