Converter valve component state prediction method and system, and medium

Abstract

The invention provides a state prediction method and system for a converter valve component, and a medium. The method comprises the steps: obtaining the real-time operation data of a to-be-tested converter valve component; on the basis of predation characteristics of a wolf pack algorithm, optimizing model parameters of the initial Mama model to obtain a target Mama model; wherein the initial Mamba model is constructed based on dimension features of historical working data of a plurality of converter valve components, and the historical working data at least comprises historical operation data and historical working state data which have a corresponding relation; and adopting the target Mama model to analyze the real-time operation data, and predicting to obtain the working operation state of the converter valve component to be tested. According to the invention, the accuracy and efficiency of predicting the working operation state of the converter valve component can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence optimization algorithms and power electronics technology, and in particular to a state prediction method, system and medium for converter valve components. Background Art

[0002] The unique insulated-gate bipolar transistor (IGBT) structure enables it to handle high voltage and high current, making it suitable for many fields such as high-voltage direct current transmission, flexible alternating current transmission and renewable energy generation. In the new power system, IGBT is the core component of the converter valve, and the stability of its state is directly related to the operating efficiency and safety of the entire power system.

[0003] In related technologies, a data-driven method is usually used to collect and analyze the working data of IGBT, which can more accurately reflect its operating status. However, this method has a strong dependence on data and poor generalization ability. Once a new and unseen fault mode is encountered, it is often unable to respond effectively. In addition, with the rapid development of artificial intelligence technology, intelligent optimization algorithms have gradually been applied to the state deduction and fault diagnosis of IGBT, but they often have problems such as complex algorithm parameter settings and high computing resource consumption, which limits their promotion and application in practical applications. Therefore, how to overcome the defects of these existing technologies to improve the accuracy and efficiency of IGBT state deduction and fault diagnosis has become a problem that needs to be solved urgently. Summary of the invention

[0004] In order to solve the problems of the prior art, the present invention proposes a state prediction method, system and medium for a converter valve component, aiming to improve the accuracy and efficiency of predicting the working state of the converter valve component to be tested.

[0005] The purpose of the present invention is achieved by adopting the following technical solutions:

[0006] In one aspect, the present invention provides a method for predicting the state of a converter valve component, the method comprising:

[0007] Obtain real-time operating data of the converter valve components to be tested;

[0008] Based on the predation characteristics of the wolf pack algorithm, the model parameters of the initial Mamba model are optimized to obtain the target Mamba model; wherein the initial Mamba model is constructed based on the dimensional characteristics of the historical working data of multiple converter valve components, and the historical working data at least includes: historical operation data and historical working status data with corresponding relationships;

[0009] The target Mamba model is used to analyze the real-time operation data and predict the working state of the converter valve component to be tested.

[0010] Optionally, the predation characteristics based on the wolf pack algorithm are used to optimize the model parameters of the initial Mamba model to obtain the target Mamba model, including:

[0011] Obtain the preset value range corresponding to each model parameter of the initial Mamba model;

[0012] Using the predation characteristics of the wolf pack algorithm, searching and optimizing the preset value range corresponding to each model parameter of the initial Mamba model, and obtaining the optimal parameter combination corresponding to the initial Mamba model;

[0013] The initial Mamba model is trained using the optimal parameter combination to obtain the target Mamba model.

[0014] Optionally, the adopting the optimal parameter combination to train the initial Mamba model to obtain the target Mamba model includes:

[0015] Using the optimal parameter combination to train the initial Mamba model to obtain the Mamba model to be adjusted;

[0016] Using the historical working data, evaluating the performance of the Mamba model to be adjusted to obtain an evaluation result;

[0017] Based on the evaluation result, the model structure and model parameters of the Mamba model to be adjusted are adjusted to obtain the target Mamba model.

[0018] Optionally, before optimizing the model parameters of the initial Mamba model based on the predation characteristics of the wolf pack algorithm to obtain the target Mamba model, the method further includes:

[0019] Based on the dimensional features of the historical work data, a Mamba model to be trained is constructed;

[0020] Based on the historical working data, the parameters of the Mamba model to be trained are trained to obtain the initial Mamba model.

[0021] Optionally, the training of the parameters of the to-be-trained Mamba model based on the historical working data to obtain the initial Mamba model includes:

[0022] Inputting the historical operation data in the historical working data into the Mamba model to be trained to predict the operation status, and obtaining the predicted operation status data;

[0023] Determine the loss between the historical working status data corresponding to the historical operating data and the predicted operating status data;

[0024] According to the loss, the model parameters of the Mamba model to be trained are adjusted until the loss of the predicted operating status data output by the adjusted initial Mamba model meets the convergence condition.

[0025] Optionally, the acquisition of historical operating data of the plurality of converter valve components includes:

[0026] Using sensors to synchronously collect voltage data, current data, temperature data and historical working status data of the multiple converter valve components during historical working processes;

[0027] The voltage data, the current data, the temperature data and the historical working state data are preprocessed respectively to obtain the historical working data.

[0028] Optionally, after the target Mamba model is used to analyze the real-time operation data and predict the working state of the converter valve component to be tested, the method further includes:

[0029] Based on the working operation status, early warning and diagnosis are performed on the potential faults of the converter valve component to be tested, and a diagnosis result of the converter valve component to be tested is obtained.

[0030] Optionally, after the potential fault of the converter valve component to be tested is warned and diagnosed based on the working state and the diagnosis result of the converter valve component to be tested is obtained, the method further includes:

[0031] When the diagnosis result indicates that the converter valve component to be tested is in an abnormal working state, an abnormality resolution suggestion matching the abnormal working state is generated.

[0032] On the other hand, the present invention also provides a state prediction system for a converter valve component, the system comprising:

[0033] An acquisition module is used to acquire real-time operating data of the converter valve components to be tested;

[0034] An optimization module, for optimizing the model parameters of the initial Mamba model based on the predation characteristics of the wolf pack algorithm, and obtaining a target Mamba model; wherein the initial Mamba model is constructed based on the dimensional characteristics of the historical working data of a plurality of converter valve components, and the historical working data at least includes: historical operation data and historical working status data having a corresponding relationship;

[0035] The prediction module is used to analyze the real-time operation data by using the target Mamba model, and predict the working state of the converter valve component to be tested.

[0036] In another aspect, the present invention further provides an electronic device, comprising: at least one processor and a memory; the memory and the processor are connected via a bus;

[0037] The memory is used to store one or more programs;

[0038] When the one or more programs are executed by the at least one processor, the state prediction method of the converter valve component as described above is implemented.

[0039] Correspondingly, the present invention also provides a readable storage medium on which an execution program is stored. When the execution program is executed, the state prediction method of the converter valve component as described above is implemented.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] The present invention provides a state prediction method for a converter valve component. Firstly, an initial Mamba model is constructed with the help of historical operation data and historical working state data of a plurality of converter valve components having a corresponding relationship, so that the constructed Mamba model can better match (adapt to) the characteristics of the IGBT working state data; secondly, the initial Mamba model is optimized with the help of the predation characteristics of the wolf pack algorithm, which helps to further improve the prediction efficiency and accuracy of the optimized Mamba model; finally, the real-time operation data of the converter valve component to be tested is predicted by the optimized Mamba model, that is, the target Mamba model, which can improve the accuracy and efficiency of predicting the working state of the converter valve component to be tested. At the same time, compared with the existing physical model or data-driven methods, the present invention can not only reduce the computational complexity, but also has good generalization ability and adaptability, so that it can better adapt to the complex and changeable working environment of the power system, and further provide strong support for the stable operation of the power system.

[0042] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions provided by the embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work, among which:

[0044] Figure 1 A schematic flow chart of a method for predicting the state of a converter valve component provided by the present invention;

[0045] Figure 2 A schematic diagram of the composition of a state prediction system for a converter valve component provided by the present invention;

[0046] Figure 3 A schematic diagram of the composition of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0047] The following will describe the embodiments of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention, not for limiting the scope of protection of the present invention.

[0048] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0049] In the following description, the terms "first\second\third" involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first\second\third" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present invention described herein can be implemented in an order other than that illustrated or described herein.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the technical field of the embodiments of the present invention. The terms used herein are only for the purpose of describing the embodiments of the present invention and are not intended to limit the embodiments of the present invention.

[0051] IGBT's unique insulated gate bipolar structure enables it to handle high voltage and high current, and is suitable for many fields such as high-voltage DC transmission, flexible AC transmission, and renewable energy generation. In new power systems, IGBT is the core component of the converter valve, and its stability is directly related to the operating efficiency and safety of the entire system. Therefore, accurately predicting the state trend of IGBT is crucial to preventing potential failures and optimizing maintenance plans.

[0052] In the related technologies, the traditional physical model is used to realize the state deduction and fault diagnosis of IGBT. Although it is classic, it often seems to be unable to cope with complex and changeable systems. At the same time, the data-driven method can more accurately reflect its operating status by collecting and analyzing the working data of IGBT, but this method has a strong dependence on data and poor generalization ability. Once it encounters a new and unseen fault mode, it often cannot effectively respond. In addition, with the rapid development of artificial intelligence technology, intelligent optimization algorithms are gradually applied to the state deduction and fault diagnosis of IGBT. These intelligent optimization algorithms can realize intelligent prediction and fault warning of IGBT operating status by simulating human thinking, which can greatly improve the reliability and safety of the system. However, in actual applications, there are often problems such as complex algorithm parameter settings and large consumption of computing resources, which limits its promotion and application in actual applications. Therefore, how to overcome the defects of these existing technologies and improve the accuracy and efficiency of IGBT state deduction and fault diagnosis has become a problem that needs to be solved urgently.

[0053] Based on the above problems, an embodiment of the present invention provides a state prediction method for a converter valve component, such as Figure 1 FIG. 1 is a flow chart of a method for predicting the state of a converter valve component provided by an embodiment of the present invention. Figure 1 The steps shown in the following are explained:

[0054] Step S101: Acquire real-time operating data of the converter valve component to be tested.

[0055] Step S102: Based on the predation characteristics of the wolf pack algorithm, the model parameters of the initial Mamba model are optimized to obtain the target Mamba model.

[0056] The initial Mamba model is constructed based on the dimensional features of historical working data of multiple converter valve components, and the historical working data at least includes: historical operation data and historical working status data with corresponding relationships.

[0057] Step S103: using the target Mamba model to analyze the real-time operation data, and predicting the working state of the converter valve component to be tested.

[0058] In some embodiments, the converter valve component may be a core component of the converter valve, such as an IGBT, and no limitation is imposed on the type of the IGBT.

[0059] In some embodiments, the "obtaining real-time operating data of the converter valve components to be tested" in the above-mentioned step S101 may be to install sensors at key positions corresponding to the converter valve components to be tested, so as to obtain the corresponding real-time operating data in real time; wherein, the real-time operating data includes but is not limited to: real-time voltage data, real-time current data and real-time temperature data of the converter valve components to be tested.

[0060] It should be noted that, in the process of obtaining the real-time operating data of the converter valve components to be tested, a suitable sampling frequency can be selected to perform real-time data collection on the converter valve components to be tested equipped with sensors, so as to improve the accuracy and consistency of data collection.

[0061] In some embodiments, before executing the above step S102 of "optimizing the model parameters of the initial Mamba model based on the predation characteristics of the wolf pack algorithm to obtain the target Mamba model", the following steps A1 and A2 may also be executed:

[0062] Step A1: construct a Mamba model to be trained based on the dimensional features of the historical work data.

[0063] Step A2: Based on the historical working data, the parameters of the Mamba model to be trained are trained to obtain the initial Mamba model.

[0064] In some embodiments, historical working data refers to historical working data of multiple converter valve components; wherein the multiple converter valve components may include the converter valve component to be tested, or may not include the converter valve component to be tested, and the multiple converter valve components and the converter valve component to be tested are components with the same attribute information. In an embodiment of the present invention, the number of components included in the multiple converter valve components may be two, three or more, and the number of components corresponding to the multiple converter valve components may be determined according to actual needs at any time, and the present invention does not impose any limitation on this.

[0065] In some embodiments, the acquisition of historical working data of multiple converter valve components can be achieved through the following process:

[0066] The first step is to use sensors to synchronously collect voltage data, current data, temperature data and historical working status data of the multiple converter valve components during their historical working processes.

[0067] The second step is to pre-process the voltage data, the current data, the temperature data and the historical working status data respectively to obtain the historical working data.

[0068] In some embodiments, the historical working data includes historical operation data and historical working status data having a corresponding relationship, wherein the historical operation data includes but is not limited to: voltage data, current data, and temperature data in the historical working process.

[0069] In some embodiments, sensors may be installed at key positions of each of the multiple converter valve components to detect and read multiple sets of voltage data, current data, temperature data, historical working status data, and other data in real time during the historical working process of the multiple converter valve components. Here, the sensors involved may have the characteristics of high precision, high stability, and fast response, so as to ensure the accuracy and real-time performance of the detected and read data. In addition, the voltage data, current data, and temperature data of the multiple converter valve components during the historical working process may be collected synchronously, so as to ensure the accuracy and consistency of the data. Here, the synchronous collection of data may be achieved with the help of a synchronous clock or a trigger signal.

[0070] It should be noted that in the above data collection process, a suitable sampling frequency can be further selected, such as: the sampling frequency should be high enough to capture the instantaneous state changes of multiple converter valve components. It should be added here that too high a sampling frequency may lead to data redundancy and reduced processing efficiency, so the sampling frequency needs to be determined according to the actual situation.

[0071] In some embodiments, the collected data can be further stored, such as: the collected data (voltage data, current data, temperature data and historical working status data of multiple converter valve components during historical working processes) can be stored in a database or file, and backed up and restored as needed for subsequent analysis and processing.

[0072] In some embodiments, since there is a nonlinear interaction relationship between parameters such as voltage data, current data, temperature data and historical working status data, and voltage data, current data, temperature data and historical working status data usually contain noise, interference and outliers, it is necessary to perform preprocessing operations such as data cleaning and normalization on them, so as to improve the data quality of the historical working data obtained subsequently.

[0073] In some embodiments, the steps of preprocessing the data (voltage data, current data, temperature data, and historical working status data) may refer to the following:

[0074] First, denoising is performed, which can remove high-frequency noise and interference signals in the data through a bandpass filter; then, outlier processing is performed, that is, outliers in the data that may be caused by sensor failure, data transmission errors, etc. are detected and processed, and replaced or deleted;

[0075] Finally, data standardization is performed and the minimum-maximum normalization method is used to convert the data into a unified scale to eliminate the impact of dimension and ensure the stability of model training.

[0076] Based on the above steps, by preprocessing the synchronously collected voltage data, current data, temperature data and historical working status data, higher quality historical working data can be obtained, thereby providing strong parameter support for subsequent analysis and prediction.

[0077] In some embodiments, since the historical working data (the historical operation data and the historical working status data with corresponding relationships) are all one-dimensional data with time as the dimension, the dimensional characteristics of the historical working data (the one-dimensional characteristics of the data) can be used to construct the Mamba model to be trained. Here, the Mamba model to be trained is constructed because the Mamba model adopts a selective state space mechanism, which can efficiently process long sequence data, so that it can be more adapted to the characteristics of the IGBT working status data.

[0078] Here, in the process of building the Mamba model to be trained, we can focus on the design of the model structure (mainly involving encoders and decoders) to ensure that it can fully explore the potential patterns of the state data of the converter valve components to be tested (IGBT is used as an example below). Among them, the encoder part mainly completes the sample feature extraction, which gradually refines the spatiotemporal features of the input sequence through multi-layer self-attention mechanisms and feedforward neural networks.

[0079] In some embodiments, regarding the design of the encoder of the Mamba model to be trained, the input time series X = {x1, x2, ..., x t} as an example; where x t represents the state vector of the IGBT at time t, and the output Z of the i-th layer of the encoder i It can be expressed by the following formula (1):

[0080] Z i =LayerNorm(Z i-1 +MultiHead(Z i-1 ))+FFN(Z i-1 ) formula (1);

[0081] Among them, Z i-1 is the output of the i-1th layer of the encoder, LayerNorm is the layer normalization operation, MultiHead is the multi-head self-attention mechanism, and FFN represents the feedforward neural network, which together constitute the basic unit of the encoder block and can effectively capture long-distance dependencies.

[0082] Correspondingly, the decoder design of the Mamba model to be trained is to reconstruct or predict the working state of the IGBT from the information transmitted by the encoder, mainly the performance at a future point in time. The decoder also adopts a multi-layer structure. Each layer not only contains a self-attention mechanism to process the internal information of the decoder, but also interacts with the output of the encoder through the cross-attention mechanism. The output Z of the jth layer of the decoder j It can be expressed by the following formula (2):

[0083]

[0084] in, represents the output of the jth layer of the decoder, Z M is the last layer output of the encoder. Here, the cross-attention mechanism can effectively utilize historical information, thereby improving prediction accuracy.

[0085] It should be noted that in order to improve the constructed Mamba model to be trained to adapt to the characteristics of IGBT state data, the selection of model parameters of the Mamba model to be trained is particularly critical. For example, the scaling factor in the self-attention in the model parameters, the size of the hidden layer and the number of layers all need to be carefully adjusted according to the data characteristics of the historical working data and the requirements of the prediction task.

[0086] Based on the above steps, the Mamba model to be trained (encoder and decoder, etc.) is constructed based on the dimensional characteristics of the historical working data. At the same time, the constructed Mamba model to be trained is trained through the historical working data, that is, optimized, which can provide an initial Mamba model that can accurately predict the IGBT state.

[0087] In some embodiments, the step A2 of "training the parameters of the Mamba model to be trained based on the historical working data to obtain the initial Mamba model" can be implemented by the following process:

[0088] The first step is to input the historical operation data in the historical work data into the Mamba model to be trained to predict the operation status and obtain the predicted operation status data.

[0089] The second step is to determine the loss between the historical working status data corresponding to the historical operating data and the predicted operating status data.

[0090] In the third step, according to the loss, the model parameters of the Mamba model to be trained are adjusted until the loss of the predicted operating status data output by the adjusted initial Mamba model meets the convergence condition.

[0091] In some embodiments, the mean square error (MSE) can be used as a loss function to measure the deviation between the predicted value output by the Mamba model to be trained and the true value, that is, the MSE is used to determine the loss between the historical working status data corresponding to the historical operating data and the predicted operating status data, as shown in the following formula (3):

[0092]

[0093] Where N is the amount of data to be trained (historical operation data in historical working data), y i and They are respectively the true value of the sample data to be trained (historical working status data) and the predicted value (predicted historical working status data) obtained by inputting the sample data to be trained into the Mamba model to be trained.

[0094] At the same time, the algorithm used to adjust and optimize the model parameters of the Mamba model to be trained according to the loss can be Adam or RMSprop, so as to accelerate convergence and improve training efficiency. Correspondingly, in the model performance evaluation stage involved in the training link, in addition to quantitative indicators (such as accuracy, precision and recall), qualitative analysis should also be considered, such as the comparison between the predicted curve and the actual curve, and the generalization ability of the model under different working conditions should be tested to ensure the reliable performance of the initial Mamba model obtained after training on unknown data. That is, based on the analysis of the IGBT working status data of the Mamba model, the encoder structure and decoder structure that match the dimensional characteristics of the historical working data are used, and the model parameters are optimized in combination with specific task requirements to construct an initial Mamba model that deeply mines the inherent laws in time series data.

[0095] Based on the above steps, the Mamba model to be trained is trained based on historical working data to obtain an initial Mamba model that deeply mines the inherent laws in the time series data, which can provide a powerful tool for state monitoring and fault prediction of IGBTs in power electronic systems.

[0096] In some embodiments, the implementation of "optimizing the model parameters of the initial Mamba model based on the predation characteristics of the wolf pack algorithm to obtain the target Mamba model" in the above step S102 can be implemented by the following steps S1021 to S1023 (not shown in the figure):

[0097] Step S1021: Obtain a preset value range corresponding to each model parameter of the initial Mamba model.

[0098] Step S1022: using the predation characteristics of the wolf pack algorithm, searching and optimizing the preset value range corresponding to each model parameter of the initial Mamba model, and obtaining the optimal parameter combination corresponding to the initial Mamba model.

[0099] Step S1023: adopt the optimal parameter combination to train the initial Mamba model to obtain the target Mamba model.

[0100] In some embodiments, the preset value range corresponding to each model parameter of the initial Mamba model is obtained, including but not limited to: a first value range corresponding to the scaling factor in the attention of the initial Mamba model, a second value range corresponding to the number of hidden layers, the number of layers corresponding to the encoder and the decoder, etc. Exemplarily, the first value range corresponding to the scaling factor is: [1, 2]; the second value range corresponding to the number of hidden layers is: [3, 10], etc.

[0101] In some embodiments, the initial parameters of the wolf pack algorithm, such as the size of the wolf pack and the number of iterations, can be customized or determined according to actual conditions, and the present invention does not impose any limitation on this.

[0102] In some embodiments, the predation characteristics of the wolf pack algorithm are determined based on the predation behavior of the simulated wolf pack, including but not limited to: wandering, summoning, siege and other intelligent behaviors, and the simulated wolf pack's predation behavior corresponds to the intelligent collaboration mechanism and information sharing mechanism of the wolf pack algorithm. Therefore, according to the intelligent collaboration mechanism and information sharing mechanism of the wolf pack algorithm, the preset value range corresponding to each model parameter of the initial Mamba model can be searched and optimized to obtain the optimal parameter combination corresponding to the initial Mamba model, thereby improving the deduction accuracy and efficiency of the optimized Mamba model.

[0103] Referring to the above description, taking the model parameters of the initial Mamba model involving: the first value range corresponding to the scaling factor is: [1, 2]; the second value range corresponding to the number of hidden layers is: [3, 10] as an example, the predation characteristics of the wolf pack algorithm are used to search and optimize the preset value range corresponding to each model parameter of the initial Mamba model, and the optimal parameter combination corresponding to the initial Mamba model is obtained: the value corresponding to the scaling factor is: 1, and the value corresponding to the number of hidden layers is 7.

[0104] It should be noted that the predation characteristics of the wolf pack algorithm are used to simulate the three stages of the wolf pack's predation: wandering stage, summoning stage, and siege stage. The preset value range corresponding to each model parameter of the initial Mamba model is searched and optimized in turn to obtain the optimal parameter combination corresponding to the initial Mamba model. For details, please refer to the following description:

[0105] First, based on the corresponding behavior of the wolf pack in the wandering stage, which corresponds to the scattered search behavior of the wolf pack, the scattered search behavior of the wolf pack algorithm can be used to extensively search for possible high-quality solutions in the parameter space (preset value range) corresponding to each model parameter of the initial Mamba model, and preliminarily locate the parameter area that has a greater impact on the performance of the initial Mamba model.

[0106] Secondly, based on the corresponding behavior of the wolf pack in the calling stage, it uses the individuals in the wolf pack to exchange information and share information to bring together the high-quality solutions found in the previous stage, that is, the wandering stage, and guide the wolf pack to gather towards a better parameter area.

[0107] Finally, based on the corresponding behavior of the wolf pack in the siege stage, that is, based on the concentrated power of the wolf pack, a detailed search and adjustment of the locked high-quality areas are carried out to find the optimal combination of model parameters, that is, the optimal parameter combination corresponding to the initial Mamba model is obtained.

[0108] In some embodiments, the "winner takes it all" alpha wolf generation rule and the "survival of the strong" wolf pack update mechanism in the wolf pack algorithm can also be introduced to improve the search efficiency and convergence speed of the optimal parameter combination of the initial Mamba model.

[0109] Based on the above steps, by applying the wolf pack algorithm to the parameter optimization of the Mamba state space model, the model complexity can be effectively reduced, the amount of calculation can be reduced, and the generalization ability of the target Mamba model can be improved. This can improve the accuracy of IGBT state deduction and enable it to better adapt to the complex and changeable working environment of the power system, providing a strong guarantee for the stable operation of the power system.

[0110] In some embodiments, the "using the optimal parameter combination to train the initial Mamba model to obtain the target Mamba model" in the above step S1023 can be implemented by the following process:

[0111] In the first step, the initial Mamba model is trained using the optimal parameter combination to obtain the Mamba model to be adjusted.

[0112] The second step is to use the historical working data to evaluate the performance of the Mamba model to be adjusted to obtain an evaluation result.

[0113] The third step is to adjust the model structure and model parameters of the Mamba model to be adjusted based on the evaluation result to obtain the target Mamba model.

[0114] In some embodiments, the performance of the Mamba model to be adjusted is evaluated, including but not limited to: evaluating accuracy, precision, and recall. The model structure and parameter settings (e.g., the scaling factor in self-attention, the size of the hidden layer, and the number of layers) can be adjusted according to the evaluation results to achieve the best deduction effect, that is, to obtain the target Mamba model corresponding to the best prediction effect.

[0115] Based on the above steps, the initial Mamba model is trained through the optimal parameter combination and the performance of the trained model is evaluated using historical working data, so that the model structure and parameter settings of the Mamba model (for example, the scaling factor in self-attention, the size of the hidden layer and the number of layers) can be adjusted according to the evaluation results to further improve the prediction effect of the final target Mamba model.

[0116] In some embodiments, after executing step S103 of "using the target Mamba model to analyze the real-time operation data and predict the working state of the converter valve component to be tested", the following step B may also be executed:

[0117] Step B: Based on the working status, early warning and diagnosis are performed on the potential faults of the converter valve component to be tested to obtain the diagnosis result of the converter valve component to be tested.

[0118] In some embodiments, the target Mamba model is applied in practice, that is, the actual working state data of the converter valve component (IGBT) to be tested is predicted or deduced, that is, the future working state of the IGBT can be predicted by inputting real-time data read by online detection (real-time operation data, including but not limited to: real-time voltage data, real-time current data and real-time temperature data), thereby providing decision support for the stable operation of the power electronic system in which the IGBT is located.

[0119] Here, after the predicted working state is obtained, the potential fault of the converter valve component to be tested can be warned and diagnosed according to the working state; illustratively, the working state of the converter valve component to be tested can be characterized as: normal working state and abnormal working state, etc. Here, based on the working state, the potential fault of the converter valve component to be tested can be determined, such as: whether there is aging phenomenon, whether the conduction time is too long, whether the corresponding temperature in the operating state is too high or too low, etc., so as to generate corresponding diagnosis results based on the potential fault, such as: too long conduction time, frequent abnormalities (abnormal operation data), etc.

[0120] Based on the above steps, according to the prediction results (predicted working status), the potential faults of the converter valve components to be tested are warned and diagnosed, so that corresponding maintenance suggestions and solutions can be provided later, thereby ensuring the normal operation of the power system where the converter valve components to be tested are located.

[0121] In some embodiments, after performing the above step B, the following step C may also be performed:

[0122] Step C: when the diagnosis result indicates that the converter valve component to be tested is in an abnormal working state, generate an abnormality solution suggestion matching the abnormal working state.

[0123] In some embodiments, if the diagnostic result indicates that the converter valve component to be tested is in an abnormal working state, then a relevant abnormal solution suggestion or plan matching the abnormal working state is generated to help the operator of the power system where the converter valve component to be tested is located to promptly resolve the abnormal situation, thereby ensuring the normal operation of the power system.

[0124] It should be noted that the form of the exception resolution suggestions can be determined according to actual needs, such as: exception report, exception data, exception analysis table, etc.

[0125] Based on the above steps, corresponding maintenance suggestions and solutions are provided in the event of abnormal working conditions, so as to help operators of the power system where the converter valve components to be tested are located to solve the abnormal situation in time, thereby ensuring the normal operation of the power system.

[0126] Although many technologies have been proposed and applied in the current field of IGBT state deduction and fault diagnosis, they still face some obvious shortcomings. Although the traditional physical model-based method is rigorous in theory, in practical applications, it is often difficult to adapt to the rapidly changing power system environment due to the high complexity of the model and the large amount of calculation. On the other hand, although the data-driven method can more accurately reflect the operating status of the IGBT, it is highly dependent on data and has poor generalization ability. Once it encounters new and unseen fault modes, it is often unable to respond effectively. In addition, although the existing intelligent optimization algorithms have achieved certain results in some scenarios, they often have problems such as complex algorithm parameter settings and high consumption of computing resources, which limits their promotion and application in practical applications.

[0127] In contrast, the present invention optimizes the Mamba model constructed based on the historical working data of multiple converter valve components (at least including: historical operation data and historical working status data with corresponding relationships) through the predation characteristics of the wolf pack algorithm, and analyzes the current operating data of the converter valve component to be tested based on the optimized Mamba model to predict the working status of the converter valve component to be tested; firstly, the initial Mamba model is constructed with the help of the historical operation data and historical working status data of multiple converter valve components with corresponding relationships, so that the constructed Mamba model can better match (adapt to) the characteristics of the IGBT working status data; secondly, the initial Mamba model is optimized with the help of the predation characteristics of the wolf pack algorithm. The initial Mamba model is helpful to further improve the prediction efficiency and accuracy of the optimized Mamba model; finally, the real-time operation data of the converter valve components to be tested are predicted by the optimized Mamba model, that is, the target Mamba model, which can improve the accuracy and efficiency of predicting the working status of the converter valve components to be tested (such as IGBT). At the same time, compared with the existing technology: the physical model or data-driven method adopted, it can not only reduce the calculation complexity, but also has good generalization ability and adaptability, so that it can better adapt to the complex and changeable (such as different scales and types) working environment of the power system, thereby providing strong support for the stable operation of the power system and improving the reliability and safety of the power system.

[0128] The present invention based on the same inventive concept also provides a state prediction system 200 for a converter valve component, such as Figure 2 As shown, the system 200 includes:

[0129] An acquisition module 210 is used to acquire real-time operation data of the converter valve components to be tested;

[0130] The optimization module 220 is used to optimize the model parameters of the initial Mamba model based on the predation characteristics of the wolf pack algorithm to obtain a target Mamba model; wherein the initial Mamba model is constructed based on the dimensional characteristics of the historical working data of multiple converter valve components, and the historical working data at least includes: historical operation data and historical working status data with corresponding relationships;

[0131] The prediction module 230 is used to analyze the real-time operation data using the target Mamba model to predict the working state of the converter valve component to be tested.

[0132] In some embodiments, the optimization module 220 includes: an acquisition submodule, used to obtain a preset value range corresponding to each model parameter of the initial Mamba model; an optimization submodule, used to use the predation characteristics of the wolf pack algorithm to search and optimize the preset value range corresponding to each model parameter of the initial Mamba model to obtain the optimal parameter combination corresponding to the initial Mamba model; a training submodule, used to use the optimal parameter combination to train the initial Mamba model to obtain the target Mamba model.

[0133] In some embodiments, the training submodule is further used to train the initial Mamba model using the optimal parameter combination to obtain the Mamba model to be adjusted; use the historical working data to evaluate the performance of the Mamba model to be adjusted to obtain an evaluation result; based on the evaluation result, adjust the model structure and model parameters of the Mamba model to be adjusted to obtain the target Mamba model.

[0134] In some embodiments, the system 200 further includes: an initial training module, used to construct a Mamba model to be trained based on the dimensional features of the historical working data; and to train the parameters of the Mamba model to be trained based on the historical working data to obtain the initial Mamba model.

[0135] In some embodiments, the initial training module is also used to input the historical operating data in the historical working data into the Mamba model to be trained to predict the operating status and obtain predicted operating status data; determine the loss between the historical working status data corresponding to the historical operating data and the predicted operating status data; and adjust the model parameters of the Mamba model to be trained according to the loss until the loss of the predicted operating status data output by the adjusted initial Mamba model meets the convergence condition.

[0136] In some embodiments, the system 200 also includes: an acquisition and processing module, which uses sensors to synchronously acquire voltage data, current data, temperature data and historical working status data of the multiple converter valve components during historical working processes; and pre-processes the voltage data, the current data, the temperature data and the historical working status data respectively to obtain the historical working data.

[0137] In some embodiments, the system 200 further includes: an early warning and diagnosis module, which is used to warn and diagnose potential faults of the converter valve component to be tested based on the working operation status, and obtain a diagnosis result of the converter valve component to be tested.

[0138] In some embodiments, the system 200 further includes: a generating module, configured to generate an abnormality resolution suggestion matching the abnormal working state when the diagnostic result indicates that the converter valve component to be tested is in an abnormal working state.

[0139] It should be noted that the description of the corresponding embodiment of the system is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. For technical details not disclosed in the system embodiment of the present invention, please refer to the description of the method embodiment of the present invention for understanding.

[0140] Based on the same inventive concept, Figure 3 As shown, the present invention also provides an electronic device, which may be a computer device, a single-chip device, an intelligent mobile device, etc. The electronic device in this embodiment may include a processor 310, a memory 320, a transceiver component 330, etc. The memory 310, the processor 320 and the transceiver component 330 are connected via a bus 340; the memory 320 can be used to store an execution program, and an exemplary execution program may include instructions; the processor 310 is used to execute the instructions stored in the memory. The memory 320 can also be used to store data, which can be called and / or modified when executing instructions.

[0141] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions in a storage medium to implement a corresponding method flow or a corresponding function, so as to implement the steps of a state prediction method for a converter valve component in the above-mentioned embodiment.

[0142] Based on the same inventive concept, the present invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory), which is a memory device in the electronic device for storing programs and data. It can be understood that the storage medium here can include both the built-in storage medium in the electronic device and the extended storage medium supported by the electronic device. The storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more execution programs (including program codes). It should be noted that the storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor loads and executes one or more instructions stored in the storage medium to implement the steps of a state prediction method for a converter valve component in the above embodiment.

[0143] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0144] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0145] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0146] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0147] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for predicting the state of a converter valve component, characterized in that: The method comprises: Obtain real-time operating data of the converter valve components to be tested; Based on the predation characteristics of the wolf pack algorithm, the model parameters of the initial Mamba model are optimized to obtain the target Mamba model; wherein the initial Mamba model is constructed based on the dimensional characteristics of the historical working data of multiple converter valve components, and the historical working data at least includes: historical operation data and historical working status data with corresponding relationships; The target Mamba model is used to analyze the real-time operation data and predict the working state of the converter valve component to be tested.

2. The method according to claim 1, characterized in that The method of optimizing the model parameters of the initial Mamba model based on the predation characteristics of the wolf pack algorithm to obtain the target Mamba model includes: Obtain the preset value range corresponding to each model parameter of the initial Mamba model; Using the predation characteristics of the wolf pack algorithm, searching and optimizing the preset value range corresponding to each model parameter of the initial Mamba model, and obtaining the optimal parameter combination corresponding to the initial Mamba model; The initial Mamba model is trained using the optimal parameter combination to obtain the target Mamba model.

3. The method according to claim 2, characterized in that The adopting the optimal parameter combination to train the initial Mamba model to obtain the target Mamba model includes: Using the optimal parameter combination to train the initial Mamba model to obtain the Mamba model to be adjusted; Using the historical working data, evaluating the performance of the Mamba model to be adjusted to obtain an evaluation result; Based on the evaluation result, the model structure and model parameters of the Mamba model to be adjusted are adjusted to obtain the target Mamba model.

4. The method according to any one of claims 1 to 3, characterized in that: Before optimizing the model parameters of the initial Mamba model based on the predation characteristics of the wolf pack algorithm to obtain the target Mamba model, the method further includes: Based on the dimensional features of the historical work data, a Mamba model to be trained is constructed; Based on the historical working data, the parameters of the Mamba model to be trained are trained to obtain the initial Mamba model.

5. The method according to claim 4, characterized in that The step of training the parameters of the to-be-trained Mamba model based on the historical working data to obtain the initial Mamba model includes: Inputting the historical operation data in the historical working data into the Mamba model to be trained to predict the operation status, and obtaining the predicted operation status data; Determine the loss between the historical working status data corresponding to the historical operating data and the predicted operating status data; According to the loss, the model parameters of the Mamba model to be trained are adjusted until the loss of the predicted operating status data output by the adjusted initial Mamba model meets the convergence condition.

6. The method according to claim 1, characterized in that The acquisition of historical working data of the plurality of converter valve components includes: Using sensors to synchronously collect voltage data, current data, temperature data and historical working status data of the multiple converter valve components during historical working processes; The voltage data, the current data, the temperature data and the historical working state data are preprocessed respectively to obtain the historical working data.

7. The method according to claim 1, characterized in that After the target Mamba model is used to analyze the real-time operation data and predict the working state of the converter valve component to be tested, the method further includes: Based on the working operation status, early warning and diagnosis are performed on the potential faults of the converter valve component to be tested, and a diagnosis result of the converter valve component to be tested is obtained.

8. The method according to claim 7, characterized in that After the potential fault of the converter valve component to be tested is warned and diagnosed based on the working state and the diagnosis result of the converter valve component to be tested is obtained, the method further includes: When the diagnosis result indicates that the converter valve component to be tested is in an abnormal working state, an abnormality resolution suggestion matching the abnormal working state is generated.

9. A state prediction system for a converter valve component, characterized in that: The system comprises: An acquisition module is used to acquire real-time operating data of the converter valve components to be tested; An optimization module, for optimizing the model parameters of the initial Mamba model based on the predation characteristics of the wolf pack algorithm, and obtaining a target Mamba model; wherein the initial Mamba model is constructed based on the dimensional characteristics of the historical working data of a plurality of converter valve components, and the historical working data at least includes: historical operation data and historical working status data having a corresponding relationship; The prediction module is used to analyze the real-time operation data by using the target Mamba model, and predict the working state of the converter valve component to be tested.

10. A readable storage medium, characterized in that: An execution program is stored thereon, and when the execution program is executed, the state prediction method of the converter valve component according to any one of claims 1 to 8 is implemented.

Images