A user electricity price execution anomaly identification and positioning method

Abstract

The present application relates to a kind of user tariff execution anomaly identification and positioning method, obtain the user profile of target user in settlement period, metering, tariff rules, rule source, metering point association and actual electricity bill data, determine reference execution parameter set and generate reference execution imprint matrix;Based on expense item type, billing input data generate reference expense item imprint bearing tensor;Candidate execution parameter domain is constructed, candidate bill and candidate execution imprint matrix are generated, and candidate execution matching cost is determined;The candidate execution parameter set with the lowest matching cost is determined as the inversion execution parameter set, and inversion expense item imprint bearing tensor is generated;According to the deviation of reference expense item imprint bearing tensor and inversion expense item imprint bearing tensor, tariff execution anomaly is identified and the abnormal position is positioned.The method can improve the identification accuracy and positioning accuracy of implicit tariff execution anomaly.

Description

Technical Field

[0001] This invention relates to a method for identifying and locating abnormal user electricity price execution, belonging to the field of data processing technology. Background Technology

[0002] Electricity billing for users typically involves processing information such as user profiles, metering data, electricity price categories, voltage levels, time-of-use pricing rules, basic electricity charge calculation methods, power factor adjustment rules, and various discounts, reductions, and surcharges. As electricity pricing rules become more detailed and user metering relationships more complex, a single user's bill often consists of multiple charges, each potentially relying on different data sources and billing rules. For example, time-of-use charges usually depend on peak and off-peak electricity consumption and time-of-use rates; basic electricity charges typically depend on contracted capacity, operating capacity, or maximum demand; power factor adjustment charges typically depend on power factor assessment standards and cost bases; and some charges may also be affected by metering point affiliation, rule effective dates, and changes to user profiles.

[0003] In actual electricity billing, inconsistencies in data synchronization or configuration may occur due to user profile changes, electricity price rule version switching, metering point adjustments, billing parameter configuration, and expense item summary processing. Abnormal electricity price execution may manifest as incorrect electricity price category use, inconsistent voltage levels, failure to switch time-of-use rules in a timely manner, incorrect execution of basic electricity pricing methods, incorrect metering point attribution, or omission of preferential or additional rules. Such anomalies can sometimes cause significant fluctuations in bill amounts. However, in some cases, the amount calculated by incorrect rules may be close to the correct amount, or the anomaly may only be reflected in a specific expense item, making it difficult to detect in a timely manner through changes in the total electricity bill.

[0004] In existing technologies, handling anomalies in user electricity pricing typically includes manual review, fixed rule verification, year-on-year and month-on-month billing analysis, and anomaly detection based on historical data. Manual review requires staff to check user files, metering point relationships, electricity price tables, and bill details item by item, resulting in low processing efficiency and difficulty in batch verification of a large number of users. Fixed rule verification usually relies on pre-configured judgment conditions and is suitable for identifying obvious anomalies such as missing fields, exceeding numerical limits, and inconsistent fee items. However, its ability to identify complex anomalies caused by the combined effects of multiple billing stages is limited. Year-on-year and month-on-month billing analysis and anomaly detection based on historical fee fluctuations mainly focus on the changing trends of fee results. When user electricity consumption behavior fluctuates significantly or the changes in abnormal amounts are not obvious, misjudgments or omissions are prone to occur. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention proposes a method for identifying and locating abnormal user electricity price execution.

[0006] The technical solution of the present invention is as follows: This invention provides a method for identifying and locating anomalies in user electricity pricing, comprising the following steps: Acquire user profile data, metering data, electricity pricing rule data, rule source data, metering point association data, and actual electricity bill data for the target user within the target billing period; Based on user profile data, electricity price rule data, and metering point association data, determine the set of benchmark execution parameters for the target user; Generate a benchmark execution imprint matrix based on the benchmark execution parameter set and rule source data; Based on the types of expense items in the actual electricity bill, the baseline execution imprint matrix, and the billing input data for each expense item, a baseline expense item imprint acceptance tensor is generated. Based on the baseline execution parameter set and the execution parameter offset operator, candidate execution parameter domains are constructed, and candidate bills and candidate execution imprint matrices are generated for each candidate execution parameter set; Based on the differences between candidate bills and actual electricity bills, the rule feasibility relationship of candidate execution parameter sets, and the imprint deviation relationship between candidate execution imprint matrices and benchmark execution imprint matrices, the candidate execution matching cost corresponding to each candidate execution parameter set is determined. The set of candidate execution parameters with the lowest matching cost is determined as the inversion execution parameter set, and the inversion cost item imprint tensor is generated based on the inversion execution parameter set. Based on the deviation between the baseline cost item imprint acceptance tensor and the inverted cost item imprint acceptance tensor, determine whether the target user has any abnormal electricity price execution and locate the abnormal location.

[0007] Preferably, the set of benchmark execution parameters includes at least one of the following: electricity consumption category parameters, voltage level parameters, time-of-use electricity price version parameters, basic electricity charge execution method parameters, power factor adjustment rule parameters, and metering point attribution parameters.

[0008] Preferably, the rule source data includes at least one of user profile source data, metering point source data, electricity price rule version source data, and bill generation source data; Among them, the user profile source data includes at least one of the following: user electricity consumption category, voltage level, contract capacity, and operating capacity; the metering point source data includes at least one of the following: metering point number, metering point multiplier, and metering point affiliation; the electricity price rule version source data includes at least one of the following: electricity price meter version number, time-of-use version number, or rule effective date; and the bill generation source data includes at least one of the following: bill generation batch, billing module identifier, or fee item generation order.

[0009] Preferably, a benchmark execution imprint matrix is ​​generated based on the benchmark execution parameter set and rule source data, including: The baseline execution parameter set and rule source data are embedded and encoded respectively to obtain execution parameter embedding vectors and rule source embedding vectors; Construct an interaction embedding vector based on the execution parameter embedding vector and the rule source embedding vector; The execution parameter embedding vector, rule source embedding vector, and interaction embedding vector are mapped to the imprint space and fused to generate the corresponding baseline imprint unit.

[0010] Preferably, the baseline cost item imprint acceptance tensor includes a cost item dimension, an execution parameter dimension, and a rule source dimension. Each baseline acceptance imprint in the tensor is used to represent the acceptance relationship of a cost item to a baseline execution parameter under a rule source dimension. The fee items include at least one of the following: electricity consumption fee, time-of-use fee, basic fee, maximum demand adjustment fee, power factor adjustment fee, additional fees, and bill summary.

[0011] Preferably, the generation of the baseline cost item imprint acceptance tensor includes: The cost item type, baseline execution parameters, and billing input data of the cost item are embedded and encoded to obtain the cost item type embedding vector, execution parameter embedding vector, and billing input embedding vector, respectively. The acceptance score of a cost item to the benchmark execution parameter is determined based on the cost item type embedding vector, execution parameter embedding vector, and billing input embedding vector. Multiple acceptance scores corresponding to the same cost item are normalized to obtain the acceptance strength. Based on the acceptance strength, the corresponding benchmark imprint unit, cost item type embedding vector and billing input embedding vector in the benchmark execution imprint matrix are mapped and fused to generate the benchmark acceptance imprint corresponding to the cost item.

[0012] Preferably, the execution parameter offset operator includes at least one of the following: electricity category replacement operator, voltage level replacement operator, time-of-use price version offset operator, basic electricity charge execution mode switching operator, power factor adjustment rule replacement operator, and metering point affiliation replacement operator; When the execution parameter offset operator is applied to the base execution parameter set, only the execution parameter corresponding to the execution parameter offset operator is changed, while the other execution parameters keep the base value unchanged. The candidate values ​​for the execution parameter offset operator are derived from at least one of the following: historical execution parameter values, adjacent voltage levels, adjacent power factor assessment standards, historical electricity price rule versions, and adjacent metering point affiliation relationships.

[0013] Preferably, the candidate execution matching cost includes bill matching cost, rule feasibility penalty item, and imprint deviation cost; Among them, the bill matching cost is determined based on the difference in expense items between the actual electricity bill data and the candidate bill, as well as the covariance matrix of expense item fluctuations. The rule feasibility penalty is determined based on whether the candidate execution parameter belongs to the set of allowed values ​​defined by the electricity price rule version data and user profile data; The imprint deviation cost is determined based on the vector similarity between each candidate imprint unit in the candidate execution imprint matrix and the corresponding benchmark imprint unit in the benchmark execution imprint matrix.

[0014] Preferably, based on the deviation between the baseline cost item imprint acceptance tensor and the inverted cost item imprint acceptance tensor, it is determined whether the target user has any abnormal electricity price execution, including: Subtract the baseline acceptance mark in the baseline cost item imprint acceptance tensor from the corresponding inversion acceptance mark in the inversion cost item imprint acceptance tensor to obtain the cost item imprint deviation tensor. The L2 norm of each imprint deviation vector in the cost item imprint deviation tensor is calculated to obtain the imprint deviation values ​​corresponding to the cost item, execution parameters and rule source dimensions. Based on the imprint deviation values ​​of different cost items under the same execution parameters and the same rule source dimension, and the dependencies between cost items, a deviation linkage matrix is ​​generated; The execution consistency gap is calculated based on the candidate execution matching cost, the cost item imprint deviation tensor, and the deviation linkage matrix, and the execution consistency gap is used to determine whether there is an abnormal electricity price execution for the target user.

[0015] Preferably, when it is determined that there is an abnormality in the electricity price execution of the target user, for each expense item, each execution parameter and each rule source dimension, the candidate initial abnormality association point score is determined based on the corresponding imprint deviation value, the active linkage contribution of the expense item to other expense items and the passive linkage influence of other expense items on the expense item. The cost item, execution parameter, and rule source dimension with the highest score of the candidate initiation anomaly association point are determined as the initiation anomaly cost item, initiation anomaly execution parameter, and initiation anomaly rule source.

[0016] The present invention has the following beneficial effects: 1. This invention determines the set of benchmark execution parameters by using user profile data, electricity price rule version data, rule application condition data, and metering point association data, and generates a benchmark execution imprint matrix. Then, the actual electricity bill is used to reverse-engineer the set of execution parameters. Therefore, anomaly identification no longer relies solely on whether the bill amount exceeds historical fluctuation ranges, but rather on determining whether the execution parameters actually reflected in the actual electricity bill deviate from the benchmark execution parameters. This enables the identification of hidden anomalies with inconspicuous bill amount changes, such as misuse of time-of-use pricing versions, mismatch of basic electricity fee execution methods, missed implementation of preferential rules, and incorrect metering point attribution.

[0017] 2. This invention uses a baseline cost item imprint bearing tensor and an inverted cost item imprint bearing tensor to structurally express the correspondence between cost items, execution parameters, and rule source dimensions, and generates a cost item imprint deviation tensor based on the deviation between the two. Therefore, the anomaly localization result no longer stops at the level of "user anomaly" or "bill anomaly," but can further locate the initial abnormal cost item, the initial abnormal execution parameter, and the initial abnormal rule source. For example, it can locate an anomaly in the time-of-use electricity price version within the time-of-use electricity price item, an abnormal execution method within the basic electricity price item, or an abnormal attribution within the metering point source.

[0018] 3. This invention generates a deviation linkage matrix based on the expense item imprint deviation tensor, and determines the initial abnormal linkage point by combining the expense item's own deviation value, active linkage contribution, and passive linkage impact. Therefore, when an execution parameter error simultaneously affects multiple expense items, it can distinguish between the initial abnormal expense item and the affected expense items, avoiding mislocation based solely on the magnitude of expense item deviation. This improves the location accuracy in complex linkage anomaly scenarios and reduces the workload of manually checking user files, metering point relationships, electricity price rule versions, and bill expense items item by item. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

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

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

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

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

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

[0025] See Figure 1 In some embodiments, a method for identifying and locating anomalies in user electricity pricing is proposed, including the following steps: Acquire user profile data, metering data, electricity pricing rule data, rule source data, metering point association data, and actual electricity bill data for the target user within the target billing period; Based on user profile data, electricity price rule data, and metering point association data, determine the set of benchmark execution parameters for the target user; Generate a benchmark execution imprint matrix based on the benchmark execution parameter set and rule source data; Based on the types of expense items in the actual electricity bill, the baseline execution imprint matrix, and the billing input data for each expense item, a baseline expense item imprint acceptance tensor is generated. Based on the baseline execution parameter set and the execution parameter offset operator, candidate execution parameter domains are constructed, and candidate bills and candidate execution imprint matrices are generated for each candidate execution parameter set; Based on the differences between candidate bills and actual electricity bills, the rule feasibility relationship of candidate execution parameter sets, and the imprint deviation relationship between candidate execution imprint matrices and benchmark execution imprint matrices, the candidate execution matching cost corresponding to each candidate execution parameter set is determined. The set of candidate execution parameters with the lowest matching cost is determined as the inversion execution parameter set, and the inversion cost item imprint tensor is generated based on the inversion execution parameter set. Based on the deviation between the baseline cost item imprint acceptance tensor and the inverted cost item imprint acceptance tensor, determine whether the target user has any abnormal electricity price execution and locate the abnormal location.

[0026] In one specific embodiment, a method for identifying and locating anomalies in user electricity pricing is proposed, including the following steps: S1. Obtain user electricity price execution data: Acquire user profile data, metering data, electricity price rule version data, metering point association data, rule source data, and actual electricity bill data for the target user within the target billing period; The user profile data includes the user's electricity consumption category, voltage level, contracted capacity, operating capacity, basic electricity fee execution method, power factor assessment standard, and industry attributes. Metering data includes total electricity consumption, peak and off-peak hourly electricity consumption, maximum demand, power factor, metering point multiplier, and metering point affiliation. Electricity pricing rule version data is used to characterize the electricity pricing and billing rules that target users should adopt within the target billing cycle and their version status. It includes at least one of the following: electricity price table version data, time-of-use version data, rate parameter version data, billing method version data, rule application condition version data, and rule change record data. Among them, electricity price table version data is used to determine the applicable electricity price table, time-of-use version data is used to determine the peak and off-peak time period division rules, rate parameter version data is used to determine the rate parameters corresponding to each fee item, billing method version data is used to determine the electricity billing calculation method, rule application condition version data is used to determine the scope of rule application, and rule change record data is used to determine the rule version switching status within the target billing cycle. Metering point association data is used to characterize the billing association relationship between a target user and one or more metering points, including at least one of the following: metering point identifier, metering point affiliation relationship, metering point type, metering point aggregation relationship, metering point conversion parameters, metering point billing application relationship, and metering point validity period; wherein, the metering point aggregation relationship is used to determine the aggregation, deduction, or allocation method of metering data from different metering points in the target user's bill. The data source for the rules includes at least one of the following: user profile data, metering point data, electricity price rule version data, and bill generation data. Among them, the user profile source data includes at least one of the following: user electricity consumption category, voltage level, contract capacity, and operating capacity; the metering point source data includes at least one of the following: metering point number, metering point multiplier, and metering point affiliation; the electricity price rule version source data includes at least one of the following: electricity price meter version number, time-of-use version number, or rule effective date; and the bill generation source data includes at least one of the following: bill generation batch, billing module identifier, or fee item generation order.

[0027] Actual electricity bill data includes electricity consumption charges, time-of-use charges, basic electricity charges, power factor adjustment charges, discount charges, additional charges, and total electricity charges. The electricity pricing data for the target users is represented as follows: ; in: Indicates the first The user in the first Electricity price execution data for each settlement period; Indicates the first The user in the first User profile data for each settlement cycle; Indicates the first The user in the first Measurement data for each settlement cycle; Indicates the first The user in the first Electricity pricing rule version data for each settlement cycle; Indicates the first The user in the first Metering point associated data for each settlement cycle; Indicates the first The user in the first Actual electricity bill data for each settlement period; Actual electricity bill data can be represented as a vector of expense items: ; in: Indicates the first The user in the first In the actual electricity bill of the current billing cycle, the first Amount of each expense item; Indicates the total number of expense items; Cost Item Dimensions Used to represent the cost components in an electricity bill that can be independently traced. For example, This indicates the electricity consumption and electricity charges. This indicates the peak / valley time-of-use electricity pricing item. This indicates the basic electricity fee item. This indicates the maximum demand adjustment item. This indicates the power factor adjustment electricity fee item. This indicates the discounted fee item. This indicates additional fees. This represents the summary item of the bill. If there are other expense items in the bill, they can be added as a new expense item dimension.

[0028] S2. Determine the set of baseline execution parameters: Based on user profile data, electricity price rule version data, and metering point association data, determine the set of benchmark execution parameters that the target user should execute within the target billing cycle.

[0029] The set of baseline execution parameters is represented as follows: ; in: Indicates the first The user in the first The set of baseline execution parameters for each settlement cycle; Indicates the first The user in the first The first settlement cycle One benchmark execution parameter; Indicates the number of execution parameters; Execution parameter dimensions Used to represent key parameters that affect the implementation of electricity prices. For example, Indicates the electricity consumption category parameter, Indicates voltage level parameters, This indicates the parameters for the time-of-use electricity pricing version. This indicates the parameters for the basic electricity charge execution method. This indicates the power factor adjustment rule parameters. This indicates the parameters to which the metering point belongs. The above correspondence is not a limitation; in actual implementation, tiered pricing parameters, surcharge parameters, or special industry pricing parameters can be added according to the billing rules.

[0030] S3. Generate the baseline execution imprint matrix: The user's baseline execution state is constructed into a parameter-rule source two-dimensional imprint matrix.

[0031] The baseline execution imprint matrix is ​​represented as follows: ; in: Indicates the first The user in the first The benchmark execution imprint matrix for each settlement cycle; Indicates the first The user in the first The first settlement cycle The benchmark execution parameter at the ... Benchmark imprint units under each rule source dimension; Indicates the number of dimensions from which the rules originate; Rule source dimension Used to distinguish the data source or rule source upon which the same execution parameter is based. For example, when... When corresponding to the user profile source, when When the corresponding metering point source is... The source of the corresponding electricity pricing rule version, when This corresponds to the source of the invoice generation. In actual implementation, the number of rule source dimensions... It can be expanded based on the source type participating in billing in the electricity pricing system.

[0032] set up Indicates user In the cycle The Middle The source data corresponding to each rule's source dimension. Specifically, if the first... If the source dimension of a rule is the user profile, then... It can be a collection of user profile fields; if it is the source of metering points, then... This can include the metering point number, multiplier, and metering point affiliation; if it's the source of the electricity pricing rule version, then... This can be the electricity meter version number, the time-of-use version number, or the rule's effective date; if it's the source of the bill generation, then... You can specify the batch, billing module identifier, or order in which bill expense items are generated.

[0033] Each reference imprint unit is generated by the following formula: ; in: Indicates the first The encoding matrix corresponding to each benchmark execution parameter; Indicates the first The encoding matrix corresponding to each rule source dimension; Indicates the first The first benchmark execution parameter and the second Interactive encoding matrix between the source dimensions of each rule; Indicates the first Embedded representation of each benchmark execution parameter; Indicates the first Embedded representation of the source dimension data for each rule; This indicates element-wise multiplication; Indicates the bias term; This represents a non-linear activation function.

[0034] This represents an embedding encoding function used to convert discrete, continuous, or combined fields into a vector representation of the same dimension. For discrete fields such as electricity category, voltage level, and electricity price version, the embedding encoding function can use a lookup table embedding method; for continuous fields such as electricity consumption, demand, capacity, and power factor, the embedding encoding function can use a normalized linear mapping method; for combined fields such as metering point affiliation and bill generation order, the embedding encoding function can use a field concatenation mapping method.

[0035] , , All of these are trainable mapping parameters. Among them, Used to the first The embedding of each execution parameter is mapped to the imprint space. Used to the first The embedding of each rule source data is mapped to the imprint space. This is used to embed and map the interaction between execution parameters and rule source data into the imprint space. If the imprint unit dimension is... And the dimensions of each embedding vector are ,but , , The output dimensions are all This ensures that the three results can be added together.

[0036] S4. Establish the baseline cost item imprint bearing tensor: Based on the types of expense items in the actual electricity bill, establish the imprinting relationship between expense items, execution parameters, and rule source dimensions to form a baseline expense item imprinting tensor.

[0037] The basis cost item imprint acceptance tensor is represented as follows: ; in: Indicates the first The user in the first The base cost item imprint acceptance volume for each settlement cycle; Indicates the first The user in the first The first settlement cycle The first expense item is related to the first The benchmark execution parameter at the ... Benchmark inheritance imprint under the source dimension of each rule; set up For the first The user in the first The first settlement cycle The billing input data corresponding to each cost item; if the first... If the expense item is a time-of-use electricity charge, then This includes peak electricity consumption, off-peak electricity consumption, average electricity consumption, valley electricity consumption, and time-of-use pricing; if it is a basic electricity charge item, then... This includes contracted capacity, operating capacity, maximum demand, and basic electricity tariff implementation method; if it is a power factor adjustment tariff item, then... This includes power factor, electricity consumption and base electricity price, and power factor reward and penalty rules; if it is a preferential fee item, then... This includes eligibility for the discount, the base amount for the discount, and the discount percentage; if it is an additional fee item, then... This includes the surcharge category, surcharge base, and surcharge rate.

[0038] No. The first expense item is related to the first The strength of acceptance of each benchmark execution parameter is expressed as follows: ; in: Indicates the first The user in the first The first settlement cycle The first expense item is related to the first The strength of acceptance of each benchmark execution parameter; Indicates the first Each expense item type; Indicates the execution parameter number that participated in the normalization; A mapping function representing the relationship between cost items and execution parameters; Specifically, the mapping function between cost items and execution parameters can be expressed as: ; in: This represents the strength of the acceptance rating vector; Represents the cost item type mapping matrix; This represents the execution parameter mapping matrix; This represents the mapping matrix of billing input data for expense items; Indicates the load-bearing strength bias term; This indicates the transpose operation.

[0039] , , All are trainable mapping matrices, used to map cost item types, execution parameters, and cost item billing input data to the same relational space. Specifically... This is used to indicate the differences in the execution parameters of different fee item types. For example, there is usually a strong correlation between the time-of-use electricity fee item and the time-of-use electricity price version parameter, and there is usually a strong correlation between the basic electricity fee item and the basic electricity fee execution method parameter. Used to indicate the ability of the execution parameters themselves to affect the cost item; This is used to represent the impact of billing input data on the acceptance relationship. For example, the peak and off-peak electricity structure in the time-of-use electricity charge item will affect its acceptance strength to the time-of-use electricity price version parameters.

[0040] Rating Vector This can be obtained through model training; during training, if historical or simulated samples show that a certain expense item is indeed mainly affected by a certain execution parameter, the model will improve the score of the corresponding relationship, increasing the strength of that expense item's relationship with that execution parameter. Increase; For example, for time-of-use electricity charges, if its billing input data... If the data includes peak electricity consumption, average electricity consumption, valley electricity consumption, and time-of-use pricing, the trained model will typically have a high degree of acceptance of the "time-of-use pricing version parameters." For the basic electricity charge item, the model will typically have a high degree of acceptance of the "basic electricity charge execution method parameters" and the "contract capacity / maximum demand related parameters." For the power factor adjustment electricity charge item, the model will typically have a high degree of acceptance of the "power factor adjustment rule parameters."

[0041] Construct corresponding benchmark bearing marks based on bearing strength: ; ; in: Indicates the first The imprint mapping matrix corresponding to each cost item; Indicates the first A cost type mapping matrix corresponding to each cost item; Indicates the first The billing input mapping matrix corresponding to each cost item; Indicates the first Each cost item corresponds to a biased acceptance item; Different cost items 、 、 These can be set independently or shared among cost items of the same category. For example, peak electricity charges, off-peak electricity charges, and valley electricity charges all belong to time-of-use electricity charges and can share some mapping matrices to reduce the number of model parameters; while the basic electricity charge and the power factor adjustment electricity charge have different billing logics and can use different mapping matrices.

[0042] The aforementioned mapping matrix can be obtained through joint training using historical normal samples and simulated abnormal samples. During training, the model adjusts its parameters by comparing the deviation between the baseline cost item imprint tensor and the inverted cost item imprint tensor. 、 、 This ensures that the baseline and inverted signatures of the same expense item remain close under normal execution conditions, while the signature deviations of the corresponding expense item, execution parameters, and rule source dimensions can be amplified under abnormal execution conditions.

[0043] S5. Construct candidate execution parameter domains: Based on the baseline set of execution parameters, candidate execution parameter domains are generated according to execution parameter offset operators that may cause abnormal electricity price execution.

[0044] The candidate execution parameter field is represented as follows: ; in: Indicates the first The user in the first Candidate execution parameter fields for each settlement cycle; Indicates the first One execution parameter offset operator; This indicates the number of candidate execution parameter sets. Its value is determined by the number of execution parameter offset operators and the number of candidate values ​​for each offset operator. To avoid the candidate execution parameter field being too large, execution parameter offset operators can be pre-filtered based on historical anomaly records, rule version change records, and user profile change records, and the number of parameters offset in each candidate execution parameter set generation can be limited.

[0045] This is used to generate a candidate set of execution parameters within a finite neighborhood of the baseline set of execution parameters. The execution parameter offset operator does not change parameters that are obviously irrelevant to the current user, but only performs limited replacement, switching, or version rollback on parameters that are related to abnormal electricity price execution. The finite neighborhood includes historically executed parameter values, adjacent voltage level parameter values, adjacent power factor assessment standards, previous or next version electricity price rules, and adjacent metering point affiliation relationships. This indicates that the execution parameter offset operator is applied to the baseline execution parameter set; when Acting on the benchmark execution parameter set At that time, only replace or modify One or more execution parameters are specified, while the remaining un-offset execution parameters retain their baseline values. For example, if For the time-of-use pricing version offset operator, then This means that only the base time-of-use tariff version is replaced with the candidate time-of-use tariff version, while other parameters such as electricity consumption category, voltage level, and basic electricity charge execution method remain unchanged.

[0046] Each candidate set of execution parameters is represented as: ; in: Indicates the first The user in the first The first settlement cycle A set of candidate execution parameters; The execution parameter offset operator includes at least one of the following: Replace the benchmark electricity consumption category with an adjacent or historical electricity consumption category; replace the benchmark voltage level with an adjacent voltage level; replace the benchmark time-of-use electricity price version with a historical time-of-use electricity price version; switch the basic electricity charge execution method between capacity-based and demand-based; replace the power factor adjustment standard with an adjacent assessment standard; replace the metering point affiliation with adjacent metering points or adjacent users; S6. Generate candidate bills and candidate execution imprint matrices: For each set of candidate execution parameters in the candidate execution parameter domain, generate candidate bills based on measurement data and user profile data: ; in: Indicates the first The user in the first The first settlement cycle Candidate bills generated from a set of candidate execution parameters; Represents a differentiable billing mapping function; Differentiable billing mapping functions are used to calculate charges based on metering data and user profile data according to a set of candidate execution parameters, outputting candidate bills with the same charge item structure as actual electricity bills. Differentiable billing mapping functions include sub-functions for time-of-use charge calculation, basic charge calculation, power factor adjustment charge calculation, discount charge calculation, surcharge calculation, and bill summary. For billing rules with segmented, tiered, or conditional judgments, conditional masks or smooth approximation functions can be used to represent them, allowing them to participate in error backpropagation during model training.

[0047] Candidate bills can be represented as: ; in: Indicates the first The user in the first The first settlement cycle The candidate bill generated from the candidate execution parameter set. Amount of each candidate expense item; Simultaneously, a candidate execution imprint matrix is ​​generated based on the candidate execution parameter set: ; in: Indicates the first The user in the first The first settlement cycle Candidate execution imprint matrices corresponding to each set of candidate execution parameters; Indicates the first The user in the first The first settlement cycle The th candidate execution parameter set The execution parameter is at the _ ... Candidate imprint units under each rule source dimension.

[0048] S7. Construct candidate execution matching costs and inversely calculate execution parameters: For each set of candidate execution parameters, calculate its candidate execution matching cost: ; ; in: Indicates the first The user in the first The first settlement cycle The candidate execution matching cost for each set of candidate execution parameters; This represents the Mahalanobis distance weighted according to the cost item fluctuation covariance matrix; Represents the cost item fluctuation covariance matrix; This represents the rule feasibility penalty term between the candidate set of execution parameters and the version of the electricity pricing rule; This represents the imprint deviation cost between the candidate execution imprint matrix and the baseline execution imprint matrix; , This represents the adjustment coefficient. , This can be achieved through parameter tuning on the validation set, or it can be adaptively adjusted based on the recognition performance of different anomaly types in historical anomaly samples. For example, when the system has a high false negative rate for rule version errors on the validation set, the parameter tuning can be improved. When the system mistakenly selects a set of candidate execution parameters that do not meet the rules but have similar billing as the inversion result, it can improve... ; Cost Item Fluctuation Covariance Matrix This is used to represent the fluctuation scale and correlation of different expense items in historical normal invoices. Since different expense items have different magnitudes and degrees of fluctuation, directly using ordinary Euclidean distance may lead to larger expense items dominating invoice differences. Therefore, this embodiment uses Mahalanobis distance to normalize invoice differences. For expense items with large historical fluctuations, This will reduce the impact of differences on matching costs; for cost items with small historical fluctuations but significant current differences, This will increase the impact of the difference on the matching cost.

[0049] The cost item fluctuation covariance matrix can be estimated based on historical normal bills: ; in: This indicates the number of historical periods used to estimate the covariance matrix; Indicates the first The user in the first A vector of billing expenses for each historical billing cycle; Indicates the first The vector mean of historical billing expense items for each user; This represents the regularization coefficient used to avoid matrix non-invertibility; Represents the identity matrix; The penalty for rule feasibility is represented as follows: ; in: Indicates the first The user in the first The first settlement cycle The first in the candidate execution parameter set One execution parameter; This indicates that, under the constraints of electricity pricing rule version data and user profile data, the first... Each execution parameter has a set of allowed values, which can be jointly determined by the billing rule table, user profile constraint table, and metering point association rule table. For example, for the electricity category parameter, the set of allowed values ​​can be determined by the user's industry attribute, voltage level, and application category; for the time-of-use pricing version parameter, the set of allowed values ​​can be determined by the pricing rule version, rule effective date, and settlement cycle; for the basic electricity charge execution method parameter, the set of allowed values ​​can be determined by whether the user is a two-part tariff user, contract capacity, and maximum demand collection conditions; for the metering point affiliation parameter, the set of allowed values ​​can be determined by the metering point profile, user number, and metering point topology. Indicates an indicator function; Indicates the first The penalty coefficient for violating the rule's feasibility for each execution parameter varies, and the impact of violations differs for different execution parameters; therefore, different penalties can be set. For example, parameters such as electricity category, voltage level, and time-of-use pricing version have a significant impact on bill generation. It can be set to a higher level; the surcharge parameter can be set with a corresponding penalty coefficient based on its impact on the statement. It can be preset by business rules, or it can be adjusted during the training process based on the impact of different parameter violations on the final anomaly location accuracy.

[0050] In particular, in one specific embodiment, It can be obtained in the following ways: First, it is generated based on historical anomaly records. For anomaly types that have occurred in the past, such as incorrect use of time-of-use pricing versions, incorrect implementation of basic electricity charges, and mismatch of power factor adjustment standards, the corresponding parameter replacement methods can be solidified into execution parameter offset operators.

[0051] Second, it is generated based on the electricity pricing rule version change record. When the electricity pricing rule is updated, the replacement relationship between the previous version, the current version, and adjacent versions can be used as the time-of-use electricity pricing version offset operator.

[0052] Third, it is generated based on user profile change records. When a user's electricity category, voltage level, contract capacity, or metering point affiliation changes, the parameter values ​​before and after the change can be used as candidate offset values ​​to generate the corresponding execution parameter offset operator.

[0053] Fourth, the set of allowed values ​​is generated according to the rules. For a given execution parameter, if its set of allowed values... There are multiple candidate values, and the baseline value can be replaced with other allowed values ​​to form a candidate offset operator.

[0054] In this way, the candidate execution parameter field is not infinitely enumerated, but generated within a limited range where erroneous execution may occur in the business, thus balancing computational efficiency and anomaly coverage.

[0055] The imprint deviation cost between the candidate execution imprint matrix and the baseline execution imprint matrix is ​​expressed as: ; The set of execution parameters for inversion is determined by the set of candidate execution parameters that has the lowest matching cost among the candidate execution parameters: ; ; in: This represents the set of inversion execution parameters obtained from the actual electricity bill. This indicates the candidate execution parameter number with the lowest execution matching cost.

[0056] S8. Generate the inversion cost item imprint carrying tensor: Based on the inverse execution parameter set obtained by inversion, generate the inversion execution imprint matrix: ; in: Indicates the first The user in the first Inverse execution imprint matrix for each settlement cycle; Indicates the first The user in the first The first settlement cycle The inversion execution parameter is at the... Inversion imprint units under the source dimension of each rule; Based on the inverted execution imprint matrix, generate the inverted cost item imprint carrying tensor: ; in: Indicates the use of the first The user in the first Inversion cost item imprint tensor for each settlement cycle; Indicates the first The user in the first The first settlement cycle The first expense item is related to the first The inversion execution parameter is at the... Inversion and inheritance imprints under the source dimension of each rule; The generation method of the inversion cost item imprint acceptance tensor is the same as that of the baseline cost item imprint acceptance tensor, except that the baseline execution parameter set is replaced with the inversion execution parameter set.

[0057] S9, Construct the cost item imprint deviation tensor: By comparing the baseline cost item imprint acceptance tensor and the inverted cost item imprint acceptance tensor, the cost item imprint deviation tensor is obtained: ; ; in: Indicates the first The user in the first Cost item imprint deviation tensor for each settlement cycle; Indicates the first The user in the first The first settlement cycle One expense item, the first The execution parameter, the first Imprint bias vector under each rule source dimension; This indicates the corresponding acceptance mark in the acceptance tensor of the benchmark cost item; This indicates the corresponding acceptance mark in the tensor of the inversion cost item acceptance mark; Furthermore, generate the expense item imprint deviation value: ; in: Indicates the first The user in the first The first settlement cycle One expense item, the first The execution parameter, the first Imprint deviation value under each rule source dimension; The imprint deviation value can express: which cost item, when receiving which execution parameter, and on which rule source dimension the deviation occurs.

[0058] For example, if A large value indicates that the time-of-use electricity charge item deviates from the source of the rule version when accepting the time-of-use electricity price version parameters, which may be due to an error in the execution of the corresponding time-of-use electricity price version.

[0059] S10. Construct the deviation linkage matrix: Based on the deviation values ​​of the expense item imprints, construct a deviation linkage matrix between expense items: ; ; in: Indicates the first The user in the first Deviation linkage matrix for each settlement cycle; Indicates the first The user in the first The first settlement cycle The imprint deviation of the first expense item is different from the first. The strength of the directional linkage between the deviations in the imprints of individual expense items; Indicates the first The user in the first The first settlement cycle The first expense item and the first The expense item in the first The execution parameter and the first The linkage coefficient under each rule source dimension; Indicates the first The user in the first The first settlement cycle One expense item, the first The execution parameter, the first Imprint deviation value under each rule source dimension; It can be pre-configured based on the dependencies between cost items in the electricity pricing rules, or it can be learned from historical anomaly samples. If cost items... Generation dependency cost item And both are affected by the first The execution parameter and the first The influence of each rule's source dimension, then Take the larger value; if there is no dependency between the two, then Take zero or a smaller value.

[0060] The correlation coefficient can also be expressed as: ; in: Indicates cost item Does it depend on the cost item? ; Indicates the first The execution parameter is at the _ ... The linkage weights under each rule's source dimension; The directionality of deviation from the linkage matrix is ​​determined by Decision. If the bill summary item depends on the time-of-use electricity item, then you can set... Greater than This indicates that deviations are more likely to be linked from time-of-use electricity items to bill summary items, rather than vice versa.

[0061] S11. Calculate execution consistency gaps and identify anomalies: Calculate the execution consistency gap based on the optimal candidate execution matching cost, the optimal candidate execution imprint deviation cost, the cost item imprint deviation tensor, and the deviation linkage matrix: ; ; in: Indicates the first The user in the first Consistency gap in execution across settlement cycles; This represents the cost of matching the optimal set of candidate execution parameters to the actual invoice. This represents the deviation cost of the optimal candidate execution imprint matrix relative to the baseline execution imprint matrix; The mixed norm of the expense item imprint deviation tensor; Denotes the F-norm; An electricity price execution anomaly is determined to exist when the consistency gap meets the following conditions: ; in: Indicates target user There are irregularities in electricity price implementation; Indicates target user There are no abnormalities in electricity price implementation; Indicates the first The user in the first The anomaly detection threshold for each settlement cycle; The anomaly detection threshold can be dynamically determined based on execution consistency gaps within the user's historical normal cycles: ; in: Indicates the first The average number of consistency gaps executed by each user within a normal historical period; No. The standard deviation of consistency gaps for each user within a normal historical period; Indicates the threshold amplification factor; The aforementioned execution consistency gap means that if the actual electricity bill is difficult to interpret using the optimal candidate execution parameter set, then... Increase; If the optimal candidate set of execution parameters can explain the actual electricity bill, but this candidate set of execution parameters deviates from the baseline set of execution parameters, then Increase; If the cost item imprint deviates across multiple execution parameters and rule source dimensions, then Increase; If the deviation from the linkage exhibits a clear direction, it indicates that there is a multi-cost item linkage deviation triggered by a certain initial abnormal correlation point. Increase.

[0062] S12. Locate the starting exception cost item, starting exception execution parameters, and starting exception rule source: The initial abnormal correlation point is determined based on the cost item imprint deviation tensor and the deviation linkage matrix.

[0063] The initial anomaly association point is represented as: ; in: Indicates the initial abnormal cost item; Indicates the parameters for initial exception execution; Indicates the source of the initial exception rule; Indicates the first The user in the first The first settlement cycle The imprint deviation of the first expense item is different from the first. The strength of the directional linkage between the deviations in the imprints of individual expense items; Indicates the contribution coefficient of proactive collaboration; This represents the passive linkage suppression coefficient; the active linkage contribution coefficient is used to improve the correlation score of cost items that can deviate from other cost items; the passive linkage suppression coefficient is used to reduce the correlation score of cost items that deviate mainly due to the influence of other cost items. Both can be obtained through training on historical anomaly samples or preset based on cost item dependencies. If you want the model to be more inclined to locate original cost items rather than aggregated cost items, you can appropriately increase this value. This reduces the score of bill summary items that are strongly affected by other expense items; if you want the model to focus more on anomalies that can trigger linkages between multiple expense items, you can appropriately increase the score. .

[0064] For example, if the time-of-use pricing version is executed incorrectly, causing the time-of-use electricity fee item to deviate and further affecting the bill summary item, then the time-of-use electricity fee item has a higher active linkage contribution to the bill summary item, and the time-of-use electricity fee item is more likely to be identified as the starting abnormal fee item; while the bill summary item also deviates, it is mainly affected by the time-of-use electricity fee item, so its passive linkage item is higher, and the final score will be suppressed.

[0065] Final anomaly localization results: ; in: Indicates the first The user in the first Anomaly location results for electricity price execution in each settlement cycle; For example: like This is for time-of-use electricity charges. For time-of-use pricing version parameters, If the source is a rule version, then the error is identified as an execution error in the time-of-use pricing version. like This is a basic electricity charge item. These are the parameters for the basic electricity charge execution method. If the source is a user profile, the issue is identified as a configuration error in the basic electricity fee execution method. like For the power factor adjustment electricity bill item, For power factor adjustment rule parameters, If the source is a metering point, then the issue is identified as an abnormality in the execution of the power factor adjustment rule. In one specific embodiment, to make the model trainable, it can be trained jointly using historical normal samples and simulated abnormal samples.

[0066] 1. Construct training samples Generate normal training samples based on historical normal bills: ; in: Indicates a normal sample; Offset is performed based on the baseline execution parameter set to generate a simulated abnormal execution parameter set: ; in: This represents the set of parameters for simulating exception execution. Indicates the first Class execution parameter offset operator; Generate a simulated abnormal bill: ; in: This represents a simulated abnormal bill; Simulated anomaly samples carry labels of real initial anomaly related points: ; in: Indicates the actual initial abnormal cost item; Indicates the actual initial exception execution parameters; Indicates the actual source of the initial exception rule.

[0067] 2. Training loss function The total training loss of the model is expressed as: ; in: This represents the total training loss of the model; This represents the loss for matching the candidate execution parameter domains. Indicates the loss from bill restructuring; Indicates that the expense item is marked as a loss. This indicates the loss in locating the initial anomaly-related point; This indicates the loss due to deviation from the linkage consistency; , , , This represents the adjustment coefficient, which is a hyperparameter. The candidate execution parameter domain matching loss is: ; in: This represents the set number of candidate execution parameters corresponding to the actual sample; This represents the matching cost of the actual set of candidate execution parameters; The loss from bill restructuring is: ; in: Indicates the loss from bill restructuring; The expense item for the loss of acceptance is: ; in: Indicates the first Should this expense item be covered by the first one? The execution parameter is at the _ ... The imprint of each rule's source dimension; This indicates the corresponding imprint deviation value; The initial anomaly correlation point location loss is: ; ; in: Indicates the score of the candidate initial anomaly association point; Indicates the actual initial abnormal cost item; Indicates the actual initial exception execution parameters; Indicates the actual source of the initial exception rule; The loss due to deviation from linkage consistency is: ; in: This indicates the loss due to deviation from the linkage consistency; This represents the deviation linkage matrix generated by the model; This represents the linkage supervision matrix generated based on the actual anomaly type.

[0068] In some embodiments, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method as described in any embodiment of the present invention.

[0069] In some embodiments, a computer-readable storage medium is provided on which a computer program is stored, which, when executed by a processor, implements the method as described in any embodiment of the present invention.

[0070] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0071] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0072] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0073] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0074] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for identifying and locating anomalies in user electricity pricing, characterized in that, Includes the following steps: Acquire user profile data, metering data, electricity pricing rule data, rule source data, metering point association data, and actual electricity bill data for the target user within the target billing period; Based on user profile data, electricity price rule data, and metering point association data, determine the set of benchmark execution parameters for the target user; Generate a benchmark execution imprint matrix based on the benchmark execution parameter set and rule source data; Based on the types of expense items in the actual electricity bill, the baseline execution imprint matrix, and the billing input data for each expense item, a baseline expense item imprint acceptance tensor is generated. Based on the baseline execution parameter set and the execution parameter offset operator, candidate execution parameter domains are constructed, and candidate bills and candidate execution imprint matrices are generated for each candidate execution parameter set; Based on the differences between candidate bills and actual electricity bills, the rule feasibility relationship of candidate execution parameter sets, and the imprint deviation relationship between candidate execution imprint matrices and benchmark execution imprint matrices, the candidate execution matching cost corresponding to each candidate execution parameter set is determined. The set of candidate execution parameters with the lowest matching cost is determined as the inversion execution parameter set, and the inversion cost item imprint tensor is generated based on the inversion execution parameter set. Based on the deviation between the baseline cost item imprint acceptance tensor and the inverted cost item imprint acceptance tensor, determine whether the target user has any abnormal electricity price execution and locate the abnormal location.

2. The method for identifying and locating abnormal user electricity price execution according to claim 1, characterized in that, The set of benchmark execution parameters includes at least one of the following: electricity consumption category parameters, voltage level parameters, time-of-use electricity price version parameters, basic electricity charge execution method parameters, power factor adjustment rule parameters, and metering point attribution parameters.

3. The method for identifying and locating abnormal user electricity price execution according to claim 1, characterized in that, The rule source data includes at least one of user profile source data, metering point source data, electricity price rule version source data, and bill generation source data; Among them, the user profile source data includes at least one of the following: user electricity consumption category, voltage level, contract capacity, and operating capacity; the metering point source data includes at least one of the following: metering point number, metering point multiplier, and metering point affiliation; the electricity price rule version source data includes at least one of the following: electricity price meter version number, time-of-use version number, or rule effective date; and the bill generation source data includes at least one of the following: bill generation batch, billing module identifier, or fee item generation order.

4. The method for identifying and locating abnormal user electricity price execution according to claim 3, characterized in that, A benchmark execution imprint matrix is ​​generated based on the benchmark execution parameter set and rule source data, including: The baseline execution parameter set and rule source data are embedded and encoded respectively to obtain execution parameter embedding vectors and rule source embedding vectors; Construct an interaction embedding vector based on the execution parameter embedding vector and the rule source embedding vector; The execution parameter embedding vector, rule source embedding vector, and interaction embedding vector are mapped to the imprint space and fused to generate the corresponding baseline imprint unit.

5. The method for identifying and locating abnormal user electricity price execution according to claim 1, characterized in that, The baseline cost item imprint acceptance tensor includes a cost item dimension, an execution parameter dimension, and a rule source dimension. Each baseline acceptance imprint in the tensor is used to represent the acceptance relationship of a cost item to a baseline execution parameter under a rule source dimension. The fee items include at least one of the following: electricity consumption fee, time-of-use fee, basic fee, maximum demand adjustment fee, power factor adjustment fee, additional fees, and bill summary.

6. The method for identifying and locating abnormal user electricity price execution according to claim 5, characterized in that, Generate the baseline cost item imprint acceptance tensor, including: The cost item type, baseline execution parameters, and billing input data of the cost item are embedded and encoded to obtain the cost item type embedding vector, execution parameter embedding vector, and billing input embedding vector, respectively. The acceptance score of a cost item to the benchmark execution parameter is determined based on the cost item type embedding vector, execution parameter embedding vector, and billing input embedding vector. Multiple acceptance scores corresponding to the same cost item are normalized to obtain the acceptance strength. Based on the acceptance strength, the corresponding benchmark imprint unit, cost item type embedding vector and billing input embedding vector in the benchmark execution imprint matrix are mapped and fused to generate the benchmark acceptance imprint corresponding to the cost item.

7. The method for identifying and locating abnormal user electricity price execution according to claim 1, characterized in that, The execution parameter offset operator includes at least one of the following: electricity category replacement operator, voltage level replacement operator, time-of-use price version offset operator, basic electricity charge execution mode switching operator, power factor adjustment rule replacement operator, and metering point affiliation replacement operator; When the execution parameter offset operator is applied to the base execution parameter set, only the execution parameter corresponding to the execution parameter offset operator is changed, while the other execution parameters keep the base value unchanged. The candidate values ​​for the execution parameter offset operator are derived from at least one of the following: historical execution parameter values, adjacent voltage levels, adjacent power factor assessment standards, historical electricity price rule versions, and adjacent metering point affiliation relationships.

8. The method for identifying and locating abnormal user electricity price execution according to claim 1, characterized in that, The candidate execution matching cost includes the bill matching cost, rule feasibility penalty, and imprint deviation cost; Among them, the bill matching cost is determined based on the difference in expense items between the actual electricity bill data and the candidate bill, as well as the covariance matrix of expense item fluctuations. The rule feasibility penalty is determined based on whether the candidate execution parameter belongs to the set of allowed values ​​defined by the electricity price rule version data and user profile data; The imprint deviation cost is determined based on the vector similarity between each candidate imprint unit in the candidate execution imprint matrix and the corresponding benchmark imprint unit in the benchmark execution imprint matrix.

9. The method for identifying and locating abnormal user electricity price execution according to claim 1, characterized in that, Based on the deviation between the baseline cost item imprint acceptance tensor and the inverted cost item imprint acceptance tensor, determine whether the target user has any abnormal electricity price execution, including: Subtract the baseline acceptance mark in the baseline cost item imprint acceptance tensor from the corresponding inversion acceptance mark in the inversion cost item imprint acceptance tensor to obtain the cost item imprint deviation tensor. The L2 norm of each imprint deviation vector in the cost item imprint deviation tensor is calculated to obtain the imprint deviation values ​​corresponding to the cost item, execution parameters and rule source dimensions. Based on the imprint deviation values ​​of different cost items under the same execution parameters and the same rule source dimension, and the dependencies between cost items, a deviation linkage matrix is ​​generated; The execution consistency gap is calculated based on the candidate execution matching cost, the cost item imprint deviation tensor, and the deviation linkage matrix, and the execution consistency gap is used to determine whether there is an abnormal electricity price execution for the target user.

10. A method for identifying and locating abnormal user electricity price execution according to claim 9, characterized in that, When it is determined that there is an abnormality in the electricity price execution of the target user, for each expense item, each execution parameter and each rule source dimension, the candidate initial abnormality association point score is determined based on the corresponding imprint deviation value, the active linkage contribution of the expense item to other expense items, and the passive linkage influence of other expense items on the expense item. The cost item, execution parameter, and rule source dimension with the highest score of the candidate initiation anomaly association point are determined as the initiation anomaly cost item, initiation anomaly execution parameter, and initiation anomaly rule source.

Images