Method for predicting residual life of high-voltage power module based on optimized neural network and multi-parameter feature fusion and related device

Abstract

The invention provides a high-voltage power module residual life prediction method based on an optimized neural network and multi-parameter feature fusion and a related device. The method comprises the following steps: acquiring a multi-dimensional state parameter related to a degradation state of a high-voltage power module; inputting the multi-dimensional state parameters into a pre-trained prediction model to obtain a residual life prediction result of the high-voltage power module; wherein the structure of the prediction model comprises a plurality of LSTM layers, the prediction model correspondingly extracts the time sequence characteristics of the state parameters of one dimension through each LSTM layer, and after weighted fusion is carried out, a prediction result is output. According to the method, the prediction accuracy and reliability can be improved by optimizing the neural network structure and parameters and combining various key state parameters, so that safe and stable operation of a power electronic system is ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of life prediction of power electronic devices, and specifically relates to a method and related device for predicting the remaining life of a high-voltage power module based on optimized neural network and multi-parameter feature fusion. Background Art

[0002] High-voltage power modules are widely used in the field of power electronics, such as DC transmission, railway traction, electric vehicles, etc. The failure rate of high-voltage power modules in power electronic systems is quite high. Taking the converter system as an example, high-voltage power module failures account for 31% of the total failures, which is higher than the 17.5% of capacitors and 14.5% of driver boards. The reliability of high-voltage power modules in complex environments and operating conditions is often lower than expected. Internal fatigue may accelerate the aging of high-voltage power modules and shorten their lifespan, thereby threatening the safe operation of equipment. Therefore, accurately predicting the remaining life of high-voltage power modules is crucial to ensure the reliability and safety of the system.

[0003] At present, the existing high-voltage power module remaining life prediction technology often fails to fully consider the influence of multiple state parameters, resulting in inaccurate prediction results. At the same time, the traditional prediction model has limited generalization ability when facing different operating conditions and is difficult to adapt to complex and changeable actual situations. These problems limit the development of high-voltage power module remaining life prediction technology and hinder its widespread application in actual engineering. Summary of the invention

[0004] In view of this, the present invention aims to provide a method and related device for predicting the remaining life of a high-voltage power module based on the fusion of optimized neural network and multi-parameter features, which can improve the accuracy and reliability of the prediction by optimizing the neural network structure and parameters and combining multiple key state parameters, thereby ensuring the safe and stable operation of the power electronic system.

[0005] In order to achieve the above object, the technical solution provided by the present invention is as follows:

[0006] In a first aspect, the present invention provides a method for predicting the remaining life of a high-voltage power module based on an optimized neural network and multi-parameter feature fusion, comprising the following steps:

[0007] Acquire multi-dimensional state parameters related to the degradation state of the high-voltage power module;

[0008] Input the multi-dimensional state parameters into the pre-trained prediction model to obtain the remaining life prediction result of the high-voltage power module;

[0009] Among them, the structure of the prediction model includes multiple LSTM layers. The prediction model extracts the time series features of the state parameters of one dimension through each LSTM layer, and outputs the prediction results after weighted fusion.

[0010] Furthermore, obtaining multi-dimensional state parameters related to the degradation state of the high-voltage power module also includes:

[0011] The multi-dimensional state parameters are preprocessed using the empirical mode decomposition algorithm, and the residual trend term signal after decomposition is retained as valid data to predict the remaining life of the high-voltage power module.

[0012] Furthermore, the training process of the prediction model includes:

[0013] Use sliding windows to reconstruct data of multi-dimensional state parameters and calculate the remaining life of each sample;

[0014] Update the structures of all samples to the corresponding characteristic parameters and remaining lifespan, and form a database for optimizing the neural network;

[0015] A multi-level LSTM network is trained based on the database of optimized neural networks to obtain a multi-parameter fusion prediction model.

[0016] Furthermore, the remaining lifetime is determined according to the difference between the final lifetime of the high-voltage power module and the lifetime position of the corresponding sample.

[0017] Furthermore, the structure of the prediction model includes:

[0018] Input layer, data deconstruction layer, multi-level LSTM layer, self-attention layer, fully connected layer and output layer, among which, the data deconstruction layer is used to divide the multi-dimensional state parameters according to different dimensions to obtain multiple one-dimensional arrays; each LSTM network in the multi-level LSTM layer extracts the time series features of a one-dimensional array; the self-attention layer is used to weightedly fuse the features extracted by multiple LSTM networks; the fully connected layer is used to generate the predicted remaining life according to the fused features.

[0019] Furthermore, the hyperparameters of the prediction model are determined using the particle swarm algorithm.

[0020] Furthermore, the multi-dimensional state parameters include: turn-off voltage peak, turn-on voltage drop and gate threshold voltage.

[0021] In a second aspect, the present invention provides a high-voltage power module remaining life prediction device based on optimized neural network and multi-parameter feature fusion, comprising:

[0022] A state parameter acquisition module, used to acquire multi-dimensional state parameters related to the degradation state of the high-voltage power module;

[0023] A prediction module, used to input multi-dimensional state parameters into a pre-trained prediction model to obtain a prediction result of the remaining life of the high-voltage power module;

[0024] Among them, the structure of the prediction model includes multiple LSTM layers. The prediction model extracts the time series features of the state parameters of one dimension through each LSTM layer, and outputs the prediction results after weighted fusion.

[0025] Accordingly, the present invention also provides a computer device, the device comprising a processor and a memory:

[0026] The memory is used to store the computer program and send the instructions of the computer program to the processor;

[0027] The processor executes a method for predicting the remaining life of a high-voltage power module based on optimized neural network and multi-parameter feature fusion as described in the first aspect according to the instructions of the computer program.

[0028] Correspondingly, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, a method for predicting the remaining life of a high-voltage power module based on the fusion of an optimized neural network and multi-parameter features as in the first aspect is implemented.

[0029] In summary, the present invention provides a method and related device for predicting the remaining life of a high-voltage power module based on optimizing a neural network and fusing multiple parameter features, including obtaining multi-dimensional state parameters related to the degradation state of the high-voltage power module; inputting the multi-dimensional state parameters into a pre-trained prediction model to obtain a prediction result of the remaining life of the high-voltage power module; wherein the structure of the prediction model includes multiple LSTM layers, and the prediction model extracts the time series features of the state parameters of one dimension through each LSTM layer, and outputs the prediction result after weighted fusion. The present invention can improve the accuracy and reliability of the prediction by optimizing the structure and parameters of the neural network and combining multiple key state parameters, thereby ensuring the safe and stable operation of the power electronic system. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0031] Figure 1 A flowchart of a method for predicting the remaining life of a high-voltage power module based on an optimized neural network and multi-parameter feature fusion provided by an embodiment of the present invention;

[0032] Figure 2 A diagram showing the development of the turn-off voltage peak value along with degradation provided by an embodiment of the present invention;

[0033] Figure 3 A diagram of data preprocessing results provided by an embodiment of the present invention;

[0034] Figure 4 A schematic diagram of data reconstruction provided by an embodiment of the present invention;

[0035] Figure 5 A data set partition diagram provided by an embodiment of the present invention;

[0036] Figure 6 A multi-level LSTM network architecture diagram for multi-parameter fusion provided in an embodiment of the present invention;

[0037] Figure 7 A flowchart of a particle swarm algorithm provided by an embodiment of the present invention;

[0038] Figure 8 A block diagram of a high-voltage power module remaining life prediction device based on optimized neural network and multi-parameter feature fusion provided by an embodiment of the present invention;

[0039] Fig. 9 A block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] See also Figure 1 This embodiment provides a method for predicting the remaining life of a high-voltage power module based on an optimized neural network and multi-parameter feature fusion, comprising the following steps:

[0042] S1: Obtain multi-dimensional state parameters related to the degradation state of the high-voltage power module.

[0043] It should be noted that this step collects a series of data related to the operating conditions of the high-voltage power module, and these data parameters can reflect the degradation of the module over time.

[0044] S2: Input the multi-dimensional state parameters into the pre-trained prediction model to obtain the remaining life prediction result of the high-voltage power module;

[0045] Among them, the structure of the prediction model includes multiple LSTM layers. The prediction model extracts the time series features of the state parameters of one dimension through each LSTM layer, and outputs the prediction results after weighted fusion.

[0046] It should be noted that the collected multi-dimensional state parameters will be input into a pre-trained prediction model. This prediction model is based on the LSTM (Long Short-Term Memory) structure, which is a special recurrent neural network (RNN) that is particularly suitable for processing sequence data because it can remember long-term dependencies and has good processing capabilities for both short-term and long-term memories.

[0047] The data extracted by the LSTM layer and weighted fusion will be used to generate the remaining life prediction value of the high-voltage power module. This can help maintenance personnel understand in advance when the module may need to be repaired or replaced, thereby avoiding service interruptions or equipment damage caused by sudden failures.

[0048] This embodiment provides a method for predicting the remaining life of a high-voltage power module based on optimizing a neural network and fusing multiple parameter features. By optimizing the neural network structure and parameters and combining multiple key state parameters, the accuracy and reliability of the prediction are improved, thereby ensuring the safe and stable operation of the power electronic system.

[0049] In some embodiments, the multi-dimensional state parameter includes: turn-off voltage peak, turn-on voltage drop and gate threshold voltage.

[0050] The state of high-voltage power modules will deteriorate after long-term use, which will cause the multi-dimensional health status characterization parameters to shift. For example, after degradation, the off-voltage peak of the high-voltage power module gradually decreases, the on-voltage drop gradually increases, and the gate threshold voltage gradually increases. Figure 2 shown.

[0051] Theoretically, the off-voltage peak value gradually decreases during the aging process. However, due to various interferences in the measurement environment, the off-voltage peak value measurement value has noise of a certain amplitude. Similarly, the measurement values ​​of the on-voltage drop and gate threshold voltage also have noise. A large amount of noise will affect the accuracy of the remaining life prediction model. Therefore, it is necessary to perform noise reduction preprocessing on the data to improve data quality.

[0052] In a further embodiment, for the convenience of operation, this embodiment uses the mature empirical mode decomposition method to reduce the noise of the data. Taking the turn-off voltage peak as an example, four modal components and residual components are decomposed. The modal component has a low amplitude and can be considered as a noise signal. The residual signal has a good correspondence with the original signal trend and is considered to be valid data. Figure 3 shown.

[0053] This embodiment performs data preprocessing on multi-dimensional state parameters such as turn-off voltage peak, on-state voltage drop, gate threshold voltage, etc. of the high-voltage power module, reduces data noise through empirical mode decomposition, improves data quality, and establishes an optimized neural network data set, which is divided into data sets. The training data set is used to train the optimized neural network, and the test data set is used to verify the effectiveness.

[0054] In some embodiments, the training process of the prediction model includes:

[0055] S21: Use a sliding window to reconstruct the data of multi-dimensional state parameters and calculate the remaining life corresponding to each sample.

[0056] Since the degradation of high-voltage power modules follows the principle of time series, and the degradation develops from light to heavy, it is necessary to use time series data to build a database. Select N data windows and perform M-step sliding windows to obtain a database. In this embodiment, N is set to 30, and the number of sliding window steps M is set to 1. The specific principle is Figure 4 shown.

[0057] In addition, the remaining life span corresponding to each sample needs to be calculated.

[0058] S22: The structures of all samples are updated to the corresponding characteristic parameters and remaining lifespan, and a database for optimizing the neural network is formed.

[0059] After the above processing, a large number of samples can be obtained. The structure of each sample is characteristic parameters and remaining life. The specific shape of the sample characteristic parameters is a 30*3 matrix, and the corresponding remaining life is 1*1 data. Finally, the database for optimizing the neural network can be built.

[0060] S23: Train a multi-level LSTM network based on the database of optimized neural networks to obtain a multi-parameter fusion prediction model.

[0061] In order to effectively train and test the prediction model, the data set needs to be divided. Data with a proportion of R is randomly selected and classified into the training data set category, and data with a proportion of 1-R is classified into the test data set category, where R is greater than 0 and less than 1. In this embodiment, R is set to 0.7. The data set is divided as follows Figure 5 shown.

[0062] In a further embodiment, a method for calculating the remaining life is proposed, and the calculation formula is as follows:

[0063]

[0064] In the formula, TEOL represents the final life of the high-voltage power module, T represents the life position of the sample, and TRUL represents the remaining life corresponding to the sample.

[0065] In some embodiments, since the degradation state of the high-voltage power module is jointly characterized by multi-dimensional parameters, establishing a prediction network with multi-parameter fusion can improve the prediction accuracy. Figure 6 shown.

[0066] Taking the above three types of state parameters as an example, the three types of state parameters in the sample, namely the turn-off voltage peak, the on-state voltage drop and the gate threshold voltage, are deconstructed and divided into three one-dimensional arrays. Each dimensional array is input into an LSTM network to extract the features of the time series data. The features extracted from each dimensional data are input into the self-attention layer, the features are weighted and fused, and then the weighted fused features are input into the fully connected layer to finally obtain the predicted remaining life. The network is trained using the training set data, and the weights and biases of each neural network are corrected through back propagation during the training process.

[0067] The model uses the Sigmoid function as the activation function of the three gates, with values ​​between [0,1]; and uses the tanh function as the activation function for candidate information screening, with values ​​between [-1,1]. The expressions of the two activation functions are:

[0068]

[0069]

[0070] This embodiment makes full use of multi-dimensional parameters that characterize the state, such as the turn-off voltage peak, the on-state voltage drop, and the gate threshold voltage, adds an attention mechanism to the optimized neural network, and weightedly fuses the multi-dimensional parameter feature information to more accurately predict the remaining service life of the high-voltage power module.

[0071] In a further embodiment, a method for improving the hyperparameters of the LSTM network based on an adaptive particle swarm algorithm is proposed.

[0072] The process of identifying thermal parameters by particle swarm algorithm is as follows: Figure 7 shown.

[0073] First, the particle randomly selects an initial position and velocity in the solution space, and calculates the objective function fob corresponding to the position. The individual optimal solution and the global optimal solution of this iteration, that is, the position of the minimum fob, are recorded. Then, based on its initial velocity, its understanding of the optimal solution, and the understanding of the optimal solution by other particles, the particle's velocity and position are updated, the state space equation is recalculated, and the fob is recalculated to determine whether the transient response curve obtained by its thermal network model is closer to the transient thermal response curve of the finite element. This process is repeated until the iteration ends. The particle update strategy is as follows

[0074]

[0075]

[0076] In the formula, k is the number of iterations, v k , v k-1 is the speed of the kth and k-1th iterations; x k and x k -1 are the positions of the kth and k-1th iterations respectively; x p,k-1 represents the optimal position of the particle before k-1 iterations; x g,k-1 represents the optimal position of the global particle before k-1 iterations. 1 is the individual learning factor. The larger the factor, the more the particle tends to choose the optimal position of the individual. 2 is the social learning factor. The larger the factor, the more likely the particle is to choose the global optimal position. p is the inertia weight. The larger the value, the more likely the particle is to maintain its original speed and direction and explore non-appearing positions rather than turning to the optimal solution. 1 and r 2 is a random number on [0,1].

[0077] At the same time, in order to reduce the risk of the particle swarm algorithm falling into the local optimum, the particle swarm algorithm is improved by using an adaptive linear decreasing weight strategy. The particle swarm algorithm using the linear decreasing weight strategy means that during the entire search process, the inertia weight decreases linearly with the increase in the number of population iterations. In the initial stage of iteration, a larger inertia weight is used to increase the iteration speed and reduce the probability of the system falling into the local optimum. In the later stage of iteration, a smaller inertia weight is used to improve the accuracy. The ω used here p yes

[0078]

[0079] In the formula, k represents the number of iterations, k max Represents the maximum number of iterations.

[0080] In this patent, the parameter settings of the particle swarm algorithm and the range settings of the optimization parameters are shown in Table 1. The loss function f of the particle swarm algorithm ob Defined as the root mean square error (RMSE) between the test set predicted by the multi-level LSTM network with multi-parameter fusion and the true value

[0081]

[0082] Where n is the number of test set data points, y i is the actual value of the remaining service life of the power module, t i It is the predicted value of the remaining service life of the power module.

[0083] Table 1 Particle swarm algorithm parameters and LSTM hyperparameter ranges

[0084]

[0085] After global optimization by particle swarm algorithm, the optimal hyperparameter combination is shown in Table 2.

[0086] Table 2 LSTM optimal hyperparameter combination

[0087]

[0088] Based on the above operations, an optimized neural network with multi-parameter feature fusion and excellent performance can be obtained. The effectiveness of the model is verified using a test set. The prediction is repeated 10 times, and the average prediction accuracy of the remaining life of the high-voltage power module is as high as 98.6%.

[0089] This embodiment designs an optimized neural network of a long short-term memory network (LSTM) with improved hyperparameters based on an adaptive particle swarm algorithm, so that the performance of the LSTM network is optimized.

[0090] Based on the same inventive concept, the embodiment of the present application also provides a high-voltage power module remaining life prediction device based on optimized neural network and multi-parameter feature fusion for implementing the above-mentioned high-voltage power module remaining life prediction method based on optimized neural network and multi-parameter feature fusion. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in the embodiment of the high-voltage power module remaining life prediction device based on optimized neural network and multi-parameter feature fusion provided below can be found in the above limitations on the high-voltage power module remaining life prediction method based on optimized neural network and multi-parameter feature fusion, which will not be repeated here.

[0091] See also Figure 8 This embodiment provides a high-voltage power module remaining life prediction device based on optimized neural network and multi-parameter feature fusion, including:

[0092] A state parameter acquisition module, used to acquire multi-dimensional state parameters related to the degradation state of the high-voltage power module;

[0093] A prediction module, used to input multi-dimensional state parameters into a pre-trained prediction model to obtain a prediction result of the remaining life of the high-voltage power module;

[0094] Among them, the structure of the prediction model includes multiple LSTM layers. The prediction model extracts the time series features of the state parameters of one dimension through each LSTM layer, and outputs the prediction results after weighted fusion.

[0095] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0096] Reference Fig. 9 An embodiment of the present invention further provides a computer device, comprising: a memory and a processor and a computer program stored in the memory. When the computer program is executed on the processor, a high-voltage power module remaining life prediction method based on optimized neural network and multi-parameter feature fusion as described in any one of the above methods is implemented.

[0097] The computer device may be a desktop computer, a notebook, a PDA, a cloud server or other computing device. The computer device may include, but is not limited to, a processor and a memory. Those skilled in the art will understand that Fig. 9 It is only an example of a computer device and does not constitute a limitation of the computer device. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, it may also include input and output devices, network access devices, etc.

[0098] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0099] In some embodiments, the memory may be an internal storage unit of the computer device, such as a hard disk or memory of the computer device. In other embodiments, the memory may also be an external storage device of the computer device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the computer device. Further, the memory may include both an internal storage unit and an external storage device of the computer device. The memory is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program. The memory may also be used to temporarily store data that has been output or is to be output.

[0100] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for predicting the remaining life of a high-voltage power module based on the fusion of an optimized neural network and multi-parameter features as described in any one of the above methods is implemented.

[0101] In this embodiment, if the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium can at least include: any entity or device that can carry the computer program code to the camera / terminal device, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. For example, USB flash drive, mobile hard disk, disk or optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0102] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0103] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0104] In the embodiments disclosed in the present application, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0105] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the remaining life of a high-voltage power module based on optimized neural network and multi-parameter feature fusion, characterized in that: The steps include: Acquire multi-dimensional state parameters related to the degradation state of the high-voltage power module; Inputting the multi-dimensional state parameters into a pre-trained prediction model to obtain a remaining life prediction result of the high-voltage power module; Among them, the structure of the prediction model includes multiple LSTM layers, and the prediction model extracts the time series characteristics of the state parameters of one dimension through each LSTM layer, and outputs the prediction results after weighted fusion.

2. The method for predicting the remaining life of a high-voltage power module based on optimized neural network and multi-parameter feature fusion according to claim 1 is characterized in that: Obtain multi-dimensional state parameters related to the degradation state of the high-voltage power module, including: The multi-dimensional state parameters are preprocessed using an empirical mode decomposition algorithm, and the decomposed residual trend term signal is retained as valid data to predict the remaining life of the high-voltage power module.

3. The method for predicting the remaining life of a high-voltage power module based on optimized neural network and multi-parameter feature fusion according to claim 1 is characterized in that: The training process of the prediction model includes: Reconstructing the data of the multi-dimensional state parameters by using a sliding window, and calculating the remaining life corresponding to each sample; Update the structures of all samples to the corresponding characteristic parameters and remaining lifespan, and form a database for optimizing the neural network; A multi-level LSTM network is trained based on the database of the optimized neural network to obtain the prediction model of multi-parameter fusion.

4. The method for predicting the remaining life of a high-voltage power module based on optimized neural network and multi-parameter feature fusion according to claim 3 is characterized in that: The remaining life is determined according to the difference between the final life of the high-voltage power module and the life position of the corresponding sample.

5. The method for predicting the remaining life of a high-voltage power module based on optimized neural network and multi-parameter feature fusion according to claim 1 or 3, characterized in that: The structure of the prediction model includes: An input layer, a data deconstruction layer, a multi-level LSTM layer, a self-attention layer, a fully connected layer and an output layer, wherein the data deconstruction layer is used to divide the multi-dimensional state parameters according to different dimensions to obtain multiple one-dimensional arrays; each LSTM network in the multi-level LSTM layer extracts a time series feature of the one-dimensional array; the self-attention layer is used to weightedly fuse the features extracted by multiple LSTM networks; and the fully connected layer is used to generate a predicted remaining life according to the fused features.

6. The method for predicting the remaining life of a high-voltage power module based on optimized neural network and multi-parameter feature fusion according to claim 5 is characterized in that: The hyperparameters of the prediction model are determined using a particle swarm algorithm.

7. The method for predicting the remaining life of a high-voltage power module based on optimized neural network and multi-parameter feature fusion according to claim 1 is characterized in that: The multi-dimensional state parameters include: turn-off voltage peak, turn-on voltage drop and gate threshold voltage.

8. A high-voltage power module remaining life prediction device based on optimized neural network and multi-parameter feature fusion, characterized in that: include: A state parameter acquisition module, used to acquire multi-dimensional state parameters related to the degradation state of the high-voltage power module; A prediction module, used for inputting the multi-dimensional state parameters into a pre-trained prediction model to obtain a prediction result of the remaining life of the high-voltage power module; Among them, the structure of the prediction model includes multiple LSTM layers, and the prediction model extracts the time series characteristics of the state parameters of one dimension through each LSTM layer, and outputs the prediction results after weighted fusion.

9. A computer device, characterized in that: The device comprises a processor and a memory: The memory is used to store a computer program and send instructions of the computer program to the processor; The processor executes a method for predicting the remaining life of a high-voltage power module based on optimized neural network and multi-parameter feature fusion as described in any one of claims 1-7 according to the instructions of the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements a method for predicting the remaining life of a high-voltage power module based on optimized neural network and multi-parameter feature fusion according to any one of claims 1 to 7.

Images