Broadband mutual inductor error prediction method, device, equipment and medium

Abstract

The invention relates to the technical field of mutual inductor online monitoring, and provides a broadband mutual inductor error prediction method and device, equipment and a medium, and the method comprises the steps: firstly obtaining a historical influence parameter data set of a broadband mutual inductor; secondly, screening the plurality of influence parameters based on an extreme random tree algorithm and a historical influence parameter data set to obtain a target influence parameter; determining a target influence parameter data set based on the target influence parameter and the historical influence parameter data set; then, constructing an error prediction model; training the error prediction model based on the target influence parameter data set to obtain a trained error prediction model; and finally, obtaining a real-time influence parameter data set of the broadband mutual inductor. And determining an error prediction value of the broadband mutual inductor based on the real-time influence parameter data set and the trained error prediction model. According to the invention, the influence parameter and the error parameter of the mutual inductor are analyzed, the change trend of the error parameter is predicted, and the error state of the mutual inductor is monitored in real time.

Description

Technical Field

[0001] The present disclosure relates to the technical field of online monitoring of transformers, and in particular to a method, device, equipment and medium for predicting wide-band transformer errors. Background Art

[0002] With the development of smart grids, mutual inductors are increasingly used in smart substations and have become one of the important equipment supporting the digital construction of power grids. They provide high-precision measurement data for metering and protection equipment. The rapid development of new energy has made the power system face more complex power signals. Especially in the process of grid connection, the volatility and instability of new energy have aggravated the complexity of grid signals, resulting in fluctuations in frequency and harmonic content. Traditional electromagnetic mutual inductors can no longer meet these new requirements due to their limited frequency response and dynamic range. Therefore, wideband mutual inductors have emerged as the times require and have become the inevitable direction for the construction of smart grids and new energy grids. In order to adapt to the access of new energy, mutual inductors need to have wide-band and wide-dynamic range measurement capabilities while ensuring long-term operational stability.

[0003] In the related art, the error analysis and performance detection of the transformer mainly rely on regular offline calibration and long-term online monitoring. Although these methods can detect errors, they still have limitations: offline calibration requires the use of standard transformers, and the cycle is long and the equipment is complex, resulting in high costs and inefficient workflows. Although online monitoring can monitor the operating status in real time, it is still difficult to meet the requirements of broadband transformers due to the lack of high-precision equipment and standard sources. In addition, the existing broadband current transformer calibration system mainly relies on standard source calibration, but the standard source is easily disturbed by environmental factors such as temperature, magnetic field and vibration, which affects the measurement accuracy, and can only be operated offline, and cannot achieve real-time or efficient on-site calibration. Summary of the invention

[0004] The embodiments of the present disclosure at least provide a wideband transformer error prediction method, device, equipment and medium, which analyzes the influencing parameters and error parameters of the transformer, predicts the change trend of the error parameters, and realizes real-time monitoring of the transformer error state.

[0005] The embodiment of the present disclosure provides a method for predicting a wide-band mutual inductor error, including: Acquire a historical influencing parameter data set of a wideband mutual inductor; wherein the historical influencing parameter data set includes historical parameter data subsets at different times, and each of the historical parameter data subsets includes a mutual inductor error at a corresponding time and data corresponding to a plurality of influencing parameters; Based on the extreme random tree algorithm and the historical influencing parameter data set, multiple influencing parameters are screened to obtain target influencing parameters; and based on the target influencing parameters and the historical influencing parameter data set, a target influencing parameter data set is determined; wherein the target influencing parameter data set includes target parameter data subsets at different times, and each target parameter data subset includes a mutual inductor error at a corresponding time and data corresponding to multiple target influencing parameters respectively; Constructing an error prediction model; and training the error prediction model based on the target influencing parameter data set to obtain a trained error prediction model; Acquire a real-time influencing parameter data set of the wideband mutual inductor; and determine an error prediction value of the wideband mutual inductor based on the real-time influencing parameter data set and the trained error prediction model; wherein the real-time influencing parameter data set includes real-time data corresponding to multiple target influencing parameters.

[0006] In some possible embodiments, the screening of multiple influencing parameters based on the extreme random tree algorithm and the historical influencing parameter data set includes: Based on the extreme random tree algorithm, a plurality of influencing parameters and segmentation thresholds are randomly selected to construct nodes of a decision tree, so as to generate a plurality of decision trees with different structures and prediction capabilities; and each decision tree is trained based on the historical influencing parameter data set; Based on the weight calculation formula and the training results of each decision tree, the weight of each decision tree is determined respectively; and based on the error rate calculation formula and the training results of each decision tree, the error rate of each influencing parameter in each decision tree on the mutual inductor error prediction is determined respectively; Based on the weight of each decision tree, the error rate of each influencing parameter in each decision tree on the prediction of the mutual inductor error, and the importance score formula, respectively calculate the importance score of each influencing parameter; The target influencing parameters are determined based on the preset screening conditions and the importance scores of each influencing parameter.

[0007] In some possible embodiments, the weight calculation formula includes: ; in, Represented as the weight of the jth decision tree; It is represented by the accuracy of the jth decision tree; M is represented by the total number of decision trees; Expressed as the accuracy of the kth decision tree; The error rate calculation formula includes: ; in, It is expressed as the error rate corresponding to the i-th influencing parameter of the j-th decision tree; Represented as the prediction result of the decision tree; Represented as the true label; Expressed as an indicator function, when When it is 1, otherwise it is 0; It is represented as the data subset obtained by splitting the i-th influencing parameter in the j-th decision tree; The importance score formula includes: ; in, It is represented as the importance score of the i-th influencing parameter; n is represented as the total amount of data corresponding to the i-th influencing parameter.

[0008] In some possible embodiments, the training of the error prediction model based on the target influencing parameter data set includes: Obtain an error prediction model to be trained; wherein the error prediction model to be trained includes a CNN module, a BiLSTM module, an AM module and an error prediction module; Constructing a model training set based on the target influencing parameter data set; wherein the model training set includes a training data set, a verification data set and a sample data set; Based on the training data set, the verification data set and the dung beetle optimization algorithm, the error prediction model to be trained is first trained to obtain optimal hyperparameters of the error prediction model to be trained; The error prediction model to be trained is parameterized according to the optimal hyperparameters to obtain a target error prediction model to be trained; and the target error prediction model to be trained is subjected to a second training based on the training data set and the sample data set to obtain the trained error prediction model.

[0009] In some possible embodiments, the first training of the error prediction model to be trained based on the training data set, the verification data set and the dung beetle optimization algorithm includes: Step 1: Set the number of model iterations and multiple dung beetles, and randomly generate a set of hyperparameters for each dung beetle; Step 2: configuring the error prediction model to be trained based on the hyperparameters corresponding to each dung beetle to obtain the error prediction model to be trained corresponding to each dung beetle; Step 3: for the error prediction model to be trained corresponding to each dung beetle, the error prediction model to be trained is trained based on the training data set, and the trained error prediction model is verified based on the verification data set, and the prediction error corresponding to the dung beetle is calculated; Step 4: Calculate the fitness value of each dung beetle based on the fitness function and the prediction error corresponding to the dung beetle; and adjust the hyperparameters corresponding to each dung beetle based on the fitness value of each dung beetle and the update rule of the dung beetle optimization algorithm; Step 5: Return to step 2 above until the training result meets the preset requirements and / or the number of training times reaches the number of iterations, and obtain the optimal hyperparameters of the error prediction model.

[0010] In some possible embodiments, the fitness function is expressed as: ; in, It is expressed as the fitness value of the i-th dung beetle; It is represented as the prediction error of the error prediction model corresponding to the i-th dung beetle on the validation set; The update rule is expressed as: ; in, Represents the updated position of the i-th dung beetle in the next iteration; It is represented as the position of the i-th dung beetle in the t-th iteration; Indicates the position of the individual with the best fitness among all dung beetles; Represented as random positions; and Expressed as a step size coefficient, it controls the movement amplitude of the dung beetle.

[0011] In some possible embodiments, the performing a second training on the target error prediction model to be trained based on the training data set and the sample data set includes: Step a: Based on the CNN module, local features are extracted for each training data in the training data set to obtain local feature vectors corresponding to each training data; based on the BiLSTM module, the temporal dependency between the local feature vectors corresponding to each training data is captured to obtain a temporal feature vector corresponding to each training data; and, for each training data, the local feature vector and the temporal feature vector corresponding to the training data are subjected to attention weighting processing based on the AM module to obtain a fused feature vector corresponding to the training data; and, based on the error prediction module and the fused feature vector, a predicted mutual inductor error corresponding to the training data is determined; Step b: determining a loss value between a predicted mutual inductor error corresponding to the training data and a mutual inductor error corresponding to the training data in the sample data set based on a preset loss function, and adjusting a model parameter of the target error prediction model to be trained based on the loss value; Repeat steps a to b above until the training results meet the preset requirements to obtain the trained error prediction model.

[0012] The embodiment of the present disclosure provides a wide-band mutual inductor error prediction device, comprising: A data acquisition module, used to acquire a historical influencing parameter data set of a wideband mutual inductor; wherein the historical influencing parameter data set includes historical parameter data subsets at different times, and each of the historical parameter data subsets includes a mutual inductor error at a corresponding time and data corresponding to a plurality of influencing parameters; A parameter screening module, used for screening multiple influencing parameters based on an extreme random tree algorithm and the historical influencing parameter data set to obtain a target influencing parameter; and determining a target influencing parameter data set based on the target influencing parameter and the historical influencing parameter data set; wherein the target influencing parameter data set includes target parameter data subsets at different times, each of the target parameter data subsets including a mutual inductor error at a corresponding time and data corresponding to the multiple target influencing parameters respectively; A model training module is used to construct an error prediction model; and train the error prediction model based on the target influencing parameter data set to obtain a trained error prediction model; An error prediction module is used to obtain a real-time influencing parameter data set of the wide-band mutual inductor; and determine the error prediction value of the wide-band mutual inductor based on the real-time influencing parameter data set and the trained error prediction model; wherein the real-time influencing parameter data set includes real-time data corresponding to multiple target influencing parameters.

[0013] In some possible embodiments, the parameter screening module is specifically used to: Based on the extreme random tree algorithm, a plurality of influencing parameters and segmentation thresholds are randomly selected to construct nodes of a decision tree, so as to generate a plurality of decision trees with different structures and prediction capabilities; and each decision tree is trained based on the historical influencing parameter data set; Based on the weight calculation formula and the training results of each decision tree, the weight of each decision tree is determined respectively; and based on the error rate calculation formula and the training results of each decision tree, the error rate of each influencing parameter in each decision tree on the mutual inductor error prediction is determined respectively; Based on the weight of each decision tree, the error rate of each influencing parameter in each decision tree on the prediction of the mutual inductor error, and the importance score formula, respectively calculate the importance score of each influencing parameter; The target influencing parameters are determined based on the preset screening conditions and the importance scores of each influencing parameter.

[0014] In some possible embodiments, the weight calculation formula includes: ; in, Represented as the weight of the jth decision tree; It is represented by the accuracy of the jth decision tree; M is represented by the total number of decision trees; Expressed as the accuracy of the kth decision tree; The error rate calculation formula includes: ; in, It is expressed as the error rate corresponding to the i-th influencing parameter of the j-th decision tree; Represented as the prediction result of the decision tree; Represented as the true label; Expressed as an indicator function, when When it is 1, otherwise it is 0; It is represented as the data subset obtained by splitting the i-th influencing parameter in the j-th decision tree; The importance score formula includes: ; in, It is represented as the importance score of the i-th influencing parameter; n is represented as the total amount of data corresponding to the i-th influencing parameter.

[0015] In some possible embodiments, the model training module is specifically used to: Obtain an error prediction model to be trained; wherein the error prediction model to be trained includes a CNN module, a BiLSTM module, an AM module and an error prediction module; Constructing a model training set based on the target influencing parameter data set; wherein the model training set includes a training data set, a verification data set and a sample data set; Based on the training data set, the verification data set and the dung beetle optimization algorithm, the error prediction model to be trained is first trained to obtain optimal hyperparameters of the error prediction model to be trained; The error prediction model to be trained is parameterized according to the optimal hyperparameters to obtain a target error prediction model to be trained; and the target error prediction model to be trained is subjected to a second training based on the training data set and the sample data set to obtain the trained error prediction model.

[0016] In some possible embodiments, the model training module is specifically used to: Step 1: Set the number of model iterations and multiple dung beetles, and randomly generate a set of hyperparameters for each dung beetle; Step 2: configuring the error prediction model to be trained based on the hyperparameters corresponding to each dung beetle to obtain the error prediction model to be trained corresponding to each dung beetle; Step 3: for the error prediction model to be trained corresponding to each dung beetle, the error prediction model to be trained is trained based on the training data set, and the trained error prediction model is verified based on the verification data set, and the prediction error corresponding to the dung beetle is calculated; Step 4: Calculate the fitness value of each dung beetle based on the fitness function and the prediction error corresponding to the dung beetle; and adjust the hyperparameters corresponding to each dung beetle based on the fitness value of each dung beetle and the update rule of the dung beetle optimization algorithm; Step 5: Return to step 2 above until the training result meets the preset requirements and / or the number of training times reaches the number of iterations, and obtain the optimal hyperparameters of the error prediction model.

[0017] In some possible embodiments, the fitness function is expressed as: ; in, It is expressed as the fitness value of the i-th dung beetle; It is represented as the prediction error of the error prediction model corresponding to the i-th dung beetle on the validation set; The update rule is expressed as: ; in, Represents the updated position of the i-th dung beetle in the next iteration; It is represented as the position of the i-th dung beetle in the t-th iteration; Indicates the position of the individual with the best fitness among all dung beetles; Represented as random positions; and Expressed as a step size coefficient, it controls the movement amplitude of the dung beetle.

[0018] In some possible embodiments, the model training module is specifically used to perform: Step a: Based on the CNN module, local features are extracted for each training data in the training data set to obtain local feature vectors corresponding to each training data; based on the BiLSTM module, the temporal dependency between the local feature vectors corresponding to each training data is captured to obtain a temporal feature vector corresponding to each training data; and, for each training data, the local feature vector and the temporal feature vector corresponding to the training data are subjected to attention weighting processing based on the AM module to obtain a fused feature vector corresponding to the training data; and, based on the error prediction module and the fused feature vector, a predicted mutual inductor error corresponding to the training data is determined; Step b: determining a loss value between a predicted mutual inductor error corresponding to the training data and a mutual inductor error corresponding to the training data in the sample data set based on a preset loss function, and adjusting a model parameter of the target error prediction model to be trained based on the loss value; Repeat steps a to b above until the training results meet the preset requirements to obtain the trained error prediction model.

[0019] An embodiment of the present disclosure provides a computer device, including: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the computer device is running, the processor communicates with the memory via the bus, and when the machine-readable instructions are executed by the processor, a wide-band mutual inductor error prediction method as described in any possible implementation manner described above is performed.

[0020] An embodiment of the present disclosure provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the wide-band mutual inductor error prediction method as described in any possible implementation manner described above is implemented.

[0021] The wide-band transformer error prediction method, device, equipment and medium provided in the embodiments of the present disclosure first obtain a historical influencing parameter data set of the wide-band transformer; next, based on the extreme random tree algorithm and the historical influencing parameter data set, multiple influencing parameters are screened to determine the target influencing parameters that have a more significant impact on the transformer error; on this basis, based on these target influencing parameters and the original historical influencing parameter data set, the target influencing parameter data set is further refined; then, a model for error prediction is constructed, and the model is trained using the refined target influencing parameter data set to finally obtain a trained error prediction model; finally, the real-time influencing parameter data set of the wide-band transformer is obtained, and the error prediction value of the wide-band transformer is determined based on the real-time influencing parameter data set and the trained error prediction model.

[0022] In this way, the present disclosure can identify the key factors that have a greater impact on the wideband transformer error by using the extreme random tree algorithm to screen and refine the influencing parameters, thereby effectively reducing the complexity of the error prediction model and improving the model's training efficiency and prediction accuracy; at the same time, the present disclosure can achieve dynamic prediction of the transformer error by using the error prediction model to monitor and analyze the influencing parameter data of the transformer in real time, significantly improving the stability and reliability of the power system. In this way, the present disclosure can effectively reduce power system failures caused by inaccurate error prediction, optimize power dispatching, improve the operating efficiency of equipment, and provide strong support for the efficient operation and maintenance of smart grids.

[0023] In order to make the above-mentioned objectives, features and advantages of the present disclosure more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following briefly introduces the drawings required to be cited in the embodiments. The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and are used together with the specification to illustrate the technical solutions of the present disclosure. It should be understood that the following drawings only illustrate certain embodiments of the present disclosure and should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 A flow chart of a wide-band mutual inductor error prediction method provided by an embodiment of the present disclosure is shown; Figure 2 A flow chart of an influencing parameter screening method provided by an embodiment of the present disclosure is shown; Figure 3 A flowchart of an error prediction model training method provided by an embodiment of the present disclosure is shown; Figure 4 A flowchart of a first training method for an error prediction model provided by an embodiment of the present disclosure is shown; Figure 5 A flowchart of a second training method for an error prediction model provided by an embodiment of the present disclosure is shown; Figure 6 A schematic structural diagram of a wide-band mutual inductor error prediction device provided by an embodiment of the present disclosure is shown; Figure 7 A schematic diagram of the structure of a computer device provided by an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical scheme and advantages of the embodiments of the present disclosure clearer, the technical scheme in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. The components of the embodiments of the present disclosure generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the drawings is not intended to limit the scope of the present disclosure for protection, but merely represents the selected embodiments of the present disclosure. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present disclosure.

[0027] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0028] The term "and / or" herein only describes an association relationship, indicating that three relationships may exist. For example, A and / or B may represent three situations: A exists alone, A and B exist at the same time, and B exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C may represent including any one or more elements selected from the set consisting of A, B, and C.

[0029] To facilitate understanding of this embodiment, the execution subject of the wideband mutual inductor error prediction method provided by the embodiment of the present disclosure is first introduced in detail. The execution subject of the wideband mutual inductor error prediction method provided by the embodiment of the present disclosure is a computer device. The computer device can be a server. Among them, the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud storage, big data and artificial intelligence platforms.

[0030] The wideband mutual inductor error prediction method provided by the embodiment of the present application is described in detail below in conjunction with the accompanying drawings. Figure 1 FIG. 1 is a flowchart of a method for predicting a wideband mutual inductor error provided by an embodiment of the present disclosure, the method comprising the following S101 to S104: S101, obtaining a historical influencing parameter data set of a wide-band transformer.

[0031] It is understandable that wideband transformers are widely used in power systems as a key device that can measure the current or voltage conversion ratio within a wide frequency range. Their main functions include providing accurate current or voltage monitoring in metering, protection, and control. The historical influencing parameter data set contains multiple historical data that may have an impact on the transformer error over a period of time. These data not only include the transformer error of the transformer at each time node, but also cover the data corresponding to other influencing parameters related to it. Here, the transformer error refers to the measurement error caused by technical, operational or environmental factors in the sensor or measuring device. Influencing parameters may include temperature, humidity, vibration, load changes, etc. These influencing parameters usually have a potential impact on the accuracy of the equipment during long-term operation. For example, drastic changes in temperature may cause the expansion or contraction of the internal material of the transformer, resulting in errors and affecting its electrical performance; changes in humidity may cause poor electrical contact, thereby affecting the measurement accuracy and generating errors; load changes will directly change the working state of the transformer, resulting in increased errors.

[0032] S102, screening multiple influencing parameters based on the extreme random tree algorithm and the historical influencing parameter data set to obtain target influencing parameters; and determining the target influencing parameter data set based on the target influencing parameters and the historical influencing parameter data set.

[0033] Here, the extreme random tree algorithm is an ensemble learning method that improves the generalization ability of the model by generating a large number of decision trees and performing random training on these trees. Specifically, during the training process of the algorithm, each tree will randomly select specific features and data samples, thereby avoiding over-reliance on certain specific features and improving the accuracy and robustness of the model. In the present disclosure, the extreme random tree algorithm is mainly used to select target parameters that have a significant impact on the prediction of transformer errors from multiple influencing parameters.

[0034] Specifically, refer to Figure 2 As shown, a method for screening influencing parameters proposed in the present disclosure specifically includes the following S201-204: S201, randomly selecting multiple influencing parameters and segmentation thresholds based on the extreme random tree algorithm to construct nodes of a decision tree to generate multiple decision trees with different structures and prediction capabilities; and training each decision tree based on the historical influencing parameter data set.

[0035] It is understandable that the extreme random tree algorithm, as an integrated learning method, has significant randomness and efficiency compared with the traditional decision tree algorithm. This method can construct the nodes of the decision tree by randomly selecting multiple influencing parameters (features) and segmentation thresholds, thereby generating multiple decision trees with different structures and predictive capabilities. The generation process of these decision trees not only reduces the variance of the model, but also enhances the robustness of the model.

[0036] Specifically, when constructing each decision tree using the extreme random tree algorithm, a part of the influencing parameters is randomly selected to divide the historical data corresponding to the influencing parameters, and in the division process of each node, the split point is randomly selected. In this way, the structure of each tree can be different, thereby forming multiple differentiated decision trees in ensemble learning. The differentiated structure of each tree can ensure that the model is not prone to overfitting when facing diversified data, thereby improving the accuracy and generalization ability of the prediction. At the same time, the extreme random tree algorithm not only takes into account the characteristic information of the parameters, but also helps the model learn the complex nonlinear relationship between the parameters and the error through the splitting process of the tree structure.

[0037] During the training process, these decision trees are trained based on the historical influencing parameter data set, which enables each decision tree to learn the nonlinear relationship between each influencing parameter and the error. That is, the training process of the extreme random tree can enable each decision tree to capture the contribution of different influencing parameters to the transformer error.

[0038] S202, based on the weight calculation formula and the training results of each decision tree, respectively determine the weight of each decision tree; and, based on the error rate calculation formula and the training results of each decision tree, respectively determine the error rate of each influencing parameter in each decision tree on the mutual inductor error prediction.

[0039] Exemplarily, in order to ensure the predictive performance of the model and the contribution of each decision tree to the final result, the influence of each decision tree can be quantified by the weight calculation formula and the error rate calculation formula. Here, the weight calculation formula determines its importance in the entire ensemble learning model by evaluating the accuracy of each decision tree. Specifically, by calculating the accuracy of each decision tree, its predictive ability can be reflected. If a decision tree has a high accuracy, it means that it contributes more to the prediction result, so its weight should also be high; conversely, a decision tree with a lower accuracy should be assigned a smaller weight.

[0040] Here, the weight calculation formula can be expressed as: ; in, Represented as the weight of the jth decision tree; It is represented by the accuracy of the jth decision tree; M is represented by the total number of decision trees; It is expressed as the accuracy of the kth decision tree.

[0041] It can be understood that the error rate calculation formula is used to evaluate the prediction error of each decision tree, so as to further analyze the contribution of each influencing parameter to the error in each decision tree. In this way, it is not only possible to identify which influencing parameters have a greater contribution to the prediction of the transformer error, but also to reveal which parameters may be the main source of the error. Specifically, by calculating the error rate generated by each decision tree when using a specific influencing parameter, it is possible to understand the degree of influence of different parameters on the prediction error. Among them, the influencing parameter corresponding to a high error rate may not provide effective information, so it can be considered that its contribution to the error prediction is small; while the influencing parameter with a low error rate indicates that it plays a more important role in the prediction.

[0042] Here, the error rate calculation formula can be expressed as: ; in, It is expressed as the error rate corresponding to the i-th influencing parameter of the j-th decision tree; Represented as the prediction result of the decision tree; Represented as the true label; Expressed as an indicator function, when Take 1 when it is, otherwise take 0; It is represented as the data subset obtained by splitting the i-th influencing parameter in the j-th decision tree.

[0043] S203, respectively calculating the importance score of each influencing parameter based on the weight of each decision tree, the error rate of each influencing parameter in each decision tree on the prediction of the mutual inductor error, and the importance score formula.

[0044] It can be understood that the weight and error rate of each decision tree reflect the contribution of the tree to the entire prediction process and the performance of each influencing parameter in the tree. Based on the weight of each decision tree and the error rate of each influencing parameter in each tree for the prediction of the transformer error, the importance score of each influencing parameter can be calculated using the importance score formula to quantify the role of each influencing parameter in the prediction of the transformer error. The higher the score, the greater the contribution of the feature to the model prediction.

[0045] Here, the importance score formula can be expressed as: ; in, It is represented as the importance score of the i-th influencing parameter; n is represented as the total amount of data corresponding to the i-th influencing parameter.

[0046] S204: Determine the target influencing parameter based on the preset screening conditions and the importance score of each influencing parameter.

[0047] Specifically, after obtaining the importance scores of each influencing parameter, the target influencing parameter with a greater impact on the error parameter can be further screened out according to the preset screening conditions. Here, the preset screening conditions can be to sort them and select the influencing parameters with the top importance scores as the target influencing parameters; or to set a score threshold and retain the influencing parameters with importance scores higher than the score threshold as the target influencing parameters; no specific limitation is made here.

[0048] In the disclosed embodiment, decision tree modeling and training are performed on the historical influencing parameter data set through the extreme random tree algorithm, and the importance score of each influencing parameter is calculated in combination with weight and error rate analysis. The target influencing parameter is screened out according to preset conditions, which can effectively identify the parameters that have important contributions to the mutual inductor error prediction, thereby providing more accurate feature selection for subsequent error prediction model training, avoiding interference from too many irrelevant factors, and helping to improve the accuracy and robustness of the prediction model.

[0049] It is understandable that after the target influencing parameters are screened out, the target influencing parameter data set can be constructed by selecting data corresponding to each target influencing parameter from the historical influencing parameter data set. The target influencing parameter data set includes target parameter data subsets at different times, and each target parameter data subset includes a mutual inductor error and data corresponding to multiple target influencing parameters.

[0050] S103, constructing an error prediction model; and training the error prediction model based on the target influencing parameter data set to obtain a trained error prediction model.

[0051] In order to predict the mutual inductor error, the present disclosure constructs an error prediction model and trains it through a target influencing parameter data set to obtain a trained error prediction model.

[0052] For example, refer to Figure 3 As shown, the above-trained error prediction model can be obtained through training in steps S301 to S304: S301, obtaining an error prediction model to be trained.

[0053] Among them, the error prediction model to be trained includes a CNN module, a BiLSTM module, an AM module and an error prediction module. Specifically, the convolutional neural network (CNN) is a deep learning model currently widely used in image processing and sequence data analysis. In the error prediction model proposed in the present disclosure, the CNN module can effectively extract local patterns or features in the input data, especially when processing data with structured patterns, it can better identify complex local features. BiLSTM is an extension of the long short-term memory network (LSTM), which can simultaneously consider the past and future information of the data, and is particularly suitable for processing time series data. In the error prediction model, the use of BiLSTM can help the model better capture the long-range dependencies in the time series data, while considering the forward and reverse trends of the data, thereby improving the ability to predict error changes. At the same time, the feature fusion module is replaced with the attention mechanism (AM) module. The AM module dynamically adjusts the weights of different features so that the model can focus more on key features and improve the pertinence and effectiveness of feature representation.

[0054] S302: construct a model training set based on the target influencing parameter data set.

[0055] Specifically, based on the target influencing parameter data set, a model training set can be further constructed. The training set is usually divided into three parts: training data set, validation data set and sample data set. The training data set is used in the learning process of the model, that is, the model updates its own parameters by learning the characteristics and rules in this part of the data; the validation data set is used to evaluate the performance of the model during the training process, and help adjust the model parameters and hyperparameters to avoid overfitting; the sample data set is used to evaluate the generalization ability and performance of the model after the model training process.

[0056] S303: Perform a first training on the error prediction model to be trained based on the training data set, the verification data set and the dung beetle optimization algorithm to obtain optimal hyperparameters of the error prediction model to be trained.

[0057] Here, the dung beetle optimization algorithm is an optimization algorithm that simulates the behavior of dung beetles in nature. It searches for the optimal solution to the problem by simulating the process of dung beetles rolling dung balls to find the best position. In the present disclosure, the dung beetle optimization algorithm is applied to the hyperparameter optimization of the model, which helps the model find the optimal hyperparameters by automatically adjusting the hyperparameters during the training process. Hyperparameters refer to those parameters that need to be manually set before model training, such as learning rate, regularization coefficient, etc., which usually have an important impact on the performance of the model.

[0058] For example, refer to Figure 4As shown, when the first training is performed on the error prediction model to be trained, the following steps S3031~S3035 may be included: S3031, set the number of model iterations and multiple dung beetles, and randomly generate a set of hyperparameters for each dung beetle.

[0059] Here, the number of iterations refers to the number of times the dung beetle performs searches and updates in the dung beetle optimization algorithm. Usually, each iteration represents the algorithm's exploration and optimization process of the solution space. In each iteration, the dung beetle decides whether to update its position by evaluating the performance of the current hyperparameter combination, thereby gradually approaching the optimal solution. The setting of the number of iterations can affect the search accuracy and convergence speed of the algorithm. Dung beetles represent individuals in the dung beetle optimization algorithm. Each dung beetle will explore the hyperparameter space and find the optimal hyperparameter combination. In order to avoid the limitations of hyperparameter search, a set of hyperparameters can be randomly generated for each dung beetle, which will become the starting values ​​in subsequent training. Through random initialization, each dung beetle can search in different areas, increasing the possibility of the algorithm jumping out of the local optimal solution and improving the chance of eventually finding the global optimal solution.

[0060] S3032: configuring the error prediction model to be trained based on the hyperparameters corresponding to each dung beetle to obtain the error prediction model to be trained corresponding to each dung beetle.

[0061] Specifically, according to the hyperparameters of each dung beetle, the structure and learning rules of the error prediction model are adjusted to conform to the individual settings of the current dung beetle (i.e., the hyperparameters corresponding to the dung beetle). In this way, it can be ensured that the model corresponding to each dung beetle can be trained according to its unique hyperparameters and generate an initial, potential model to be trained. Different dung beetles may have different effects on the model architecture and training methods.

[0062] S3033, for the error prediction model to be trained corresponding to each dung beetle, train the error prediction model to be trained based on the training data set, and verify the trained error prediction model based on the verification data set, and calculate the prediction error corresponding to the dung beetle.

[0063] It is understandable that after obtaining the error prediction model to be trained corresponding to each dung beetle, the training data set can be used to train the error prediction model corresponding to each dung beetle, aiming to let the model learn how to minimize the error prediction through data input. Then, the trained model is verified using the validation data set to ensure that the performance of the model on new data is not overfit. The validation data set is used to evaluate the performance of the model outside the training set, and its purpose is to test the generalization ability of the model. At this time, the prediction error corresponding to each dung beetle can be calculated to determine the impact of the hyperparameters configured for each dung beetle on the model performance. Among them, the method for calculating the prediction error can use the mean square error calculation or the cross entropy loss calculation method, etc., which is not specifically limited here.

[0064] S3034, respectively calculating the fitness value of each dung beetle based on the fitness function and the prediction error corresponding to the dung beetle; and adjusting the hyperparameters corresponding to each dung beetle based on the fitness value of each dung beetle and the update rule of the dung beetle optimization algorithm.

[0065] Here, the fitness function is used to evaluate the performance of each dung beetle. The smaller the prediction error, the larger the fitness value, indicating that the hyperparameter configuration of the dung beetle is better. According to the fitness value of each dung beetle, the hyperparameters (or positions) of each dung beetle can be adjusted according to the update rule of the dung beetle optimization algorithm to ensure that the optimization process tends to the optimal solution. Among them, the update rule of the optimization algorithm is to continuously adjust the hyperparameters by imitating the process of dung beetles looking for food to achieve the goal of reducing errors.

[0066] Specifically, the fitness function can be expressed as: ; in, It is expressed as the fitness value of the i-th dung beetle; It is represented as the prediction error of the error prediction model corresponding to the i-th dung beetle on the validation set.

[0067] Specifically, the update rule can be expressed as: ; in, Represents the updated position of the i-th dung beetle in the next iteration; It is represented as the position of the i-th dung beetle in the t-th iteration; Indicates the position of the individual with the best fitness among all dung beetles; Represented as random positions; and Expressed as a step size coefficient, it controls the movement amplitude of the dung beetle.

[0068] S3035, return to the above step S3032, until the training result meets the preset requirements and / or the number of training times reaches the number of iterations, and the optimal hyperparameters of the error prediction model are obtained.

[0069] Exemplarily, after the hyperparameter update and model training are completed, the process returns to step S3032 and continues iterative training until the preset performance requirements are met or the set maximum number of iterations is reached. Through multiple iterations, the dung beetle optimization algorithm can find a set of optimal hyperparameter combinations in the hyperparameter space, thereby improving the prediction accuracy of the error prediction model. In this way, after sufficient training and adjustment, the error prediction model obtained can have the optimal hyperparameter configuration and can effectively perform error prediction.

[0070] Here, the training results that meet the preset requirements can be expressed as the error value, prediction accuracy, the difference between the training error and the verification error, etc. reaching the set tolerance threshold; if the training results do not meet the preset requirements when the number of training times reaches the number of iterations, a better hyperparameter can be found from the historical hyperparameters and then output as the optimal hyperparameter. In this way, even if the preset goals cannot be fully met during the training process, a more superior hyperparameter configuration can still be selected based on previous training experience, so that the final output model still has a high prediction performance, reducing the waste of computing resources.

[0071] Exemplarily, in the process of optimizing the hyperparameters of a model using the dung beetle algorithm, the present disclosure introduces a global search factor to optimize the search strategy. By improving the global search strategy, the global search capability of the algorithm can be effectively enhanced and the impact of local optimal solutions can be reduced.

[0072] Specifically, in the global search phase, dung beetles will randomly select multiple targets and explore in the solution space in order to cover the entire solution space and find as many potential high-quality solutions as possible. The improved global search strategy includes: first, increasing the step size of dung beetle individuals so that they can jump out of the current local optimal area and avoid falling into the local optimal solution; second, introducing the global attraction factor (GAF), which is inversely proportional to the quality of the current search solution, to ensure that dung beetle individuals have a certain preference to explore potential high-quality areas during the search process. Here, the global search formula is as follows: ; ; in, Indicated as the new position of the dung beetle; represents the current position of the dung beetle; GAF represents the global attraction factor; Represented as the objective function value at the current position; Represents the location of the global optimal solution.

[0073] The improvement of global search mainly enhances the exploration ability of dung beetles by dynamically adjusting the step size, introducing global attraction factors, and mutation operations. These improvements can effectively prevent dung beetles from falling into local optimal solutions, expand the search range, and thus improve the global optimization ability of the algorithm. Through reasonable step size adjustment and global attraction mechanism, the improved global search can more efficiently find high-quality solutions in the solution space and accelerate the convergence process.

[0074] S304, configuring the parameters of the error prediction model to be trained according to the optimal hyperparameters to obtain a target error prediction model to be trained; and performing a second training on the target error prediction model to be trained based on the training data set and the sample data set to obtain the trained error prediction model.

[0075] It is understandable that after obtaining the optimal hyperparameters, they can be applied to the error prediction model to be trained for further parameter configuration. After the configuration is completed, the model can be trained in the second phase based on the training data set, and the performance of the model can be verified during the training process through the sample data set to evaluate whether the model has good prediction ability, thereby obtaining a trained error prediction model.

[0076] For example, refer to Figure 5 As shown, when the second training is performed on the target error prediction model to be trained, the following steps S3041-S3043 may be included: S3041, based on the CNN module, respectively extract local features of each training data in the training data set to obtain local feature vectors corresponding to each training data; based on the BiLSTM module, capture the temporal dependency between the local feature vectors corresponding to each training data to obtain a temporal feature vector corresponding to each training data; and, for each training data, based on the AM module, perform attention weighted processing on the local feature vector and the temporal feature vector corresponding to the training data to obtain a fused feature vector corresponding to the training data; and, based on the error prediction module and the fused feature vector, determine the predicted mutual inductor error corresponding to the training data.

[0077] Here, the CNN module is used to extract local features of each data in the training data to obtain the local feature vector corresponding to each training data; then, the BiLSTM module is used to capture the temporal dependency between the local feature vectors, and BiLSTM can extract richer temporal features from the forward and backward time series information; further, the AM module is used to perform weighted fusion of the local feature vector and the temporal feature vector; finally, the fused feature vector will be sent as input to the error prediction module to determine the predicted mutual inductor error corresponding to the training data.

[0078] S3042, determining a loss value between a predicted mutual inductor error corresponding to the training data and a mutual inductor error corresponding to the training data in the sample data set based on a preset loss function, and adjusting model parameters of the target error prediction model to be trained based on the loss value.

[0079] Here, during the model training process, the difference between the predicted error and the actual transformer error (i.e., the loss value) is calculated through a preset loss function. The loss value is used to guide the adjustment of the model parameters so as to gradually optimize the prediction performance of the model. Among them, the loss function may include mean square error (MSE), cross entropy loss, absolute error (MAE), etc., which are not specifically limited here.

[0080] S3043, repeat the above steps S3041~S3042 until the training results meet the preset requirements, and obtain the trained error prediction model.

[0081] Specifically, the second training process of the entire model is repeated until the training result meets the preset requirements, and a trained error prediction model is obtained.

[0082] S104, obtaining a real-time influencing parameter data set of the wideband mutual inductor; and determining an error prediction value of the wideband mutual inductor based on the real-time influencing parameter data set and the trained error prediction model.

[0083] Among them, the real-time influencing parameter data set includes real-time data corresponding to multiple target influencing parameters. By obtaining the real-time influencing parameter data set of the wide-band transformer and inputting these data into the trained error prediction model, the error value of the transformer can be predicted in real time. The purpose is to effectively monitor the working status of the transformer and timely discover potential errors and fault risks, thereby providing reliable decision support for equipment maintenance and optimization.

[0084] The wideband transformer error prediction method, device, equipment and medium provided in the embodiments of the present disclosure can identify the key factors that have a greater impact on the wideband transformer error by screening and refining the influencing parameters using the extreme random tree algorithm, thereby effectively reducing the complexity of the error prediction model and improving the training efficiency and prediction accuracy of the model; at the same time, the present disclosure can realize the dynamic prediction of the transformer error by using the error prediction model to monitor and analyze the influencing parameter data of the transformer in real time, significantly improving the stability and reliability of the power system. In this way, the present disclosure can effectively reduce power system failures caused by inaccurate error prediction, optimize power dispatching, improve the operating efficiency of equipment, and provide strong support for the efficient operation and maintenance of smart grids.

[0085] Those skilled in the art will appreciate that, in the above method of specific implementation, the order in which the steps are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of the steps should be determined by their functions and possible internal logic.

[0086] Based on the same inventive concept, a wide-band transformer error prediction device corresponding to the wide-band transformer error prediction method is also provided in the embodiment of the present disclosure. Since the principle of solving the problem by the device in the embodiment of the present disclosure is similar to the above-mentioned wide-band transformer error prediction method in the embodiment of the present disclosure, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0087] Reference Figure 6 FIG. 6 is a schematic diagram of a wideband mutual inductor error prediction device 600 provided in an embodiment of the present disclosure, wherein the device comprises: The data acquisition module 601 is used to acquire a historical influencing parameter data set of a wideband transformer; wherein the historical influencing parameter data set includes historical parameter data subsets at different times, and each of the historical parameter data subsets includes a transformer error at a corresponding time and data corresponding to a plurality of influencing parameters; The parameter screening module 602 is used to screen multiple influencing parameters based on the extreme random tree algorithm and the historical influencing parameter data set to obtain the target influencing parameter; and determine the target influencing parameter data set based on the target influencing parameter and the historical influencing parameter data set; wherein the target influencing parameter data set includes target parameter data subsets at different times, and each of the target parameter data subsets includes the mutual inductor error at the corresponding time and data corresponding to the multiple target influencing parameters respectively; The model training module 603 is used to construct an error prediction model; and train the error prediction model based on the target influencing parameter data set to obtain a trained error prediction model; The error prediction module 604 is used to obtain a real-time influencing parameter data set of the wide-band mutual inductor; and determine the error prediction value of the wide-band mutual inductor based on the real-time influencing parameter data set and the trained error prediction model; wherein the real-time influencing parameter data set includes real-time data corresponding to multiple target influencing parameters.

[0088] In some possible embodiments, the parameter screening module 602 is specifically used to: Based on the extreme random tree algorithm, a plurality of influencing parameters and segmentation thresholds are randomly selected to construct nodes of a decision tree, so as to generate a plurality of decision trees with different structures and prediction capabilities; and each decision tree is trained based on the historical influencing parameter data set; Based on the weight calculation formula and the training results of each decision tree, the weight of each decision tree is determined respectively; and based on the error rate calculation formula and the training results of each decision tree, the error rate of each influencing parameter in each decision tree on the mutual inductor error prediction is determined respectively; Based on the weight of each decision tree, the error rate of each influencing parameter in each decision tree on the prediction of the mutual inductor error, and the importance score formula, respectively calculate the importance score of each influencing parameter; The target influencing parameters are determined based on the preset screening conditions and the importance scores of each influencing parameter.

[0089] In some possible embodiments, the weight calculation formula includes: ; in, Represented as the weight of the jth decision tree; It is represented by the accuracy of the jth decision tree; M is represented by the total number of decision trees; Expressed as the accuracy of the kth decision tree; The error rate calculation formula includes: ; in, It is expressed as the error rate corresponding to the i-th influencing parameter of the j-th decision tree; Represented as the prediction result of the decision tree; Represented as the true label; Expressed as an indicator function, when When it is 1, otherwise it is 0; It is represented as the data subset obtained by splitting the i-th influencing parameter in the j-th decision tree; The importance score formula includes: ; in, It is represented as the importance score of the i-th influencing parameter; n is represented as the total amount of data corresponding to the i-th influencing parameter.

[0090] In some possible embodiments, the model training module 603 is specifically used to: Obtain an error prediction model to be trained; wherein the error prediction model to be trained includes a CNN module, a BiLSTM module, an AM module and an error prediction module; Constructing a model training set based on the target influencing parameter data set; wherein the model training set includes a training data set, a verification data set and a sample data set; Based on the training data set, the verification data set and the dung beetle optimization algorithm, the error prediction model to be trained is first trained to obtain optimal hyperparameters of the error prediction model to be trained; The error prediction model to be trained is parameterized according to the optimal hyperparameters to obtain a target error prediction model to be trained; and the target error prediction model to be trained is subjected to a second training based on the training data set and the sample data set to obtain the trained error prediction model.

[0091] In some possible embodiments, the model training module 603 is specifically used to: Step 1: Set the number of model iterations and multiple dung beetles, and randomly generate a set of hyperparameters for each dung beetle; Step 2: configuring the error prediction model to be trained based on the hyperparameters corresponding to each dung beetle to obtain the error prediction model to be trained corresponding to each dung beetle; Step 3: for the error prediction model to be trained corresponding to each dung beetle, the error prediction model to be trained is trained based on the training data set, and the trained error prediction model is verified based on the verification data set, and the prediction error corresponding to the dung beetle is calculated; Step 4: Calculate the fitness value of each dung beetle based on the fitness function and the prediction error corresponding to the dung beetle; and adjust the hyperparameters corresponding to each dung beetle based on the fitness value of each dung beetle and the update rule of the dung beetle optimization algorithm; Step 5: Return to step 2 above until the training result meets the preset requirements and / or the number of training times reaches the number of iterations, and obtain the optimal hyperparameters of the error prediction model.

[0092] In some possible embodiments, the fitness function is expressed as: ; in, It is expressed as the fitness value of the i-th dung beetle; It is represented as the prediction error of the error prediction model corresponding to the i-th dung beetle on the validation set; The update rule is expressed as: ; in, Represents the updated position of the i-th dung beetle in the next iteration; It is represented as the position of the i-th dung beetle in the t-th iteration; Indicates the position of the individual with the best fitness among all dung beetles; Represented as random positions; and Expressed as a step size coefficient, it controls the movement amplitude of the dung beetle.

[0093] In some possible embodiments, the model training module 603 is specifically used to perform: Step a: Based on the CNN module, local features are extracted for each training data in the training data set to obtain local feature vectors corresponding to each training data; based on the BiLSTM module, the temporal dependency between the local feature vectors corresponding to each training data is captured to obtain a temporal feature vector corresponding to each training data; and, for each training data, the local feature vector and the temporal feature vector corresponding to the training data are subjected to attention weighting processing based on the AM module to obtain a fused feature vector corresponding to the training data; and, based on the error prediction module and the fused feature vector, a predicted mutual inductor error corresponding to the training data is determined; Step b: determining a loss value between a predicted mutual inductor error corresponding to the training data and a mutual inductor error corresponding to the training data in the sample data set based on a preset loss function, and adjusting a model parameter of the target error prediction model to be trained based on the loss value; Repeat steps a to b above until the training results meet the preset requirements to obtain the trained error prediction model.

[0094] Based on the same technical concept, the embodiment of the present disclosure also provides a computer device. Figure 7 , which is a schematic diagram of the structure of a computer device 700 provided in an embodiment of the present disclosure, including a processor 701, a memory 702, and a bus 703. The memory 702 is used to store execution instructions, including a memory 7021 and an external memory 7022; the memory 7021 is also called an internal memory, which is used to temporarily store the operation data in the processor 701 and the data exchanged with the external memory 7022 such as a hard disk. The processor 701 exchanges data with the external memory 7022 through the memory 7021.

[0095] In the embodiment of the present application, the memory 702 is specifically used to store the application code for executing the solution of the present application, and the execution is controlled by the processor 701. That is, when the computer device 700 is running, the processor 701 communicates with the memory 702 through the bus 703, so that the processor 701 executes the application code stored in the memory 702, and then executes the wide-band mutual inductor error prediction method described in any of the aforementioned embodiments.

[0096] Among them, the memory 702 can be, but is not limited to, random access memory (Random Access Memory, RAM), read only memory (Read Only Memory, ROM), programmable read-only memory (Programmable Read-Only Memory, PROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, EPROM), electrically erasable read-only memory (Electric Erasable Programmable Read-Only Memory, EEPROM), etc.

[0097] Processor 701 may be an integrated circuit chip with signal processing capabilities. The above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The disclosed methods, steps and logic block diagrams in the embodiments of the present invention may be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0098] It is to be understood that the structure illustrated in the embodiment of the present application does not constitute a specific limitation on the computer device 700. In other embodiments of the present application, the computer device 700 may include more or fewer components than shown in the figure, or combine some components, or split some components, or arrange the components differently. The components shown in the figure may be implemented in hardware, software, or a combination of software and hardware.

[0099] The embodiment of the present disclosure also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the wide-band mutual inductor error prediction method described in the above method embodiment are executed. The storage medium can be a volatile or non-volatile computer-readable storage medium.

[0100] The embodiments of the present disclosure also provide a computer program product, which carries a program code. The instructions included in the program code can be used to execute the steps of the wide-band mutual inductor error prediction method described in the above method embodiment. For details, please refer to the above method embodiment, which will not be repeated here.

[0101] The computer program product may be implemented in hardware, software or a combination thereof. In one optional embodiment, the computer program product is implemented as a computer storage medium. In another optional embodiment, the computer program product is implemented as a software product, such as a software development kit (SDK).

[0102] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the several embodiments provided in the present disclosure, it should be understood that the disclosed system and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0103] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0104] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0105] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0106] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present disclosure, which are used to illustrate the technical solutions of the present disclosure, rather than to limit them. The protection scope of the present disclosure is not limited thereto. Although the present disclosure is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed in the present disclosure, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A method for predicting a wideband mutual inductor error, characterized in that: include: Acquire a historical influencing parameter data set of a wideband mutual inductor; wherein the historical influencing parameter data set includes historical parameter data subsets at different times, and each of the historical parameter data subsets includes a mutual inductor error at a corresponding time and data corresponding to a plurality of influencing parameters; Based on the extreme random tree algorithm and the historical influencing parameter data set, multiple influencing parameters are screened to obtain target influencing parameters; and based on the target influencing parameters and the historical influencing parameter data set, a target influencing parameter data set is determined; wherein the target influencing parameter data set includes target parameter data subsets at different times, and each target parameter data subset includes a mutual inductor error at a corresponding time and data corresponding to multiple target influencing parameters respectively; Constructing an error prediction model; and training the error prediction model based on the target influencing parameter data set to obtain a trained error prediction model; Acquire a real-time influencing parameter data set of the wideband mutual inductor; and determine an error prediction value of the wideband mutual inductor based on the real-time influencing parameter data set and the trained error prediction model; wherein the real-time influencing parameter data set includes real-time data corresponding to multiple target influencing parameters.

2. The method according to claim 1, characterized in that The screening of multiple influencing parameters based on the extreme random tree algorithm and the historical influencing parameter data set includes: Based on the extreme random tree algorithm, a plurality of influencing parameters and segmentation thresholds are randomly selected to construct nodes of a decision tree, so as to generate a plurality of decision trees with different structures and prediction capabilities; and each decision tree is trained based on the historical influencing parameter data set; Based on the weight calculation formula and the training results of each decision tree, the weight of each decision tree is determined respectively; and based on the error rate calculation formula and the training results of each decision tree, the error rate of each influencing parameter in each decision tree on the mutual inductor error prediction is determined respectively; Based on the weight of each decision tree, the error rate of each influencing parameter in each decision tree on the prediction of the mutual inductor error, and the importance score formula, respectively calculate the importance score of each influencing parameter; The target influencing parameters are determined based on the preset screening conditions and the importance scores of each influencing parameter.

3. The method according to claim 2, characterized in that The weight calculation formula includes: ； in, Represented as the weight of the jth decision tree; It is represented by the accuracy of the jth decision tree; M is represented by the total number of decision trees; Expressed as the accuracy of the kth decision tree; The error rate calculation formula includes: ； in, It is expressed as the error rate corresponding to the i-th influencing parameter of the j-th decision tree; Represented as the prediction result of the decision tree; Represented as the true label; Expressed as an indicator function, when Take 1 when it is, otherwise take 0; It is represented as the data subset obtained by splitting the i-th influencing parameter in the j-th decision tree; The importance score formula includes: ； in, It is represented as the importance score of the i-th influencing parameter; n is represented as the total amount of data corresponding to the i-th influencing parameter.

4. The method according to claim 1, characterized in that: The training of the error prediction model based on the target influencing parameter data set includes: Obtain an error prediction model to be trained; wherein the error prediction model to be trained includes a CNN module, a BiLSTM module, an AM module and an error prediction module; Constructing a model training set based on the target influencing parameter data set; wherein the model training set includes a training data set, a verification data set and a sample data set; Based on the training data set, the verification data set and the dung beetle optimization algorithm, the error prediction model to be trained is first trained to obtain optimal hyperparameters of the error prediction model to be trained; The error prediction model to be trained is parameterized according to the optimal hyperparameters to obtain a target error prediction model to be trained; and the target error prediction model to be trained is subjected to a second training based on the training data set and the sample data set to obtain the trained error prediction model.

5. The method according to claim 4, characterized in that The first training of the error prediction model to be trained based on the training data set, the verification data set and the dung beetle optimization algorithm includes: Step 1: Set the number of model iterations and multiple dung beetles, and randomly generate a set of hyperparameters for each dung beetle; Step 2: configuring the error prediction model to be trained based on the hyperparameters corresponding to each dung beetle to obtain the error prediction model to be trained corresponding to each dung beetle; Step 3: for the error prediction model to be trained corresponding to each dung beetle, the error prediction model to be trained is trained based on the training data set, and the trained error prediction model is verified based on the verification data set, and the prediction error corresponding to the dung beetle is calculated; Step 4: Calculate the fitness value of each dung beetle based on the fitness function and the prediction error corresponding to the dung beetle; and adjust the hyperparameters corresponding to each dung beetle based on the fitness value of each dung beetle and the update rule of the dung beetle optimization algorithm; Step 5: Return to step 2 above until the training result meets the preset requirements and / or the number of training times reaches the number of iterations, and obtain the optimal hyperparameters of the error prediction model.

6. The method according to claim 5, characterized in that The fitness function is expressed as: ； in, It is represented as the fitness value of the i-th dung beetle; It is represented as the prediction error of the error prediction model corresponding to the i-th dung beetle on the validation set; The update rule is expressed as: ； in, Represents the updated position of the i-th dung beetle in the next iteration; It is represented as the position of the i-th dung beetle in the t-th iteration; Indicates the position of the individual with the best fitness among all dung beetles; Represented as random positions; and Expressed as a step size coefficient, it controls the movement amplitude of the dung beetle.

7. The method according to claim 4, characterized in that The performing a second training on the target error prediction model to be trained based on the training data set and the sample data set includes: Step a: Based on the CNN module, local features are extracted for each training data in the training data set to obtain local feature vectors corresponding to each training data; based on the BiLSTM module, the temporal dependency between the local feature vectors corresponding to each training data is captured to obtain a temporal feature vector corresponding to each training data; and, for each training data, the local feature vector and the temporal feature vector corresponding to the training data are subjected to attention weighting processing based on the AM module to obtain a fused feature vector corresponding to the training data; and, based on the error prediction module and the fused feature vector, a predicted mutual inductor error corresponding to the training data is determined; Step b: determining a loss value between a predicted mutual inductor error corresponding to the training data and a mutual inductor error corresponding to the training data in the sample data set based on a preset loss function, and adjusting a model parameter of the target error prediction model to be trained based on the loss value; Repeat steps a to b above until the training results meet the preset requirements to obtain the trained error prediction model.

8. A broadband mutual inductor error prediction device, characterized in that: include: A data acquisition module, used to acquire a historical influencing parameter data set of a wideband mutual inductor; wherein the historical influencing parameter data set includes historical parameter data subsets at different times, and each of the historical parameter data subsets includes a mutual inductor error at a corresponding time and data corresponding to a plurality of influencing parameters; A parameter screening module, used for screening multiple influencing parameters based on an extreme random tree algorithm and the historical influencing parameter data set to obtain a target influencing parameter; and determining a target influencing parameter data set based on the target influencing parameter and the historical influencing parameter data set; wherein the target influencing parameter data set includes target parameter data subsets at different times, each of the target parameter data subsets including a mutual inductor error at a corresponding time and data corresponding to the multiple target influencing parameters respectively; A model training module is used to construct an error prediction model; and train the error prediction model based on the target influencing parameter data set to obtain a trained error prediction model; An error prediction module is used to obtain a real-time influencing parameter data set of the wide-band mutual inductor; and determine the error prediction value of the wide-band mutual inductor based on the real-time influencing parameter data set and the trained error prediction model; wherein the real-time influencing parameter data set includes real-time data corresponding to multiple target influencing parameters.

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

10. A computer device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

Images