A power distribution network digitalization level comprehensive evaluation method, system and device based on multi-index weighting and a storage medium

Abstract

The application discloses a kind of based on multi-index weighted distribution network digitization level comprehensive evaluation method, system, equipment and storage medium, it is related to digital evaluation technical field, method includes: obtaining multi-source heterogeneous data generation time sequence fingerprint and multi-index vector;Based on data integrity rate and time delay dynamic correction index weight;To the decoupling pruning of multi-index vector obtains independent index set;Calculate conflict compensation weight between indexes;Combination weight is generated and is fused to construct weighted decision matrix, finally output distribution network digitization level quantitative evaluation report;The application realizes the efficient compression and standardization characterization of massive heterogeneous data, reduces the calculation pressure, eliminates the evaluation distortion caused by low-quality data, solves the index repeated scoring problem, provides scientific, robust decision support for distribution network resource allocation.

Description

Technical Field

[0001] This invention relates to the field of digital evaluation technology, and in particular to a method, system, equipment and storage medium for comprehensive evaluation of the digitalization level of distribution networks based on multi-index weighting. Background Technology

[0002] With the construction of new power systems, distribution networks are transforming from traditional automation to full digitalization. Establishing a scientific evaluation system is crucial for quantifying the effectiveness of construction. Existing technologies have incorporated combined models such as the analytic hierarchy process (AHP) and entropy weight method, but digital evaluation of distribution networks still faces serious systemic bottlenecks in practical applications: traditional systems struggle to handle massive amounts of time-series data generated by tens of millions of measuring points, leading to lags in real-time calculation and storage, and failing to support minute-level dynamic indicator evaluation; simultaneously, due to the wide distribution and complex environment of end-point devices, data acquisition often faces issues of incompleteness and latency, and this blurred vision can easily lead to evaluation distortion in the absence of a reliable identification mechanism; furthermore, there is a high degree of correlation among digital indicators in distribution networks, and traditional methods often ignore indicator redundancy and conflicts, resulting in duplicate scoring, making it difficult to accurately identify weak links in the transformation process. Summary of the Invention

[0003] In view of the above-mentioned problems, the present invention provides a method, system, equipment and storage medium for comprehensive evaluation of the digitalization level of distribution networks based on multi-index weighting.

[0004] Therefore, the technical problem solved by this invention is that existing digital evaluation methods for power distribution networks are difficult to efficiently process massive amounts of heterogeneous time-series data, are easily affected by data quality defects leading to evaluation distortion, and cannot effectively eliminate redundant correlations between indicators, resulting in duplicate scoring.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting, including: Acquire multi-source heterogeneous time-series data from all measurement points in the power distribution network, and use a symbolic aggregation algorithm to convert the multi-source heterogeneous time-series data into a sequence composed of discrete symbols to generate a time-series fingerprint. Based on the time-series fingerprint, generate a multi-index vector composed of feature components. The integrity rate and response delay of multi-source heterogeneous time-series data are monitored to generate confidence factors; the weights of the digital evaluation indicators of the distribution network are dynamically corrected using the confidence factors to obtain the corrected indicator weights. The multi-indicator vectors are decoupled and pruned to generate a decoupled index set. Calculate the contrast intensity and conflict coefficient of each indicator in the decoupled indicator set, and generate conflict compensation weights; The combined weights are generated by integrating the corrected indicator weights and the conflict compensation weights. The weighted decision matrix is ​​constructed by combining the multi-indicator vector and the decoupled indicator set, and finally a quantitative evaluation report on the digitalization level of the distribution network is generated.

[0006] As a preferred scheme for a comprehensive evaluation method of the digitalization level of distribution networks based on multi-index weighting, wherein: The process involves acquiring multi-source heterogeneous time-series data from measurement points across the entire distribution network, converting this data into a sequence of discrete symbols using a symbolic aggregation algorithm to generate a time-series fingerprint, and then generating a multi-index vector composed of feature components based on the time-series fingerprint. Collect multi-source heterogeneous time-series data reflecting the physical operation, user electricity consumption, distributed energy and communication status of the entire power distribution network, and construct four evaluation dimensions based on the data sources: clean and low-carbon, safe and reliable, efficient and flexible, and digital and intelligent. The collected time-series data are normalized and segmented and aggregated. The aggregated mean is mapped to discrete symbols using equal probability intervals based on Gaussian distribution, and time-series fingerprints are generated by arranging them in time sequence. The frequency of symbols in the statistical time-series fingerprint is used to generate frequency features, the information entropy is calculated to generate complexity features, the transition probability of adjacent symbols is extracted to generate dynamic association features, the three types of features are concatenated and mapped to the corresponding evaluation dimensions according to the data source type to obtain a multi-index vector.

[0007] As a preferred scheme for a comprehensive evaluation method of the digitalization level of distribution networks based on multi-index weighting, wherein: The completeness and response delay of the monitored multi-source heterogeneous time-series data are used to generate confidence factors. These confidence factors are then used to dynamically adjust the weights of the distribution network digitization evaluation indicators, resulting in the adjusted indicator weights, including: Real-time monitoring of the integrity rate and response delay of multi-source heterogeneous time-series data, wherein the integrity rate is the ratio of the actual number of valid data points received to the theoretical number of data points that should be received, and the response delay is the time difference between the data being sent from the test point and entering the evaluation system; The integrity rate is mapped to the first intermediate factor using the S-shaped growth function, and the response delay is mapped to the second intermediate factor using the negative exponential decay function.

[0008] As a preferred scheme for a comprehensive evaluation method of the digitalization level of distribution networks based on multi-index weighting, wherein: The monitoring of the completeness and response delay of multi-source heterogeneous time-series data generates a confidence factor; the confidence factor is used to dynamically adjust the weights of the distribution network digitalization evaluation indicators to obtain the adjusted indicator weights, which also includes: Multiply the first intermediate factor by the second intermediate factor to generate the confidence factor; The initial weights of each evaluation indicator are obtained. After multiplying the initial weights by the corresponding confidence factors and normalizing them, the dynamically adjusted indicator weights are obtained.

[0009] As a preferred scheme for a comprehensive evaluation method of the digitalization level of distribution networks based on multi-index weighting, wherein: The process of decoupling and pruning multiple indicator vectors to generate a decoupled indicator set includes: Multi-index vectors are mapped to a high-dimensional space to represent the nonlinear interaction relationships between different dimensions. The degree of correlation between each dimension is calculated by using a self-attention mechanism, and redundant dimensions are identified and projected onto an orthogonal subspace for decoupling. Assign a learnable gated mask to each decoupled dimension. After sparsification constraints, set the mask parameters of dimensions below a preset threshold to zero and multiply them by the vector to eliminate invalid dimensions. The original mapping vector is reconstructed by using the pruned vector. When the loss of reconstructed information meets the consistency requirement, a non-zero dimension index is output. The non-zero dimension index is used to determine the decoupled index set.

[0010] As a preferred scheme for a comprehensive evaluation method of the digitalization level of distribution networks based on multi-index weighting, wherein: The calculation of the contrast strength and conflict coefficient of each indicator within the decoupled indicator set, and the generation of conflict compensation weights, includes: Obtain the numerical sequence of each indicator in the decoupled indicator set within the evaluation period, and calculate the standard deviation of the numerical sequence of each indicator as the comparative strength of the corresponding indicator. Calculate the correlation coefficient between any two indicators in the set, and then sum the differences between the correlation coefficient of each indicator and all other indicators and the value of 1 to generate the conflict coefficient of the corresponding indicator. Multiply the contrast intensity of each indicator by the conflict coefficient to obtain the comprehensive information content of each indicator; The total information content of all indicators is normalized so that the sum of the weights of each indicator is 1, thus generating conflict compensation weights.

[0011] As a preferred scheme for a comprehensive evaluation method of the digitalization level of distribution networks based on multi-index weighting, wherein: The fused and corrected indicator weights and conflict compensation weights generate combined weights, which, combined with multi-indicator vectors and the decoupled indicator set, construct a weighted decision matrix, ultimately generating a quantitative evaluation report on the digitalization level of the distribution network, including: Extract the weight components corresponding to the non-zero dimension index from the corrected index weights, multiply and fuse them with the generated conflict compensation weights, and normalize them to obtain the combined weights. The feature components at corresponding positions are extracted from the multi-index vector using non-zero dimension indexes to construct the feature matrix to be evaluated; The combined weights are multiplied with the feature matrix to be evaluated to obtain the original scoring matrix, which is then normalized to generate a weighted decision matrix. Based on the weighted decision matrix, a quantitative evaluation report on the digitalization level of the distribution network is generated, which includes scores for each dimension, a comprehensive score, and evaluation improvement suggestions.

[0012] Secondly, the present invention provides a comprehensive evaluation system for the digitalization level of a distribution network based on multi-index weighting, comprising: The fingerprint construction module is used to acquire multi-source heterogeneous time-series data from measurement points across the entire power distribution network, and to convert the multi-source heterogeneous time-series data into a sequence composed of discrete symbols using a symbolic aggregation algorithm to generate a time-series fingerprint. Based on the time-series fingerprint, a multi-index vector composed of feature components is generated. The weight correction module is used to monitor the integrity rate and response delay of multi-source heterogeneous time-series data and generate confidence factors; the confidence factors are used to dynamically correct the weights of the digital evaluation indicators of the distribution network to obtain the corrected indicator weights. The decoupling and pruning module is used to decouple and prune multi-index vectors based on correlation, and generate a decoupled index set. The conflict compensation module is used to calculate the contrast strength and conflict coefficient of each indicator in the decoupled indicator set and generate conflict compensation weights. The integrated evaluation module is used to generate combined weights by integrating the corrected index weights and conflict compensation weights, and to construct a weighted decision matrix by combining multiple index vectors and the decoupled index set, ultimately generating a quantitative evaluation report on the digitalization level of the distribution network.

[0013] Thirdly, the present invention provides a computer device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of a comprehensive evaluation method for the digitalization level of distribution networks based on multi-index weighting.

[0014] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of a comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting.

[0015] The beneficial effects of this invention are as follows: By introducing a symbolic aggregation algorithm, this invention transforms massive amounts of multi-source heterogeneous time-series data from the distribution network into distinctive discrete symbol sequences, i.e., time-series fingerprints. This processing method greatly reduces the storage and computational pressure on the system for tens of millions of measurement points' raw data, effectively solving the problem of real-time evaluation lag caused by excessive data volume, and ensuring the efficiency and timeliness of the evaluation process. By establishing a confidence evaluation mechanism based on data integrity rate and response delay, the monitored communication quality is dynamically mapped to correction factors using an S-shaped growth function and a negative exponential decay function, and the initial weights of various indicators are adjusted accordingly. This directly addresses the complex environment of end-point equipment and the challenges of data acquisition. The "blurred vision" pain point caused by incomplete or delayed data enables the assessment system to adaptively identify and mitigate the negative impact of low-quality data, ensuring the objectivity and robustness of the quantitative report on digitalization level. By employing a sparse decoupled neural network to perform spatial mapping and conflict compensation processing on multi-indicator vectors, and utilizing self-attention mechanisms and mask pruning techniques to identify and remove redundant dimensions between indicators, the system overcomes the defects of highly correlated digital indicators that easily lead to duplicate scoring. Through in-depth "slimming" and conflict correction of the indicator set, the final evaluation results can more clearly reveal the true state of the distribution network during the transformation process, making it easier for managers to identify weak links in digitalization construction. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is an overall flowchart of a comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting, provided by the present invention. Detailed Implementation

[0018] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0019] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting, including: S1: Acquire multi-source heterogeneous time-series data from all measurement points in the distribution network, and use a symbolic aggregation algorithm to convert the multi-source heterogeneous time-series data into a sequence composed of discrete symbols to generate a time-series fingerprint. Based on the time-series fingerprint, generate a multi-index vector composed of feature components. S2: Monitor the completeness and response delay of multi-source heterogeneous time-series data to generate confidence factors; use the confidence factors to dynamically correct the weights of the digital evaluation indicators of the distribution network to obtain the corrected indicator weights; S3: Perform correlation decoupling and pruning on the multi-indicator vectors to generate a decoupled indicator set; S4: Calculate the contrast strength and conflict coefficient of each indicator in the decoupled indicator set, and generate conflict compensation weights. S5: The combined weights are generated by integrating the corrected indicator weights and the conflict compensation weights. The weighted decision matrix is ​​constructed by combining the multi-indicator vector and the decoupled indicator set, and finally a quantitative evaluation report on the digitalization level of the distribution network is generated.

[0020] It should be noted that, through steps S1-S5, this invention constructs a complete evaluation chain from data preprocessing, quality adaptive weighting, indicator redundancy elimination to conflict compensation fusion, realizing efficient compression and standardized representation of multi-source heterogeneous data of the distribution network. It can dynamically identify and weaken the interference of low-quality data on the evaluation results, fundamentally eliminate the problem of duplicate scoring caused by high correlation between indicators, and finally output quantitative scores and decision suggestions that take into account objectivity and independence, providing scientific and robust support for the evaluation of the effectiveness of digital transformation of the distribution network and the optimal allocation of resources.

[0021] Example 2, refer to Figure 1 As an embodiment of the present invention, based on the previous embodiment, a comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting is provided, including: In this embodiment, step S1 above involves acquiring multi-source heterogeneous time-series data from all measurement points in the distribution network, and using a symbolic aggregation algorithm to convert the multi-source heterogeneous time-series data into a sequence composed of discrete symbols to generate a time-series fingerprint. Generating a multi-index vector composed of feature components based on the time-series fingerprint includes: Acquire multi-source heterogeneous time-series data from measurement points across the entire distribution network, including electrical quantity time-series data reflecting the physical operating status of the power grid, metering data reflecting user-side electricity consumption behavior and green electricity consumption, monitoring data reflecting the operation of distributed power sources and energy storage systems, and communication performance and terminal online status data reflecting the operational quality of digital infrastructure; the digital evaluation indicators for the distribution network include clean and low-carbon indicators, safety and reliability indicators, high efficiency and flexibility indicators, and digital intelligence indicators.

[0022] Specifically, basic data collection is achieved through sensor terminals and monitoring equipment deployed throughout the distribution network. In practice, distribution terminals, feeder terminals, and substation terminals are used to collect real-time electrical quantity time-series data reflecting the physical operating status of the power grid. These data include parameters such as voltage amplitude, current magnitude, active power, reactive power, and system frequency at each monitoring node. The data is collected at a millisecond-level sampling frequency to form a basic time series reflecting the fluctuation characteristics of the distribution network operation.

[0023] To acquire user-side data, metering data reflecting user electricity consumption behavior and green electricity consumption is collected through smart meters and advanced measurement systems. During implementation, total electricity consumption and time-of-use electricity consumption data are extracted every 15 minutes or 1 hour, with a focus on tracking users' subscription and actual consumption of renewable energy sources such as photovoltaics and wind power. For the operation of distributed power sources and energy storage systems, monitoring data is obtained through inverter monitoring interfaces and energy storage management systems. Real-time recording is maintained of distributed photovoltaic output power, instantaneous wind power generation, and energy storage battery charging and discharging power, remaining charge status, cycle count, and other operational data. Simultaneously, communication performance and terminal online status data are extracted from the communication network management system and terminal management platform. In practice, signal strength, data packet transmission success rate, end-to-end average communication latency, and packet loss rate are recorded at each data collection point. Furthermore, real-time statistics are maintained on the online rate, offline frequency, and number of abnormal alarms for hardware devices such as distribution automation terminals and smart meters. After completing the acquisition of multi-source heterogeneous time-series data, the acquired data is normalized and classified to construct four dimensions of digital evaluation indicators: clean and low-carbon indicators using distributed power output and green electricity consumption data; safety and reliability indicators using electrical quantity time-series data and equipment fault alarm records; high efficiency and flexibility indicators using energy storage operation data and load response data; and digital intelligence indicators using communication performance and terminal online status data. These four types of indicators serve as the top-level framework of the evaluation system.

[0024] It should be noted that by deeply integrating multi-source data such as power grid physical operation, user-side electricity consumption behavior, distributed energy dynamics, and communication infrastructure performance, the limitations of traditional evaluation methods that only focus on power grid physical performance or single business indicators are overcome, ensuring the comprehensiveness and timeliness of the evaluation data sources. At the same time, by constructing an indicator system covering four dimensions—clean and low-carbon, safe and reliable, efficient and flexible, and digital and intelligent—the degree to which digital technology empowers energy structure optimization and power grid operation efficiency is scientifically reflected.

[0025] A symbolic aggregation algorithm is used to convert multi-source heterogeneous time-series data into a sequence composed of discrete symbols, generating a time-series fingerprint; Specifically, the acquired multi-source heterogeneous time series data is normalized, divided into several equal-length sub-intervals according to a preset time window, and the mean of the data in each sub-interval is calculated to generate a segmented aggregated approximate sequence. A symbol set containing discrete levels is preset, and a probability interval threshold corresponding to each discrete symbol is established based on Gaussian distribution. The mean of the segmented aggregated approximate sequence is mapped to the probability interval threshold and converted into the corresponding discrete symbols. The symbols are then arranged in chronological order to generate a time-series fingerprint.

[0026] The normalization process uses the zero-mean normalization method, which subtracts the arithmetic mean of each original time series data from the series and then divides it by the standard deviation of the series, thereby converting all data into a standard normal distribution series with a mean of 0 and a standard deviation of 1. The normalized time-series data is divided into several consecutive and non-overlapping time windows along the time axis. For example, for a daily load curve containing 96 sampling points, if the sub-interval length is set to 8, it is divided into 12 sub-intervals.

[0027] Calculate the arithmetic mean of all sampling points in each sub-interval, compressing the original high-dimensional data into a numerical sequence consisting of 12 means; The number of levels in the symbol set is set; in this embodiment, it is set to 5 levels, corresponding to symbols 1, 2, 3, 4, and 5. Since the data has been processed into a standard normal distribution, the standard normal distribution table is used to find the dividing point under the probability density curve, so that the interval corresponding to each symbol is equal in probability. For example, the horizontal axis of the normal distribution is divided into 5 equal probability intervals, resulting in 4 specific numerical thresholds as judgment boundaries. These boundaries divide the continuous numerical space into non-overlapping discrete regions. The process of mapping the mean of the segmented aggregated approximate sequence to the probability interval threshold and converting it into corresponding discrete symbols involves comparing the calculated sub-interval mean with the preset probability interval threshold one by one. In specific implementation, it is determined which preset probability interval the mean falls into. For example, if the mean of a certain sub-interval is less than the first threshold (set to -0.84 in this embodiment), it is mapped to symbol 1; if it is between the first threshold and the second threshold (set to -0.25 in this embodiment), it is mapped to symbol 2, and so on. The mapped symbol sequence is combined into a string or a fixed-length vector to form the time series fingerprint of this segment of multi-source heterogeneous time series data. It should be noted that by using zero-mean processing and segmented aggregation, sampling noise is effectively filtered out while eliminating the influence of dimensions. Based on the equiprobability symbol mapping of the normal distribution, the time-series fingerprint can accurately depict the evolution of power grid operation in a simplified form, providing standardized feature input for the subsequent quantitative evaluation of digitalization level.

[0028] The frequency of various discrete symbols in the time-series fingerprint is counted, a symbol distribution histogram is constructed, and frequency feature components are generated; the symbol sequence information entropy of the time-series fingerprint is calculated, and complexity feature components are generated; the transition probability matrix between adjacent symbols in the time-series fingerprint is extracted, and dynamic correlation feature components are generated; the frequency feature components, complexity feature components and dynamic correlation feature components are concatenated and spliced ​​to generate a multi-index vector corresponding to the dimensions of clean and low-carbon, safe and reliable, efficient and flexible and digital intelligence indicators.

[0029] In practice, for the five preset discrete-level symbols, the total number of occurrences of symbol 1, symbol 2, symbol 3, symbol 4, and symbol 5 in the entire time-series fingerprint is counted. Then, the number of occurrences of each symbol is divided by the total length of the sequence to obtain the probability distribution of each symbol. Based on these probability values, a symbol distribution histogram is constructed, and the five probability values ​​are arranged in order of symbol level to form a 5-dimensional numerical vector. This vector represents the frequency feature component reflecting the static characteristics of the data distribution. Multiply the probability value of each symbol by its own logarithm, sum the products of all symbols, and take the negative value. The resulting value is the entropy value of the time series data, which is used to characterize the fluctuation complexity of the distribution network operation status. A sliding window with a step size of 1 is used to scan the temporal fingerprint, recording the transitions of all adjacent symbol pairs. For example, the specific number of transitions from symbol 1 to symbol 2 is counted and divided by the total number of occurrences of symbol 1 to calculate the conditional probability of transitioning from state 1 to state 2. By traversing all possible transition combinations, a 5x5 transition probability matrix is ​​constructed. The 25 probability elements in this matrix are extracted row-wise and arranged sequentially to generate dynamic correlation feature components reflecting the state evolution trend. The frequency feature components, complexity feature components, and dynamic correlation feature components are concatenated in a preset order to form a higher-dimensional feature vector. Based on the evaluation business requirements, the concatenated feature vector is mapped and aligned with four indicator dimensions: clean and low-carbon, safe and reliable, efficient and flexible, and digital intelligence. Specifically, for different types of data sources, the feature vector corresponding to electrical quantity data will be used as a component of the safe and reliable dimension, while the feature vector corresponding to communication status data will be used as a component of the digital intelligence dimension, generating a multi-indicator vector. It should be noted that by integrating static statistics, fluctuation entropy values, and dynamic evolution logic, a multi-dimensional and in-depth characterization of the distribution network's operational features has been achieved. Compared to traditional methods, this not only quantifies the uncertainty of operational states but also accurately identifies the temporal correlation patterns of state evolution, significantly improving the discriminative power of feature components. By scientifically mapping high-dimensional technical features with business evaluation dimensions, a data foundation with strong logical support is provided for the accurate quantification of digitalization levels.

[0030] In this embodiment, step S2 above monitors the completeness and response delay of multi-source heterogeneous time-series data to generate a confidence factor; the confidence factor is used to dynamically correct the weights of the distribution network digitalization evaluation indicators, resulting in the corrected indicator weights, which include: The integrity rate and response delay of multi-source heterogeneous time-series data acquired in real time are monitored. The integrity rate is defined as the ratio of the number of valid data points actually received to the theoretical number of data points that should be received within a preset period. The response delay is defined as the time difference between the data being sent from the test point and entering the evaluation method. The S-shaped growth function is called to map the monitored integrity rate to the first intermediate factor; the negative exponential decay function is called to map the monitored response delay to the second intermediate factor; the first intermediate factor and the second intermediate factor are multiplied to generate the confidence factor corresponding to each digital evaluation index of the distribution network. Obtain the initial weights of each evaluation indicator, multiply the initial weights by the confidence factor to obtain the corrected intermediate variables, and generate the corrected indicator weights.

[0031] Specifically, the quality of multi-source heterogeneous time-series data acquired in real time is monitored, including two dimensions: completeness rate and response latency. In this embodiment, it is preferred to set one hour as a cycle and count the number of valid data points actually received by each monitoring point within the cycle. The number of points that should theoretically be received is calculated based on the sampling frequency of the sensing device, and the completeness rate is obtained by dividing the actual number of points by the theoretical number of points.

[0032] At the same time, the original timestamp of the data message when it was sent at the measurement point is extracted, and the time when it arrives at the evaluation platform is recorded. The time difference between the two is calculated as the response delay. For each type of evaluation indicator, the corresponding integrity rate value and response delay value are calculated respectively for the specific data source on which it depends. The S-shaped growth function is invoked to map the monitored completeness rate to the first intermediate factor. The S-shaped growth function maps the completeness rate between 0 and 1 to a score space between 0 and 1; when the completeness rate is low, the first intermediate factor output by the function grows slowly and approaches 0; when the completeness rate reaches a preset threshold, the output value increases rapidly; when the completeness rate is close to 1, the output value tends to stabilize and approach 1; the negative exponential decay function is invoked to map the monitored response delay to the second intermediate factor.

[0033] It should be noted that the negative exponential decay function takes response delay as its independent variable, and its output value decreases exponentially as the delay increases. When the response delay is extremely small, the second intermediate factor of the function output is close to 1, indicating excellent real-time performance of the data; as the delay gradually increases, the output value decreases rapidly.

[0034] After obtaining the first and second intermediate factors, they are multiplied to generate the confidence factors for each digital evaluation index of the power distribution network. Since the values ​​of the two intermediate factors are both between 0 and 1, the generated confidence factors are also between 0 and 1. The confidence factors comprehensively reflect the completeness and real-time nature of the data source: the confidence factor approaches 1 only when the data is both complete and timely; if any of the indicators performs poorly, the confidence factor will decrease significantly, thereby reducing the interference of low-quality data on the evaluation results at the source. The initial weights of each evaluation indicator are obtained, and then multiplied by the corresponding confidence factors. These initial weights are typically pre-set values ​​based on historical experience. The corrected intermediate variables are obtained through multiplication. To ensure that the sum of all indicator weights still meets the normalization requirements, all corrected intermediate variables are summed, and each intermediate variable is divided by this sum to generate the corrected indicator weights. It should be noted that by introducing a dynamic weight correction mechanism based on multi-source data quality, the nonlinear impact of data integrity rate and response delay on evaluation credibility is accurately characterized by the S-shaped growth function and negative exponential decay function. This design ensures that under actual working conditions such as sensor failure, signal packet loss, or communication delay, the evaluation system can automatically identify and weaken the weight contribution of low-confidence data, so that the quantitative conclusions are always anchored to high-quality core data sources. This effectively avoids evaluation distortion caused by defects in the underlying data and provides adaptive dynamic support for the scientific assessment of the digitalization level of the distribution network.

[0035] In this embodiment, step S3 above involves decoupling and pruning the multi-index vectors to generate a decoupled index set, including: A sparse decoupling neural network is used to perform correlation decoupling and pruning on multi-index vectors; The sparse decoupling neural network includes: a feature interaction mapping layer, which uses a fully connected perceptron structure to spatially map multi-index vectors to generate hidden layer vectors; a self-attention decoupling layer, which uses a multi-head self-attention mechanism to calculate the correlation matrix between each dimension within the hidden layer vectors and identify redundant dimensions, mapping the hidden layer vectors to an orthogonal subspace to generate decoupled multi-index vectors; a differentiable mask pruning layer, which matches gate mask parameters to the decoupled multi-index vectors, performs sparsification constraints and sets them to zero, and then performs a dot product operation with the decoupled multi-index vectors to generate pruned multi-index vectors; and a reconstruction verification output layer, which reconstructs and verifies the pruned multi-index vectors, outputting non-zero dimension indices when the information consistency requirement is met, thus determining the decoupled index set.

[0036] Specifically, the feature interaction mapping layer takes the multi-index vector as the network input, and through the fully connected perceptron structure, uses the preset weight matrix and bias vector to perform a linear mapping transformation on the input multi-index vector. It also uses a non-linear activation function to reconstruct the space, transforming it from the original feature space to a high-dimensional hidden space, and generating a hidden layer vector that can represent the complex non-linear interaction relationship between the indicators. The implementation process of the self-attention decoupling layer involves inputting the generated hidden layer vector into a multi-head self-attention module; using multiple independent and parallel attention heads to calculate the degree of correlation between different dimensions within the hidden layer vector, thereby constructing a correlation matrix between dimensions; identifying redundant dimensions and highly coupled feature pairs by analyzing the numerical distribution of the correlation matrix; and projecting the hidden layer vector into a set of mutually orthogonal subspaces based on the singular value decomposition principle.

[0037] The implementation process of the differentiable mask pruning layer involves matching each dimension of the decoupled multi-index vector with a gated mask parameter between 0 and 1. In practice, sparsity penalty constraints are applied to these mask parameters during iterative training, forcing the mask parameters corresponding to dimensions that contribute very little to the evaluation results to shrink towards 0. When the mask parameter is lower than a preset minimum threshold (set to 0.01 in this embodiment), it is forcibly set to zero. The processed mask parameter is then multiplied element-wise with the decoupled multi-index vector to remove invalid dimensions and generate a pruned multi-index vector with reduced dimensionality. The implementation process of reconstructing and validating the output layer involves a closed-loop verification of the effectiveness of the pruned features. The original hidden layer vector is reconstructed using the pruned multi-index vector, and the information loss before and after reconstruction is compared. Under the premise of meeting a preset information consistency threshold (i.e., reconstruction accuracy higher than 95%), the position indices of non-zero numerical dimensions in the current vector are recorded and output. These indices are then used for rapid retrieval in the original feature set to ultimately determine the decoupled index set used for quantitative evaluation. It should be noted that by employing a sparse decoupled neural network, deep separation between indicator dimensions is achieved through feature interaction mapping and self-attention mechanism. Orthogonal projection technology is used to eliminate correlation interference and redundant information that are common in traditional indicator systems, fundamentally avoiding the phenomenon of repeated calculations in the evaluation process. Combined with differentiable mask pruning technology, the indicator dimensions can be significantly compressed while retaining core evaluation information, and interference items with minimal contribution to the results can be eliminated, thereby making the final set of indicators more representative and independent.

[0038] In this embodiment, the calculation of the contrast intensity and conflict coefficient of each index within the decoupled index set in step S4 above, and the generation of conflict compensation weights, includes: Calculate the contrast strength of each indicator in the decoupled indicator set. The contrast strength is the standard deviation of each indicator in the decoupled indicator set. Calculate the correlation coefficients between indicators in the decoupled indicator set, and sum the differences between the correlation coefficients of each indicator and other indicators and 1 to generate the conflict coefficient; The contrast intensity and the conflict coefficient are multiplied to obtain the comprehensive information content of each indicator; the comprehensive information content is normalized to generate conflict compensation weights. Specifically, obtain the numerical sequence of each indicator in the decoupled indicator set within the evaluation period; for each specific indicator, first calculate the arithmetic mean of the sequence, then calculate the square of the difference between each value and the mean, sum these squares and divide by the total number of samples, and finally obtain the standard deviation of the indicator sequence by taking the square root. The standard deviation is used as the contrast strength. Using linear correlation analysis, the degree of correlation between any two index sequences within the set is calculated; by dividing the covariance of the two index sequences by the product of their respective standard deviations, the correlation coefficient value between -1 and +1 is obtained. For a specific indicator in the set, its correlation coefficients with all other indicators are extracted one by one. The difference between 1 and these correlation coefficients is calculated, and all differences are summed. The resulting sum is the conflict coefficient of that indicator. The conflict coefficient measures the independence and inconsistency between the indicator and other indicators. The larger the sum, the stronger the conflict between the indicator and other indicators, and the more new information it provides in the evaluation system. The standard deviation of each indicator is multiplied by its corresponding conflict coefficient. The product is the comprehensive information content of the indicator under the current evaluation dimension. The comprehensive information content of all indicators in the decoupled indicator set is summed to obtain the total information content. The comprehensive information content of each indicator is divided by the total information content to make the sum of the conflict compensation weights of all indicators equal to 1. It should be noted that the standard deviation is used to objectively quantify the ability of each indicator to distinguish fluctuations among different evaluation objects, ensuring that indicators carrying rich discriminative information can obtain reasonable weight allocation. At the same time, through correlation analysis and conflict accumulation, feature components with high independence and low redundancy are effectively identified and amplified, overcoming the evaluation distortion caused by overlapping indicator information in traditional weighting methods. This compensation mechanism based on the statistical characteristics of the data itself makes the weight allocation more in line with the real distribution characteristics of the digitalization level of the distribution network.

[0039] In this embodiment, the combined weights generated in step S5 above, which integrate the corrected index weights and conflict compensation weights, are combined with the multi-index vector and the decoupled index set to construct a weighted decision matrix. The final quantitative evaluation report on the digitalization level of the distribution network includes: Extract the weight components corresponding to the non-zero dimension index from the corrected indicator weights, and multiply and fuse them with the conflict compensation weights and normalize them to generate combined weights; use the non-zero dimension index to extract the corresponding feature components from the multi-indicator vectors to construct the feature matrix to be evaluated; perform matrix multiplication on the combined weights and the feature matrix to be evaluated to obtain the original scoring matrix; normalize the original scoring matrix to generate a weighted decision matrix. Specifically, using the non-zero dimension index output by the sparse decoupled neural network, the weight values ​​of the indicators that have made significant contributions to the evaluation results and have been deredundant are accurately selected from the weights of various evaluation indicators that have been corrected by the confidence factor; the selected weight components are multiplied and fused with the conflict compensation weights, and the corresponding two weight values ​​are multiplied together to obtain an intermediate weight value that takes into account both data quality and information conflict. The product-fused values ​​are normalized to generate combined weights. In practice, the sum of all intermediate weights is calculated and each intermediate weight is divided by the sum to ensure that the sum of the combined weights of all indicators is equal to 1. The frequency features, complexity features and dynamic correlation features of the corresponding index positions in the original multi-indicator vector are extracted.

[0040] If the evaluation involves multiple time periods or multiple distribution areas, these feature components are arranged by row or column to form a multi-dimensional feature matrix. Each element in the feature matrix represents a quantitative representation of the distribution network in a specific digital dimension.

[0041] The combined weights are multiplied with the feature matrix to be evaluated to obtain the original scoring matrix. In practice, the combined weights are treated as a row vector and summed by dot product with the feature matrix to be evaluated. This process essentially weights and aggregates the various characteristics of the distribution network based on the importance and independent contribution of each indicator. The resulting original scoring matrix reflects the preliminary quantitative scores of the distribution network in various dimensions such as clean and low-carbon, safe and reliable, efficient and flexible, and digital intelligence. A standard scoring range is preset, such as 0 to 100 points. Based on the distribution range of the original scores, a linear transformation function is used to map them to this standard range. The resulting weighted decision matrix not only contains the specific scores of each dimension, but also the comprehensive numerical score obtained through matrix operations. It can intuitively reflect the differences in the level of digital development of different distribution network objects or the same object at different times. Based on a weighted decision matrix, a quantitative evaluation report on the digitalization level of the power distribution network is generated. Pre-set report templates are automatically invoked, transforming the values ​​in the weighted decision matrix into intuitive radar charts, bar charts, or trend curves. Simultaneously, based on the score's range (e.g., above 90 points is excellent, 80 to 90 points is good), corresponding evaluation conclusions and improvement suggestions are automatically generated. For example, if the digital intelligence dimension score is low, the report will automatically link to underlying communication performance data, pointing out specific communication weaknesses, thus completing a closed-loop evaluation process from raw time-series data to high-level decision-making recommendations.

[0042] It should be noted that by weight fusion and matrix aggregation, the evaluation process is automated and objective. In weighting, data quality and information independence are taken into account, effectively reducing subjective bias. Matrix operations are used to efficiently process high-dimensional features, ensuring the rigor of score mapping. The final report transforms complex scores into intuitive charts and improvement suggestions, which can identify weak links in digital construction and provide highly operable decision support for the optimized configuration and differentiated operation and maintenance of the power distribution network.

[0043] Example 3: The above is an illustrative scheme of a comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting, as described in this embodiment. It should be noted that the technical solution of a comprehensive evaluation system for the digitalization level of a distribution network based on multi-index weighting and the technical solution of the aforementioned comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting belong to the same concept. Details not described in detail in the technical solution of the comprehensive evaluation system for the digitalization level of a distribution network based on multi-index weighting in this embodiment can be found in the description of the aforementioned comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting.

[0044] This embodiment also provides a comprehensive evaluation system for the digitalization level of distribution networks based on multi-index weighting, including: The fingerprint construction module is used to acquire multi-source heterogeneous time-series data from measurement points across the entire power distribution network, and to convert the multi-source heterogeneous time-series data into a sequence composed of discrete symbols using a symbolic aggregation algorithm to generate a time-series fingerprint. Based on the time-series fingerprint, a multi-index vector composed of feature components is generated. The weight correction module is used to monitor the integrity rate and response delay of multi-source heterogeneous time-series data and generate confidence factors; the confidence factors are used to dynamically correct the weights of the digital evaluation indicators of the distribution network to obtain the corrected indicator weights. The decoupling and pruning module is used to decouple and prune multi-index vectors based on correlation, and generate a decoupled index set. The conflict compensation module is used to calculate the contrast strength and conflict coefficient of each indicator in the decoupled indicator set and generate conflict compensation weights. The integrated evaluation module is used to generate combined weights by integrating the corrected index weights and conflict compensation weights, and to construct a weighted decision matrix by combining multiple index vectors and the decoupled index set, ultimately generating a quantitative evaluation report on the digitalization level of the distribution network.

[0045] This embodiment also provides a computer device applicable to a comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement a comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting, as proposed in the above embodiments.

[0046] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting as proposed in the above embodiment.

[0047] The storage medium proposed in this embodiment belongs to the same inventive concept as the comprehensive evaluation method for the digitalization level of distribution networks based on multi-index weighting proposed in the above embodiment. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0048] Example 4 is an embodiment of the present invention, which provides a comprehensive evaluation method for the digitalization level of distribution networks based on multi-index weighting. In order to verify the beneficial effects of the present invention, a simulation experiment is conducted for scientific demonstration.

[0049] This embodiment applies the method of the present invention to a smart power distribution park in a city, referred to as Park A; In the digital evaluation of Park A, the evaluation system first obtains multi-source heterogeneous time-series data of the area throughout the day by measuring points across the entire region.

[0050] Taking the daily load curve of a certain feeder segment as an example, the original sampling frequency is one point every 15 minutes, with a total of 96 sampling points throughout the day. Using a symbolic aggregation algorithm, the data is first normalized to zero mean, and then divided into 12 equal-length sub-intervals, each containing 8 points. After calculating the mean of each interval, it is mapped to 5 preset discrete level symbols {1,2,3,4,5}. The complex power curve is transformed into a time-series fingerprint similar to "122355421122".

[0051] Based on temporal fingerprints, the probability of symbol occurrence (i.e., frequency feature), information entropy (i.e., complexity feature), and state transition matrix (i.e., dynamic association feature) are extracted to form the initial multi-indicator vector of the park under the dimension of security and reliability. Real-time monitoring of the data quality of the communication terminals in this area. Assuming that theoretically 96 data points should be received within an evaluation period, and 92 valid data points were actually received, the data integrity rate is calculated to be 95.8% (92 divided by 96). Simultaneously, the average response latency from data transmission from the terminal to entry into the system is measured to be 220 milliseconds. The sigmoid growth function is used to map the 95.8% integrity rate to a first intermediate factor (approximately 0.93), and the negative exponential decay function is used to map the 220 millisecond latency to a second intermediate factor (approximately 0.84). Multiplying these two factors yields a confidence factor of 0.7812 (0.93 * 0.84) for this dimension. This factor is then used to correct and normalize the initial weight of 0.25, generating the corrected indicator weights. The generated multi-index vectors are input into a sparse decoupled neural network for processing. In the feature interaction mapping layer, the vectors are projected into the hidden space. The self-attention decoupling layer identifies a high correlation between "distributed power supply power" and "voltage deviation" in the park by calculating the correlation matrix. The differentiable mask pruning layer matches mask parameters for these dimensions, sets redundant dimension parameters with minimal contribution to 0, and performs dot product operations to remove them. The reconstruction verification output layer ensures that the information reconstruction accuracy is still higher than 95% after pruning and outputs the decoupled non-zero dimension index. For the decoupled set of indicators, the comparative strength of each indicator is calculated. Statistically, the standard deviation of the "energy storage charging and discharging status" indicator in this park is 0.18, indicating a high comparative strength. Further, the correlation coefficient between this indicator and other indicators is calculated, and the differences are summed to obtain the conflict coefficient, reflecting the amount of new information it carries. The comparative strength is multiplied by the conflict coefficient to obtain the comprehensive information content of the indicator. After normalization, a conflict compensation weight is generated. The corrected weight is multiplied and fused with this conflict compensation weight to obtain a combined weight that balances data quality and indicator independence. Feature components are extracted from the multi-index vector based on the non-zero dimension index to construct the feature matrix to be evaluated. The combined weight vector is then multiplied with this matrix to obtain the original scoring matrix. The original scores are mapped to a standard range of 0 to 100 points to obtain the comprehensive score for the digitalization level of the park's power distribution network. For example, the final score is 84.6 points, with the high efficiency and flexibility dimension receiving 90 points due to high energy storage utilization, and the digital intelligence dimension receiving 75 points due to some latency issues. A quantitative evaluation report is automatically generated using a template.

[0052] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting, characterized in that, include: Acquire multi-source heterogeneous time-series data from all measurement points in the power distribution network, and use a symbolic aggregation algorithm to convert the multi-source heterogeneous time-series data into a sequence composed of discrete symbols to generate a time-series fingerprint. Based on the time-series fingerprint, generate a multi-index vector composed of feature components. The integrity rate and response delay of multi-source heterogeneous time-series data are monitored to generate confidence factors; the weights of the digital evaluation indicators of the distribution network are dynamically corrected using the confidence factors to obtain the corrected indicator weights. The multi-indicator vectors are decoupled and pruned to generate a decoupled index set. Calculate the contrast intensity and conflict coefficient of each indicator in the decoupled indicator set, and generate conflict compensation weights; The combined weights are generated by integrating the corrected indicator weights and the conflict compensation weights. The weighted decision matrix is ​​constructed by combining the multi-indicator vector and the decoupled indicator set, and finally a quantitative evaluation report on the digitalization level of the distribution network is generated.

2. The comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting as described in claim 1, characterized in that, The process involves acquiring multi-source heterogeneous time-series data from measurement points across the entire power distribution network, converting this data into a sequence of discrete symbols using a symbolic aggregation algorithm to generate a time-series fingerprint, and then generating a multi-index vector composed of feature components based on the time-series fingerprint. Collect multi-source heterogeneous time-series data reflecting the physical operation, user electricity consumption, distributed energy and communication status of the entire power distribution network, and construct four evaluation dimensions based on the data sources: clean and low-carbon, safe and reliable, efficient and flexible, and digital and intelligent. The collected time-series data are normalized and segmented and aggregated. The aggregated mean is mapped to discrete symbols using equal probability intervals based on Gaussian distribution, and time-series fingerprints are generated by arranging them in time sequence. The frequency of symbols in the statistical time-series fingerprint is used to generate frequency features, the information entropy is calculated to generate complexity features, the transition probability of adjacent symbols is extracted to generate dynamic association features, the three types of features are concatenated and mapped to the corresponding evaluation dimensions according to the data source type to obtain a multi-index vector.

3. The comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting as described in claim 2, characterized in that, The completeness and response delay of the monitored multi-source heterogeneous time-series data are used to generate confidence factors. These confidence factors are then used to dynamically adjust the weights of the distribution network digitization evaluation indicators, resulting in the adjusted indicator weights, including: Real-time monitoring of the integrity rate and response delay of multi-source heterogeneous time-series data, wherein the integrity rate is the ratio of the actual number of valid data points received to the theoretical number of data points that should be received, and the response delay is the time difference between the data being sent from the test point and entering the evaluation system; The integrity rate is mapped to the first intermediate factor using the S-shaped growth function, and the response delay is mapped to the second intermediate factor using the negative exponential decay function.

4. The comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting as described in claim 3, characterized in that, The monitoring of the completeness rate and response delay of multi-source heterogeneous time-series data generates a confidence factor; the confidence factor is used to dynamically adjust the weights of the distribution network digitalization evaluation indicators to obtain the adjusted indicator weights, which also includes: Multiply the first intermediate factor by the second intermediate factor to generate the confidence factor; The initial weights of each evaluation indicator are obtained. After multiplying the initial weights by the corresponding confidence factors and normalizing them, the dynamically adjusted indicator weights are obtained.

5. The comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting as described in claim 4, characterized in that, The process of decoupling and pruning multiple indicator vectors to generate a decoupled indicator set includes: Multi-index vectors are mapped to a high-dimensional space to represent the nonlinear interaction relationships between different dimensions. The degree of correlation between each dimension is calculated by using a self-attention mechanism, and redundant dimensions are identified and projected onto an orthogonal subspace for decoupling. Assign a learnable gated mask to each decoupled dimension. After sparsification constraints, set the mask parameters of dimensions below a preset threshold to zero and multiply them by the vector to eliminate invalid dimensions. The original mapping vector is reconstructed by using the pruned vector. When the loss of reconstructed information meets the consistency requirement, a non-zero dimension index is output. The non-zero dimension index is used to determine the decoupled index set.

6. The comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting as described in claim 5, characterized in that, The calculation of the contrast strength and conflict coefficient of each indicator within the decoupled indicator set, and the generation of conflict compensation weights, includes: Obtain the numerical sequence of each indicator in the decoupled indicator set within the evaluation period, and calculate the standard deviation of the numerical sequence of each indicator as the comparative strength of the corresponding indicator. Calculate the correlation coefficient between any two indicators in the set, and then sum the differences between the correlation coefficient of each indicator and all other indicators and the value of 1 to generate the conflict coefficient of the corresponding indicator. Multiply the contrast intensity of each indicator by the conflict coefficient to obtain the comprehensive information content of each indicator; The total information content of all indicators is normalized so that the sum of the weights of each indicator is 1, thus generating conflict compensation weights.

7. The comprehensive evaluation method for the digitalization level of a distribution network based on multi-index weighting as described in claim 6, characterized in that, The fused and corrected indicator weights and conflict compensation weights generate combined weights, which, combined with multi-indicator vectors and the decoupled indicator set, construct a weighted decision matrix, ultimately generating a quantitative evaluation report on the digitalization level of the distribution network, including: Extract the weight components corresponding to the non-zero dimension index from the corrected index weights, multiply and fuse them with the generated conflict compensation weights, and normalize them to obtain the combined weights. The feature components at corresponding positions are extracted from the multi-index vector using non-zero dimension indexes to construct the feature matrix to be evaluated; The combined weights are multiplied with the feature matrix to be evaluated to obtain the original scoring matrix, which is then normalized to generate a weighted decision matrix. Based on the weighted decision matrix, a quantitative evaluation report on the digitalization level of the distribution network is generated, which includes scores for each dimension, a comprehensive score, and evaluation improvement suggestions.

8. A comprehensive evaluation system for the digitalization level of a distribution network based on multi-index weighting, using the method described in any one of claims 1 to 7, characterized in that, include: The fingerprint construction module is used to acquire multi-source heterogeneous time-series data from measurement points across the entire power distribution network, and to convert the multi-source heterogeneous time-series data into a sequence composed of discrete symbols using a symbolic aggregation algorithm to generate a time-series fingerprint. Based on the time-series fingerprint, a multi-index vector composed of feature components is generated. The weight correction module is used to monitor the integrity rate and response delay of multi-source heterogeneous time-series data and generate confidence factors; the confidence factors are used to dynamically correct the weights of the digital evaluation indicators of the distribution network to obtain the corrected indicator weights. The decoupling and pruning module is used to decouple and prune multi-index vectors based on correlation, and generate a decoupled index set. The conflict compensation module is used to calculate the contrast strength and conflict coefficient of each indicator in the decoupled indicator set and generate conflict compensation weights. The integrated evaluation module is used to generate combined weights by integrating the corrected index weights and conflict compensation weights, and to construct a weighted decision matrix by combining multiple index vectors and the decoupled index set, ultimately generating a quantitative evaluation report on the digitalization level of the distribution network.

9. A computer device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores computer-executable instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1 to 7.

Images