Converter valve multi-element monitoring platform data processing method and device, medium and equipment

Abstract

The invention relates to the technical field of electrical equipment state monitoring, and discloses a converter valve multi-element monitoring platform data processing method and device, a medium and equipment.The method comprises the steps that firstly, simulation data samples of converter valve multi-physical fields under all working conditions are obtained, and then a least square generative adversarial network is constructed and trained; the method comprises the steps of obtaining a simulation data sample, obtaining a multi-physics field parameter sample, combining the simulation data sample with the multi-physics field parameter sample to obtain an extended data sample, training and evaluating a preset evaluation model by using the preprocessed extended data sample to generate a target evaluation model, and finally establishing a fault prediction model based on the extended data sample and the target evaluation model. And processing the multi-physical field data of the converter valve by using the fault prediction model to obtain a converter valve fault prediction result. According to the method, the diversity of the data, the calculation capability of the model and the prediction precision of the model are effectively improved, and efficient processing and fault prediction of the multi-physical field data of the converter valve are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical equipment status monitoring, and in particular to a data processing method and device, medium and equipment for a converter valve multi-element monitoring platform. Background Art

[0002] As the core equipment of DC power transmission control, the converter valve is used to realize the energy conversion between three-phase AC voltage and DC terminal connection. Its working reliability directly affects the operational stability of the transmission project. In a complex electromagnetic environment, thousands of power submodules inside the converter valve are commutated and controlled. During operation, key modules are subjected to continuous high temperature, transient electromagnetic interference and strong mechanical stress. There is a high probability that power devices will be damaged or control system failures will occur. Therefore, the key to the scientific operation and maintenance of DC transmission projects lies in the effective monitoring of the status of the converter valve.

[0003] The existing technology has the following two problems in the status monitoring and fault prediction of converter valves in DC transmission projects: first, the number of real fault samples that can be directly obtained is limited, which restricts the model's ability to learn fault characteristics; second, the operation of key components of converter valves involves multiple physical field parameters, and there are complex influence relationships between multiple physical field parameters. Existing models are difficult to handle highly nonlinear and complex systems. At the same time, the generalization ability of the model is insufficient, which seriously affects the accuracy of fault prediction. Summary of the invention

[0004] In view of this, the present application provides a data processing method and device, medium, and equipment for a multi-element monitoring platform for a converter valve. The main purpose is to solve the technical problem in the prior art that the accuracy of converter valve fault prediction is low due to the limitation of the number of real fault samples and the inability of the model to handle highly nonlinear and complex multi-physical field systems.

[0005] According to a first aspect of the present invention, a method for processing data of a converter valve multivariate monitoring platform is provided, the method comprising:

[0006] Using a preset converter valve multi-physics field simulation model to obtain simulation data samples of the converter valve multi-physics field under all working conditions;

[0007] Constructing a least squares generative adversarial network, training the least squares generative adversarial network based on the simulation data sample to obtain a multi-physics field parameter sample, and combining the simulation data sample with the multi-physics field parameter sample to obtain an extended data sample;

[0008] Preprocessing the extended data samples, using the preprocessed extended data samples to train and evaluate a preset evaluation model to generate a target evaluation model;

[0009] A fault prediction model is established based on the extended data sample and the target evaluation model, multi-physical field data of the converter valve is collected, and the multi-physical field data of the converter valve is processed using the fault prediction model to obtain a converter valve fault prediction result.

[0010] Optionally, the using a preset converter valve multi-physics field simulation model to obtain simulation data samples of the converter valve multi-physics field under all working conditions includes:

[0011] Determine a converter valve to be simulated, and obtain components to be analyzed and corresponding parameter information in the converter valve to be simulated;

[0012] Determine a preset converter valve multi-physics field simulation model based on the components to be analyzed and the corresponding parameter information, set a simulation scene in the converter valve multi-physics field simulation model and define boundary conditions and material properties, perform simulation calculations, and obtain simulation results;

[0013] The simulation results are analyzed and processed to obtain simulation data samples of the multi-physical fields of the converter valve under all working conditions.

[0014] Optionally, constructing a least squares generative adversarial network includes:

[0015] Determine an original adversarial network, and obtain a loss function of a generator and a loss function of a discriminator in the original adversarial network;

[0016] Using the least squares method to process the loss function of the generator and the loss function of the discriminator respectively, to obtain the least squares loss function of the generator and the least squares loss function of the discriminator;

[0017] Training the generator and the discriminator respectively to update the generator and the discriminator;

[0018] Based on the updated generator and discriminator, optimization is performed with the goal of minimizing the least squares loss function of the generator and the least squares loss function of the discriminator to generate a least squares generative adversarial network.

[0019] Optionally, the training the generator and the discriminator respectively to update the generator and the discriminator includes:

[0020] Acquire a real data set and a preset noise distribution of the converter valve, collect a batch of real samples from the real data set, and collect a batch of noise vectors from the noise distribution;

[0021] Inputting the noise vector into the generator to generate a batch of fake samples;

[0022] updating the generator based on the fake samples;

[0023] The real sample and the fake sample are respectively input into the discriminator to update the discriminator.

[0024] Optionally, the preprocessing the extended data sample includes:

[0025] The extended data samples are normalized and standardized.

[0026] Optionally, the using the preprocessed extended data samples to train and evaluate a preset evaluation model to generate a target evaluation model includes:

[0027] Performing feature extraction on the preprocessed extended data samples to obtain feature data samples, wherein the feature data samples include electrical data, thermal data and mechanical data;

[0028] Based on the preset weights corresponding to the electrical data, the thermal data, and the mechanical data, respectively, a weighted sum of the characteristic data samples is calculated, and based on the weighted sum of the characteristic data samples, a state level of the extended data sample is determined, and a data label is added to the data sample based on the state level;

[0029] Dividing the extended data samples into a training set, a validation set, and a test set according to the data labels, and setting basic parameters of a preset evaluation model;

[0030] The preset evaluation model is trained using the training set, and then the preset evaluation model is cross-validated using the validation set to perform hyperparameter tuning, and the preset evaluation model is evaluated using the test set, and the preset evaluation model that passes the evaluation is marked as the target evaluation model.

[0031] Optionally, the using the fault prediction model to process the multi-physical field data of the converter valve to obtain a converter valve fault prediction result includes:

[0032] Processing the multi-physical field data of the converter valve using the fault prediction model to obtain a data processing result;

[0033] Determining the weak point of the converter valve according to the data processing result, and installing a sensor on the weak point of the converter valve to obtain actual monitoring data of the converter valve;

[0034] The data processing result and the actual monitoring data are integrated to obtain a state evaluation result of the converter valve.

[0035] According to a second aspect of the present invention, a data processing device for a converter valve multivariate monitoring platform is provided, the device comprising:

[0036] A data acquisition module, used to acquire simulation data samples of the multi-physical field of the converter valve under all working conditions by using a preset multi-physical field simulation model of the converter valve;

[0037] A data extension module, used for constructing a least squares generative adversarial network, training the least squares generative adversarial network based on the simulation data sample to obtain a multi-physics field parameter sample, and combining the simulation data sample with the multi-physics field parameter sample to obtain an extended data sample;

[0038] A model evaluation module, used to preprocess the extended data samples, train and evaluate a preset evaluation model using the preprocessed extended data samples, and generate a target evaluation model;

[0039] The result output module is used to establish a fault prediction model based on the extended data sample and the target evaluation model, collect multi-physical field data of the converter valve, and use the fault prediction model to process the multi-physical field data of the converter valve to obtain a converter valve fault prediction result.

[0040] According to a third aspect of the present invention, there is provided a storage medium on which a computer program is stored, and when the program is executed by a processor, the data processing method of the multi-element monitoring platform of the converter valve is implemented.

[0041] According to a fourth aspect of the present invention, there is provided a computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned data processing method for the multi-element monitoring platform of the converter valve when executing the program.

[0042] The present invention provides a data processing method, device, medium, and equipment for a multi-element monitoring platform for a converter valve. First, a preset converter valve multi-physical field simulation model is used to obtain simulation data samples of the multi-physical field of the converter valve under all working conditions; then a least squares generative adversarial network is constructed, and the least squares generative adversarial network is trained based on the simulation data samples to obtain multi-physical field parameter samples, and the simulation data samples are combined with the multi-physical field parameter samples to obtain extended data samples; then the extended data samples are preprocessed, and the preprocessed extended data samples are used to train and evaluate a preset evaluation model to generate a target evaluation model; finally, a fault prediction model is established based on the extended data samples and the target evaluation model, and the multi-physical field data of the converter valve is collected, and the multi-physical field data of the converter valve is processed using the fault prediction model to obtain a converter valve fault prediction result. The above method takes into account that the operation of the key components of the converter valve involves a variety of physical field parameters, including electromagnetic fields, temperature rise and stress distribution, forming a highly nonlinear and complex system. Then, combined with the multi-physical field quantity simulation data samples of the converter valve under all working conditions, the state evaluation intelligent algorithm is continuously trained, and the least squares generative adversarial network is generated by deep learning. The trained evaluation model is used to finally generate a fault prediction model, which provides a basis for the state identification and graded fault prediction of the converter valve. The least squares generative adversarial network greatly expands the original data set by generating new multi-physical field parameter samples, supplements the deficiencies in the data set while ensuring data diversity, and improves the prediction accuracy and generalization ability of the model. The trained target evaluation model has strong nonlinear modeling capabilities and efficient computing performance, and can better cope with this multi-physical field optimization problem. The above method effectively improves the diversity of data, the computing power of the model and the prediction accuracy of the model, and realizes efficient processing and fault prediction of the multi-physical field data of the converter valve.

[0043] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0045] Figure 1 A schematic flow chart of a data processing method for a converter valve multi-element monitoring platform provided by an embodiment of the present invention is shown;

[0046] Figure 2A schematic diagram of generating a least squares generative adversarial network in another data processing method for a converter valve multivariate monitoring platform provided by an embodiment of the present invention is shown;

[0047] Figure 3 A flow chart of a method for generating a target evaluation model in a data processing method for a converter valve multivariate monitoring platform provided by an embodiment of the present invention is shown;

[0048] Figure 4 A schematic diagram of an XGBoost model in a data processing method for a converter valve multivariate monitoring platform provided by an embodiment of the present invention is shown;

[0049] Figure 5 A principle logic flow chart of generating a target evaluation model in a data processing method for a converter valve multivariate monitoring platform provided by an embodiment of the present invention is shown;

[0050] Figure 6 A principle logic flow chart of a data processing method for a converter valve multi-element monitoring platform provided by an embodiment of the present invention is shown;

[0051] Figure 7 A schematic diagram of the structure of a data processing device for a converter valve multi-element monitoring platform provided by an embodiment of the present invention is shown;

[0052] Figure 8 A schematic diagram of the device structure of a computer device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0053] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0054] The present application embodiment provides a method for processing data of a converter valve multi-element monitoring platform, such as Figure 1 As shown, the method comprises the following steps:

[0055] 101. Use the preset converter valve multi-physics field simulation model to obtain simulation data samples of the converter valve multi-physics field under all working conditions.

[0056] Among them, the multi-physics field simulation model of the converter valve is a mathematical model constructed based on physical principles and engineering experience. It can simulate the behavior of multi-physics fields such as electric-magnetic-thermal-mechanical under different working conditions. The model comprehensively considers a variety of physical phenomena and can provide parameters such as current, voltage, temperature, and stress to fully reflect the working status of the converter valve; full working conditions means covering all possible working conditions of the converter valve, including normal operation, partial load, full load, overload, fault conditions, etc., and under different working conditions, various physical parameters, such as current, voltage, temperature, etc., will have different values ​​and change trends; by running different working conditions through the simulation model, the corresponding multi-physics field data is generated, and then the multi-physics field parameters under each working condition are recorded to form multiple groups of simulation data samples.

[0057] In the embodiments of the present application, some extreme operating conditions that are difficult to encounter in actual operation can be simulated through simulation, thereby generating data covering various operating conditions, ensuring the diversity and comprehensiveness of the data set; the simulation data can be generated under controlled conditions to ensure the quality and consistency of the data; and a large amount of data can be quickly generated through the multi-physical field simulation model of the converter valve to form a simulation data sample, which contains a large amount of detailed physical parameter information, which helps to have a deeper understanding of the behavior of the converter valve.

[0058] 102. Construct a least squares generative adversarial network, train the least squares generative adversarial network based on simulation data samples, obtain multi-physics field parameter samples, and combine the simulation data samples with the multi-physics field parameter samples to obtain extended data samples.

[0059] Among them, the least squares generative adversarial network is a variant of GAN (Generative Adversarial Network). It trains the generator and discriminator by minimizing the square difference of the loss function. Compared with the traditional adversarial network, the least squares generative adversarial network can produce a more stable learning process and reduce the problem of mode collapse. Specifically, the least squares generative adversarial network can provide a basis for converter valve state identification and graded fault prediction.

[0060] In an embodiment of the present application, a least squares generative adversarial network is used to train with simulated data samples to generate a large number of realistic multi-physics parameter samples, which significantly increases the amount of training data. The simulated data samples are expanded with the multi-physics parameter samples to obtain extended data samples. The extended data samples obtained can cover more possible situations, including some extreme situations that are difficult to obtain in actual data, thereby improving the diversity of the data. More data helps to improve the generalization ability of the machine learning model, and diversified data helps the model to better handle various working conditions and abnormal situations and improve its robustness. At the same time, a rich data set can reduce the risk of overfitting the model during training.

[0061] 103. Preprocess the extended data samples, use the preprocessed extended data samples to train and evaluate the preset evaluation model, and generate a target evaluation model.

[0062] Preprocessing is a very important step in the machine learning process. It processes data to obtain more high-quality data. The preprocessed data is then used to train and evaluate the preset evaluation model to improve the generalization ability of the model. Finally, the target evaluation model is selected and determined from the trained and evaluated evaluation models for use in the subsequent fault prediction process.

[0063] In this embodiment, the extended data samples contain more diverse data, which helps the model learn more comprehensive features. The extended data samples are then preprocessed to ensure data accuracy. More data helps improve the generalization ability of the model and reduce the risk of overfitting of the model during training. Models trained based on high-quality data can provide more accurate status assessment and fault prediction results, providing a scientific basis for operation and maintenance decisions. The target evaluation model can achieve real-time status monitoring and promptly identify potential problems.

[0064] 104. A fault prediction model is established based on the extended data sample and the target evaluation model, and the multi-physical field data of the converter valve is collected. The multi-physical field data of the converter valve is processed using the fault prediction model to obtain the converter valve fault prediction result.

[0065] In this embodiment, the extended data samples include the original data samples obtained from the simulation model and the multi-physical field parameter samples generated by the least squares generative adversarial network, and are combined with the evaluation model with the best performance selected after training and evaluation to form a fault prediction model to further optimize and adjust the multi-physical field data of the real-time converter valve so that it can be specifically used for fault prediction tasks. Finally, the results output by the fault prediction model can be used to determine the current state of the converter valve and predict possible faults, so that the operation and maintenance personnel can take corresponding maintenance measures according to the prediction results to prevent the occurrence of potential faults.

[0066] The present invention provides a data processing method, device, storage medium and computer equipment for a multi-element monitoring platform for a converter valve. First, a preset converter valve multi-physical field simulation model is used to obtain simulation data samples of the multi-physical field of the converter valve under all working conditions; then a least squares generative adversarial network is constructed, and the least squares generative adversarial network is trained based on the simulation data samples to obtain multi-physical field parameter samples, and the simulation data samples are combined with the multi-physical field parameter samples to obtain extended data samples; then the extended data samples are preprocessed, and the preprocessed extended data samples are used to train and evaluate a preset evaluation model to generate a target evaluation model; finally, a fault prediction model is established based on the extended data samples and the target evaluation model, and the multi-physical field data of the converter valve is collected, and the multi-physical field data of the converter valve is processed by the fault prediction model to obtain a converter valve fault prediction result. The above method takes into account that the operation of the key components of the converter valve involves a variety of physical field parameters, including electromagnetic fields, temperature rise and stress distribution, forming a highly nonlinear and complex system. Then, combined with the multi-physical field quantity simulation data samples of the converter valve under all working conditions, the state evaluation intelligent algorithm is continuously trained, and the least squares generative adversarial network is generated by deep learning. The trained evaluation model is used to finally generate a fault prediction model, which provides a basis for the state identification and graded fault prediction of the converter valve. The least squares generative adversarial network greatly expands the original data set by generating new multi-physical field parameter samples, supplements the deficiencies in the data set while ensuring data diversity, and improves the prediction accuracy and generalization ability of the model. The trained target evaluation model has strong nonlinear modeling capabilities and efficient computing performance, and can better cope with this multi-physical field optimization problem. The above method effectively improves the diversity of data, the computing power of the model and the prediction accuracy of the model, and realizes efficient processing and fault prediction of the multi-physical field data of the converter valve.

[0067] Specifically, in the above embodiment, a preset multi-physical field simulation model of a converter valve is used to obtain simulation data samples of the multi-physical field of the converter valve under all operating conditions, including first determining the converter valve to be simulated, obtaining the components to be analyzed and the corresponding parameter information in the converter valve to be simulated, and then determining the preset multi-physical field simulation model of the converter valve based on the components to be analyzed and the corresponding parameter information, setting a simulation scene and defining boundary conditions and material properties in the multi-physical field simulation model of the converter valve, performing simulation calculations, obtaining simulation results, and finally analyzing and processing the simulation results to obtain simulation data samples of the multi-physical field of the converter valve under all operating conditions.

[0068] In this implementation, the Henan Lingbao Power Transmission Project KPE Taking the 4500-72 thyristor converter valve as an example, determine the key components that need to be analyzed and their related electric-magnetic-thermal-mechanical parameters, and then select the appropriate multi-physics simulation software according to the needs. If there is an existing converter valve tower simulation model, it can be used directly. If not, it is necessary to build a model based on the actual design drawings and technical specifications, and then define the boundary conditions and material properties, including setting the working environment, such as defining the electrical, mechanical and thermal environment in the simulation, including power supply, voltage, current, cooling system parameters and ambient medium temperature, and specifying material properties, inputting electromagnetic properties, thermal conductivity, heat capacity ratio and mechanical strength of each component material; then set up the simulation scene, including defining the operating conditions, that is, considering different load conditions, ambient temperature changes and other factors, setting a variety of typical conditions for simulation, and loading the excitation source, applying corresponding electrical signals or other external excitations to the simulation model to simulate the working state under actual operating conditions; after completing the above work content, run the simulation program, and observe the changing trends of various indicators during the simulation process to ensure that the calculation is carried out normally, and finally complete the data collection and processing to obtain the simulation data samples of the multi-physics field of the converter valve under all working conditions.

[0069] Specifically, in the above embodiment, a least squares generative adversarial network is constructed, including first determining the original adversarial network, obtaining the loss function of the generator and the loss function of the discriminator in the original adversarial network, and then using the least squares method to process the loss function of the generator and the loss function of the discriminator respectively to obtain the least squares loss function of the generator and the least squares loss function of the discriminator, and then training the generator and the discriminator respectively to update the generator and the discriminator, and finally, based on the updated generator and discriminator, optimizing with the goal of minimizing the least squares loss function of the generator and the least squares loss function of the discriminator to generate a least squares generative adversarial network.

[0070] In this embodiment, the least squares generative adversarial network is a deep learning model for generating data, which consists of two adversarial networks, a generator and a discriminator. The optimization goal of the least squares generative adversarial network is to generate realistic data by training the generator so that the discriminator cannot distinguish between real data and generated data. In the original adversarial network, the loss function of the generator uses the cross entropy loss. Since the goal of the generator is to maximize the error probability of the discriminator for the generated samples and to ensure the accuracy of the monitoring platform, deeper network learning is required, which leads to the generator encountering the problem of gradient disappearance in the early training stage. The least squares loss function used by the least squares generative adversarial network can alleviate this problem because its gradient is not easy to disappear, thereby making the gradient of the generator more stable.

[0071] The least squares method generates adversarial networks. Figure 2 As shown, specifically, the loss function of the generator in the traditional adversarial network is:

[0072]

[0073] Where z is a random vector sampled from the noise distribution pz, G(z) is the sample output by the generator, and D(G(z)) is the score of the discriminator on the generated sample;

[0074] The loss function of the generator in the traditional adversarial network is:

[0075]

[0076] Where x is a real sample sampled from the real data distribution pdata, D(x) is the score of the discriminator for the real sample, and 1_D(G(z)) is the score for the generated sample;

[0077] Then the least square method is used to improve the above loss function;

[0078] The loss function of the generator in the least squares generative adversarial network is:

[0079]

[0080] The loss function of the discriminator in the least squares generative adversarial network is:

[0081]

[0082] Finally, the least squares generative adversarial network is determined based on the generator loss function improved by the least squares method and the discriminator loss function.

[0083] Furthermore, the generator and the discriminator are trained separately to update the generator and the discriminator, including first obtaining a real data set and a preset noise distribution of the converter valve, collecting a batch of real samples in the real data set, and collecting a batch of noise vectors in the noise distribution, and then inputting the noise vectors into the generator to generate a batch of fake samples, and then updating the generator based on the fake samples, and finally inputting the real samples and the fake samples into the discriminator respectively to update the discriminator.

[0084] In this embodiment, the training process of the discriminator includes sampling a batch of real samples from a real data set and sampling a batch of noise vectors from a noise distribution, generating fake samples through a generator, calculating the loss of the discriminator, and updating the parameters of the discriminator so that it can better distinguish between real samples and generated samples, while the training of the generator includes sampling a batch of noise vectors from a noise distribution, generating fake samples through a generator, calculating the loss of the generator, and updating the parameters of the generator so that the generated samples are more difficult for the discriminator to identify as fake samples.

[0085] The present application embodiment provides a method for generating a target evaluation model, such as Figure 3 As shown, the method comprises the following steps:

[0086] 201. Preprocess the extended data samples.

[0087] In this embodiment, the evaluation model used in this application is the XGBoost model. XGBoost (eXtremeGradient Boosting) is a machine learning algorithm based on gradient boosting decision trees (GBDT). XGBoost has the advantages of high efficiency, accuracy, low overfitting, and good interpretability in processing large-scale classification tasks. The principle of the algorithm is as follows Figure 4 The specific contents are as follows:

[0088]

[0089] Where T is the number of trees; fk corresponds to the score of the kth independent tree;

[0090] The regularized objective function is:

[0091]

[0092] Among them, l is a differentiable convex loss function, which is used to express the predicted value and the true value y i The error between them; Ω is the complexity function; Tk is the number of leaf nodes in the tree; γ is the penalty coefficient, and its value range is [0, 1]; λ is the regular term coefficient.

[0093] The XGBoost model itself does not need to standardize or normalize the data because it is a tree-based algorithm that is insensitive to the scale of features and is not greatly affected by the scale of features. However, for the special case of huge feature scales of electromagnetic, magnetic, thermal and mechanical multi-physics field simulation data, the XGBoost model needs to be normalized and standardized.

[0094] Standardization converts the data into a distribution with a mean of 0 and a standard deviation of 1. The formula is:

[0095]

[0096] Among them, X is the original data, μ is the mean, and σ is the standard deviation;

[0097] Normalization scales the data to the range of [0, 1], and proposes a classification maximum normalization method with the formula:

[0098]

[0099] Among them, X c is one type of raw data, It is feature X c The maximum value of . ε is a constant to prevent division by zero;

[0100] Although preprocessing the data will increase the number of data processing steps, in the converter valve monitoring that integrates multi-physics field simulation, the data that needs to be obtained must first be accurate to ensure that the weak points of the key components of the converter valve are accurately found and the sensors are arranged.

[0101] 202. Use the preprocessed extended data samples to train and evaluate the preset evaluation model to generate a target evaluation model.

[0102] Specifically, feature extraction is performed on the preprocessed extended data samples to obtain feature data samples, wherein the feature data samples include electrical data, thermal data and mechanical data; based on the preset weights corresponding to the electrical data, thermal data and mechanical data, a weighted sum of the feature data samples is calculated, and based on the weighted sum of the feature data samples, a state level of the extended data samples is determined, and data labels are added to the data samples based on the state level; the extended data samples are divided into a training set, a validation set and a test set according to the data labels, and basic parameters of a preset evaluation model are set; the preset evaluation model is trained using the training set, and the preset evaluation model is cross-validated using the validation set to perform hyperparameter tuning, and the preset evaluation model is evaluated using the test set, and the preset evaluation model that passes the evaluation is marked as a target evaluation model.

[0103] In this embodiment, the preprocessed extended data samples are obtained, and then useful features are extracted from the extended data samples according to the valve tower, valve assembly, and component classification, that is, feature data samples are obtained, and labels can be added to each feature data sample to indicate the four states of normal, attention, abnormal, and severe. Considering that the model has three categories of output results, the weighted sum of the three categories is proposed as the output result. Specifically, f(x1), f(x2), and f(x3) are set to represent the three categories of data: electricity, heat, and force, and their weighted sum is calculated:

[0104] f total (x)=ω1f1(x)+ω2f2(x)+ω3f3(x),

[0105] Based on the weighted sum f total (x) to classify:

[0106]

[0107] Then, the extended data samples are divided into training set, validation set and test set according to the data labels. Usually, 70% of the data is used as the training set, 15% of the data is used as the validation set, and 15% of the data is used as the test set. Ensure that the sample size of the training set is large enough to include a variety of working conditions and states to ensure the generalization ability of the model. Use the training set to train the XGBoost model, perform hyperparameter tuning as needed, and use the test set to evaluate the performance of the model. Use accuracy, recall rate, and F1 score to measure the performance of the model, such as Figure 5 Ensure that the test set is independent from the training set and does not contain overlapping samples to avoid overfitting.

[0108] Specifically, in the above embodiment, the multi-physical field data of the converter valve is processed by using the fault prediction model to obtain the converter valve fault prediction result, including firstly processing the multi-physical field data of the converter valve by using the fault prediction model to obtain the data processing result, then determining the weak point of the converter valve according to the data processing result, and installing sensors at the weak point of the converter valve to obtain the actual monitoring data of the converter valve, and finally fusing the data processing result and the actual monitoring data to obtain the state evaluation result of the converter valve.

[0109] In this embodiment, the multi-physical field data of the converter valve is processed using a fault prediction model to obtain a data processing result, and then the weak points of the converter valve are determined based on the data processing result, specifically including analyzing the data processing result to determine which parts or parameters show a higher fault risk or abnormal situation, and using the feature importance score provided by the model to identify the key parameters and parts that have the greatest impact on fault prediction; then, sensors are installed on the identified weak points to monitor the multi-physical field parameters of these key parts in real time, and then the multi-physical field data in actual operation is continuously collected through the sensors, and the data processing results and the actual monitoring data are fused to obtain the state evaluation results of the converter valve, and finally, based on the fused data, the state evaluation results of the converter valve are generated for judging the current state, predicting future fault risks, and providing maintenance suggestions, wherein the results provided by the fault prediction model can provide specific fault prediction information and maintenance suggestions for operation and maintenance personnel, which helps to make more scientific and reasonable decisions.

[0110] In summary, the principle and logic flow of the data processing method of the converter valve multi-element monitoring platform provided in this application is as follows: Figure 6 shown.

[0111] Further, as Figure 1 The specific implementation of the method, the embodiment of the present application provides a data processing device for a converter valve multi-element monitoring platform, such as Figure 7 As shown, the device includes: a data acquisition module 301, a data expansion module 302, a model evaluation module 303 and a result output module 304.

[0112] The data acquisition module 301 is used to acquire simulation data samples of the multi-physical field of the converter valve under all working conditions by using a preset multi-physical field simulation model of the converter valve;

[0113] A data extension module 302 is used to construct a least squares generative adversarial network, train the least squares generative adversarial network based on the simulation data sample to obtain a multi-physics field parameter sample, and combine the simulation data sample with the multi-physics field parameter sample to obtain an extended data sample;

[0114] The model evaluation module 303 is used to preprocess the extended data samples, train and evaluate the preset evaluation model using the preprocessed extended data samples, and generate a target evaluation model;

[0115] The result output module 304 is used to establish a fault prediction model based on the extended data sample and the target evaluation model, collect the multi-physical field data of the converter valve, and use the fault prediction model to process the multi-physical field data of the converter valve to obtain the converter valve fault prediction result.

[0116] In a specific application scenario, the data acquisition module 301 can be used to determine the converter valve to be simulated, obtain the components to be analyzed and the corresponding parameter information in the converter valve to be simulated; determine the preset converter valve multi-physical field simulation model based on the components to be analyzed and the corresponding parameter information, set the simulation scene in the converter valve multi-physical field simulation model and define boundary conditions and material properties, perform simulation calculations, and obtain simulation results; analyze and process the simulation results to obtain simulation data samples of the converter valve multi-physical field under all working conditions.

[0117] In a specific application scenario, the data expansion module 302 can be specifically used to determine the original adversarial network, obtain the loss function of the generator and the loss function of the discriminator in the original adversarial network; use the least squares method to process the loss function of the generator and the loss function of the discriminator respectively to obtain the least squares loss function of the generator and the least squares loss function of the discriminator; train the generator and the discriminator respectively to update the generator and the discriminator; based on the updated generator and discriminator, optimize with the goal of minimizing the least squares loss function and the least squares loss function to generate a least squares generative adversarial network.

[0118] In a specific application scenario, the data expansion module 302 can also be used to obtain a real data set and a preset noise distribution of the converter valve, collect a batch of real samples in the real data set, and collect a batch of noise vectors in the noise distribution; input the noise vectors into the generator to generate a batch of false samples; update the generator based on the false samples; input the real samples and the false samples into the discriminator respectively to update the discriminator.

[0119] In a specific application scenario, the model evaluation module 303 is specifically used to normalize and standardize the extended data samples.

[0120] In a specific application scenario, the model evaluation module 303 can also be used to perform feature extraction on the preprocessed extended data samples to obtain feature data samples, wherein the feature data samples include electrical data, thermal data and mechanical data; based on the preset weights corresponding to the electrical data, thermal data and mechanical data respectively, the weighted sum of the feature data samples is calculated, and the state level of the extended data samples is determined based on the weighted sum of the feature data samples, and data labels are added to the data samples based on the state level; the extended data samples are divided into a training set, a validation set and a test set according to the data labels, and the basic parameters of a preset evaluation model are set; the preset evaluation model is trained using the training set, and the preset evaluation model is cross-validated using the validation set to perform hyperparameter tuning, and the preset evaluation model is evaluated using the test set, and the preset evaluation model that passes the evaluation is marked as a target evaluation model.

[0121] In a specific application scenario, the result output module 304 can be specifically used to process the multi-physical field data of the converter valve using the fault prediction model to obtain data processing results; determine the weak points of the converter valve based on the data processing results, and install sensors at the weak points of the converter valve to obtain actual monitoring data of the converter valve; and fuse the data processing results and the actual monitoring data to obtain a status evaluation result of the converter valve.

[0122] It should be noted that for other corresponding descriptions of the functional units involved in the data processing device for a converter valve multi-element monitoring platform provided in this embodiment, reference can be made to Figure 1 The corresponding description in will not be repeated here.

[0123] Based on the above Figure 1 The method shown, accordingly, this embodiment also provides a storage medium, on which a computer program is stored, and when the program is executed by a processor, the data processing method of the multi-element monitoring platform of the converter valve is implemented.

[0124] Based on this understanding, the technical solution of the present application can be embodied in the form of a software product. The software product to be identified can be stored in a non-volatile storage medium (which can be a CD-ROM, USB flash drive, mobile hard disk, etc.), including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute the data processing method of the multi-element monitoring platform of the converter valve in each implementation scenario of the present application.

[0125] Based on the above Figure 1 The method shown, and Figure 7 In order to achieve the above-mentioned purpose, the embodiment of the data processing device for the multi-element monitoring platform of the converter valve shown in the figure is as follows: Figure 8 As shown, this embodiment also provides a physical device for data processing of a multi-element monitoring platform for a converter valve, the device includes a communication bus, a processor, a memory and a communication interface, and may also include an input and output interface and a display device, wherein each functional unit can communicate with each other through the bus. The memory stores a computer program, and the processor is used to execute the program stored in the memory and execute the data processing method of the multi-element monitoring platform for a converter valve in the above embodiment.

[0126] Optionally, the physical device may also include a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, etc. The user interface may include a display, an input unit such as a keyboard, etc., and the optional user interface may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface), etc.

[0127] Those skilled in the art will appreciate that the data processing entity device structure of a converter valve multi-element monitoring platform provided in this embodiment does not constitute a limitation on the entity device, and may include more or fewer components, or a combination of certain components, or different component arrangements.

[0128] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware of the physical device and the software resources to be identified, and supports the operation of the information processing program and other software and / or programs to be identified. The network communication module is used to realize the communication between the components inside the storage medium, and the communication with other hardware and software in the information processing physical device.

[0129] Through the description of the above implementation methods, the technical personnel in this field can clearly understand that the present application can be implemented by means of software plus the necessary general hardware platform, or by hardware. By applying the technical solution of the present application, firstly, the preset multi-physical field simulation model of the converter valve is used to obtain the simulation data samples of the multi-physical field of the converter valve under all working conditions; then, a least squares generative adversarial network is constructed, and the least squares generative adversarial network is trained based on the simulation data samples to obtain the multi-physical field parameter samples, and the simulation data samples are combined with the multi-physical field parameter samples to obtain the extended data samples; then, the extended data samples are preprocessed, and the pre-processed extended data samples are used to train and evaluate the preset evaluation model to generate the target evaluation model; finally, a fault prediction model is established based on the extended data samples and the target evaluation model, the multi-physical field data of the converter valve is collected, and the multi-physical field data of the converter valve is processed by the fault prediction model to obtain the converter valve fault prediction result. The above method takes into account that the operation of the key components of the converter valve involves a variety of physical field parameters, including electromagnetic fields, temperature rise and stress distribution, forming a highly nonlinear and complex system. Then, combined with the multi-physical field quantity simulation data samples of the converter valve under all working conditions, the state evaluation intelligent algorithm is continuously trained, and the least squares generative adversarial network is generated by deep learning. The trained evaluation model is used to finally generate a fault prediction model, which provides a basis for the state identification and graded fault prediction of the converter valve. The least squares generative adversarial network greatly expands the original data set by generating new multi-physical field parameter samples, supplements the deficiencies in the data set while ensuring data diversity, and improves the prediction accuracy and generalization ability of the model. The trained target evaluation model has strong nonlinear modeling capabilities and efficient computing performance, and can better cope with this multi-physical field optimization problem. The above method effectively improves the diversity of data, the computing power of the model and the prediction accuracy of the model, and realizes efficient processing and fault prediction of the multi-physical field data of the converter valve.

[0130] Those skilled in the art will appreciate that the accompanying drawings are only schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily necessary for implementing the present application. Those skilled in the art will appreciate that the modules in the devices in the implementation scenario can be distributed in the devices of the implementation scenario according to the description of the implementation scenario, or can be changed accordingly and located in one or more devices different from the present implementation scenario. The modules of the above-mentioned implementation scenario can be combined into one module, or can be further split into multiple submodules.

[0131] The above serial numbers of this application are only for description and do not represent the advantages and disadvantages of the implementation scenarios. The above disclosure is only a few specific implementation scenarios of this application, but this application is not limited to them, and any changes that can be thought of by technicians in this field should fall within the scope of protection of this application.

Claims

1. A data processing method for a converter valve multivariate monitoring platform, characterized in that: include: Using a preset converter valve multi-physics field simulation model to obtain simulation data samples of the converter valve multi-physics field under all working conditions; Constructing a least squares generative adversarial network, training the least squares generative adversarial network based on the simulation data sample to obtain a multi-physics field parameter sample, and combining the simulation data sample with the multi-physics field parameter sample to obtain an extended data sample; Preprocessing the extended data samples, using the preprocessed extended data samples to train and evaluate a preset evaluation model to generate a target evaluation model; A fault prediction model is established based on the extended data sample and the target evaluation model, multi-physical field data of the converter valve is collected, and the multi-physical field data of the converter valve is processed using the fault prediction model to obtain a converter valve fault prediction result.

2. The method according to claim 1, characterized in that: The method of obtaining simulation data samples of the multi-physical field of the converter valve under all working conditions by using a preset multi-physical field simulation model of the converter valve includes: Determine a converter valve to be simulated, and obtain components to be analyzed and corresponding parameter information in the converter valve to be simulated; Determine a preset converter valve multi-physics field simulation model based on the components to be analyzed and the corresponding parameter information, set a simulation scene in the converter valve multi-physics field simulation model and define boundary conditions and material properties, perform simulation calculations, and obtain simulation results; The simulation results are analyzed and processed to obtain simulation data samples of the multi-physical fields of the converter valve under all working conditions.

3. The method according to claim 1, characterized in that The least squares generative adversarial network is constructed, comprising: Determine an original adversarial network, and obtain a loss function of a generator and a loss function of a discriminator in the original adversarial network; Using the least squares method to process the loss function of the generator and the loss function of the discriminator respectively, to obtain the least squares loss function of the generator and the least squares loss function of the discriminator; Training the generator and the discriminator respectively to update the generator and the discriminator; Based on the updated generator and discriminator, optimization is performed with the goal of minimizing the least squares loss function of the generator and the least squares loss function of the discriminator to generate a least squares generative adversarial network.

4. The method according to claim 1, characterized in that: The training of the generator and the discriminator respectively to update the generator and the discriminator comprises: Acquire a real data set and a preset noise distribution of the converter valve, collect a batch of real samples from the real data set, and collect a batch of noise vectors from the noise distribution; Inputting the noise vector into the generator to generate a batch of fake samples; updating the generator based on the fake samples; The real sample and the fake sample are respectively input into the discriminator to update the discriminator.

5. The method according to claim 1, characterized in that The preprocessing of the extended data sample comprises: The extended data samples are normalized and standardized.

6. The method according to claim 1, characterized in that The method of training and evaluating a preset evaluation model using the preprocessed extended data samples to generate a target evaluation model includes: Performing feature extraction on the preprocessed extended data samples to obtain feature data samples, wherein the feature data samples include electrical data, thermal data and mechanical data; Based on the preset weights corresponding to the electrical data, the thermal data, and the mechanical data, respectively, a weighted sum of the characteristic data samples is calculated, and based on the weighted sum of the characteristic data samples, a state level of the extended data sample is determined, and a data label is added to the data sample based on the state level; Dividing the extended data samples into a training set, a validation set, and a test set according to the data labels, and setting basic parameters of a preset evaluation model; The preset evaluation model is trained using the training set, and then the preset evaluation model is cross-validated using the validation set to perform hyperparameter tuning, and the preset evaluation model is evaluated using the test set, and the preset evaluation model that passes the evaluation is marked as the target evaluation model.

7. The method according to claim 1, characterized in that The method of processing the multi-physical field data of the converter valve by using the fault prediction model to obtain a converter valve fault prediction result includes: Processing the multi-physical field data of the converter valve using the fault prediction model to obtain a data processing result; Determining the weak point of the converter valve according to the data processing result, and installing a sensor on the weak point of the converter valve to obtain actual monitoring data of the converter valve; The data processing result and the actual monitoring data are integrated to obtain a state evaluation result of the converter valve.

8. A data processing device for a converter valve multi-element monitoring platform, characterized in that: The device comprises: A data acquisition module, used to acquire simulation data samples of the multi-physical field of the converter valve under all working conditions by using a preset multi-physical field simulation model of the converter valve; A data extension module, used for constructing a least squares generative adversarial network, training the least squares generative adversarial network based on the simulation data sample to obtain a multi-physics field parameter sample, and combining the simulation data sample with the multi-physics field parameter sample to obtain an extended data sample; A model evaluation module, used to preprocess the extended data samples, train and evaluate a preset evaluation model using the preprocessed extended data samples, and generate a target evaluation model; The result output module is used to establish a fault prediction model based on the extended data sample and the target evaluation model, collect multi-physical field data of the converter valve, and use the fault prediction model to process the multi-physical field data of the converter valve to obtain a converter valve fault prediction result.

9. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images