Substation misoperation risk assessment method and system based on adaptive learning

Abstract

The invention discloses a transformer substation misoperation risk assessment method and system based on adaptive learning, and relates to the technical field of transformer substations, and the method comprises the steps: collecting and preprocessing multi-dimensional feature data related to the electrical misoperation risk of a transformer substation; performing adaptive feature selection on the preprocessed feature data by using a TabNet algorithm, evaluating the importance of different features through an attention mechanism, and dynamically adjusting the weights of the features; performing deep learning modeling on the selected features by adopting a CatBoost algorithm, processing a nonlinear relationship between category features and numerical features, and establishing a risk level judgment model; and in combination with output results of the TabNet and CatBoost algorithms, dividing risk levels of electrical misoperation of the transformer substation, setting a risk level quantitative standard and a dynamically optimized risk level division criterion, and completing assessment of the electrical misoperation risk of the transformer substation. According to the method, the risk level judgment criterion can be adjusted in real time, and the problem that a traditional static evaluation method cannot adapt to switching operation dynamic changes is effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of substations, and in particular to a method and system for assessing the risk of substation misoperation based on adaptive learning. Background Art

[0002] In the field of substation electrical misoperation risk level assessment, the currently commonly used assessment methods have significant technical limitations. Traditional risk level determination methods mainly rely on single models such as fuzzy comprehensive evaluation or hierarchical analysis method. These methods cannot accurately describe the boundary characteristics between risk levels when quantifying and dividing risk levels, resulting in greater uncertainty in the risk level determination results. In particular, when determining the risk level division criteria, existing technologies rely too much on expert experience for subjective judgment and lack data-driven scientific quantitative basis.

[0003] In practical applications, substations have accumulated a large amount of risk level assessment data and classification judgment cases, which contain rich risk level characteristics and distribution patterns. However, the existing risk level assessment model lacks the ability to deeply learn and mine historical data, and cannot automatically extract the distinguishing features of different risk levels from massive data. At the same time, the risk level determination of substations is a dynamically evolving process, and the existing assessment method uses a static risk level classification standard, which cannot dynamically adjust the risk level determination criteria according to the actual operating conditions, resulting in a deviation between the risk level determination results and the actual risk status.

[0004] In addition, the current risk level assessment methods often use a unified quantitative standard and a fixed level threshold when dividing risk levels. This rigid level division mechanism cannot adapt to the differences in risk characteristics of different substations and under different operating conditions. This makes the risk level determination results lack pertinence and adaptability, and it is difficult to meet the actual needs of substation risk classification management and control. Summary of the invention

[0005] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0006] In view of the above existing problems, the present invention is proposed.

[0007] Therefore, the present invention provides a method and system for assessing the risk of substation misoperation based on adaptive learning, which can solve the problems mentioned in the background technology.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0009] In a first aspect, the present invention provides a method for assessing the risk of misoperation of a substation based on adaptive learning, which includes collecting and preprocessing multi-dimensional feature data related to the risk of electrical misoperation of a substation;

[0010] The TabNet algorithm is used to perform adaptive feature selection on the preprocessed feature data, evaluate the importance of different features through the attention mechanism, and dynamically adjust the weight of the features;

[0011] The CatBoost algorithm is used to perform deep learning modeling on the selected features, process the nonlinear relationship between categorical and numerical features, and establish a risk level determination model;

[0012] Combining the output results of TabNet and CatBoost algorithms, the risk levels of electrical misoperation in substations are divided, and the risk level quantification standard and dynamically optimized risk level division criteria are set to complete the assessment of the risk of electrical misoperation in substations.

[0013] As a preferred solution of the substation misoperation risk assessment method based on adaptive learning according to the present invention, the preprocessing includes:

[0014] The target statistical coding method is used to convert the categorical feature data into numerical features;

[0015] The multi-dimensional feature data includes data of personnel, equipment, environment, management and system dimensions;

[0016] The collected numerical characteristic data are standardized, the mean and standard deviation of the substation characteristic data are calculated, and the original characteristic values ​​are converted into standardized characteristic values. The calculation formula is as follows:

[0017] X norm =(X raw -μ sta ) / σ sta

[0018] Among them, X norm is the standardized characteristic value of the substation, X raw is the original characteristic value of the substation, μ sta is the average value of substation characteristics, σ sta is the standard deviation of substation characteristics;

[0019] The collected categorical feature data are processed using the target coding method. By calculating the probability distribution of different risk levels under each category, the categorical features are converted into numerical features with predictive value. The target coding formula for categorical features is as follows:

[0020] Risk cat=(N risk +α smooth ) / (N total +α smooth ×K level )

[0021] Among them, Risk ca is the risk probability value of the feature category, N risk is the number of risk events under this category, N total is the total number of operations in this category, α smooth is the smoothing coefficient, K level is the number of risk levels.

[0022] As a preferred solution of the substation misoperation risk assessment method based on adaptive learning according to the present invention, the adaptive feature selection includes:

[0023] Based on the feature selector parameters at the current moment and the hidden state at the previous moment, a nonlinear transformation is performed through the Sparsemax function to generate a feature selection mask. The calculation formula for feature mask generation is as follows:

[0024] Mask feat =Sparsemax(W sel ×State prev )

[0025] Among them, Mask feat is the feature selection mask, W sel is the feature selection weight matrix, State prev is the previous state information;

[0026] Apply the feature transformation network to transform the original features into more expressive feature representations. The feature transformation formula is as follows:

[0027] Trans feat =Mask feat ×Net feat (X input )

[0028] Among them, Trans feat is the transformed feature, X input is the input feature, Net feat It is the feature transformation network.

[0029] As a preferred solution of the substation misoperation risk assessment method based on adaptive learning according to the present invention, the deep learning modeling includes:

[0030] In the CatBoost model construction stage, an objective function for risk level determination is designed. The objective function includes a prediction error term and a regularization term. The risk prediction objective function is as follows:

[0031] Loss total =Σ(Risk true -Risk pred )2+λ reg ×Σ||W feat ||2

[0032] Among them, Loss total is the total loss value, Risk true is the actual risk level, Risk pred To predict the risk level, λ reg is the regularization parameter, W feat is the feature weight;

[0033] The CatBoost algorithm's sequence feature encoding and gradient boosting mechanism are used to perform deep learning and modeling on the features selected and transformed by TabNet, process the nonlinear relationship between categorical and numerical features, and construct a risk level determination model.

[0034] As a preferred solution of the method for assessing the risk of substation misoperation based on adaptive learning according to the present invention, the risk level of substation electrical misoperation is divided into:

[0035] The feature importance scores of the TabNet and CatBoost models are fused, and the final feature weights are calculated based on the feature importance weighting mechanism. The feature importance fusion formula is as follows:

[0036] Weight final =β mix ×Weight tabnet +(1-β mix )×Weight boost

[0037] Among them, Weight final is the final feature weight, β mix is the fusion coefficient, Weight tabnet is the TabNet feature weight, Weight boost is the CatBoost feature weight.

[0038] As a preferred solution of the method for assessing the risk of substation misoperation based on adaptive learning according to the present invention, the risk level of substation electrical misoperation is divided into:

[0039] Calculate the final risk score based on the fused feature weights;

[0040] Adopting the adaptive threshold method, and determining the risk level threshold based on the statistical characteristics of historical data, the risk level threshold calculation formula is as follows:

[0041] Thresh level =μ risk +k adj ×σ risk

[0042] Among them, Thresh level is the risk level threshold, μ risk is the average risk score, k adj is the adjustment coefficient, σ risk is the standard deviation of the risk score.

[0043] As a preferred solution of the method for assessing the risk of electrical misoperation of a substation based on adaptive learning according to the present invention, the assessment of the risk of electrical misoperation of a substation includes:

[0044] Based on the calculated risk score and the determined risk level threshold, the risk level of electrical misoperation of the substation is divided into different levels to complete the risk assessment.

[0045] In a second aspect, the present invention provides a substation misoperation risk assessment system based on adaptive learning, which includes: a data acquisition module, a feature selection module, a deep learning module and a risk assessment module;

[0046] The data acquisition module is used to collect and preprocess multi-dimensional feature data related to the risk of electrical misoperation of the substation;

[0047] The feature selection module is used to perform adaptive feature selection on the preprocessed feature data using the TabNet algorithm, evaluate the importance of different features through the attention mechanism, and dynamically adjust the weight of the features;

[0048] The deep learning module is used to use the CatBoost algorithm to perform deep learning modeling on the selected features, process the nonlinear relationship between categorical and numerical features, and establish a risk level determination model;

[0049] The risk assessment module is used to combine the output results of TabNet and CatBoost algorithms, divide the risk level of substation electrical misoperation, set risk level quantification standards and dynamically optimized risk level division criteria, and complete the assessment of substation electrical misoperation risks.

[0050] In a third aspect, the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, the steps of a substation misoperation risk assessment method based on adaptive learning are implemented.

[0051] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, the steps of a substation misoperation risk assessment method based on adaptive learning are implemented.

[0052] Compared with the prior art, the beneficial effect of the present invention is that the scientificity and accuracy of risk level determination are significantly improved by innovatively combining TabNet and CatBoost algorithms. Compared with traditional single evaluation models such as fuzzy comprehensive evaluation and hierarchical analysis method, this method breaks through the limitation of over-reliance on expert experience and realizes an important transformation from qualitative evaluation to data-driven quantitative evaluation.

[0053] In terms of risk level assessment, this method realizes the accurate classification of risk levels of substation switching operations through the adaptive selection mechanism of TabNet; the system can automatically learn the boundary features between different risk levels and establish scientific risk level quantitative standards; at the same time, the deep learning ability of the CatBoost algorithm can effectively handle complex scenarios in switching operations, making the determination of each risk level more accurate; this innovative design of dual-path feature fusion not only improves the model's risk identification ability for different switching operation scenarios, but also enhances the reliability of risk level determination results; in practical applications, the adaptive risk level assessment system established by this method can dynamically optimize the risk level classification standard according to the substation switching operation data. By designing the feature importance weighting mechanism and the adaptive threshold method, the system can adjust the risk level determination criteria in real time, effectively solving the problem that the traditional static assessment method cannot adapt to the dynamic changes of switching operations. For example, in the switching operation scenarios such as main transformer maintenance, the system can dynamically adjust the risk level determination threshold according to the risk distribution law of historical data to achieve more accurate risk level classification; in addition, this method shows good operability and promotion value in the practice of substation switching operation risk management. Through standardized data preprocessing procedures and a systematic risk level determination mechanism, the risk assessment process of switching operations is made more standardized and objective; the system can automatically learn the distribution characteristics of different risk levels from historical switching operation cases, establish data-driven risk level determination standards, and provide a scientific basis for the risk prevention and control of substation switching operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0055] Figure 1 A method flow chart of a method and system for assessing the risk of substation misoperation based on adaptive learning provided by an embodiment of the present invention;

[0056] Figure 2 An internal structural diagram of a computer device of a substation misoperation risk assessment method and system based on adaptive learning provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0057] In order to make the above-mentioned purposes, features and advantages of the present invention more understandable, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0058] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0059] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0060] Example 1

[0061] Reference Figure 1-Figure 2 , which is the first embodiment of the present invention, and provides a method for assessing the risk of substation misoperation based on adaptive learning, comprising:

[0062] Before describing the embodiments of the present application in detail, some related concepts are first explained for the sake of clarity.

[0063] TabNet: TabNet is a deep learning model for tabular data, which is particularly suitable for feature selection and classification tasks of structured data. In the present invention, TabNet is used for adaptive feature selection. Through the attention mechanism, TabNet can evaluate the contribution of different features to the risk level determination, and dynamically adjust the importance weights of the features, so as to screen out key features that have a significant impact on the risk level determination. Compared with traditional feature selection methods, TabNet can automatically learn the importance of features, avoid the limitations of manually setting weights, and improve the accuracy and robustness of feature selection.

[0064] CatBoost: CatBoost is a machine learning algorithm based on gradient boosting decision trees (GBDT), which is particularly good at processing mixed data of categorical features and numerical features. In the present invention, CatBoost is used for risk level determination modeling. Through its unique sequence feature encoding and gradient boosting mechanism, CatBoost can effectively handle the nonlinear relationship between categorical and numerical features and establish an accurate risk level determination model. CatBoost uses the target statistical coding method when processing categorical features to convert categorical features into numerical features, effectively avoiding the information loss caused by improper processing of categorical features in traditional methods. At the same time, its gradient boosting mechanism can capture the complex relationship between features and improve the prediction accuracy of the model.

[0065] Sparsemax: Sparsemax is an activation function for multi-classification tasks, similar to Softmax, but the generated results are more sparse, that is, more output values ​​are zero. In the present invention, Sparsemax is used to generate a feature selection mask. Through the Sparsemax function, the system can generate a sparse feature selection mask that indicates which features contribute most to the risk level determination. The mask generated by Sparsemax can highlight key features more clearly, reduce the interference of noise features, and improve the accuracy of feature selection and the interpretability of the model.

[0066] Target Encoding: Target Encoding is a method of converting categorical features into numerical features, which is achieved by calculating the statistical values ​​(such as mean, frequency, etc.) of each category on the target variable (such as risk level). In the present invention, target statistical coding is used to process categorical features. By calculating the probability distribution of different risk levels under each category, the categorical features are converted into numerical features with predictive value. Target statistical coding can preserve the relationship between categorical features and target variables, avoid the information loss caused by improper processing of categorical features in traditional methods, and improve the predictive ability of features.

[0067] This application provides a method that can effectively solve the above-mentioned problems. Next, multiple embodiments will be combined to explain in detail how to implement the substation misoperation risk assessment method based on adaptive learning;

[0068] Figure 1 A method flow chart of a substation misoperation risk assessment method and system based on adaptive learning is shown, including:

[0069] S1: Collect and preprocess multi-dimensional feature data related to the risk of electrical misoperation in substations;

[0070] Furthermore, multi-dimensional feature data includes but is not limited to operator skill level, equipment operating time, system operating status and environmental conditions; by comprehensively collecting multi-dimensional feature data, this step can ensure the integrity and diversity of basic data for subsequent analysis and provide comprehensive information support for risk assessment.

[0071] Standardization of numerical feature data includes:

[0072] The target statistical coding method is used to convert categorical feature data into numerical features.

[0073] Collect multi-dimensional feature data related to electrical misoperation in substations, including data on personnel, equipment, environment, management, and systems.

[0074] The collected numerical characteristic data are standardized, specifically by calculating the mean and standard deviation of the substation characteristic data, and converting the original characteristic values ​​into standardized characteristic values. The standardization formula is:

[0075] X norm =(X raw -μ sta ) / σ sta

[0076] Among them, X norm is the standardized characteristic value of the substation, X raw is the original characteristic value of the substation, μ sta is the average value of substation characteristics, σ sta is the standard deviation of substation characteristics;

[0077] The collected categorical feature data are processed using the target coding method. By calculating the probability distribution of different risk levels under each category, the categorical features are converted into numerical features with predictive value. The target coding formula for categorical features is:

[0078] Risk cat =(N risk +α smooth ) / (N total+α smooth ×K level )

[0079] Among them, Risk ca is the risk probability value of the feature category, N risk is the number of risk events under this category, N total is the total number of operations in this category, α smooth is the smoothing coefficient, K level is the number of risk levels. Through target encoding, this step effectively retains the conditional probability relationship between categorical features and risk levels, avoids information loss caused by improper processing of categorical features in traditional methods, and improves the predictive value of feature data.

[0080] S2: Use the TabNet algorithm to perform adaptive feature selection on the preprocessed feature data, evaluate the importance of different features through the attention mechanism, and dynamically adjust the weight of the features;

[0081] Furthermore, a TabNet feature selection network is constructed. Based on the feature selector parameters at the current moment and the hidden state at the previous moment, a nonlinear transformation is performed through the Sparsemax function to generate a feature selection mask. The feature mask generation formula is:

[0082] Mask feat =Sparsemax(W sel ×State prev )

[0083] Among them, Mask feat is the feature selection mask, W sel is the feature selection weight matrix, State prev is the previous state information; through the feature selection network, this step can adaptively select and transform features, automatically capture the key features and their interactive relationships that have a significant impact on the risk level determination, avoid the limitations of artificially set weights, and improve the accuracy and robustness of feature selection.

[0084] Apply the feature transformation network to transform the original features into more expressive feature representations. The feature transformation formula is:

[0085] Trans feat =Mask feat ×Net feat (X input )

[0086] Among them, Trans feat is the transformed feature, X input is the input feature, Net featIt is a feature transformation network. Through the feature transformation network, this step can transform the original features into a more expressive form, enhance the model's ability to interpret the features, and improve the accuracy of subsequent risk level determination.

[0087] S3: Use the CatBoost algorithm to perform deep learning modeling on the selected features, process the nonlinear relationship between categorical and numerical features, and establish a risk level determination model.

[0088] Furthermore, in the CatBoost model construction stage, an objective function for risk level determination is designed. The objective function includes a prediction error term and a regularization term. The risk prediction objective function is:

[0089] Loss total =Σ(Risk true -Risk pred )2+λ reg ×Σ||W feat ||2

[0090] Among them, Loss total is the total loss value, Risk true is the actual risk level, Risk pred To predict the risk level, λ reg is the regularization parameter, W feat is the feature weight; by designing the objective function, this step can effectively balance the prediction error and complexity of the model, prevent overfitting, and improve the generalization ability and prediction accuracy of the model.

[0091] Using the sequence feature encoding and gradient boosting mechanism of the CatBoost algorithm, deep learning and modeling are performed on the features selected and transformed by TabNet, the nonlinear relationship between categorical and numerical features is processed, and a risk level determination model is constructed. Through the CatBoost algorithm, this step can effectively process the nonlinear relationship between categorical and numerical features, establish an accurate risk level determination model, and improve the model's adaptability to complex scenarios and the accuracy of risk identification.

[0092] S4: Combining the output results of TabNet and CatBoost algorithms, the risk level of electrical misoperation of substations is accurately divided. By setting scientific and reasonable risk level quantification standards and dynamically optimized risk level division criteria, the risk of electrical misoperation of substations is assessed.

[0093] Furthermore, the feature importance scores of the TabNet and CatBoost models are fused, and the final feature weights are calculated based on the feature importance weighting mechanism. The feature importance fusion formula is:

[0094] Weightfinal =β mix ×Weight tabnet +(1-β mix )×Weight boost

[0095] Among them, Weight final is the final feature weight, β mix is the fusion coefficient, Weight tabnet is the TabNet feature weight, Weight boost is the CatBoost feature weight. Specifically, by integrating the feature importance scores of the two models, this step can fully utilize the advantages of the two algorithms, improve the accuracy and robustness of feature weight allocation, and ensure the reliability of subsequent risk score calculation.

[0096] Based on the fused feature weights, the final risk score is calculated, which reflects the severity of the risk of electrical misoperation in the substation. By calculating the risk score based on the fused feature weights, this step can more comprehensively assess the risk, reduce misjudgments caused by single model deviations, and improve the accuracy and credibility of risk assessment.

[0097] Adopting the adaptive threshold method, the risk level threshold is determined according to the statistical characteristics of historical data. The risk level threshold calculation formula is:

[0098] Thresh level =μ risk +k adj ×σ risk

[0099] Among them, Thresh level is the risk level threshold, μ risk is the average risk score, k adj is the adjustment coefficient, σ risk is the standard deviation of risk scores. Through the adaptive threshold method, this step can dynamically adjust the risk level classification standard according to historical data, solving the problem that the traditional static assessment method cannot adapt to changes in actual operating conditions, and improving the flexibility and accuracy of risk level classification.

[0100] Furthermore, according to the calculated risk score and the determined risk level threshold, the risk level of electrical misoperation of the substation is divided into different levels to complete the risk assessment. Through dynamic risk level classification, this step can more accurately reflect the actual risk status of electrical misoperation of the substation, provide a scientific basis for the operation and management of the substation, and effectively support the formulation and implementation of risk prevention and control measures.

[0101] Furthermore, this embodiment also provides a substation misoperation risk assessment system based on adaptive learning, including: a data acquisition module, a feature selection module, a deep learning module and a risk assessment module;

[0102] The data acquisition module is used to collect and preprocess multi-dimensional feature data related to the risk of electrical misoperation in substations;

[0103] The feature selection module is used to perform adaptive feature selection on preprocessed feature data using the TabNet algorithm, evaluate the importance of different features through the attention mechanism, and dynamically adjust the weight of the features;

[0104] The deep learning module is used to perform deep learning modeling on the selected features using the CatBoost algorithm, process the nonlinear relationship between categorical and numerical features, and establish a risk level determination model;

[0105] The risk assessment module is used to combine the output results of TabNet and CatBoost algorithms, divide the risk level of electrical misoperation of substations, set risk level quantification standards and dynamically optimized risk level division criteria, and complete the assessment of the risk of electrical misoperation of substations.

[0106] This embodiment also provides a computer device, which may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 2 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for assessing the risk of misoperation of a substation based on adaptive learning is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0107] This embodiment further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0108] Collect and preprocess multi-dimensional feature data related to the risk of electrical misoperation in substations;

[0109] The TabNet algorithm is used to perform adaptive feature selection on the preprocessed feature data, evaluate the importance of different features through the attention mechanism, and dynamically adjust the weight of the features;

[0110] The CatBoost algorithm is used to perform deep learning modeling on the selected features, process the nonlinear relationship between categorical and numerical features, and establish a risk level determination model;

[0111] Combining the output results of TabNet and CatBoost algorithms, the risk levels of electrical misoperation in substations are divided, and the risk level quantification standard and dynamically optimized risk level division criteria are set to complete the assessment of the risk of electrical misoperation in substations.

[0112] Example 2

[0113] Reference Figure 1 - Figure 2 , which is the second embodiment of the present invention, and this embodiment provides a substation misoperation risk assessment method based on adaptive learning. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0114] Take the risk level judgment of the switching operation during the maintenance of the main transformer #1 of a 500kV substation as an example. This operation involves complex operation processes such as busbar transfer and main transformer outlet switch operation.

[0115] Table 1. Input feature information table

[0116] Dimensions Feature Item Eigenvalue Personnel Dimension Skill Level intermediate Experience 5 years Recent training time 3 months ago Recent error records 1 minor operational error in the past six months Device Dimensions Main transformer running time 8 years Last maintenance time 6 months ago Equipment status score 85 points (out of 100) Environmental Dimension Weather conditions cloudy day visibility good Ambient temperature 25℃ Relative humidity 65％ Special weather none Management Dimension Operation Ticket Review Level 3 Audit On-site monitoring Provide full-time guardian Operational drills Completed simulation exercise System Dimension System operating status Heavy load operation Protection device status Normal operation Operational complexity Involving 3 intervals

[0117] Table 2. Algorithm feature processing results

[0118]

[0119] Table 3. Feature weight distribution results

[0120]

[0121]

[0122] Table 4. Risk level determination results table

[0123] Risk Level Score range Threshold determination method Historical data distribution Low risk <0.3 μ1+1.0σ1 45％ Medium risk 0.3-0.6 μ2+1.2σ2 35％ High risk 0.6-0.8 μ3+1.5σ3 15％ Very high risk >0.8 μ4+2.0σ4 5％

[0124] This case was judged as a medium risk level through the calculation result of the final risk score of 0.68, which fully demonstrated the superiority of the dual-path feature adaptive learning framework in determining the risk level of substation switching operations, especially its adaptability when dealing with multi-dimensional features and dynamic scenarios.

[0125] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0126] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiments of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0127] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0128] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0129] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1The steps for the functions specified in one or more boxes.

[0130] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0131] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for assessing the risk of substation misoperation based on adaptive learning, characterized in that: include, Collect and preprocess multi-dimensional feature data related to the risk of electrical misoperation in substations; The TabNet algorithm is used to perform adaptive feature selection on preprocessed feature data, evaluate the importance of different features through the attention mechanism, and dynamically adjust the weight of the features; The CatBoost algorithm is used to perform deep learning modeling on the selected features, process the nonlinear relationship between categorical and numerical features, and establish a risk level determination model; Combining the output results of TabNet and CatBoost algorithms, the risk levels of electrical misoperation in substations are divided, and the risk level quantification standard and dynamically optimized risk level division criteria are set to complete the assessment of the risk of electrical misoperation in substations.

2. The method for assessing substation misoperation risk based on adaptive learning according to claim 1, characterized in that: The pretreatment includes: The target statistical coding method is used to convert the categorical feature data into numerical features; The multi-dimensional feature data includes data of personnel, equipment, environment, management and system dimensions; The collected numerical characteristic data are standardized, the mean and standard deviation of the substation characteristic data are calculated, and the original characteristic values ​​are converted into standardized characteristic values. The calculation formula is as follows: X norm =(X raw -m sta ) / s sta Among them, X norm is the standardized characteristic value of the substation, X raw is the original characteristic value of the substation, μ sta is the average value of substation characteristics, σ sta is the standard deviation of substation characteristics; The collected categorical feature data are processed using the target coding method. By calculating the probability distribution of different risk levels under each category, the categorical features are converted into numerical features with predictive value. The target coding formula for categorical features is as follows: Risk cat =(N risk +a smooth ) / (N total +a smooth ×K level ) Among them, Risk ca is the risk probability value of the feature category, N risk is the number of risk events under this category, N total is the total number of operations in this category, α smooth is the smoothing coefficient, K level is the number of risk levels.

3. The method for assessing substation misoperation risk based on adaptive learning as claimed in claim 2, characterized in that: The adaptive feature selection includes: Based on the feature selector parameters at the current moment and the hidden state at the previous moment, a nonlinear transformation is performed through the Sparsemax function to generate a feature selection mask. The calculation formula for feature mask generation is as follows: Mask feat =Sparsemax(W sel ×State prev ) Among them, Mask feat is the feature selection mask, W sel is the feature selection weight matrix, State prev is the previous state information; Apply the feature transformation network to transform the original features into more expressive feature representations. The feature transformation formula is as follows: Trans feat =Mask feat ×Net feat (X input ) Among them, Trans feat is the transformed feature, X input is the input feature, Net feat It is the feature transformation network.

4. The method for assessing substation misoperation risk based on adaptive learning as claimed in claim 3, characterized in that: The deep learning modeling includes: In the CatBoost model construction stage, an objective function for risk level determination is designed. The objective function includes a prediction error term and a regularization term. The risk prediction objective function is as follows: Loss total =Σ(Risk true -Risk pred )2+λ reg ×Σ||W feat ||2 Among them, Loss total is the total loss value, Risk true is the actual risk level, Risk pred To predict the risk level, λ reg is the regularization parameter, W feat is the feature weight; The CatBoost algorithm's sequence feature encoding and gradient boosting mechanism are used to perform deep learning and modeling on the features selected and transformed by TabNet, process the nonlinear relationship between categorical and numerical features, and construct a risk level determination model.

5. The method for assessing substation misoperation risk based on adaptive learning as claimed in claim 4, characterized in that: The risk levels of electrical misoperation of substations include: The feature importance scores of the TabNet and CatBoost models are fused, and the final feature weights are calculated based on the feature importance weighting mechanism. The feature importance fusion formula is as follows: Weight final =b mix ×Weight tabnet +(1-β mix )×Weight boost Among them, Weight final is the final feature weight, β mix is the fusion coefficient, Weight tabnet is the TabNet feature weight, Weight boost is the CatBoost feature weight.

6. The method for assessing substation misoperation risk based on adaptive learning as claimed in claim 5, characterized in that: The risk level of electrical misoperation of substations also includes: Calculate the final risk score based on the fused feature weights; Adopting the adaptive threshold method, and determining the risk level threshold based on the statistical characteristics of historical data, the risk level threshold calculation formula is as follows: Thresh level =μ risk +k adj ×σ risk Among them, Thresh level is the risk level threshold, μ risk is the average risk score, k adj is the adjustment coefficient, σ risk is the standard deviation of the risk score.

7. The method for assessing substation misoperation risk based on adaptive learning according to claim 6, characterized in that: The assessment of the risk of electrical misoperation of the substation includes: Based on the calculated risk score and the determined risk level threshold, the risk level of electrical misoperation of the substation is divided into different levels to complete the risk assessment.

8. A substation misoperation risk assessment system based on adaptive learning, based on the substation misoperation risk assessment method based on adaptive learning according to any one of claims 1 to 7, characterized in that: It includes data collection module, feature selection module, deep learning module and risk assessment module; The data acquisition module is used to collect and preprocess multi-dimensional feature data related to the risk of electrical misoperation of the substation; The feature selection module is used to perform adaptive feature selection on the preprocessed feature data using the TabNet algorithm, evaluate the importance of different features through the attention mechanism, and dynamically adjust the weight of the features; The deep learning module is used to perform deep learning modeling on the selected features using the CatBoost algorithm, process the nonlinear relationship between categorical and numerical features, and establish a risk level determination model; The risk assessment module is used to combine the output results of TabNet and CatBoost algorithms, divide the risk level of substation electrical misoperation, set risk level quantification standards and dynamically optimized risk level division criteria, and complete the assessment of substation electrical misoperation risks.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the substation misoperation risk assessment method based on adaptive learning according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the substation misoperation risk assessment method based on adaptive learning according to any one of claims 1 to 7 are implemented.

Images