A high-altitude power grid insulation gap compensation method and system based on gradient boosting

Abstract

This invention relates to a gradient-lifting-based method and system for compensating insulation gaps in high-altitude power grids, belonging to the field of insulation characteristic testing technology. The method includes the following steps: collecting real-time environmental parameters and measured insulation gap values ​​of the high-altitude power grid; after preprocessing, dividing the data into training and validation sets using stratified sampling; inputting the feature matrix of the training set into a preset standard insulation correction model to output a benchmark insulation gap value, and calculating the initial residual based on its deviation from the measured value; constructing a gradient-lifting optimization model, using the benchmark value as the initial value and residual correction as the objective, and iteratively training until the convergence condition is met, outputting the total correction amount; fusing the benchmark value and the total correction amount to obtain a predicted value, and outputting this predicted value as the final insulation compensation result when its deviation rate from the measured value is less than a preset engineering accuracy threshold. This invention achieves a deep integration of physical mechanisms and data-driven algorithms, improving the dynamic response speed and accuracy of insulation gap prediction in high-altitude environments.

Description

Technical Field

[0001] This invention relates to a method and system for compensating insulation gaps in high-altitude power grids based on gradient enhancement, belonging to the field of insulation characteristic testing technology. Background Technology

[0002] In power systems, the safe operation of power grids in high-altitude areas faces unique environmental challenges. Insulation performance is a core indicator for ensuring the stable operation of transmission lines and substation equipment. In high-altitude environments, the combined effects of multiple factors such as low air pressure, drastic temperature differences, strong ultraviolet radiation, and humidity fluctuations can significantly affect the breakdown characteristics of insulating media. For example, a decrease in air pressure leads to a decrease in air insulation strength, with the insulation breakdown voltage decreasing by approximately 8%-10% for every 1000 meters increase in altitude. Strong ultraviolet radiation accelerates the aging of insulating materials such as insulators, further weakening insulation performance. Therefore, it is necessary to accurately calculate insulation gaps to avoid accidents such as insulation breakdown and line tripping, ensuring the reliability of power grid supply.

[0003] In existing technologies, the calculation and compensation methods for insulation gaps in high-altitude power grids are mainly divided into two categories: one is the traditional standard empirical formula method, with the IEC-60071 series as the core, and the other is the traditional machine learning and static modeling method, with the core being the construction of a mapping model between environmental parameters and measured insulation gaps to achieve correction. In practical applications, for the traditional standard empirical formula method, different combinations of air pressure and humidity at the same altitude lead to significant differences in insulation performance, and the static formula cannot cover such dynamic fluctuations. Moreover, it is only applicable to standard altitudes and standard meteorological environments. In extremely high-altitude areas and under special meteorological conditions, the correction error is large, resulting in redundant or insufficient insulation gap design. For the traditional machine learning and static modeling methods, the high-altitude environmental parameter samples are sparsely distributed and highly heterogeneous. The air pressure and altitude in different regions have significantly different coupling correlation characteristics, which makes the model prone to overfitting and the prediction accuracy drops significantly. Furthermore, the parameters are fixed after the model is trained, and it is impossible to dynamically adjust the correction strategy according to real-time environmental parameters, making it difficult to match the real-time fluctuation characteristics of the high-altitude environment.

[0004] Therefore, there is an urgent need for a gradient-based method and system for compensating insulation gaps in high-altitude power grids, which has the advantages of dynamic correction and compensation and error reduction. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention proposes a method and system for compensating insulation gaps in high-altitude power grids based on gradient enhancement.

[0006] The technical solution of the present invention is as follows: On the one hand, this invention proposes a method for compensating insulation gaps in high-altitude power grids based on gradient improvement, comprising the following steps: Collect real-time environmental parameters and measured insulation gap values ​​of power grids in high-altitude areas; The real-time environmental parameters are preprocessed to construct a parameter set; the parameter set is divided into a training set and a validation set using stratified sampling, and the feature matrix of the training set is extracted. The feature matrix is ​​input into the preset standard insulation correction model, and the reference insulation gap value is output; the deviation between the reference insulation gap value and the measured insulation gap value is calculated to obtain the insulation gap residual; The pre-trained gradient maximization optimization model uses a baseline insulation gap value as the initial value and aims to correct the insulation gap residual for iterative training. When the preset convergence condition is met, the iterative training stops and the total correction amount is output. The total correction amount and the reference insulation gap value are combined to obtain the corrected insulation gap value; the corrected insulation gap value is compared with the measured insulation gap value, and the deviation rate of the insulation gap value is output; the deviation rate is judged to be smaller than the preset threshold. When the deviation rate is less than the preset engineering accuracy threshold, the corrected insulation gap value is output as the high-altitude insulation compensation result.

[0007] Preferably, the parameter set includes altitude, and at least two of air pressure, temperature, humidity, and ultraviolet intensity.

[0008] Preferably, the preset standard insulation correction model adopts the standard empirical formula established in the IEC-60071 standard.

[0009] Preferably, the gradient boosting optimization model is trained according to the following steps: The reference insulation gap value is set as the initial predicted value of the gradient boosting optimization model; The sum of squared residuals of the insulation gap prediction is defined as the loss function of the model; the sum of squared residuals of the insulation gap prediction is the sum of the squared deviations between the initial predicted value and the measured insulation gap, and a regularization strategy is configured. The training set is input into the gradient boosting optimization model. The prediction residual is fitted based on the loss function to calculate the first round of insulation gap optimization correction. This correction is then added to the initial prediction value to obtain the prediction value after the first round of iteration. After each round of iteration, the loss value of the validation set is calculated based on the prediction value of the current round and the measured insulation gap value of the validation set. The loss value of the validation set is the sum of squares of the deviations between the current round's validation set prediction value and the corresponding measured insulation gap value. Calculate the decrease in the validation set loss value within the preset statistical window. The loss value of the current round of validation set is compared with the preset convergence threshold, and the decrease rate of the loss value within the statistical window is compared with the preset rate. Based on the comparison results, iterative operations are performed.

[0010] Preferably, the iterative operation based on the comparison results specifically includes the following steps: If the validation set loss value is greater than the preset convergence threshold and the decrease in loss value is greater than the preset magnitude, continue to the next iteration; at the same time, the regularization strategy remains unchanged, and the prediction residual is fitted based on the loss function of the current round and the feature matrix of the training set, the new round of insulation gap optimization correction is calculated and accumulated to the current prediction value; If the validation set loss value is greater than the preset convergence threshold and the decrease in the loss value is less than or equal to the preset magnitude, continue to the next iteration; at the same time, the regularization strategy remains unchanged, and the fitting correction amount, the accumulated prediction value, the validation set loss value and the decrease magnitude are repeatedly calculated. If the validation set loss value is less than or equal to the preset convergence threshold, and the decrease in loss value is greater than the preset magnitude, continue to the next iteration; at the same time, adjust the regularization strategy, fit the prediction residual based on the adjusted regularization strategy, the current loss function and the training set feature matrix, calculate the new round of insulation gap optimization correction amount and accumulate it to the current prediction value, and recalculate the validation set loss value and the magnitude of decrease. If the validation set loss value is less than or equal to the preset convergence threshold and the decrease in loss value is less than or equal to the preset magnitude, stop iterative training; accumulate the insulation gap optimization correction amount of all rounds in the iteration process to obtain the total correction amount.

[0011] Preferably, the regularization strategy is selected from one of the following: upper limit constraint on the weights of a single base learner and constraint on the number of base learners, wherein: The upper limit constraint on the weights of the single-base learner is expressed by the formula: ; In the formula, Indicates the base learner's first The upper limit of weights in each iteration. Indicates the base learner's first The initial weights for each iteration. Indicates the upper limit threshold of the weight. This represents the weight decay coefficient; The constraint on the number of base learners is expressed by the formula: ; In the formula, Indicates the first The maximum number of base cells allowed in a round of iteration. Indicates the first The maximum number of base cells allowed in a round of iteration. This represents the upper limit of the number of base learners. This indicates the percentage decrease in quantity.

[0012] Preferably, the deviation rate is compared with a preset threshold. When the deviation rate of the insulation gap value is greater than or equal to the preset engineering accuracy threshold, a backtracking adjustment is triggered, backtracking to the gradient boosting optimization model, and adjusting the statistical window, specifically: When the decrease in the validation set loss value within the statistical window is greater than the preset range, and the deviation rate of the insulation gap value is less than the preset engineering accuracy threshold, the window adjustment step size is set to a positive value. When the decrease in the validation set loss value within the statistical window is less than or equal to the preset range, and the deviation rate of the insulation gap value is greater than or equal to the preset engineering accuracy threshold, the window adjustment step size is set to a negative value. In other cases, the window adjustment step size is set to zero, and the statistical window size remains unchanged. Based on the current statistical window size, the window adjustment step size of the corresponding round is added to obtain the updated statistical window size.

[0013] On the other hand, the present invention also proposes a high-altitude power grid insulation gap compensation system based on gradient improvement, comprising the following modules: Parameter acquisition module: Collects real-time environmental parameters and measured insulation gap values ​​of power grids in high-altitude areas; Preprocessing module: preprocesses real-time environmental parameters to construct a parameter set; uses stratified sampling to divide the parameter set into a training set and a validation set, and extracts the feature matrix of the training set; The residual calculation module inputs the feature matrix into the preset standard insulation correction model and outputs the reference insulation gap value; it calculates the deviation between the reference insulation gap value and the measured insulation gap value to obtain the insulation gap residual. Total correction calculation module: Constructs a gradient boosting optimization model, takes the baseline insulation gap value as the initial value, and performs iterative training with the goal of correcting the insulation gap residual; when the preset convergence condition is met, iterative training stops and the total correction is output; The compensation result output module combines the total correction amount and the reference insulation gap value to obtain the corrected insulation gap value; it compares the corrected insulation gap value with the measured insulation gap value and outputs the deviation rate of the insulation gap value; it judges the magnitude of the deviation rate and the preset threshold. When the deviation rate is less than the preset engineering accuracy threshold, it outputs the corrected insulation gap value as the high-altitude insulation compensation result.

[0014] In another aspect, the present invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method as described in any embodiment of the present invention.

[0015] In another aspect, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any embodiment of the present invention.

[0016] The present invention has the following beneficial effects: (1) This invention is a method and system for compensating insulation gaps in high-altitude power grids based on gradient boosting. It employs a two-layer modeling logic that combines anchoring physical benchmarks with precise correction using a gradient boosting algorithm. By using the benchmark insulation gap value output by the engineering-verified standard empirical formula as the initial prediction value of the gradient boosting optimization model, it leverages the robustness of the standard formula to ensure a minimum prediction accuracy under extreme altitudes and special weather conditions, while the gradient boosting algorithm focuses on nonlinear coupling errors not covered by the fitted benchmark value. This design effectively solves the technical pain points of traditional pure empirical formula methods being unable to adapt to dynamic environmental fluctuations and pure machine learning models lacking physical mechanism support and prone to overfitting. It achieves a deep integration of physical laws and data-driven approaches, significantly improving the stability and accuracy of high-altitude insulation gap prediction.

[0017] (2) This invention is a method and system for compensating insulation gaps in high-altitude power grids based on gradient boosting. Through dynamic iterative training of the gradient boosting optimization model throughout the entire process, coupled with a dual regularization strategy of upper limit constraint on single-base learner weights and constraint on the number of base learners, a dynamic optimization mechanism integrating iterative fitting, loss monitoring, and strategy adjustment is constructed. Unlike the shortcomings of traditional static machine learning models where parameters are fixed after training, in this design, each iteration adjusts the iteration strategy based on the validation set loss value and the rate of loss decrease within the statistical window. Regularization-related parameters are dynamically adapted with the iteration process, effectively suppressing the risk of overfitting caused by sparse and heterogeneous samples in high-altitude environments, while ensuring that the model can respond to environmental parameter fluctuations in real time. This innovative design enables the model to have excellent generalization ability in different altitude ranges and voltage levels, successfully solving the problems of poor adaptability and insufficient dynamic response of traditional models.

[0018] (3) This invention is a method and system for compensating insulation gaps in high-altitude power grids based on gradient boosting. By constructing a closed-loop control mechanism that links deviation rate verification, backtracking adjustment, and window optimization, when the insulation gap deviation rate does not meet the preset engineering accuracy threshold, it backtracks to the gradient boosting optimization model and dynamically adjusts the size of the statistical window based on the decrease in the loss value of the validation set within the statistical window and the current deviation rate. The window adjustment follows the logic of precise adaptation. When the loss decreases rapidly, a positive value adjustment is used to increase the window to smooth noise and fully explore the optimization space. When the loss decreases slowly and the accuracy is insufficient, a negative value adjustment is used to reduce the window boosting response sensitivity and capture small optimization potential. In other cases, the window size is kept stable to balance performance. This design differs from the coarse mode of traditional methods that only output results after a single modeling. It ensures that the model continuously approaches the engineering accuracy requirements. At the same time, the deviation rate uses relative error logic to achieve unified accuracy judgment across scenarios. Combined with safety redundancy verification, it directly avoids the risk of insulation breakdown and waste of equipment resources, reducing the difficulty of engineering implementation and operational risks. Attached Figure Description

[0019] Figure 1 This is a flowchart of the high-altitude power grid insulation gap compensation method based on gradient improvement proposed in Embodiment 1 of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0022] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0023] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0024] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0025] Example 1: See Figure 1 This embodiment proposes a gradient-based method for compensating insulation gaps in high-altitude power grids, comprising the following steps: S100: Collect real-time environmental parameters and measured insulation gap values ​​of power grids in high-altitude areas; The real-time environmental parameters include altitude, air pressure, temperature, humidity, and ultraviolet radiation intensity.

[0026] S200. Remove outliers from the real-time environmental parameters, fill in missing values ​​using interpolation, and perform normalization to obtain preprocessed real-time environmental parameters, which constitute a parameter set. The parameter set includes altitude, and at least two of air pressure, temperature, humidity, and ultraviolet intensity; It should be noted that the insulation strength of the insulation gap is not determined by a single real-time environmental parameter, but is dominated by altitude and influenced by the synergistic coupling of multiple real-time environmental parameters. By constraining the feature dimension and the integrity of the sample set, a data foundation is provided for the entire process of two-layer modeling, dynamic iteration and accuracy verification in the scheme. This ensures that the input features can cover the core influencing factors of high-altitude insulation gaps, and also guarantees the effective use and distribution consistency of sample data.

[0027] Furthermore, the parameter set is divided into a training set and a validation set using stratified sampling. The number of samples in the parameter set is The number of training set samples is The number of samples in the validation set is And satisfy ; Preprocessed The samples are arranged in rows, and for each sample, features including altitude and two or more other environmental parameters are extracted and arranged in columns to construct the feature matrix of the training set, denoted as . .

[0028] S300, the feature matrix of the training set Input the preset standard insulation correction model and output the reference insulation gap value; The standard insulation correction model utilizes the standard empirical formulas established in the IEC-60071 standard. It should be noted that the traditional formula can already cover the "linear principal effect" of high-altitude insulation gaps. The error between the fitted baseline insulation gap value and the true value is usually small. Gradient boosting does not need to learn the full mapping from zero. It only needs to focus on the "nonlinear coupling effect" not covered by the baseline insulation gap value, such as the synergistic effect of high altitude, low air pressure and high humidity, which greatly reduces the number of base learners and iteration rounds required. If the initial predicted value deviates too much from the true value, the gradient boosting may cause subsequent corrections to oscillate due to the large error, making it difficult to converge to the global optimum. However, the baseline insulation gap value, as a high-precision initial value, has a more concentrated error vector distribution and smaller fluctuations, enabling the base learner to accurately fit the error pattern and making the iteration process more stable. Traditional formulas have been validated in engineering projects at different altitudes and voltage levels, demonstrating strong adaptability to extreme scenarios and making them less prone to performance collapse due to missing local samples. The reference insulation gap serves as a benchmark, ensuring that the model can still output minimum accuracy in edge scenarios not covered by the training set, thus avoiding the problem of insufficient generalization ability of pure machine learning models.

[0029] S400. Construct a gradient boosting model, using the baseline insulation gap value as the initial predicted value for the gradient boosting model, denoted as... ; The sum of squared insulation gap prediction errors is calculated based on the baseline insulation gap value and used as the loss function value of the gradient boosting optimization model, expressed by the formula: ; In the formula, This represents the sum of squares of the predicted insulation gap error. Represents the first predicted value One sample, The first value represents the measured insulation gap value. One sample, Indicates the number of samples in the parameter set; Regularization strategies are employed to suppress overfitting in gradient boosting models, based on the loss function. and the feature matrix of the training set By fitting the predicted error output, the optimized correction amount for the insulation gap is obtained, expressed by the formula: ; ; In the formula, Indicates the first The insulation gap optimization correction amount in the round iteration. This represents the fitting function of the base learner in the gradient boosting model. The feature matrix of the training set is represented by... Indicates the first The regularization coefficient of the round iteration, Indicates the first Prediction error after rounds of iteration This indicates the total correction amount. This represents the first measured insulation gap value in the verification set. One sample; Furthermore, the loss value is calculated based on the baseline insulation gap value and the optimized correction amount for the insulation gap, expressed by the formula: ; In the formula, Indicates the first The loss value of the validation set in each iteration. The first value represents the predicted insulation gap value in the verification set. One sample, The first value represents the measured insulation gap value in the training set. One sample; The magnitude of the decrease in the loss value is calculated based on the loss value, expressed by the formula: ; In the formula, Represents the statistics window The decrease in the lower loss value Represents a statistics window. Indicates the first The loss value of the validation set in each iteration. Indicates the first The loss value of the validation set in each iteration; Determine the loss value on the validation set. With preset convergence threshold The size relationship and statistical window In the middle, the decrease in the loss value With preset amplitude The size relationship, specifically: a. When ,and Meanwhile, the gradient boosting model continues to iterate, and the regularization strategy remains unchanged to avoid prematurely adjusting the regularization constraint on the model's optimization space and strive to further reduce the error. b. When ,and Meanwhile, the gradient boosting model continues to iterate, the regularization strategy remains unchanged, and the optimization space is fully utilized to approach the convergence threshold. c. When ,and Meanwhile, the gradient boosting model continues to iterate, adjusting the regularization strategy to suppress overfitting while maintaining accuracy; In this embodiment, the adjustment regularization strategy includes either a single-base learner weight upper limit constraint or a base learner number constraint, wherein: The upper limit constraint on the weights of a single-base learner is expressed by the formula: ; In the formula, Indicates the base learner's first The upper limit of weights in each iteration. Indicates the base learner's first The initial weights for each iteration. Indicates the upper limit threshold of the weight. This represents the weight decay coefficient; Based on the magnitude of the loss decrease, a feature complexity factor is introduced to adjust the weight decay coefficient in the upper limit constraint of the single-base learner weights in real time, expressed by the formula: ; ; In the formula, Indicates the first The weight decay coefficient after each iteration. Indicates the first The weight decay coefficient of each iteration. Indicates the preset amplitude. Represents the feature complexity factor. express; Indicates the magnitude of the decrease in the loss value; It should be noted that by limiting the contribution of each base learner to the total correction, the overfitting risk of the gradient boosting algorithm is suppressed, ensuring the stability of the iterative process and the safety of engineering compensation, avoiding a single base learner dominating the optimization direction, and balancing the model's optimization and generalization capabilities; simultaneously, the weight decay coefficient... The dynamic change is an adaptive adjustment mechanism for the weight constraints of the single-base learner. By combining the decrease in the loss value with the feature complexity factor, the intensity of the weight decay is adjusted in real time, so that the weight constraints are upgraded from fixed rules to dynamic adaptation. This ensures the model's performance in the optimization stage and suppresses overfitting in a timely manner in the stable stage.

[0030] The constraint on the number of base learners is expressed by the formula: ; In the formula, Indicates the first The maximum number of base cells allowed in a round of iteration. Indicates the first The maximum number of base cells allowed in a round of iteration. This represents the upper limit of the number of base learners. Indicates the percentage decrease in quantity; The change in the decay ratio of the base learner quantity constraint function is expressed by the following formula: ; ; ; In the formula, Indicates the first The rate of quantity decay after each iteration Indicates the first The current quantity decay rate in each iteration. This represents the difference in loss between the training set and the validation set. Indicates the benchmark of the loss mean. Indicates the sample size factor. Indicates the first The loss value of the validation set in each iteration. Indicates the first The loss value of the training set in each iteration; It should be noted that by dynamically limiting the maximum number of base learners allowed in the gradient boosting algorithm, the overall complexity of the model is controlled, suppressing the risk of overfitting from the total quantity dimension, while ensuring iterative efficiency and real-time engineering performance. This complements the single base learner weight constraint, together forming a double regularization barrier for gradient boosting optimization, adapting to the needs of dynamic compensation for high-altitude insulation. At the same time, by combining the loss difference between the training set and the validation set and the sample size factor, the decay intensity of the number of base learners is adjusted in real time, upgrading the quantity constraint from fixed decay to precise adaptation. This ensures the learning ability of the model during the optimization phase and tightens the constraint in time when the risk of overfitting occurs.

[0031] d. When ,and When the gradient boosting model stops iterating, it outputs the total correction amount, expressed by the formula: ; In the formula, This indicates the total correction amount. This indicates the iteration stop cycle.

[0032] S500. Combine the total correction amount with the reference insulation gap value to obtain the corrected insulation gap value, expressed by the formula: ; In the formula, This indicates the corrected insulation gap value. Indicates the reference insulation gap value. Indicates the total correction amount; The corrected insulation gap value is compared with the measured insulation gap value, and the deviation rate of the insulation gap value is output, expressed by the formula: ; In the formula, This indicates the deviation rate of the insulation gap value. This indicates the measured insulation gap value.

[0033] S501, Further, determine the deviation rate of the insulation gap value. With preset threshold Size, specifically: when When correcting the insulation gap value, please specify the corrected value. Less than or more than the measured insulation gap value If there is a risk of insulation breakdown or waste of equipment resources, a backtracking adjustment must be triggered to eliminate the safety hazard; backtrack to the gradient optimization model and adjust the statistical window and regularization strategy; It should be noted that the update function of the statistical window is a dynamic adaptive form, based on the rate of decrease of the loss value. Deviation rate from insulation gap value The feedback adjustment is expressed by the formula: ; In the formula, Indicates the first The updated statistics window size. Indicates the first The current statistical window size of the wheel, Represented as the first The window size is adjusted after the wheel is updated; Among them, the After the wheel is updated, the window size can be adjusted. Based on the magnitude of the loss value Deviation rate of insulation gap value Adjustments will be made, specifically: a. When ,and hour, ; It should be noted that increasing the statistical window size when the loss decreases rapidly can smooth out short-term noise and prevent the loss value from decreasing too much. To prevent noise interference from misjudging the decline as a slowdown, we must ensure that the model can fully explore the optimization space and does not stop iterating prematurely. b. When ,and hour, ; It should be noted that when the rate of loss reduction slows down but the accuracy of on-site data still needs optimization, reducing the size of the statistical window can increase the rate of loss reduction. The sensitivity of the statistical window is enhanced to detect minute drops in loss and avoid sluggish response due to an excessively large statistical window, thus missing the last opportunity for optimization. c. When other situations occur ; It should be noted that the statistical window size remains unchanged to balance sensitivity and stability and avoid fluctuations in judgment caused by frequent adjustments.

[0034] when If the error is within the safety redundancy range and can be directly used for dynamic compensation to ensure the safe operation of the power grid, then the corrected insulation gap value will be directly output. This is a result of high-altitude insulation compensation.

[0035] Example 2: This embodiment proposes a gradient-based high-altitude power grid insulation gap compensation system, which includes the following modules: Parameter acquisition module: Collects real-time environmental parameters and measured insulation gap values ​​of power grids in high-altitude areas; Preprocessing module: preprocesses real-time environmental parameters to construct a parameter set; uses stratified sampling to divide the parameter set into a training set and a validation set, and extracts the feature matrix of the training set; The residual calculation module inputs the feature matrix into the preset standard insulation correction model and outputs the reference insulation gap value; it calculates the deviation between the reference insulation gap value and the measured insulation gap value to obtain the insulation gap residual. Total correction calculation module: Constructs a gradient boosting optimization model, takes the baseline insulation gap value as the initial value, and performs iterative training with the goal of correcting the insulation gap residual; when the preset convergence condition is met, iterative training stops and the total correction is output; The compensation result output module combines the total correction amount and the reference insulation gap value to obtain the corrected insulation gap value; it compares the corrected insulation gap value with the measured insulation gap value and outputs the deviation rate of the insulation gap value; it judges the magnitude of the deviation rate and the preset threshold. When the deviation rate is less than the preset engineering accuracy threshold, it outputs the corrected insulation gap value as the high-altitude insulation compensation result.

[0036] Example 3: This embodiment proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the method described in any embodiment of the present invention.

[0037] Example 4: This embodiment proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in any embodiment of the present invention.

[0038] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0039] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0040] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0041] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0042] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for compensating insulation gaps in high-altitude power grids based on gradient improvement, characterized in that, Includes the following steps: Collect real-time environmental parameters and measured insulation gap values ​​of power grids in high-altitude areas; The real-time environmental parameters are preprocessed to construct a parameter set; the parameter set is divided into a training set and a validation set using stratified sampling, and the feature matrix of the training set is extracted. Input the feature matrix into the preset standard insulation correction model and output the reference insulation gap value; The deviation between the reference insulation gap value and the measured insulation gap value is calculated to obtain the insulation gap residual; The pre-trained gradient retrieval optimization model is iteratively trained with the baseline insulation gap value as the initial value and the goal of correcting the insulation gap residual. When the preset convergence condition is met, stop iterative training and output the total correction amount; The total correction amount and the reference insulation gap value are combined to obtain the corrected insulation gap value; The corrected insulation gap value is compared with the measured insulation gap value, and the deviation rate of the insulation gap value is output. The deviation rate is judged to be smaller than the preset threshold. When the deviation rate is less than the preset engineering accuracy threshold, the corrected insulation gap value is output as the high-altitude insulation compensation result.

2. The high-altitude power grid insulation gap compensation method based on gradient improvement according to claim 1, characterized in that, The parameter set includes altitude, and at least two of air pressure, temperature, humidity, and ultraviolet radiation intensity.

3. The high-altitude power grid insulation gap compensation method based on gradient improvement according to claim 1, characterized in that, The preset standard insulation correction model adopts the standard empirical formulas established in the IEC-60071 standard.

4. The method for compensating insulation gaps in high-altitude power grids based on gradient improvement according to claim 1, characterized in that, The gradient boosting optimization model is trained according to the following steps: The reference insulation gap value is set as the initial predicted value of the gradient boosting optimization model; The sum of squared residuals of the insulation gap prediction is defined as the loss function of the model; the sum of squared residuals of the insulation gap prediction is the sum of the squared deviations between the initial predicted value and the measured insulation gap, and a regularization strategy is configured. The training set is input into the gradient boosting optimization model. The prediction residual is fitted based on the loss function to calculate the first round of insulation gap optimization correction. This correction is then added to the initial prediction value to obtain the prediction value after the first round of iteration. After each round of iteration, the loss value of the validation set is calculated based on the prediction value of the current round and the measured insulation gap value of the validation set. The loss value of the validation set is the sum of squares of the deviations between the current round's validation set prediction value and the corresponding measured insulation gap value. Calculate the decrease in the validation set loss value within the preset statistical window. The loss value of the current round of validation set is compared with the preset convergence threshold, and the decrease rate of the loss value within the statistical window is compared with the preset rate. Based on the comparison results, iterative operations are performed.

5. The high-altitude power grid insulation gap compensation method based on gradient improvement according to claim 4, characterized in that, The iterative operation based on the comparison results specifically includes the following steps: If the validation set loss value is greater than the preset convergence threshold and the decrease in loss value is greater than the preset magnitude, continue to the next iteration; at the same time, the regularization strategy remains unchanged, and the prediction residual is fitted based on the loss function of the current round and the feature matrix of the training set, the new round of insulation gap optimization correction is calculated and accumulated to the current prediction value; If the validation set loss value is greater than the preset convergence threshold and the decrease in the loss value is less than or equal to the preset magnitude, continue to the next iteration; at the same time, the regularization strategy remains unchanged, and the fitting correction amount, the accumulated prediction value, the validation set loss value and the decrease magnitude are repeatedly calculated. If the validation set loss value is less than or equal to the preset convergence threshold, and the decrease in loss value is greater than the preset magnitude, continue to the next iteration; at the same time, adjust the regularization strategy, fit the prediction residual based on the adjusted regularization strategy, the current loss function and the training set feature matrix, calculate the new round of insulation gap optimization correction amount and accumulate it to the current prediction value, and recalculate the validation set loss value and the magnitude of decrease. If the validation set loss value is less than or equal to the preset convergence threshold and the decrease in loss value is less than or equal to the preset magnitude, stop iterative training; accumulate the insulation gap optimization correction amount of all rounds in the iteration process to obtain the total correction amount.

6. The high-altitude power grid insulation gap compensation method based on gradient improvement according to claim 4, characterized in that, The regularization strategy is selected from one of the following: upper limit constraint on the weights of a single base learner and constraint on the number of base learners, wherein: The upper limit constraint on the weights of the single-base learner is expressed by the formula: ； In the formula, Indicates the base learner's first The upper limit of weights in each iteration. Indicates the base learner's first The initial weights for each iteration. Indicates the upper limit threshold of the weight. This represents the weight decay coefficient; The constraint on the number of base learners is expressed by the formula: ； In the formula, Indicates the first The maximum number of base cells allowed in a round of iteration. Indicates the first The maximum number of base cells allowed in a round of iteration. This represents the upper limit of the number of base learners. This indicates the percentage decrease in quantity.

7. The high-altitude power grid insulation gap compensation method based on gradient improvement according to claim 4, characterized in that, The deviation rate is compared to a preset threshold. When the deviation rate of the insulation gap value is greater than or equal to the preset engineering accuracy threshold, a backtracking adjustment is triggered, reverting to the gradient boosting optimization model and adjusting the statistical window. Specifically: When the decrease in the validation set loss value within the statistical window is greater than the preset range, and the deviation rate of the insulation gap value is less than the preset engineering accuracy threshold, the window adjustment step size is set to a positive value. When the decrease in the validation set loss value within the statistical window is less than or equal to the preset range, and the deviation rate of the insulation gap value is greater than or equal to the preset engineering accuracy threshold, the window adjustment step size is set to a negative value. In other cases, the window adjustment step size is set to zero, and the statistical window size remains unchanged. Based on the current statistical window size, the window adjustment step size of the corresponding round is added to obtain the updated statistical window size.

8. A high-altitude power grid insulation gap compensation system based on gradient improvement, characterized in that, Includes the following modules: Parameter acquisition module: Collects real-time environmental parameters and measured insulation gap values ​​of power grids in high-altitude areas; Preprocessing module: preprocesses real-time environmental parameters to construct a parameter set; uses stratified sampling to divide the parameter set into a training set and a validation set, and extracts the feature matrix of the training set; Residual calculation module: Input the feature matrix into the preset standard insulation correction model and output the reference insulation gap value; The deviation between the reference insulation gap value and the measured insulation gap value is calculated to obtain the insulation gap residual; Total correction calculation module: Constructs a gradient boosting optimization model, takes the baseline insulation gap value as the initial value, and performs iterative training with the goal of correcting the insulation gap residual; when the preset convergence condition is met, iterative training stops and the total correction is output; Compensation result output module: Combines the total correction amount and the reference insulation gap value to obtain the corrected insulation gap value; The corrected insulation gap value is compared with the measured insulation gap value, and the deviation rate of the insulation gap value is output. The deviation rate is judged to be smaller than the preset threshold. When the deviation rate is less than the preset engineering accuracy threshold, the corrected insulation gap value is output as the high-altitude insulation compensation result.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 7.

Images