Key feature screening method and device for transient voltage stability evaluation
By screening key features through contribution weighted calculation and contribution growth rate, the difficulty of feature selection in transient voltage stability assessment in new power systems is solved, and efficient and lightweight feature selection is achieved, which is suitable for online application scenarios.
Patent Information
- Application Number
- CN202511084489.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-04
- Publication Date
- 2025-10-17
AI Technical Summary
In the process of transient voltage stability assessment in new power systems, existing technologies have difficulty in selecting key features, and existing methods rely on specific models, lack versatility, and are difficult to apply online.
The contribution weighted calculation and contribution growth rate methods are adopted to quantify the output change value of the transient voltage stability assessment model, calculate the local and comprehensive contribution of the features, and combine the preset threshold to screen the key features to avoid the iterative search process.
The method improves the versatility and real-time performance, is applicable to different transient voltage assessment models, ensures the simplicity of the feature set and the assessment accuracy, and solves the problems of poor adaptability and long calculation time of traditional methods in power systems.
Smart Images
Figure CN120804644A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the field of power transmission, and particularly relates to a key feature screening method and device for transient voltage stability evaluation. BACKGROUND
[0002] With the construction of a new energy system and the access of large-scale wind power and photovoltaic power, the randomness of sources, networks and loads in a new power system gradually increases, and the transient voltage stability evaluation has strong uncertainty. Since wind power and photovoltaic power are connected to the power grid through power electronic devices, multiple types of state variables in the system have a complex influence on the transient voltage stability evaluation. In the process of transient voltage stability evaluation, the selection of key features of the power system faces great difficulties.
[0003] For the key feature screening problem, scholars at home and abroad have explored various methods. For example, conditional mutual information and maximum correlation minimum redundancy criterion have been widely used to screen the optimal feature subset. In recent years, machine learning algorithms have also been applied to this, such as using decision tree rules to extract or support vector machine recursive feature selection to reduce dimensionality. However, these methods depend on specific models and have insufficient universality. Although Tabu search technology and filter type feature selection methods are also used to screen the optimal feature set, they need to be assisted by search algorithms and are difficult to apply online, which has limitations in the analysis of voltage stability of power systems. SUMMARY
[0004] To solve the above problems, the application provides a key feature screening method and device for transient voltage stability evaluation.
[0005] To achieve the above purpose, the application provides the following technical solutions: A key feature screening method for transient voltage stability evaluation, the method comprising: obtaining a plurality of input features of a preset transient voltage stability evaluation model; quantifying the output result of the transient voltage stability evaluation model, and calculating the model output change value before and after excluding any input feature; respectively calculating the first feature combination number of all input features and the second feature combination number after excluding any input feature; taking the ratio of the second feature combination number to the first feature combination number as the weight of the model output change value, and determining the local contribution degree of any input feature; accumulating the sum value of the local contribution degree of the input feature in different combination conditions of other input features as the comprehensive contribution degree of the excluded input feature; determining the key feature according to the contribution degree growth rate and a preset threshold.
[0006] Optionally, the calculation formula of the comprehensive contribution degree is: ; wherein T is the number of input features in the input model; is the number of feature combinations without considering the input feature i; and are the model output results without excluding and excluding the input feature i in any feature combination of the T input features, M is the total number of input features, and there are combinations of the remaining features excluding any feature i.
[0007] Optionally, the output of the transient voltage stability assessment model includes the deviation voltage at the assessment time, and the quantization of the output result of the transient voltage stability assessment model includes: determining a voltage deviation rate according to the deviation voltage, integrating the voltage deviation rate to obtain a voltage trajectory deviation integral, and taking the voltage trajectory deviation integral as the quantization result of the transient voltage stability assessment model.
[0008] Optionally, determining the contribution degree growth rate based on the comprehensive contribution degree includes: sequentially sorting all input features in descending order according to the comprehensive contribution degree to obtain a feature sequence; for any input feature corresponding to the feature sequence, determining a first cumulative contribution rate of the input feature and all previous input features according to the ratio of the first comprehensive contribution degree sum value of the input feature and all previous input features to the second comprehensive contribution degree sum value of all input features; determining a second cumulative contribution rate of the next sequence input feature and all previous input features, and determining the contribution degree growth rate according to the first cumulative contribution rate and the second cumulative contribution rate; the next sequence input feature is the input feature after the last input feature corresponding to the first cumulative contribution rate in the feature sequence.
[0009] Optionally, the calculation formula of the first cumulative contribution rate is: ; wherein M is the total number of input features, j is the sequence number of any input feature in the feature sequence, and i is the i-th input feature.
[0010] Optionally, determining the contribution degree growth rate according to the first cumulative contribution rate and the second cumulative contribution rate includes: ; wherein is the contribution degree growth rate, is the first cumulative contribution rate, is the second cumulative contribution rate, M is the total number of input features, j is the sequence number of any input feature in the feature sequence which is greater than or equal to 2.
[0011] Optionally, determining the key feature according to the contribution growth rate and the preset threshold value comprises: wherein j is the sequence number of any input feature in the feature sequence which is greater than or equal to 2, k is the sequence number of the last key feature in the feature sequence when k takes the maximum value, is the preset threshold value.
[0012] A key feature screening device for transient voltage stability evaluation, the device comprises: an acquisition module configured to acquire a plurality of input features of a preset transient voltage stability evaluation model; a quantification module configured to quantify an output result of the transient voltage stability evaluation model and calculate a model output change value before and after input of each input feature; a calculation module configured to calculate a first feature combination number of all input features and a second feature combination number after exclusion of any input feature respectively; take a ratio of the second feature combination number to the first feature combination number as a weight of the model output change value, determine a local contribution degree of any input feature; and accumulate a sum value of the local contribution degrees of other input features as a comprehensive contribution degree of the fixed input feature in the case of exclusion of any input feature; a determination module configured to determine the key feature according to the contribution growth rate and the preset threshold value.
[0013] A computer readable storage medium, the storage medium stores a computer program, the computer program is executed by a processor to realize the above-mentioned key feature screening method for transient voltage stability evaluation.
[0014] A computer device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor executes the program to realize the above-mentioned key feature screening method for transient voltage stability evaluation.
[0015] The key feature screening method for transient voltage stability evaluation provided by the application has the following beneficial effects: Firstly, by contribution weight calculation and contribution growth rate calculation, the dependence on specific algorithms such as Tabu search is eliminated, which is applicable to different transient voltage evaluation models, and the universality of the method is significantly improved; secondly, the feature contribution calculation mechanism based on different feature combination weighted summation avoids the complex iterative search process of traditional methods, and is more suitable for power system online application scenarios with high real-time requirements; in addition, the comprehensive contribution degree is adopted, the situation of different feature combinations is considered, and the features are dynamically selected combined with the preset threshold, which not only ensures the simplicity of the feature set but also maintains the evaluation accuracy under different working conditions; based on the above steps, while keeping the algorithm lightweight, the core pain points of the traditional feature selection method in the power system scene, such as poor adaptability and long calculation time, are effectively solved. The method provides an efficient and reliable feature selection scheme for power system transient voltage stability evaluation through the organic combination of probability statistics and serialization processing. BRIEF DESCRIPTION OF DRAWINGS
[0016] In order to more clearly illustrate the embodiments of the present application and the design scheme thereof, the drawings required by the present embodiments will be briefly introduced as follows. The drawings in the following description are only part of the embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.
[0017] Figure 1 A flowchart of a key feature screening method for transient voltage stability evaluation according to an exemplary embodiment of the present application is provided.
[0018] Figure 2 A block diagram of a key feature screening device for transient voltage stability evaluation according to an exemplary embodiment of the present application is provided. DETAILED DESCRIPTION
[0019] In order to make those skilled in the art better understand the technical solutions of the present application and can be implemented, the present application will be described in detail below in combination with the drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present application, and cannot limit the protection scope of the present application.
[0020] The technical solutions provided by the embodiments of the present application will be described in detail below in combination with the drawings.
[0021] Firstly, the present application provides a key feature screening method for transient voltage stability evaluation, specifically as shown in Figure 1 The method comprises the following steps: S101, obtaining a plurality of input features of a preset transient voltage stability evaluation model.
[0022] In the present application, in order to facilitate the evaluation of the contribution degree of the input features, for the trained machine learning model, i.e. the preset transient voltage stability evaluation model, the model can be analyzed based on the contribution degree in this step, and an interpretable artificial intelligence model representing the contribution degree is constructed as follows: ; Where g(·) is an interpretable artificial intelligence model; M is the number of all input features; φ m is the contribution value of the mth feature; φ0 is a constant.
[0023] S102, quantifying the output result of the transient voltage stability evaluation model, and calculating the change value of the model output before and after excluding any input feature.
[0024] In this step, the output result of the model needs to be quantified first. The output result of the model needs to be able to represent the transient voltage stability degree, for example, the output of the transient voltage stability evaluation model can include stability margin, instability probability or safety boundary distance, etc., which is a continuous numerical value, representing the transient voltage stability degree of the system under a certain operating condition.
[0025] The above outputs for representing the transient voltage stability degree all need to be calculated according to the deviation voltage at the current prediction time, i.e. the difference between the voltage at the prediction time and the rated voltage, so the output of the transient voltage stability evaluation model includes the deviation voltage at the evaluation time.
[0026] For the quantification of the output result, the voltage deviation rate can be determined according to the deviation voltage, the voltage trajectory deviation integral is obtained by integrating the voltage deviation rate, and the voltage trajectory deviation integral is taken as the quantification result of the transient voltage stability evaluation model.
[0027] In addition, in this step, different input feature combinations represent different operating conditions, and in the case of excluding any input feature, the operating condition corresponding to the input feature is fixed, i.e. in order to reflect the contribution degree of any input feature to the model output, the influence of the input feature on the model output under different operating conditions needs to be considered, the condition corresponding to the input feature is fixed, and then the output results obtained based on different operating conditions (different combinations of other input features) may also be different.
[0028] Therefore, the condition corresponding to any input feature can be fixed, and the model output can be determined based on the multiple combinations of other input features (i.e. multiple operating conditions corresponding to other input features), i.e. in the case of excluding any input feature, the outputs of the multiple combinations of other input features are determined.
[0029] Based on the quantized model output, the output obtained by inputting all input features into the model, and the output obtained by inputting other input features into the model after excluding any input feature, the absolute value of the output difference obtained in different input conditions with only the excluded any input feature as the difference is taken as the model output change value.
[0030] S103, respectively calculate the first feature combination number of all input features and the second feature combination number after excluding any input feature; take the ratio of the second feature combination number to the first feature combination number as the weight of the model output change value to determine the local contribution degree of any input feature.
[0031] In this step, for the number of all input features, denoted as M, for the number of input features of the input model, denoted as T. Since different input feature combinations represent different working conditions, that is, in the case of considering the order, the model M features have M! combinations, that is, the first feature combination number is M!, and the second feature combination number after excluding any input feature is .
[0032] Taking as the weight of the model output change value, the local contribution degree of the any input feature in a certain working condition is .
[0033] S104, accumulate the sum of the local contribution degrees of the any input feature in different combination conditions of other input features as the comprehensive contribution degree of the excluded input feature.
[0034] For the trained transient voltage stability evaluation model, to calculate the contribution degree of the i-th feature to the output result, the contribution degrees of all possible feature combinations need to be calculated and then weighted and summed. Then the contribution degree of feature i in the nonlinear model is:
[0035] Wherein, T is the number of input features in the input model; is the number of feature combinations without considering the input feature i; and are the model output results without excluding and excluding the input feature i in any feature combination of T input features; M is the total number of input features, and after excluding any feature i, the remaining features have combinations.
[0036] In summary, based on the calculated feature contribution degree in the interpretable model, the contribution degree of each input feature to the model output result in the transient voltage stability evaluation model can be effectively quantified φ iThe contribution degree of each input feature to the evaluation result is determined, so that the key feature affecting the evaluation result in the transient voltage stability feature is determined. The greater the contribution degree of a single input feature, the greater the influence of the input feature on the transient voltage evaluation result. Thus, by comparing the relative sizes of the contribution degrees φi corresponding to the input features, the key feature affecting the transient voltage stability evaluation among the input features can be determined.
[0037] S105, determining the key feature according to the contribution degree growth rate and the preset threshold.
[0038] In this step, the contribution degree growth rate needs to be determined first.
[0039] For example, all input features are sorted in descending order according to the comprehensive contribution degree, and a feature sequence is obtained.
[0040] For example, in a new energy grid-connected system, the contribution degrees of the input features are calculated by using an interpretable artificial intelligence method. Generally, for M input feature quantities of the system, M contribution degrees of the input features to the model output are obtained, which are arranged in descending order according to their absolute values as:
[0041] Taking the above feature sequence as an example, the matrix and correspond to the contribution degrees of the key input feature quantities of the system ( φ 1, φ 2,…, φ m ) and the non-key input feature quantities ( φ m+1 ,…, φ M ), respectively, according to different growth rates.
[0042] Secondly, for any input feature in the feature sequence, the first cumulative contribution rate of the input feature and all previous input features is determined according to the ratio of the first contribution degree and the value of the input feature and all previous input features to the second contribution degree and the value of all input features.
[0043] For example, the key features affecting the transient voltage stability evaluation are determined by screening the features. That is, the number of contribution degrees in the matrix is determined.
[0044] In one embodiment, the contribution degrees of the non-key features correspond to the element values in the matrix , which are much smaller than the elements in the matrix , but not equal to zero. The contribution degree increment of the contribution degrees of the first j features in all M feature contribution values is defined as : ; wherein, M is the total number of input features, j is the sequence number of any input feature in the feature sequence which is greater than or equal to 2, and i is the i th input feature.
[0045] Then, the second cumulative contribution rate of the next sequence input feature and all previous input features is determined, and the contribution growth rate is determined according to the first cumulative contribution rate and the second cumulative contribution rate.
[0046] wherein, the next sequence input feature is the input feature after the last input feature corresponding to the first cumulative contribution rate in the feature sequence.
[0047] The contribution growth rate is further defined as The contribution growth rate can be determined according to the difference between the first cumulative contribution rate and the second cumulative contribution rate: ; The contribution growth rate can also be determined according to the ratio of the difference between the first cumulative contribution rate and the second cumulative contribution rate to the second cumulative contribution rate: ; wherein, is the first cumulative contribution rate, is the second cumulative contribution rate, M is the total number of input features, and j is the sequence number of any input feature in the feature sequence.
[0048] Finally, the key feature is determined according to the contribution growth rate and a preset threshold value.
[0049] The number of system key features is determined based on the contribution growth rate, that is, the number of elements in the matrix . ; wherein, j is the sequence number of any input feature in the feature sequence which is greater than or equal to 2, k is the sequence number of the last key feature in the feature sequence when k takes the maximum value, is a preset threshold value.
[0050] Since the first k key feature quantities in the feature sequence contain much more system feature information than the non-key feature information of the system, the contribution growth rate must have a mutation point at the contribution of the k th input feature, that is, the contribution will drop sharply, and the last input feature before the drop is the last key feature.
[0051] By using the above method, the contribution degree weighted calculation and the contribution degree growth rate calculation are adopted, the dependence on specific algorithms such as Tabu search is got rid of, the method is suitable for different transient voltage evaluation models, and the universality of the method is significantly improved; secondly, the feature contribution degree calculation mechanism based on the weighted sum of different feature combinations avoids the complex iterative search process of the traditional method, and is more suitable for the power system online application scene with high real-time requirement; in addition, the comprehensive contribution degree is adopted, the conditions of different feature combinations are considered, the features are dynamically screened in combination with the preset threshold, the simplicity of the feature set is ensured, and the evaluation accuracy under different working conditions is maintained; based on the above steps, while keeping the algorithm lightweight, the core pain points of the traditional feature selection method in the power system scene, such as poor adaptability and long calculation time, are effectively solved. The method combines probability statistics and serialization processing, and provides an efficient and reliable feature selection scheme for power system transient voltage stability evaluation.
[0052] Secondly, the application further provides a key feature screening device for transient voltage stability evaluation, as shown in the accompanying drawings, comprising: Figure 2 The acquisition module 201 is used for acquiring a plurality of input features of a preset transient voltage stability evaluation model.
[0053] The quantification module 202 is used for quantifying the output result of the transient voltage stability evaluation model, and calculating the model output change value before and after each input feature is input.
[0054] The calculation module 203 is used for respectively calculating the first feature combination number of all input features and the second feature combination number after excluding any input feature; the ratio of the second feature combination number to the first feature combination number is taken as the weight of the model output change value, and the local contribution degree of any input feature is determined; in the case of excluding any input feature, the sum value of the local contribution degrees of other input features is accumulated as the comprehensive contribution degree of the fixed input feature.
[0055] The determination module 204 is used for determining the key feature according to the contribution degree growth rate and the preset threshold.
[0056] By adopting the above-mentioned device, through the weighted calculation of contribution and the calculation of contribution growth rate, the dependence on specific algorithms such as Tabu search is eliminated, and it is applicable to different transient voltage assessment models, which significantly improves the versatility of the method; secondly, the feature contribution calculation mechanism based on the weighted summation of different feature combinations avoids the complex iterative search process of traditional methods and is more suitable for online application scenarios of power systems with high real-time requirements; in addition, the comprehensive contribution is adopted, taking into account the situations of different feature combinations, and combining the preset threshold to dynamically screen features, which not only ensures the simplicity of the feature set but also maintains the assessment accuracy under different working conditions; based on the above steps, while keeping the algorithm lightweight, it effectively solves the core pain points of traditional feature selection methods in power system scenarios, such as poor adaptability and long calculation time. This method provides an efficient and reliable feature selection solution for transient voltage stability assessment of power systems through the organic combination of probability statistics and serialization processing.
[0057] The present invention also provides a computer-readable storage medium, which stores a computer program, which can be used to execute the above Figure 1 The steps of the critical characteristics screening method for transient voltage stability assessment are provided.
[0058] The present invention also provides a computer device. At the hardware level, the computer device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory. Of course, it may also include hardware required for other services. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to achieve the above Figure 1 The steps of the critical characteristics screening method for transient voltage stability assessment are provided.
[0059] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0060] The present invention is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams.Figure 1 apparatuses that implement the functions specified in the flowchart Figure 1
[0061] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flowchart Figure 1 apparatuses that implement the functions specified in the flowchart Figure 1
[0062] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions that are executed on the computer or other programmable apparatus provide steps for implementing the flowchart Figure 1 apparatuses that implement the functions specified in the flowchart Figure 1
[0063] It should be noted that the above detailed description is merely exemplary and is not intended to limit the present application in any way. Thus, although the present application has been described in detail with respect to particular embodiments, it should not be construed as limited thereto. Those skilled in the art will recognize that modifications and variations can be made to the present application without departing from the spirit or scope of the application. Accordingly, it will be clear that the application is intended to cover all modifications and variations of this application that fall within the scope of the application. It is therefore intended that the application cover all such modifications and variations as fall within the scope of the application. No claim element herein is to be construed under the provisions of 35 U.S.C. 112, sixth paragraph, unless the word "means" is specifically used. No claim element is therefore to be construed as meaning plus function unless the phrase "means for" is explicitly used. The use of the common abbreviation "e.g.", unless specifically stated, means that what follows is an example of something. Thus, in descriptions of some embodiments, "e.g." can sometimes, but does not always, indicate that what follows is an example of something.
Claims
1. A method for screening key features of transient voltage stability assessment, characterized in that: The method comprises: obtaining multiple input features of a preset transient voltage stability assessment model; quantifying the output of the transient voltage stability assessment model, and calculating the change in the model output before and after any input feature is excluded; Calculate the number of first feature combinations of all input features and the number of second feature combinations after excluding any input feature; Using the ratio of the number of the second feature combinations to the number of the first feature combinations as the weight of the model output change value to determine the local contribution of any input feature; The sum of the local contributions of any input feature in different combinations of other input features is accumulated as the comprehensive contribution of the excluded input feature; A contribution growth rate is determined based on the comprehensive contribution, and a key feature is determined according to the contribution growth rate and a preset threshold.
2. A method for screening key features of transient voltage stability assessment according to claim 1, characterized in that: The calculation formula for the comprehensive contribution is: ； Where T is the number of input features in the input model; is the number of feature combinations without considering the input feature i; and They are the model output results when any feature combination of T input features is not excluded and when input feature i is excluded. M is the total number of input features. Excluding any feature i, the remaining features have combinations.
3. The method for screening key features of transient voltage stability assessment according to claim 1, characterized in that: The output of the transient voltage stability assessment model includes the deviation voltage at the assessment time. Quantifying the output results of the transient voltage stability assessment model includes: A voltage deviation rate is determined according to the deviation voltage, the voltage deviation rate is integrated to obtain a voltage trajectory deviation integral, and the voltage trajectory deviation integral is used as a quantified result of a transient voltage stability assessment model.
4. The method for screening key features of transient voltage stability assessment according to claim 1, characterized in that: Determining the contribution growth rate based on comprehensive contribution includes: Sort all input features in descending order according to their comprehensive contribution to obtain a feature sequence; For any input feature corresponding to the feature sequence, determine a first cumulative contribution rate of the input feature and all preceding input features based on a ratio of the first comprehensive contribution sum of the input feature and all preceding input features to the second comprehensive contribution sum of all input features; Determine the second cumulative contribution rate of the next sequence input feature and all previous sequence input features, and determine the contribution growth rate based on the first cumulative contribution rate and the second cumulative contribution rate; the next sequence input feature is the input feature following the last input feature corresponding to the first cumulative contribution rate in the feature sequence.
5. A method for screening key features of transient voltage stability assessment according to claim 4, characterized in that: The calculation formula for the first cumulative contribution rate is: ； Among them, M is the total number of input features, j is the sequence number of any input feature in the feature sequence, and i is the i-th input feature.
6. A method for screening key features of transient voltage stability assessment according to claim 4, characterized in that: Determining the contribution growth rate according to the first cumulative contribution rate and the second cumulative contribution rate includes: ； in, is the contribution growth rate, is the first cumulative contribution rate, is the second cumulative contribution rate, M is the total number of input features, and j is the number of sequences of any input feature in the feature sequence that is greater than or equal to 2.
7. The method for screening key features of transient voltage stability assessment according to claim 1, characterized in that: Determining the key features based on the contribution growth rate and the preset threshold includes: Among them, j is the sequence number of any input feature whose sequence is greater than or equal to 2 in the feature sequence. When k takes the maximum value, it is the sequence number of the last key feature in the feature sequence. is the preset threshold.
8. A key feature screening device for transient voltage stability assessment, characterized in that: The device comprises: An acquisition module, used for acquiring multiple input features of a preset transient voltage stability assessment model; a quantification module, configured to quantify the output of the transient voltage stability assessment model and calculate the change in the model output before and after each input feature is input; a calculation module, configured to respectively calculate the number of first feature combinations of all input features and the number of second feature combinations after excluding any input feature; use the ratio of the second feature combination number to the first feature combination number as the weight of the model output change value to determine the local contribution of any input feature; and when excluding any input feature, accumulate the sum of the local contributions of other input features as the comprehensive contribution of the fixed input feature; The determination module is used to determine the key feature according to the contribution growth rate and a preset threshold.
9. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and 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, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 7 when executing the program.