Classification pricing method and system for 10 kV distribution network non-power-cut operation cost

Abstract

The invention discloses a 10 kV distribution network non-power-cut operation cost classification pricing method and system, and relates to the technical field of non-power-cut operation classification pricing, and the method comprises the following steps: collecting the line data of a target operation, and carrying out the initial classification of the non-power-cut operation; grading different types of jobs according to the initial classification result; according to the grading result, constructing a corresponding dynamic resource input strategy table, and matching corresponding resource standard consumption according to the job grade to obtain a resource reference input value; determining a reference unit price according to the initial classification result, correcting the reference unit price according to the resource reference input value, generating a classification pricing table, and forming a traceable cost accounting record; according to the invention, the problems that the operation risk difference, the dynamic change of resource investment and the non-traceability of the cost accounting cannot be accurately reflected by the non-power-cut operation cost pricing are solved, and the precision, transparency and management efficiency of the distribution network operation cost accounting are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of non-power operation classification pricing, more particularly, the present application relates to a 10kV distribution network non-power operation cost classification pricing method and system. BACKGROUND

[0002] With the promotion of the intelligent construction of distribution network, the 10kV distribution network non-power operation technology is widely used in power system operation and maintenance. This kind of operation can realize the maintenance, reconstruction or expansion of distribution line under the premise of not affecting the power supply of users through live operation, bypass lapping, mobile power supply access and other ways, which significantly improves the power supply reliability and service quality. However, due to the diversity of operation types, the complexity of operation environment and the large difference of resource investment involved in non-power operation, the existing cost pricing method cannot effectively reflect the risk level, resource consumption and comprehensive influence on power supply system of different operations, resulting in problems such as cost accounting deviation, uneven resource allocation and inaccurate cost control in power operation and maintenance management.

[0003] At present, the commonly used pricing method of distribution network non-power operation in the industry is mainly based on the fixed quota pricing method or the experience analogy method. The fixed quota pricing method calculates the operation cost by pre-setting the standard consumption of manpower, materials and tools and instruments of various operations, combined with the unified unit price standard. This method is suitable for conventional and standardizable operation scenarios, but in the operation involving high risk, complex environment or dynamic change of resource investment, there are the following main defects: The existing fixed quota standard is mostly based on typical operation samples, lacks data support from multi-source systems (such as dispatching system, equipment archives, GIS, operation records, etc.), and is difficult to quantify the influence of different line environments, load levels and safety conditions on operation cost; the safety risk and technical complexity of non-power operation directly affect the manpower investment and operation time, while the existing method usually only considers the workload and type of work, and does not have the risk quantification ability, which cannot accurately reflect the actual economic value of high-risk operation; and the cost records in the existing system are mostly stored in a manual or static table manner, which cannot track the version or backtrack the parameters of the pricing process, and is difficult to meet the requirements of modern power grid lean management and audit. In view of the above problems, the present application provides a solution. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a 10kV distribution network non-power operation cost classification pricing method and system, which solves the problems that the non-power operation cost pricing cannot accurately reflect the operation risk difference, the dynamic change of resource investment and the non-traceable cost accounting, and improves the precision, transparency and management efficiency of distribution network operation cost accounting.

[0005] To achieve the above purpose, the present application provides the following technical solutions: Firstly, this application provides a classification and pricing method for live-line work in 10kV distribution networks. The method includes: collecting line data for the target work and initially classifying the live-line work; classifying different types of work according to the initial classification results; constructing a corresponding dynamic resource input strategy table according to the classification results, and matching the corresponding standard resource consumption according to the work level to obtain the resource benchmark input value; determining the benchmark unit price according to the initial classification results, and correcting the benchmark unit price according to the resource benchmark input value to generate a classification and pricing table, forming a traceable cost accounting record.

[0006] In one embodiment, line data of the target operation is collected, and the power-on-demand operation is initially classified. Specifically, the data source is determined based on the line data, and each data source is parsed to obtain a data source metadata set; based on the field information in the data source metadata set, semantic vectors of fields and tags are extracted, the semantic similarity between each field and tag is calculated, and a semantic tag dictionary is established; based on the established semantic tag dictionary, if the semantic similarity between a field and a tag exceeds a preset threshold, a semantic mapping relationship table is established; based on the semantic mapping relationship table, the data from each system is preprocessed and merged to generate a unified tagged dataset; and the power-on-demand operation is initially classified based on the unified tagged dataset.

[0007] In one embodiment, the initial classification of live-line work is performed based on a unified labeled dataset. Specifically, this involves: extracting key feature parameters from the unified labeled dataset and standardizing them to obtain standardized feature vectors; plotting the sum of squares within clusters as a function of the number of clusters for the standardized feature vectors, and selecting the inflection point as the optimal number of clusters; assigning feature weights to each feature dimension, the feature weights being calculated using accident rate, work complexity index, and user impact index; calculating the weighted Euclidean distance based on the optimal number of clusters and feature weights, and performing clustering to obtain preliminary classification results; comparing the preliminary classification results with historical work records to calculate the classification deviation; if the classification deviation is greater than a preset deviation threshold, adjusting the feature weights according to the deviation direction; and re-performing clustering based on the adjusted feature weights to generate stable preliminary classification results.

[0008] In one embodiment, based on the initial classification results, different types of tasks are classified, specifically as follows: Based on stable preliminary classification results, the input feature parameter set for the corresponding task is extracted; the input feature parameter set is normalized to form an input vector for fuzzy inference; a fuzzy membership function is established for the input vector, and the membership value of each input vector under different risk levels is calculated to obtain a fuzzy representation set of task features; a fuzzy rule base is constructed based on the fuzzy representation set, and the input parameters are matched and fuzzy synthesized using a fuzzy inference mechanism to obtain a fuzzy risk output set; the fuzzy risk output set is defuzzified to obtain risk input variables for Bayesian inference; and the risk input variables are classified using a Bayesian network.

[0009] In one embodiment, the risk input variables are classified using a Bayesian network. Specifically, a Bayesian network model for classifying job risks is established based on the risk input variables. The risk input variables are input into the Bayesian network model to obtain the posterior probabilities of different job levels, and the initial level to which the target job belongs is determined based on the maximum a posteriori probability method. Confidence correction is performed on the initial level, and the confidence score of the risk level is calculated. If the confidence score of the risk level is lower than a preset threshold, fuzzy rule adaptive optimization is triggered to locally update the membership function parameters until it stops. The final classification result after confidence correction and the corresponding confidence score of the risk level are output.

[0010] In one embodiment, a corresponding dynamic resource input strategy table is constructed based on the classification results. Specifically, based on a reinforcement learning algorithm, a state vector is constructed according to the classification results, and the action space is defined as the combination of input amounts for various types of resources; a reward function is established that includes job completion efficiency, resource consumption cost, and safety risk indicators; a resource allocation strategy is established and iteratively trained based on the state vector, action space, and reward function; during job execution, resource input actions are selected based on the current state vector and the trained resource allocation strategy to generate a dynamic resource input strategy table corresponding to the job level.

[0011] In one embodiment, the resource baseline input value is obtained by matching the corresponding standard resource consumption according to the task level. Specifically, this involves: extracting the corresponding resource combination and recommended input of various resources from the dynamic resource input strategy table according to the target task level to form an initial resource input structure corresponding to the level; constructing a resource category index matrix based on the initial resource input structure; standardizing the resource category index matrix to obtain a normalized matrix; calculating the entropy value and information utility of various resources based on the normalized matrix; calculating the entropy weight of various resources based on the information utility; and weighting and accumulating the standard input of various resources in the initial resource input structure based on the entropy weight to obtain the resource baseline input value.

[0012] In one embodiment, the benchmark unit price is determined based on the initial classification results. Specifically, a two-dimensional benchmark unit price matrix is ​​constructed based on the job type and corresponding job complexity index in the initial classification results, and the initial value of each cell in the benchmark unit price matrix is ​​filled to form a benchmark price lookup table; for any job to be priced, the benchmark unit price is obtained by matching it with the benchmark unit price matrix.

[0013] In one embodiment, the benchmark unit price is adjusted based on the resource benchmark input value to generate a categorized pricing table and form a traceable cost accounting record. Specifically, this involves: obtaining the actual resource input value and calculating the deviation based on the resource benchmark input value to obtain the resource deviation; obtaining the operation impact coefficient and risk level confidence level, and adjusting the benchmark unit price based on the resource deviation; generating a categorized pricing table based on the adjusted benchmark unit price and the corresponding operation data; recording the categorized pricing table and establishing a cost traceability chain structure to form a traceable cost accounting record.

[0014] Secondly, this application provides a classification and pricing system for 10kV distribution network live-line work fees. The system includes: a classification module for collecting line data of the target work and performing initial classification of the live-line work; a grading module for grading different types of work based on the initial classification results; a resource benchmark input value acquisition module for constructing a corresponding dynamic resource input strategy table based on the grading results and matching the corresponding resource standard consumption according to the work level to obtain the resource benchmark input value; and a classification and pricing table generation module for determining the benchmark unit price based on the initial classification results, correcting the benchmark unit price based on the resource benchmark input value, generating a classification and pricing table, and forming a traceable cost accounting record.

[0015] As can be seen from the above technical solutions, the embodiments of this application have the following advantages: By constructing a comprehensive system encompassing multi-source heterogeneous data acquisition, semantic fusion, feature extraction, risk grading, and cost classification and pricing, this method achieves intelligent, dynamic, and traceable calculation of live-line work costs. Building upon traditional quota-based pricing, this approach introduces a data-driven mechanism, utilizing semantic mapping and cluster analysis to improve the accuracy of work classification. It achieves high-confidence risk grading through the fusion of fuzzy inference and Bayesian networks, and further optimizes resource allocation strategies using reinforcement learning and entropy weight analysis, thus ensuring that the benchmark unit price adjustment more closely aligns with actual work consumption and risk levels. Furthermore, a cost traceability chain structure is established, giving each cost record a unique hash identifier and version tracking mechanism, enabling transparent and traceable cost management throughout the entire process. Overall, this solution effectively improves the accuracy of cost accounting, resource allocation efficiency, and scientific economic decision-making in live-line work, promoting the safety and lean management of power grid operation and maintenance. Attached Figure Description

[0016] Figure 1 This is a schematic flowchart illustrating a classification and pricing method for 10kV distribution network live-line work fees, provided as an embodiment of this application.

[0017] Figure 2 This is a schematic diagram of a classification and pricing system for 10kV distribution network live-line work fees, provided as an embodiment of this application.

[0018] Figure 3 A line graph showing the results of the operation risk classification provided in this application embodiment. Detailed Implementation

[0019] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] Reference Figure 1 As shown in the diagram, this invention provides a flowchart for a classification and pricing method for 10kV distribution network live-line work fees, which includes the following steps: S1: Collect line data for the target operation and perform initial classification of uninterrupted power supply operations.

[0021] In this embodiment, the line data of the target operation is collected, and the uninterrupted power supply operations are initially classified, specifically as follows: Acquire the line data for the target uninterrupted power supply operation, the line data including line number, area number, voltage level and operating status; Based on the route data, determine the operation route number, the area to which it belongs, and the corresponding data sources for the scheduling, GIS, detection, and manual data entry systems. Then, parse each data source, record the data field name, data type, update time granularity, unit format, and data reliability identifier to obtain the data source metadata set. The data source metadata set is a collection of structural features of each data source. It records information such as field names, data types, unit systems, update time granularity, coordinate systems, and reliability of each data source, providing a complete structured description for subsequent semantic mapping.

[0022] Based on the field information in the data source metadata set, a pre-trained word embedding model is used to extract the semantic vector representations of fields and tags, and the semantic similarity (cosine similarity) between the semantic vector of each field and the semantic vector of the tag is calculated. Among them, the pre-trained word embedding model is a language model (such as Word2Vec or BERT) trained on a large-scale corpus. By converting field names and label text into semantic vectors, the system can calculate the semantic similarity between them, thereby achieving intelligent semantic alignment of fields across systems.

[0023] Establish a semantic tag dictionary, where each semantic tag corresponds to the business semantics required for the classification of live-line work. The business semantics include line number, voltage level, load current, terrain slope, and power supply user level. The semantic tag dictionary is established through a combination of manual review and automatic semantic expansion. When a new data source is connected, the system automatically updates the semantic tag dictionary based on the root word matching results of the field names, thereby realizing the dynamic expansion of the tag library. Based on the establishment of a semantic tag dictionary, if the semantic similarity between a field and a tag exceeds a preset similarity threshold, a mapping relationship is established to form a semantic mapping relationship table; Based on the semantic mapping table, data from the scheduling system, GIS system, field detection equipment and manual data entry system are preprocessed and merged to generate a unified labeled dataset. The preprocessing includes mapping fields with the same semantics but different names in different systems to a unified semantic label; performing unit conversion for numerical fields with different unit systems; achieving time alignment for data that is not synchronized in time by aggregating through a sliding time window; and performing coordinate system unification for spatial coordinate data.

[0024] Initial classification of live-line work is performed based on a unified labeled dataset.

[0025] It should be noted that by completing the entire process from data source identification and semantic mapping to standardized fusion, the automatic collection and unified processing of multi-source heterogeneous data for live-line work can be achieved, thereby constructing a high-quality, computable, and traceable basic dataset. This not only ensures the consistency and integrity of the data and enables accurate matching and standardization of data from different systems, but also improves the level of automation in processing, reduces manual intervention, and provides reliable data support for subsequent operation feature extraction, classification analysis, risk assessment, and cost pricing.

[0026] Furthermore, the live-line work is initially classified based on the unified labeled dataset, specifically as follows: Based on a unified labeled dataset, key feature parameters for classifying uninterrupted power supply operations are extracted. These key feature parameters include environmental complexity, load intensity, safety margin, and user impact index. Among them, environmental complexity reflects the construction complexity and potential operational difficulty of the environment along the target route, including factors such as passage obstacles, terrain slope and weather conditions. The environmental complexity is obtained by calculating the obstacle density per unit length based on the number of buildings, trees or other obstacles along the route, calculating the average slope along the route using a GIS digital elevation model, and weighting the obstacle density and average slope with wind speed, rainfall and humidity according to preset weights to obtain the environmental complexity. Load intensity represents the ratio of the current load level of a line to its rated capacity, reflecting the operating pressure of energized lines during operation; safety margin represents the ratio of the safe distance between energized parts of the line and the operator to the standard distance during operation, reflecting the safety margin of the operation; user impact index measures the degree of impact that live-line work may have on power supply users, especially important load users, obtained by statistically analyzing the ratio of the total number of users within the scope of the operation's impact to the number of important load users through the user information system.

[0027] The key feature parameters are standardized to obtain standardized feature vectors, which provide a unified input for weighted clustering. For the standardized feature vectors, the Elbow method is used to plot the curve of the sum of squares within a cluster as a function of the number of clusters, and the inflection point is selected as the optimal number of clusters; Each feature dimension is assigned a feature weight, which is calculated using the accident rate, job complexity index, and user impact index of historical job data. The specific calculation formula for the feature weights is as follows:

[0028]

[0029]

[0030] In the formula, For accident rate, Let feature dimension i correspond to the number of operational accidents. Let i be the total number of jobs corresponding to feature dimension i. As an indicator of task complexity, Let j be the original value of the j-th historical job on feature dimension i. Let i be the total number of historical job samples in feature dimension i. The minimum value among historical assignments for feature dimension i. The maximum value of feature dimension i in historical jobs. For feature weights, User Influence Index , , These are the weighting coefficients, To normalize the denominator, the weighted values ​​of all four features are summed to ensure that the final sum of the feature weights is 1, thus achieving normalization.

[0031] Based on the optimal number of clusters and feature weights, the weighted Euclidean distance between each sample and the center of each cluster is calculated, and clustering is performed to obtain preliminary classification results. The preliminary classification results include routine low-risk live-line work, medium-load bypass work, high-complexity mobile power supply work, and high-risk integrated collaborative work. The specific formula for calculating the weighted Euclidean distance is as follows:

[0032] In the formula, For weighted Euclidean distance, Let j be the standardized feature vector of the j-th job sample. Let be the center vector of the k-th cluster. Let j be the value of the j-th sample on the i-th feature. The value of the k-th cluster center on the i-th feature.

[0033] The preliminary classification results are compared with the actual historical records to calculate the classification bias, which is the difference between the weighted Euclidean distance and the historical classification center vector. The classification bias is compared with a preset bias threshold. If the classification bias is greater than the preset bias threshold, the feature weights are adjusted according to the direction of the bias to favor features that improve classification accuracy. The specific calculation formula for adjusting the feature weights is as follows:

[0034] In the formula, The adjusted feature weights, The adjustment coefficients, used to control the magnitude of weight changes, are typically set based on expert experience or historical data. For classification bias, If it is a sign function, then ,but If the feature weights increase, ,but The feature weights decrease.

[0035] Based on the adjusted feature weights, clustering is re-executed to generate stable preliminary classification results.

[0036] It should be noted that by extracting key features (environmental complexity, load intensity, safety margin, and user impact index) from a unified labeled dataset, standardizing the data, calculating feature weights based on historical operation data, and performing initial classification based on a weighted clustering method, and then dynamically adjusting the weights through bias feedback to achieve adaptive optimization, live-line work can be accurately classified into different types according to risk and complexity. The advantages are that it ensures the scientificity and quantifiability of the classification results, so that the risk, load, and user impact of each type of work are reasonably reflected, and it also realizes data-driven dynamic optimization, which can effectively guide work scheduling, improve safety, reduce work risks, and provide a reliable and traceable standardized basis for subsequent cost pricing.

[0037] S2, based on the initial classification results, classify different types of tasks into different levels.

[0038] In this embodiment, based on the initial classification results, different types of tasks are classified into different levels, specifically as follows: Based on the stable initial classification results, the input feature parameter set of the corresponding operation is extracted. The input feature parameter set includes the operation voltage level, load intensity, operation environment complexity, personnel cooperation level, and safety risk level. The input feature parameter set is normalized to form an input vector for fuzzy inference; A fuzzy membership function is established for the input vector. The membership function includes three risk level intervals: low, medium, and high. The membership value of each input vector under different risk levels is calculated through the membership function to obtain the fuzzy representation set of the operation features. The membership function is calculated using the following formula:

[0039] In the formula, This represents the membership degree value. For the input vector, The membership function center value corresponds to a typical value for a certain fuzzy level (low, medium, high). is the diffusion coefficient of the membership function, representing the tolerance range of the fuzzy level.

[0040] It should be noted that when Deviation The farther away, the higher the degree of membership. The smaller the value, the higher the degree of non-compliance with that level. The larger the value, the smoother the membership function, indicating that the level is more tolerant of input parameters; The smaller the value, the more sensitive the level is to deviations.

[0041] A fuzzy rule base is constructed based on the fuzzy representation set. The input parameters are then matched and synthesized using a fuzzy inference mechanism to obtain the fuzzy risk output set of the task. The construction of the fuzzy rule base includes extracting fuzzy cluster centers from historical job datasets using the fuzzy C-means (FCM) algorithm to generate basic rules. The fuzzy inference mechanism adopts Mamdani-type fuzzy inference.

[0042] The fuzzy risk output set is defuzzified, and the centroid method is used to calculate the comprehensive risk value to obtain the risk input variables for Bayesian inference. Risk input variables are classified using a Bayesian network.

[0043] Furthermore, such as Figure 3 As shown, risk input variables are classified using a Bayesian network, specifically as follows: Based on risk input variables A Bayesian network model for classifying operational risks is established, wherein the Bayesian network model includes a node layer, a dependency layer, and a result layer. The node layer represents the input vector of the task and its fuzzy output, as well as the risk input variable; the dependency layer represents the conditional dependencies between variables through mutual information analysis; and the result layer outputs the task level nodes. The job level nodes are Level I, Level II, Level III, and Level IV, with each level including a preset value range.

[0044] The risk input variable is fed into a Bayesian network model, and the posterior probability of different job levels is calculated through conditional probability inference. And determine the initial level of the target operation based on the maximum a posteriori probability method; The maximum a posteriori probability method is specifically calculated using the following formula:

[0045] In the formula, The final job grade corresponding to the maximum posterior probability. Let be the posterior probability.

[0046] Confidence correction is applied to the initial risk level, the variance of the corresponding posterior probability is calculated, and the risk level confidence score is obtained. If the risk level confidence score is lower than a preset threshold, fuzzy rule adaptive optimization is triggered, adjusting the membership function parameters. Perform partial updates to improve model confidence. If the confidence level of the risk level increases by less than a set value in two consecutive iterations, then stop the optimization. The specific formula for calculating the confidence level of the risk grade is as follows:

[0047] In the formula, For the risk level confidence level, Let be the variance of the posterior probability.

[0048] Among them, triggering fuzzy rule adaptive optimization involves adjusting the membership function parameters. The local update involves calculating the contribution of the fuzzy output membership value of each rule to the input vector, and finding the parameter corresponding to the rule with the largest contribution. As the object of local updates, it has the greatest impact on the result. The specific formula for calculating its contribution is as follows:

[0049] In the formula, As for contribution level, Let be the membership value of the i-th fuzzy rule under the input vector X. The output risk value for the i-th rule. The sum of the outputs of all rules is used as a normalization factor.

[0050] Output the final classification result after confidence correction and the corresponding risk level confidence level.

[0051] It should be noted that by combining fuzzy inference with Bayesian networks, high-precision, interpretable, and adaptive processing of job classification is achieved. Specifically, fuzzy inference can effectively handle the uncertainty and continuity of input feature parameters, quantifying input vectors into membership degrees to form a fuzzy representation set, thus taking into account the diversity and fuzziness of job features. Subsequently, the Bayesian network performs conditional probability inference on the fuzzified risk input variables, which can capture the dependencies between different job features and achieve probabilistic classification. At the same time, the optimal level is selected through the maximum a posteriori probability method to ensure the statistical reliability of the classification results. Furthermore, confidence correction and adaptive optimization are introduced, enabling the model to dynamically adjust the fuzzy membership function parameters, improving the stability and credibility of classification. Thus, when facing complex, dynamic, and multidimensional job scenarios, it possesses accuracy, robustness, and traceability.

[0052] S3. Based on the classification results, construct the corresponding dynamic resource input strategy table, and match the corresponding standard resource consumption according to the operation level to obtain the resource baseline input value.

[0053] In this embodiment, a corresponding dynamic resource allocation strategy table is constructed based on the hierarchical results, specifically as follows: Based on the reinforcement learning algorithm, a state vector is constructed according to the classification results. The state vector includes the current job level, job complexity index, number of personnel involved in the job and skill distribution, and availability of equipment and tools required for the job. The action space is defined as the combination of input quantities of various resources. The action space includes the number of personnel and job roles, the number of work equipment and tools, the amount of safety protection resources invested, and the work cycle and time arrangement. Establish a reward function, which includes task completion efficiency, resource consumption cost, and safety risk indicators; The reward function is calculated using the following formula:

[0054] In the formula, For the reward function value, To improve the efficiency of completing tasks, For resource consumption costs, As a safety risk indicator, , , These are the weighting coefficients.

[0055] Establish a resource allocation strategy, which is a combination of resource inputs corresponding to input state vectors and output actions. Use a reward function to iteratively train the strategy and optimize the long-term cumulative reward. During job execution, resource input actions are selected based on the current state vector and the trained resource allocation strategy, generating a dynamic resource input strategy table corresponding to the job level. Each level includes recommended resource combinations and input amounts, forming a dynamic resource allocation scheme that can be directly applied.

[0056] The dynamic resource allocation strategy table is a table that maps the job level to the recommended resource combination and allocation amount corresponding to the real-time job status, and can dynamically generate the optimal resource allocation scheme according to environmental changes.

[0057] Furthermore, by matching the corresponding standard resource consumption according to the task level, the baseline resource input value is obtained, specifically as follows: Based on the level of the target task, the corresponding resource combination and the recommended input amount of each type of resource are extracted from the dynamic resource input strategy table to form the initial resource input structure corresponding to the level. Construct a resource category index matrix based on the initial resource input structure. ,in, This represents the standardized input value of resource type i in the j-th job sample; The resource category index matrix is ​​standardized to obtain a normalized matrix. ,in, In the formula, The values ​​are standardized. Let i be the minimum input value of resource i across all job samples. Let i be the maximum input value of resource type i across all job samples; Calculate the entropy and information utility of various resources based on the normalized matrix; The specific formula for calculating the information utility is as follows: ,

[0058]

[0059] In the formula, Information utility represents the discriminative power of a resource within a sample. Let i be the entropy value of the i-th type of resource. The normalization coefficient is... This represents the number of job samples.

[0060] The greater the information utility, the more significant the differences in the resource across different job samples, and the higher its contribution to job level differentiation and benchmark input calculation.

[0061] Entropy weights for various resources are calculated based on information utility. The specific formula for calculating the entropy weight is as follows:

[0062] In the formula, is the entropy weight, and n is the total number of resource categories, that is, the number of resource types considered in the baseline input calculation.

[0063] Based on entropy weighting, the standard input quantities of various resources in the initial resource input structure are weighted and accumulated to obtain the resource benchmark input value.

[0064] The initial resource input structure is a standardized combination of recommended resources for each type extracted from the dynamic resource input strategy table based on the target operation level, reflecting the preliminary allocation plan of various resources required to complete the operation at that level; the standard input amount of each type of resource is the recommended input value or standard value of a certain type of resource in the initial resource input structure at the target operation level; the resource benchmark input value represents the total amount of resources or comprehensive input level required to complete the operation at the target level.

[0065] It should be noted that the calculation of the resource baseline input value not only achieves dynamic resource matching by combining the job level and real-time job status, but also uses reinforcement learning to generate the optimal resource allocation strategy. At the same time, it objectively quantifies the contribution of various resources through the entropy weight method, realizing the scientific weighting of the initial resource input structure to the comprehensive baseline input value. This ensures that the final resource baseline input value can reflect the actual contribution of various resources under different job conditions, and can adapt to environmental changes and job complexity, thereby improving the accuracy, efficiency and traceability of resource allocation. This provides a reliable basis for job scheduling, cost accounting and safety management.

[0066] S4. Determine the benchmark unit price based on the initial classification results, and correct the benchmark unit price based on the resource benchmark input value to generate a classification pricing table and form a traceable cost accounting record.

[0067] In this embodiment, the benchmark unit price is determined based on the initial classification results, specifically as follows: Based on the job types and corresponding job complexity indicators in the initial classification results, a two-dimensional benchmark unit price matrix is ​​constructed, and the initial value of each cell in the benchmark unit price matrix is ​​filled to form a benchmark price lookup table. The benchmark unit price matrix has one dimension as job type and the other dimension as job complexity indicator. The initial value is filled by integrating the standard working hours and unit prices published by external industries with the actual cost data of historical operations within the enterprise. For any task to be priced, a benchmark unit price is obtained by matching it with a benchmark unit price matrix. The benchmark unit price is used to reflect the basic cost of different task types.

[0068] Furthermore, the benchmark unit price is adjusted based on the resource benchmark input value to generate a categorized pricing table, forming a traceable cost accounting record, specifically as follows: The actual resource input value is obtained, and the deviation is calculated by combining the actual resource input value with the resource benchmark input value. The resource deviation is obtained by obtaining the difference between the actual resource input value and the resource benchmark input value, and the resource deviation is obtained by the ratio of the difference to the resource benchmark input value. Obtain the operation impact coefficient and risk level confidence level, and adjust the benchmark unit price based on resource deviation; The specific formula for calculating the operation impact coefficient is as follows:

[0069] In the formula, The operation impact coefficient is a comprehensive indicator that reflects the degree of impact of uninterrupted power supply operations on the operational stability of the power supply system and the continuity of power supply for users. User Influence Index The experience weight of users influencing the index. The contribution of live-line work to the power supply reliability index of the target power supply line area can be monitored and obtained through the Supervisory Control and Data Acquisition (SCADA) system and the Distribution Management System (DMS). The empirical weights for the contribution of uninterrupted power supply operations to power supply reliability indicators.

[0070] The correction is made, and the specific calculation formula is as follows:

[0071] In the formula, This is the revised benchmark unit price. The base unit price, Due to resource discrepancies, , , These are the weighting coefficients.

[0072] A categorized pricing table is generated based on the revised benchmark unit price and the corresponding work data. The categorized pricing table includes work number, work type, risk level and risk level confidence level, revised benchmark unit price, work impact coefficient, and resource deviation. The categorized pricing table is stored in the expense accounting record database, and an expense traceability chain list structure is established to form a traceable expense accounting record. The expense traceability chain list structure includes recording chain list node information, generating a unique expense hash code, establishing a version tracking mechanism, and outputting an expense accounting traceability report.

[0073] The system records the node information of each expense record, including the job ID, timestamp, correction parameters, and reviewer identifier. It generates a unique expense hash code by using a hash algorithm (such as SHA-256) to generate an immutable identifier. It establishes a version tracking mechanism so that when the expense is corrected or the parameters are updated, the system automatically generates a new version record and associates it with the old version through the hash chain, achieving full traceability. It outputs an expense accounting traceability report so that the history of expense changes can be retrieved and tracked by job category, time interval, or risk level.

[0074] It should be noted that by adjusting the benchmark unit price through resource deviation, operation impact coefficient, and risk level confidence level, a dynamic adjustment mechanism can be introduced into the original quota pricing system. This allows the pricing results to reflect changes in actual operation resource consumption and risk levels in real time, effectively avoiding the problem of "overestimation or underestimation" existing in traditional fixed unit price models. The introduction of resource deviation ensures the consistency between cost accounting and actual input. The operation impact coefficient reflects the degree of impact of the operation on the stability of the power supply system and the continuity of users, making the pricing results more reflective of the value of power grid operation. The application of risk level confidence level improves the reliability and scientific nature of pricing by quantifying the uncertainty of risk assessment. The generated classification pricing table not only presents the type, risk, resource consumption, and cost adjustment results of various operations in a structured manner, achieving transparency and comparability of cost calculation, but also forms a fully traceable cost accounting record by establishing a cost traceability chain structure. This facilitates subsequent audit tracking, cost optimization, and management decision-making, providing data support for the economic assessment and lean management of live-line operations.

[0075] Reference Figure 2 As shown in the diagram, this invention provides a structural schematic of a classification and pricing system for 10kV distribution network live-line work fees. The system includes a classification module, a grading module, a resource benchmark input value acquisition module, and a classification and pricing table generation module. These modules are interconnected. The classification module is used to collect line data for the target operation and perform initial classification of live-line operations; The classification module is used to classify different types of jobs based on the initial classification results; The resource baseline input value acquisition module is used to construct a corresponding dynamic resource input strategy table based on the classification results, and match the corresponding standard resource consumption according to the operation level to obtain the resource baseline input value. The classification pricing table generation module is used to determine the benchmark unit price based on the initial classification results, and to correct the benchmark unit price based on the benchmark resource input value, thereby generating a classification pricing table and forming a traceable cost accounting record.

[0076] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0077] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0078] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. 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.

[0079] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0080] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0081] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A classification and pricing method for 10kV distribution network live-line work fees, characterized in that, include: Collect line data for the target operation and perform initial classification of uninterrupted power supply operations; Based on the initial classification results, different types of assignments are graded. Based on the classification results, a corresponding dynamic resource input strategy table is constructed, and the corresponding standard resource consumption is matched according to the operation level to obtain the resource baseline input value. The benchmark unit price is determined based on the initial classification results, and then adjusted according to the benchmark resource input value to generate a classification pricing table and form a traceable cost accounting record.

2. The classification and pricing method for 10kV distribution network live-line work fees according to claim 1, characterized in that, The process involves collecting line data for the target operation and initially classifying the uninterrupted power supply operations as follows: The data source is determined based on the line data, and each data source is parsed to obtain the data source metadata set; Based on the field information in the data source metadata set, the semantic vectors of fields and tags are extracted, the semantic similarity between each field and tag is calculated, and a semantic tag dictionary is built. Based on the establishment of a semantic tag dictionary, if the semantic similarity between a field and a tag exceeds a preset threshold, a semantic mapping relationship table is established; Based on the semantic mapping table, the data from each system are preprocessed and merged to generate a unified labeled dataset. Initial classification of live-line work is performed based on a unified labeled dataset.

3. The classification and pricing method for 10kV distribution network live-line work fees according to claim 2, characterized in that, The initial classification of live-line work based on a unified labeled dataset is as follows: Based on a unified labeled dataset, key feature parameters are extracted and standardized to obtain a standardized feature vector. For the standardized eigenvectors, plot the curve of the sum of squares within a cluster as a function of the number of clusters, and select the inflection point as the optimal number of clusters; Each feature dimension is assigned a feature weight, which is calculated using accident rate, job complexity index, and user impact index. The weighted Euclidean distance is calculated based on the optimal number of clusters and feature weights, and clustering is performed to obtain preliminary classification results; The preliminary classification results are compared with the actual records of historical operations to calculate the classification deviation; If the classification bias is greater than the preset bias threshold, the feature weights are adjusted according to the direction of the bias. Based on the adjusted feature weights, clustering is re-executed to generate stable preliminary classification results.

4. The classification and pricing method for 10kV distribution network live-line work fees according to claim 3, characterized in that, The process of grading different types of jobs based on the initial classification results is as follows: Based on the stable preliminary classification results, extract the input feature parameter set for the corresponding task; The input feature parameter set is normalized to form an input vector for fuzzy inference; A fuzzy membership function is established for the input vectors, and the membership values ​​of each input vector under different risk levels are calculated to obtain a fuzzy representation set of the operation features; A fuzzy rule base is constructed based on the fuzzy representation set. The fuzzy inference mechanism is used to perform rule matching and fuzzy synthesis on the input parameters to obtain the fuzzy risk output set. The fuzzy risk output set is defuzzified to obtain the risk input variables for Bayesian inference. Risk input variables are classified using a Bayesian network.

5. The classification and pricing method for 10kV distribution network live-line work fees according to claim 4, characterized in that, The classification of risk input variables using a Bayesian network is specifically as follows: Based on risk input variables, a Bayesian network model for classifying operational risks is established. The risk input variable is input into the Bayesian network model to obtain the posterior probability of different job levels, and the initial level of the target job is determined based on the maximum a posteriori probability method. The initial level is corrected for confidence, and the risk level confidence score is calculated. If the risk level confidence score is lower than the preset threshold, the fuzzy rule adaptive optimization is triggered, and the membership function parameters are locally updated until the process stops. Output the final classification result after confidence correction and the corresponding risk level confidence level.

6. The classification and pricing method for 10kV distribution network live-line work fees according to claim 1, characterized in that, The step of constructing a corresponding dynamic resource allocation strategy table based on the hierarchical results is as follows: Based on the reinforcement learning algorithm, a state vector is constructed according to the hierarchical results, and the action space is defined as the combination of input amounts of various resources. Establish a reward function that includes indicators such as task completion efficiency, resource consumption cost, and safety risk. Establish a resource allocation strategy and perform iterative training based on the state vector, action space, and reward function; During job execution, resource input actions are selected based on the current state vector and the trained resource allocation strategy, generating a dynamic resource input strategy table corresponding to the job level.

7. The classification and pricing method for 10kV distribution network live-line work fees according to claim 6, characterized in that, The process of matching the corresponding standard resource consumption based on the task level to obtain the baseline resource input value is as follows: Based on the level of the target task, the corresponding resource combination and the recommended input amount of each type of resource are extracted from the dynamic resource input strategy table to form the initial resource input structure corresponding to the level. Construct a resource category indicator matrix based on the initial resource input structure; The resource category index matrix is ​​standardized to obtain a normalized matrix. Calculate the entropy and information utility of various resources based on the normalized matrix; Entropy weights for various resources are calculated based on information utility. Based on entropy weighting, the standard input quantities of various resources in the initial resource input structure are weighted and accumulated to obtain the resource benchmark input value.

8. The classification and pricing method for 10kV distribution network live-line work fees according to claim 1, characterized in that, The determination of the benchmark unit price based on the initial classification results is specifically as follows: Based on the job types and corresponding job complexity indicators in the initial classification results, a two-dimensional benchmark unit price matrix is ​​constructed, and the initial value of each cell in the benchmark unit price matrix is ​​filled to form a benchmark price lookup table. For any operation to be priced, the benchmark unit price is obtained by matching it with the benchmark unit price matrix.

9. The classification and pricing method for 10kV distribution network live-line work fees according to claim 8, characterized in that, The process of adjusting the benchmark unit price based on the resource benchmark input value, generating a categorized pricing table, and forming traceable cost accounting records is as follows: Obtain the actual resource input value and calculate the deviation by combining it with the resource baseline input value to obtain the resource deviation; Obtain the operation impact coefficient and risk level confidence level, and adjust the benchmark unit price based on resource deviation; A categorized pricing table is generated based on the revised benchmark unit price and the corresponding work data; Record the categorized pricing schedule and establish a cost traceability chain structure to form a traceable cost accounting record.

10. A system for classifying and pricing 10kV distribution network live-line work fees using the method described in any one of claims 1-9, characterized in that, include: The classification module is used to collect line data for the target operation and perform initial classification of live-line operations; The classification module is used to classify different types of jobs based on the initial classification results; The resource baseline input value acquisition module is used to construct a corresponding dynamic resource input strategy table based on the classification results, and match the corresponding standard resource consumption according to the operation level to obtain the resource baseline input value. The classification pricing table generation module is used to determine the benchmark unit price based on the initial classification results, and to correct the benchmark unit price based on the benchmark resource input value, thereby generating a classification pricing table and forming a traceable cost accounting record.

Images