Power customer credit rating evaluation method, system and equipment based on mapping knowledge domain

Abstract

The invention discloses an electric power customer credit rating evaluation method, system and equipment based on a knowledge graph, particularly relates to the technical field of electric power customer credit evaluation, and is used for solving the problem of double-role customer credit misjudgment caused by splitting processing of an electricity utilization arrearage risk and power generation income capacity in the prior art. The method comprises the following steps: establishing a data mapping relation between a load side and a power side and constructing a knowledge graph association structure; analyzing the nonlinear characteristics of the voltage reactive difference adjustment curve based on the power supply side operation data to generate a hedging adjustment mechanism; load side reactive compensation behavior characteristics are extracted, and a collaborative risk level is generated through tensor fusion in combination with the load and power impedance characteristic mismatch degree; dynamically adjusting the action intensity of the hedging mechanism according to the risk level; and finally, integrating the mechanism and the association structure in the knowledge graph to form a credit evaluation result. Quantitative hedging of electricity selling income to arrearage risks is realized, and high-risk rating deviation of double-role customers is eliminated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power customer credit evaluation, more particularly, the present application relates to a power customer credit rating evaluation method, system and device based on a knowledge graph. BACKGROUND

[0002] In the field of power customer credit evaluation, the existing technology generally adopts a multi-dimensional data analysis model to construct a credit scoring system. In particular, for industrial and commercial customers participating in distributed energy transactions (such as photovoltaic enterprises), their credit assessment needs to consider both power consumption and power generation and sales behavior characteristics. The current mainstream method analyzes power consumption payment records and power generation income data independently, and calculates power consumption arrears risk and power sales income ability separately, forming two isolated credit evaluation sub-systems. Although this method can handle the credit risk of single-role customers, it does not establish a credit hedging mechanism that associates the identities of power consumers and power generators.

[0003] The defects of the existing technology are that when a customer has both power load and power supply roles, the fragmented credit evaluation model cannot quantify the risk hedging effect between electricity debt and power sales income, which leads to a misjudgment of the customer's true credit level - the arrears risk that could be offset by power sales income is fully counted, resulting in a high risk rating bias. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the present application provides a power customer credit rating evaluation method, system and device based on a knowledge graph to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0006] The power customer credit rating evaluation method based on a knowledge graph comprises:

[0007] S1, obtaining load-side operation data and power-side operation data of a power customer, and establishing a data mapping relationship between the load side and the power side based on the identity of the power customer;

[0008] S2, generating a knowledge graph containing load-side metering entities and power-side metering entities based on the data mapping relationship, and constructing an association structure of the load-side metering entities and the power-side metering entities in the knowledge graph;

[0009] S3, analyzing the nonlinear section characteristics of the grid point voltage and reactive power adjustment curve based on the power-side operation data, and generating a hedging adjustment mechanism for the arrears risk state;

[0010] S4, extract the reactive power compensation behavior characteristics in the load side operation data, synchronously analyze the mismatch degree of the load impedance characteristics corresponding to the load side metering entity and the power source impedance characteristics corresponding to the power source side metering entity, and generate a cooperative risk level through tensor fusion;

[0011] S5, adjust the action strength of the hedging adjustment mechanism according to the cooperative risk level;

[0012] S6, integrate the adjusted hedging adjustment mechanism and the associated structure in the knowledge graph to form the power customer credit level evaluation result.

[0013] Further, the load side operation data and the power source side operation data of the power customer are obtained, and a data mapping relationship between the load side and the power source side is established based on the power customer identity, including:

[0014] Obtain the load side operation data containing the electricity metering characteristics and the overdue risk state from the power marketing system;

[0015] Obtain the power source side operation data containing the power generation metering characteristics and the supply capacity state from the power dispatching system;

[0016] Through the unified social credit code in the power customer identity, a bidirectional association mapping relationship is established between the load side operation data and the power source side operation data;

[0017] The electricity metering characteristics in the load side operation data and the power generation metering characteristics in the power source side operation data are stored in time dimension alignment;

[0018] Based on the bidirectional association mapping relationship, a data mapping index table of the load side metering point and the power source side metering point is generated.

[0019] Further, based on the data mapping relationship, a knowledge graph containing the load side metering entity and the power source side metering entity is generated, and an associated structure of the load side metering entity and the power source side metering entity is constructed in the knowledge graph, including:

[0020] The load side operation data in the data mapping relationship is mapped to the load side metering entity in the knowledge graph, wherein the load side metering entity contains the electricity metering characteristics and the overdue risk state attribute;

[0021] The power source side operation data in the data mapping relationship is mapped to the power source side metering entity in the knowledge graph, wherein the power source side metering entity contains the power generation metering characteristics and the supply capacity state attribute;

[0022] According to the data mapping relationship corresponding to the power customer identity, a power consumption association relationship is created between the load side metering entity and the power source side metering entity;

[0023] A topology edge structure connecting the load-side metering entity and the power-side metering entity is formed based on the power consumption and power supply correlation;

[0024] The load-side metering entity, the power-side metering entity and the topology edge structure are persistently stored in the knowledge graph storage system.

[0025] Further, based on the nonlinear section characteristics of the grid-connected point voltage and reactive power regulation curve analyzed from the power-side operation data, a hedging adjustment mechanism for the arrears risk state is generated, including:

[0026] The grid-connected point voltage and reactive power regulation curve is extracted from the power-side operation data;

[0027] The nonlinear characteristic section in the voltage and reactive power regulation curve is identified, and the nonlinear characteristic section is a fluctuation section in which the voltage deviates from the rated value;

[0028] The voltage fluctuation amplitude characteristics and duration characteristics of the nonlinear characteristic section are analyzed;

[0029] The voltage stability margin attenuation factor is calculated based on the product relationship of the voltage fluctuation amplitude characteristics and the duration characteristics;

[0030] The margin compensation coefficient is generated according to the voltage stability margin attenuation factor;

[0031] The margin compensation coefficient is bound to the arrears risk state to construct a hedging adjustment mechanism, and the hedging adjustment mechanism includes the margin compensation parameter characteristics.

[0032] Further, the reactive power compensation behavior characteristics in the load-side operation data are extracted, the mismatch degree of the load impedance characteristics corresponding to the load-side metering entity and the power impedance characteristics corresponding to the power-side metering entity is synchronously analyzed, and the collaborative risk level is generated by tensor fusion, including:

[0033] The power factor characteristics and the reactive power fluctuation characteristics in the load-side operation data are extracted to form the reactive power compensation behavior characteristics;

[0034] The load impedance characteristics are obtained based on the load impedance characteristic data corresponding to the load-side metering entity;

[0035] The power impedance characteristics are obtained based on the power impedance characteristic data corresponding to the power-side metering entity;

[0036] The vector distance of the load impedance characteristics and the power impedance characteristics in the power frequency band is calculated as the impedance characteristic mismatch degree;

[0037] The reactive power compensation behavior characteristics and the impedance characteristic mismatch degree are constructed as a multi-dimensional feature tensor;

[0038] The multi-dimensional feature tensor is subjected to tensor contraction operation to obtain a core risk factor, and the core risk factor is converted into a collaborative risk level.

[0039] Further, the action intensity of the hedging adjustment mechanism is adjusted according to the cooperative risk level, including:

[0040] obtaining the cooperative risk level;

[0041] extracting the margin compensation parameter feature in the hedging adjustment mechanism;

[0042] determining the compensation intensity adjustment coefficient corresponding to the cooperative risk level through a preset risk level compensation mapping relationship;

[0043] updating the margin compensation parameter feature using the compensation intensity adjustment coefficient;

[0044] re-writing the updated margin compensation parameter feature into the hedging adjustment mechanism to complete the action intensity adjustment.

[0045] Further, the preset risk level compensation mapping relationship is realized by the following way:

[0046] establishing a mapping relationship table of the cooperative risk level and the compensation intensity adjustment coefficient, wherein the cooperative risk level is divided into different level intervals according to the risk degree, and each level interval corresponds to a preset compensation intensity adjustment coefficient value;

[0047] when the cooperative risk level is obtained, the corresponding compensation intensity adjustment coefficient is outputted by matching the level interval to which it belongs in the mapping relationship table.

[0048] Further, the adjusted hedging adjustment mechanism and the associated structure are integrated in the knowledge graph to form the power customer credit level evaluation result, including:

[0049] obtaining the margin compensation parameter feature in the adjusted hedging adjustment mechanism;

[0050] positioning the associated structure between the load-side metering entity and the power-side metering entity in the knowledge graph;

[0051] adding the margin compensation parameter feature as a dynamic attribute to the topological edge structure of the associated structure;

[0052] constructing a credit evaluation feature vector based on the topological edge attribute of the associated structure and the metering entity attribute;

[0053] converting the credit evaluation feature vector into the power customer credit level evaluation result through a preset credit rating conversion rule;

[0054] persistently storing the power customer credit level evaluation result to the customer credit file node of the knowledge graph.

[0055] On the other hand, the present application provides a power customer credit level evaluation system based on a knowledge graph, including:

[0056] The mapping establishment module is configured to obtain load-side operation data and power-side operation data of a power consumer, and establish a data mapping relationship between the load side and the power side based on an identity of the power consumer;

[0057] The graph construction module is configured to generate a knowledge graph containing load-side metering entities and power-side metering entities based on the data mapping relationship, and construct an association structure of the load-side metering entities and the power-side metering entities in the knowledge graph;

[0058] The mechanism generation module is configured to analyze a non-linear section feature of a grid point voltage reactive power regulation curve based on the power-side operation data, and generate a hedging adjustment mechanism for the arrears risk state;

[0059] The grade generation module is configured to extract a reactive power compensation behavior feature in the load-side operation data, synchronously analyze a mismatch degree of a load impedance characteristic corresponding to a load-side metering entity and a power source impedance characteristic corresponding to a power-side metering entity, and generate a collaborative risk grade through tensor fusion;

[0060] The adjustment updating module is configured to adjust an action strength of the hedging adjustment mechanism according to the collaborative risk grade;

[0061] The evaluation integration module is configured to integrate the adjusted hedging adjustment mechanism and the association structure in the knowledge graph, and form a power consumer credit grade evaluation result.

[0062] In another aspect, the present application provides a power consumer credit grade evaluation device based on a knowledge graph, which comprises a processor, a memory, and a program or instruction stored in the memory and executable on the processor, and the program or instruction is executed by the processor to implement the power consumer credit grade evaluation method based on the knowledge graph.

[0063] Compared with the prior art, the present application has the following beneficial effects:

[0064] 1. The present application dynamically fuses the power generation and consumption characteristics through the knowledge graph, innovatively constructs a risk hedging mechanism, significantly improves the credit evaluation precision of the dual-role consumer, and is different from the defects of the prior art that the power consumption arrears risk and the power generation income ability are processed separately. The present application establishes a dynamic association structure of the load-side metering entities and the power-side metering entities in the knowledge graph, generates a hedging adjustment mechanism based on the non-linear feature of the power-side voltage reactive power regulation curve, quantifies the active offsetting ability of the power sales income to the arrears risk, accurately captures the power generation and consumption collaborative stability risk through the mismatch degree of the load impedance characteristic and the power source impedance characteristic, makes the actual hedging effect of the power sales income to the arrears risk objectively reflected in the credit evaluation, and eliminates the high risk rating deviation from the root.

[0065] 2. Real-time response of credit risk is realized through dynamic strength adjustment, the strength of the hedging adjustment mechanism is adaptively adjusted based on the collaborative risk level generated based on tensor fusion, the hedging strength is accurately matched with the system stability state, the adjusted hedging mechanism and the associated structure are integrated in the knowledge graph, the credit evaluation feature vector of the power generation and consumption behavior linkage is formed, and finally the output result reflects not only the short-term payment ability of the user but also the long-term power generation and consumption collaborative reliability. BRIEF DESCRIPTION OF DRAWINGS

[0066] Figure 1 A flowchart of the power customer credit level evaluation method based on the knowledge graph of the present application;

[0067] Figure 2 A structural schematic diagram of the power customer credit level evaluation system based on the knowledge graph of the present application. DETAILED DESCRIPTION

[0068] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0069] Embodiment 1: Figure 1 The power customer credit level evaluation method based on the knowledge graph of the present application is given, which comprises:

[0070] S1, obtaining the load side operation data and the power source side operation data of the power customer, and establishing a data mapping relationship between the load side and the power source side based on the power customer identity;

[0071] S2, generating a knowledge graph comprising the load side metering entity and the power source side metering entity based on the data mapping relationship, and constructing the associated structure of the load side metering entity and the power source side metering entity in the knowledge graph;

[0072] S3, analyzing the non-linear section characteristics of the grid connection point voltage reactive power adjustment curve based on the power source side operation data, and generating a hedging adjustment mechanism for the arrears risk state;

[0073] S4, extracting the reactive power compensation behavior characteristics in the load side operation data, synchronously analyzing the mismatch degree of the load impedance characteristics corresponding to the load side metering entity and the power source impedance characteristics corresponding to the power source side metering entity, and generating a collaborative risk level through tensor fusion;

[0074] S5, adjusting the strength of the hedging adjustment mechanism according to the collaborative risk level;

[0075] S6, integrate the adjusted hedging adjustment mechanism and the correlation structure in the knowledge graph to form the power customer credit rating evaluation result.

[0076] S1, obtain the load-side operation data and the power-side operation data of the power customer, and establish a data mapping relationship between the load side and the power side based on the power customer identity, and the specific implementation is as follows:

[0077] The load-side operation data containing power consumption characteristics and overdue risk state are obtained from the power marketing system, and the specific implementation process is as follows: through the data interface of the power marketing system, a preset customer power consumption data query protocol is called, the power consumption characteristics in the target time period are extracted with the power customer identity as the index, the power consumption characteristics include active power value, reactive power value, voltage and current phase angle; the overdue risk state of the power customer is synchronously extracted, the overdue risk state is determined by analyzing the payment record and credit history, and the specific determination rules include: when the number of continuous overdue times exceeds a certain number threshold or the cumulative overdue amount exceeds a preset amount threshold, it is marked as a high-risk state, for example, more than 3 times of continuous overdue or more than 1000 yuan of cumulative overdue; when there is a single overdue time exceeding a certain number of days but the cumulative amount does not exceed the mark, it is marked as an attention state, for example, a single overdue time exceeding 30 days; the rest is marked as a normal state. An incremental update mechanism is adopted in the data acquisition process, and only the change data in the latest time slice is synchronized each time.

[0078] The power-side operation data containing power generation characteristics and supply capacity state are obtained from the power dispatching system, and the specific implementation process is as follows: through the open data service of the power dispatching system, a data request is sent based on the unified data access specification, the request parameters include the power customer identity and the target time range; the power generation characteristics include the grid-connected point voltage amplitude, frequency deviation value, and active power fluctuation; the supply capacity state is determined according to the availability of power generation equipment and the standby capacity, and the specific determination method includes: when the availability of power generation equipment is lower than a certain percentage and the standby capacity is insufficient for a certain percentage of total installed capacity, it is marked as a supply shortage state, for example, the availability is lower than 90% and the standby capacity is insufficient for 5%; when the availability is in a certain interval and the standby capacity is in a certain interval, it is marked as a supply balance state; the rest is marked as a supply sufficient state. When the data is obtained, invalid sampling points are automatically filtered, and only data with a quality mark of qualified is retained.

[0079] By the unified social credit code in the power customer identity, a bidirectional association mapping relationship is established between the load side operation data and the power side operation data, and the specific implementation process is as follows: the unified social credit code field in the load side operation data is extracted as the primary key, and the unified social credit code field in the power side operation data is extracted as the foreign key; a one-to-one association relationship between the primary key and the foreign key is established, and the association rule requires that the difference between the load side data timestamp and the power side data timestamp is not more than a specific time threshold, for example, 5 minutes; when there are multiple power side data matching, the record with the smallest timestamp difference is preferentially selected; when the association fails, the data completion process is triggered, and the missing unified social credit code information is supplemented through the power customer archive library. The association relationship is stored in the association mapping table of the distributed relational database, and the table contains the load side data identifier, the power side data identifier, and the unified social credit code field.

[0080] The power consumption metering features in the load side operation data and the power generation metering features in the power side operation data are stored in time dimension alignment, and the specific implementation process is as follows: taking a specific time interval as the basic time slice, for example, 15 minutes, the active power value, the reactive power value, the voltage and current phase angle of the power consumption metering features and the grid-connected point voltage amplitude, the frequency deviation value and the active output fluctuation of the power generation metering features are divided into the same time slice; the alignment rule requires that the load side data timestamp and the power side data timestamp are in the same slice interval; the time slice boundary is divided according to the whole point time; the aligned data is stored in the time series database, and each time slice contains the timestamp label, the load side feature set and the power side feature set data block.

[0081] Based on the bidirectional association mapping relationship, a data mapping index table of the load side metering point and the power side metering point is generated, and the specific implementation process is as follows: each association record in the association mapping table is traversed, and the metering point number of the load side metering point and the metering point number of the power side metering point are extracted; the metering point number and the unified social credit code are bound to form an index entry, and the index entry also records the metering point type identifier; the index table adopts a hash index structure, and the unified social credit code is used as the hash key, and the hash value points to the metering point number list; the index table updating mechanism includes: when a new association relationship is added, an entry is automatically added, and when the association relationship is removed, the entry is deleted after a specific time period, for example, 24 hours. The index table is stored in the in-memory database.

[0082] S2, based on the data mapping relationship, a knowledge graph containing the load side metering entity and the power side metering entity is generated, and an association structure of the load side metering entity and the power side metering entity is constructed in the knowledge graph, and the specific implementation is as follows:

[0083] The load side operation data in the data mapping relationship is mapped to the load side metering entity in the knowledge graph, and the specific implementation process is as follows: the metering point number of the load side metering point is extracted from the data mapping index table, and the load side metering entity is created in the knowledge graph taking the metering point number as an entity identifier; the power consumption metering features are assigned to the attribute set of the load side metering entity, including active power attribute, reactive power attribute, voltage and current phase angle attribute, wherein the attribute values are directly derived from the measurement values of the corresponding time slice in the load side operation data; the arrears risk state is assigned as a discrete attribute to the load side metering entity, and the attribute value enumeration type is set according to the level of the arrears risk state, for example, the attribute value 1 corresponds to the high risk state, the attribute value 2 corresponds to the attention state, and the attribute value 3 corresponds to the normal state. The entity unique identifier is automatically generated when the entity is created, and the creation timestamp and version number are recorded.

[0084] The power side operation data in the data mapping relationship is mapped to the power side metering entity in the knowledge graph, and the specific implementation process is as follows: the metering point number of the power side metering point is obtained from the data mapping index table, and the power side metering entity is created in the knowledge graph taking the metering point number as an entity identifier; the power generation metering features are assigned to the attribute set of the power side metering entity, including grid-connected point voltage amplitude attribute, frequency deviation value attribute, and active power fluctuation attribute, and the attribute values are derived from the measurement values of the corresponding time slice in the power side operation data; the supply capacity state is assigned as a discrete attribute to the power side metering entity, and the attribute value enumeration type is set according to the level of the supply capacity state, for example, the attribute value 1 corresponds to the supply shortage state, the attribute value 2 corresponds to the supply balance state, and the attribute value 3 corresponds to the supply abundance state. The entity unique identifier is automatically generated after the entity is created, and the creation timestamp and version number are recorded.

[0085] According to the data mapping relationship corresponding to the power customer identity, the power consumption correlation is created between the load side metering entity and the power side metering entity, and the specific implementation process is as follows: the load side metering point number and the power side metering point number corresponding to the power customer identity are queried through the data mapping index table; the load side metering entity and the power side metering entity are located in the knowledge graph; the correlation relationship from the power side metering entity to the load side metering entity is created, and the relationship type is defined as the power consumption correlation; the correlation attribute includes the correlation strength coefficient, and the initial value is set to the default value, for example, 1.0; the correlation time range is recorded when the correlation is created, and the time range is consistent with the validity period recorded in the data mapping index table.

[0086] The topology edge structure connecting the load side metering entity and the power source side metering entity is formed based on the power generation and power consumption correlation, and the specific implementation process is as follows: the power generation and power consumption correlation is instantiated as an edge object in the knowledge graph; the starting point of the topology edge structure is set as the power source side metering entity, and the ending point is the load side metering entity; the topology edge structure includes an edge type attribute, and the attribute value is fixed as the power generation and power consumption topology connection; the topology edge structure includes an edge weight attribute, and the initial value is calculated according to the correlation strength coefficient, and the calculation formula is edge weight = correlation strength coefficient * preset reference weight value; the reference weight value is determined according to the power customer type, for example, the reference weight value of industrial and commercial customers is 0.8, and the reference weight value of residential customers is 0.5; the topology edge structure is automatically added with a creation time stamp and a version number when generated.

[0087] The load side metering entity, the power source side metering entity and the topology edge structure thereof are persistently stored in the knowledge graph storage system, and the specific implementation process is as follows: the transaction processing mechanism of the graph database is used to execute the persistent operation; first, the load side metering entity is converted into a graph vertex object and stored; second, the power source side metering entity is converted into a graph vertex object and stored; finally, the topology edge structure is converted into a graph edge object and stored; a batch submission strategy is used in the persistent process, and a transaction is submitted once every specific number of operations, for example, once every 100 operations; a storage confirmation report is generated after storage is completed. The persistent data supports index query through the unified social credit code.

[0088] S3, based on the power source side operation data, the non-linear section characteristics of the grid connection point voltage and reactive power regulation curve are analyzed, and the hedging adjustment mechanism for the arrears risk state is generated, and the specific implementation is as follows:

[0089] The grid connection point voltage and reactive power regulation curve is extracted from the power source side operation data, and the specific implementation process is as follows: the power generation metering characteristics attribute set of the power source side metering entity is accessed, and the grid connection point voltage measurement value and the corresponding reactive power measurement value containing the time stamp sequence are selected; the reactive power value is taken as the horizontal coordinate, the voltage measurement value is taken as the vertical coordinate, and the original curve is drawn according to the time sequence; the original curve is subjected to data smoothing processing, and the moving average filtering algorithm is used to eliminate measurement noise, and the filtering window width is determined according to the sampling frequency, for example, a 5-minute window is used when the sampling interval is 1 minute; the smoothed curve is stored as a data structure containing a coordinate point sequence, and each coordinate point contains reactive power value, voltage value and time stamp information. In the curve extraction process, invalid data points are automatically filtered, and only points with voltage values within a reasonable range are retained, for example, points within the range of 9.5kV to 10.5kV are retained for a 10kV system.

[0090] The nonlinear feature section in the voltage reactive power difference curve is identified, and the specific implementation process is as follows: the rated voltage value is set as the reference value, for example, 10kV system sets 10.0kV as the rated value; the absolute deviation of the voltage value of each point from the rated voltage value is calculated by traversing the curve coordinate point sequence; when the voltage deviation of a plurality of consecutive points exceeds a certain threshold value, it is determined as a nonlinear feature section, for example, the deviation of three consecutive points exceeds 0.5kV; the boundary points of the nonlinear feature section are determined as the starting point where the voltage deviation first exceeds the threshold value and the end point where it returns to the threshold value. The starting time stamp, ending time stamp and maximum voltage deviation value of each identified nonlinear feature section are recorded. The identification process uses a sliding window algorithm, and the window size is consistent with the filter window.

[0091] The voltage fluctuation amplitude feature and duration feature of the nonlinear feature section are analyzed, and the specific implementation process is as follows: for each nonlinear feature section, the voltage fluctuation amplitude feature is the maximum absolute deviation value of the voltage measurement value in the section from the rated voltage value; the duration feature is calculated as the difference between the ending time stamp and the starting time stamp of the section, and the time unit is converted to minutes; when there are a plurality of nonlinear feature sections in the same curve, the fluctuation amplitude and duration of each section are calculated respectively. The analysis result is stored as a feature set, and each feature item contains the fluctuation amplitude value, the duration value and the corresponding section identifier. For the section with a duration less than a certain threshold value, for example, the duration is less than 1 minute, it is regarded as an invalid section and is not processed.

[0092] The product relationship between the voltage fluctuation amplitude feature and the duration feature is used to calculate the voltage stability margin attenuation factor, and the specific implementation process is as follows: for each valid nonlinear feature section, the voltage fluctuation amplitude feature value is multiplied by the duration feature value to obtain the original product value; the original product value is normalized, and the normalization coefficient is determined according to the system voltage level, for example, 100 is used as the reference coefficient for 10kV system, and 350 is used as the reference coefficient for 35kV system; the calculation formula of the voltage stability margin attenuation factor is: the attenuation factor is equal to the normalized product value divided by the reference coefficient, for example, the section with a fluctuation amplitude of 2kV and a duration of 5 minutes has an original product value of 10, and the normalized attenuation factor is 0.1. When there are a plurality of valid sections, the maximum attenuation factor is taken as the final result. If there is no valid nonlinear feature section, the attenuation factor is set to 0 by default.

[0093] The margin compensation coefficient is generated according to the voltage stability margin attenuation factor, and the specific implementation process is as follows: a conversion relationship between the attenuation factor and the compensation coefficient is established, and the conversion rule is that the compensation coefficient is equal to 1 minus the attenuation factor; when the attenuation factor exceeds a specific upper limit, the compensation coefficient takes a minimum value, for example, when the attenuation factor is greater than 0.8, the compensation coefficient is fixed at 0.2; when the attenuation factor is lower than a specific lower limit, the compensation coefficient takes a maximum value of 1.0; between the upper and lower limits, the compensation coefficient is calculated by using a linear interpolation method, for example, when the attenuation factor is 0.3, the compensation coefficient is 0.7; the generated margin compensation coefficient is kept to two decimal places. The conversion process is realized by using a table lookup method, and the table stores the compensation coefficient values corresponding to different attenuation factors.

[0094] The margin compensation coefficient is bound to the risk state of arrears to build a hedging adjustment mechanism, and the specific implementation process is as follows: obtaining the attribute value of the risk state of arrears from the load side metering entity; creating a hedging adjustment mechanism data structure, which includes a margin compensation parameter characteristic field and a risk state field; assigning the margin compensation coefficient to the margin compensation parameter characteristic field; assigning the attribute value of the risk state of arrears to the risk state field; establishing a dynamic association relationship between the margin compensation parameter characteristic and the risk state, and the association rule is that when the risk state changes, the margin compensation coefficient is recalculated. After the hedging adjustment mechanism is generated, a time stamp label is added and stored in a special cache area, and the data retention period of the cache area is a specific time period, for example, 24 hours. The hedging adjustment mechanism supports retrieval and query through the power customer identity.

[0095] S4, extracting the reactive power compensation behavior characteristics in the load side operation data, synchronously analyzing the mismatch degree of the load impedance characteristics corresponding to the load side metering entity and the power source impedance characteristics corresponding to the power source side metering entity, and generating a cooperative risk level by tensor fusion, and the specific implementation is as follows:

[0096] The power factor feature and the reactive power fluctuation feature are extracted from the load side operation data to form the reactive power compensation behavior feature, and the specific implementation process is as follows: accessing the power consumption measurement feature attribute set of the load side measurement entity, the power factor measurement value sequence is extracted; the statistical features of the power factor in a specific time window are calculated, including the average value, the standard deviation and the range value, for example, the power factor average value is calculated with 1 hour as the time window; the reactive power measurement value sequence is extracted synchronously, the fluctuation features of the reactive power in the same time window are calculated, including the fluctuation amplitude and the fluctuation frequency, the fluctuation amplitude takes the difference between the maximum value and the minimum value of the reactive power in the time window, and the fluctuation frequency takes the zero-crossing number per unit time, the zero-crossing refers to the number of times that the reactive power value crosses zero; the power factor statistical features and the reactive power fluctuation features are combined into the reactive power compensation behavior feature vector, the feature vector dimension is fixed as 5 dimensions, including the power factor average value, the power factor standard deviation, the power factor range, the reactive power fluctuation amplitude and the reactive power fluctuation frequency. Abnormal data points are automatically filtered in the feature extraction process, for example, points with power factor greater than 1 or less than -1 are regarded as invalid points.

[0097] The load impedance characteristic is obtained based on the load impedance characteristic data corresponding to the load side measurement entity, and the specific implementation process is as follows: analyzing the extended attribute field of the load side measurement entity, the field stores the load impedance characteristic parameters trained by historical operation data; the load impedance characteristic is represented as an impedance spectrum in complex form, including impedance values of multiple frequency points in a power frequency band, and the power frequency band is defined as a range of 48Hz to 52Hz; the latest updated version of the impedance characteristic data is automatically loaded when the data is obtained, and if no historical data is stored, an online identification algorithm is called to calculate in real time, the online identification adopts a least square fitting method, the input is voltage and current waveform sampling data, and the sampling frequency is a specific value, for example, 4kHz; the impedance characteristic data is stored in an array structure, and each element includes a frequency value, a resistance component and a reactance component. When online identification is performed, the calculation process includes establishing a voltage and current equation, and solving the impedance parameters by the least square method.

[0098] The power source impedance characteristic is obtained based on the power source impedance characteristic data corresponding to the power source side measurement entity, and the specific implementation process is as follows: accessing the power generation feature extended attribute of the power source side measurement entity, and extracting the pre-stored power source impedance characteristic parameters; the power source impedance characteristic is also represented as an impedance spectrum in complex form, covering the power frequency band of 48Hz to 52Hz; when the data is obtained, the validity period of the impedance spectrum is verified, if the data is expired, the latest power source impedance parameters are obtained in real time through a dispatching system interface, and the obtaining period is a specific time period, for example, updated every 15 minutes; the power source impedance characteristic data structure is consistent with the load impedance characteristic, including a frequency value, a resistance component and a reactance component array. When the real-time acquisition is performed, the pre-calculated impedance parameter table is read through an application program interface provided by the dispatching system.

[0099] The vector distance between the load impedance characteristic and the power supply impedance characteristic in the power frequency band is taken as the impedance characteristic mismatch degree, and the specific implementation process is as follows: sampling points are selected at preset frequency intervals in the power frequency band, and the frequency interval is determined according to the accuracy requirement, for example, 48Hz, 48.5Hz, 49Hz, 49.5Hz, 50Hz, 50.5Hz, 51Hz, 51.5Hz, 52Hz are selected with a step of 0.5Hz; the Euclidean distance between the load impedance and the power supply impedance is calculated at each sampling point, and the distance calculation formula is: the distance is equal to the square root of the square of the resistance component difference plus the square of the reactance component difference; the distance values of each sampling point are combined to form a distance vector; the distance vector is normalized, and the normalization method adopts the maximum and minimum value scaling method, and the scaling formula is: (original value-minimum value) / (maximum value-minimum value); the length of the normalized distance vector is taken as the impedance characteristic mismatch degree scalar value. The calculation result is kept to four decimal places. When some sampling point data is missing, the adjacent point interpolation is used to complete it.

[0100] The reactive compensation behavior characteristics and the impedance characteristic mismatch degree are constructed into a multi-dimensional feature tensor, and the specific implementation process is as follows: a three-dimensional tensor data structure is created, the first dimension corresponds to the five components of the reactive compensation behavior characteristics, the second dimension corresponds to the impedance characteristic mismatch degree scalar value, and the third dimension is reserved as a constant filling dimension; the five components of the reactive compensation behavior characteristic vector are assigned to the first dimension in a fixed order, and the order is: power factor average, power factor standard deviation, power factor range, reactive power fluctuation amplitude, and reactive power fluctuation frequency; the impedance characteristic mismatch degree scalar value is assigned to the second dimension; the third dimension is filled with a fixed value 1 to maintain the integrity of the tensor structure; after the tensor is constructed, the standardization processing is performed, each dimension is subtracted by the historical mean value of the dimension and divided by the historical standard deviation, and the mean value and the standard deviation are obtained based on the historical data of a specific number of days in the past, for example, the past 30 days. The final tensor size is fixed as 5x1x1. The feature value range is automatically checked when the tensor is constructed, and the features exceeding the historical maximum and minimum value range are truncated.

[0101] The core risk factor is obtained by tensor contraction operation on the multi-dimensional feature tensor, and the specific implementation process is as follows: a contraction weight matrix is defined, the matrix size is 5*1, and the weight values are predefined according to the feature importance. The weight distribution principle is to give higher weight to the features with strong risk correlation, for example, the average weight of power factor is 0.3, the standard deviation of power factor is 0.2, the range of power factor is 0.1, the fluctuation amplitude of reactive power is 0.25, and the fluctuation frequency of reactive power is 0.15. The sum of all weight values is 1.0. When performing tensor contraction operation, first perform dot product operation: multiply each weight element in the weight matrix with the corresponding feature value element in the first dimension of the tensor to obtain five product results; add the five product results to obtain the dot product operation result value. Then perform linear superposition: add the dot product operation result value to the result of multiplying the impedance characteristic mismatch degree value by a fixed weight coefficient, where the fixed weight coefficient is 0.5, and the superposition obtains the core risk factor value. Finally, perform function transformation: input the core risk factor value into the Sigmoid function for calculation, the function calculation process is to divide 1 by the sum of 1 and the negative core risk factor power of the natural constant e, and the result is mapped to the closed interval range of 0 to 1 through the function transformation; convert the transformed core risk factor into a collaborative risk level, and use an equal interval classification method, for example, divide the 0-1 range into 5 equal interval ranges, with an interval width of 0.2, specifically: 0-0.2 corresponds to level 1, 0.2-0.4 corresponds to level 2, 0.4-0.6 corresponds to level 3, 0.6-0.8 corresponds to level 4, and 0.8-1.0 corresponds to level 5. The conversion result is written into the risk attribute field of the corresponding measurement entity in the knowledge graph, and the storage format is an integer value. When the core risk factor is an abnormal value, for example, less than 0 or greater than 1, it is forced to be classified into the lowest or highest level.

[0102] S5, adjust the action strength of the hedging adjustment mechanism according to the collaborative risk level, and the specific implementation is as follows:

[0103] Get the collaborative risk level, and the specific implementation process is as follows: read the collaborative risk level value from the risk attribute field of the load side measurement entity in the knowledge graph, which is generated and stored by step S4; the collaborative risk level is an integer value, ranging in a specific interval, for example, integers from 1 to 5; verify the validity of the level value when acquiring, the verification method includes checking whether the value is an integer and whether it is within the valid range, if it is out of the valid range, trigger the exception handling process, the exception handling includes recalculating the collaborative risk level or replacing it with a default value, the default value is a specific value, for example, level 3; the acquisition operation is performed through a graph database query interface, and the query condition includes the power customer identity and the time range parameter; the query result is cached in the memory, and the cache validity period is a specific time period, for example, 5 minutes, to reduce the database access frequency. The cache update mechanism is to automatically refresh when the validity period is exceeded or data changes are detected.

[0104] The margin compensation parameter feature in the hedging adjustment mechanism is extracted, and the specific implementation process is as follows: the hedging adjustment mechanism data structure generated in the S3 step is located, and the structure is stored in a special cache area; the margin compensation parameter feature field in the hedging adjustment mechanism is accessed, the field stores a floating point value, and represents the margin compensation coefficient; the data structure integrity is verified before extraction, the integrity check includes verifying whether the time stamp is within the valid period, whether the version number is compatible with the current system, and whether the data structure checksum matches; if the data structure is damaged, a new hedging adjustment mechanism is generated by re-invoking the S3 process, and the latest available power side running data is used when generating; the extraction operation adopts a non-blocking read mode, and a read timeout of a specific value, for example, 200 milliseconds, is set to avoid system resource contention; the extracted margin compensation parameter feature is temporarily stored in a register for subsequent processing, and extraction logs are recorded.

[0105] The compensation intensity adjustment coefficient corresponding to the collaborative risk level is determined through the preset risk level compensation mapping relationship, and the specific implementation process is as follows: the preset risk level compensation mapping relationship is stored in an independent mapping configuration file, and the file format is a key-value pair list; the mapping relationship is established in the following manner: the collaborative risk level is divided into different level intervals according to the risk degree, and each level interval corresponds to a preset compensation intensity adjustment coefficient value, for example, risk level 1 corresponds to coefficient 1.2, level 2 corresponds to coefficient 1.0, level 3 corresponds to coefficient 0.8, level 4 corresponds to coefficient 0.6, and level 5 corresponds to coefficient 0.4; the determination process includes analyzing the collaborative risk level value, matching its corresponding level interval in the mapping relationship table, and the matching rule is equal value matching; after successful matching, the corresponding compensation intensity adjustment coefficient value is output; if the matching fails, the default coefficient value 1.0 is used and an error log is recorded. The mapping relationship supports dynamic updating, which needs to be verified through administrator permission when updating, and the updating operation includes version control and rollback mechanism.

[0106] The margin compensation parameter feature is updated using the compensation intensity adjustment coefficient, and the specific implementation process is as follows: the extracted margin compensation parameter feature value is multiplied by the compensation intensity adjustment coefficient, and the calculation formula is: updated parameter = original parameter x adjustment coefficient; the operation process adopts high-precision floating point calculation, and the precision is kept to four decimal places; the calculation result is subjected to boundary check, and the boundary check includes verifying whether the result is within the valid range, and the valid range is set to a specific interval, for example, 0.1 to 2.0; if the result exceeds the valid range, the amplitude limiting rule is to take 0.1 if it is less than the lower limit, and to take 2.0 if it is greater than the upper limit; the updating operation is executed under transaction protection to ensure atomicity and consistency, and the transaction fails automatically retries for a specific number of times, for example, 3 times; the updated margin compensation parameter feature is marked as the latest version, and the version number is incremented. The updating process records operation logs, including the original value, the adjustment coefficient, the updated value and the time stamp.

[0107] The updated margin compensation parameter feature is re-written into the hedging adjustment mechanism to complete the action intensity adjustment, and the specific implementation process is as follows: opening the write channel of the hedging adjustment mechanism data structure; writing the updated margin compensation parameter feature value into the margin compensation parameter feature field; updating the timestamp and version number fields of the data structure at the same time; the write operation adopts a mutual exclusion lock mechanism to prevent concurrent conflicts, and the mutual exclusion lock timeout time is set to a specific value, for example, 100 milliseconds; after the write operation is completed, a verification process is triggered, and the verification content includes field value range verification and data structure integrity verification; after the verification is passed, the hedging adjustment mechanism is marked as a ready state; finally, the hedging adjustment mechanism is stored in the cache area and the associated structure of the graph database, and the association with the load side measurement entity is maintained. After the action intensity adjustment is completed, an operation report is generated, including the parameter comparison before and after the adjustment and the adjustment effect evaluation, and the evaluation indexes include the parameter change amplitude and the adjustment time consumption.

[0108] The preset risk level compensation mapping relationship is implemented in the following way: a mapping relationship table of collaborative risk level and compensation intensity adjustment coefficient is established, the mapping relationship table is stored in a two-dimensional array structure, the first column is the lower limit value of the level interval, the second column is the upper limit value of the level interval, and the third column is the compensation intensity adjustment coefficient value; the collaborative risk level is divided into different level intervals according to the risk degree, and the interval division rule is equal-width division, for example, level 1 interval is [1, 1], level 2 interval is [2, 2], level 3 interval is [3, 3], level 4 interval is [4, 4], and level 5 interval is [5, 5]; each level interval corresponds to a preset compensation intensity adjustment coefficient value, and the coefficient value setting principle is that the higher the risk level, the lower the compensation intensity; when the collaborative risk level is obtained, its corresponding level interval is sequentially matched in the mapping relationship table, and the matching algorithm is traversal comparison, and when the level value is greater than or equal to the interval lower limit and less than or equal to the interval upper limit, it is determined that the matching is successful; the corresponding compensation intensity adjustment coefficient value is output after the matching is successful; the mapping relationship table is initialized by loading from a configuration file, and is resident in memory during running to improve query efficiency. The mapping relationship table supports hot loading, and service is not interrupted during updating.

[0109] S6、In the knowledge graph, the adjusted hedging adjustment mechanism and the associated structure are integrated to form the power customer credit level evaluation result, and the specific implementation is as follows:

[0110] The adjusted hedge adjustment mechanism is obtained. The specific implementation process is as follows: accessing the updated hedge adjustment mechanism data structure in S5 step, the structure is stored in a special cache area or a graph database; reading the value of the margin compensation parameter feature field, the value is a floating point number, indicating the adjusted margin compensation coefficient; verifying the data validity when obtaining, the validity verification includes checking whether the timestamp is within the valid period, the valid period is set as a specific time period, for example, 30 minutes, checking whether the version number matches the latest version, and checking whether the data structure checksum is consistent; if the data is invalid, triggering the reacquisition process, including loading the latest version from the persistent storage or requesting regeneration, and calling the S3 and S5 processes when regenerating; the acquisition operation is performed through a data access interface, and the interface parameters include the power customer identity and the time range; the acquired margin compensation parameter feature is temporarily stored in the cache, and the cache time is a specific length of time, for example, 1 minute, to improve the efficiency of subsequent operations. The cache refresh mechanism is to automatically update when data changes are detected.

[0111] The association structure between the load side metering entity and the power side metering entity in the knowledge graph is located. The specific implementation process is as follows: querying the load side metering entity node in the knowledge graph through the power customer identity; starting from the load side metering entity node, traversing to the power side metering entity node along the power supply and use association relationship edge; locating the topological edge structure in the association structure, the topological edge structure is stored in the knowledge graph edge attribute; the locating process adopts a graph traversal algorithm, the algorithm type is breadth-first search, the search depth is set to 1 layer, and the search timeout time is set to a specific value, for example, 500 milliseconds; the locating result includes the unique identifier and attribute set of the topological edge structure; if the locating fails, checking whether the association relationship is released, the checking method includes querying the association mapping table state field, if it is released, terminating the process and recording the alarm log, and the alarm level is set according to the business importance. The positioning operation is executed under transaction protection to ensure data consistency.

[0112] The margin compensation parameter feature is added to the topology edge structure of the association structure as a dynamic attribute, and the specific implementation process is as follows: opening the attribute editing channel of the topology edge structure; creating a new dynamic attribute field, and the field name is margin compensation parameter feature; writing the obtained margin compensation parameter feature value into the field, and the written value is kept to four decimal places; at the same time, updating the time stamp and version number of the topology edge structure, and the version number is incremented by 1 based on the original value; the optimistic lock mechanism is used to handle concurrent conflicts in the adding operation, and the conflict detection is realized by comparing the version numbers, and when the version numbers do not match, it is automatically retried for a certain number of times, for example, 3 times; after the addition is completed, the attribute verification process is triggered, and the verification content includes field value range verification and data type verification, and the value range is set to a specific interval, for example, 0.1 to 2.0; after the verification is passed, the topology edge structure is marked as an updated state. The dynamic attribute supports an automatic cleaning mechanism, and the cleaning period is a specific time period, for example, 24 hours, and the latest specific number of versions, for example, 5 versions, are retained during cleaning.

[0113] The credit evaluation feature vector is constructed based on the topology edge attribute of the association structure and the metering entity attribute, and the specific implementation process is as follows: creating a fixed-dimension feature vector container, and the dimension number is set according to the number of evaluation indexes, for example, 10 dimensions; the margin compensation parameter feature value is extracted from the topology edge structure as the first dimension feature; the arrears risk state attribute value is extracted from the load side metering entity as the second dimension feature; the supply capacity state attribute value is extracted from the power side metering entity as the third dimension feature; the edge weight attribute value is extracted from the topology edge structure as the fourth dimension feature; the power factor average value attribute value is extracted from the load side metering entity as the fifth dimension feature; the frequency deviation average value attribute value is extracted from the power side metering entity as the sixth dimension feature; other dimensions are filled with preset reference feature values, and the reference values are determined according to the customer type; after the feature vector is constructed, standardization processing is performed, and the standardization method adopts Z-score standardization, that is, each feature value is subtracted from the historical mean and then divided by the historical standard deviation; the standardization parameters are calculated based on the historical data of a specific number of days in the past, for example, 90 days, and the calculation period is a specific time every day, for example, 2 a.m. The feature vector is automatically filtered when it is constructed, and the missing values are filled with the mean value of the same type of entity.

[0114] The credit evaluation feature vector is converted into the power customer credit level evaluation result through a preset credit rating conversion rule, and the specific implementation process is as follows: the preset credit rating conversion rule is stored in the rule engine, and the rule form is a decision tree model; the decision tree model comprises a plurality of judgment nodes and leaf nodes, each judgment node corresponds to a threshold comparison of a certain dimension of the feature vector, for example, when the margin compensation parameter feature is less than 0.5, a high-risk branch is entered; the leaf node stores the credit level result, and the level range is a specific interval, for example, 1 to 10 levels, and the smaller the level value, the higher the credit risk; the conversion process comprises inputting the feature vector into the decision tree model, traversing to the leaf node along the judgment path, outputting the credit level value stored in the leaf node; the decision tree model is periodically retrained using historical data, the training period is a specific time period, for example, the training is performed on the first day of each month, and the training data retention period is a specific length of time, for example, 13 months. The conversion result is kept in integer form, and if the model output is a floating-point number, the integer is rounded. When the conversion fails, a default level value is returned, for example, level 5.

[0115] The power customer credit level evaluation result is stored in the customer credit file node of the knowledge graph, and the specific implementation process is as follows: the customer credit file node is located or created in the knowledge graph, and the node is associated with the power customer identity through the unified social credit code; the credit level evaluation result is written into the credit level field of the node; at the same time, the evaluation time stamp, version number and feature vector snapshot are stored; the batch submission strategy is adopted for the persistent operation, and the submission trigger conditions include accumulating a specific number of updates, for example, submitting once every 50 records, or reaching a specific time interval, for example, forced submission every 5 minutes; after the storage is completed, the credit file node is associated with the load side metering entity and the power side metering entity in a bidirectional manner, and the association relationship type is defined as a credit evaluation association; the persistent data supports the version backtracking function, and the number of historical versions is a specific value, for example, 12 versions, and the version storage period is a specific length of time, for example, 1 year. The storage operation generates an audit log, records the storage time, the operator and the data verification code, and the audit log retention period is a specific length of time, for example, 3 years.

[0116] Embodiment 2 Figure 2 The structure diagram of the power customer credit level evaluation system based on the knowledge graph is given, and the power customer credit level evaluation system based on the knowledge graph comprises:

[0117] The mapping establishment module is used for acquiring the load side operation data and the power side operation data of the power customer, and establishing the data mapping relationship between the load side and the power side based on the power customer identity;

[0118] The graph construction module is used for generating the knowledge graph comprising the load side metering entity and the power side metering entity based on the data mapping relationship, and constructing the association structure of the load side metering entity and the power side metering entity in the knowledge graph.

[0119] The mechanism generation module is configured to analyze and parse the characteristic of the nonlinear section of the grid point voltage reactive power adjustment curve based on the power supply side operation data, and generate a hedging adjustment mechanism for the arrears risk state;

[0120] The grade generation module is configured to extract the reactive power compensation behavior characteristics in the load side operation data, synchronously analyze the mismatch degree of the load impedance characteristics corresponding to the load side metering entity and the power supply impedance characteristics corresponding to the power supply side metering entity, and generate a cooperative risk grade through tensor fusion;

[0121] The adjustment updating module is configured to adjust the action strength of the hedging adjustment mechanism according to the cooperative risk grade;

[0122] The evaluation integration module is configured to integrate the adjusted hedging adjustment mechanism and the associated structure in the knowledge graph to form a power customer credit grade evaluation result.

[0123] Embodiment 3: A power customer credit grade evaluation device based on a knowledge graph, the device comprising: a processor, a memory, and a program or instruction stored on the memory and executable on the processor, the program or instruction being executed by the processor to implement the power customer credit grade evaluation method based on the knowledge graph.

[0124] The calculations involved in the embodiments are all dimensionless numerical calculations, and the preset parameters and threshold values in the calculations are set by a person skilled in the art according to actual conditions.

[0125] It should be noted that the present application can be deployed in the device itself to realize embedded application, or can be run on a PC or other terminal with a user interface, thereby meeting various hardware environments and use requirements.

[0126] The above-described embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented by software, the above-described embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center through wireless or wired transmission. The wired transmission includes optical fiber, twisted pair, coaxial cable, etc. The wireless transmission includes infrared, microwave, etc. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. containing one or more available medium collections. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0127] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device and module can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0128] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the modules is only a logical function division, and actual implementation can have another division manner, for example, a plurality of modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.

[0129] The modules described as separate components can or can not be physically separated, and the components displayed as modules can or can not be physical modules, which can be located in one place or distributed on a plurality of network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments.

[0130] In addition, each functional module in the various embodiments of the present application can be integrated in one processing module, or each module can exist physically independently, or two or more modules can be integrated in one module.

[0131] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to make a computer device (which can be a personal computer, a server or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0132] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0133] Finally: the above is only a preferred embodiment of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A knowledge graph-based method for evaluating the credit rating of electricity customers, characterized in that, include: S1. Obtain the load-side operation data and power-side operation data of the power customer, and establish a data mapping relationship between the load side and the power supply side based on the power customer's identity identifier; S2. Generate a knowledge graph containing load-side metering entities and power-side metering entities based on data mapping relationships, and construct the association structure between load-side metering entities and power-side metering entities in the knowledge graph. S3. Based on the nonlinear segment characteristics of the voltage and reactive power droop curve at the grid connection point, analyze the power supply side operation data to generate a hedging adjustment mechanism for the risk of overdue payment. S4. Extract reactive power compensation behavior characteristics from load-side operation data, simultaneously analyze the degree of mismatch between load impedance characteristics corresponding to load-side metering entities and power supply impedance characteristics corresponding to power supply metering entities, and generate collaborative risk levels through tensor fusion. S5. Adjust the intensity of the hedging mechanism based on the level of collaborative risk. S6. Integrate the adjusted hedging and regulation mechanisms and related structures in the knowledge graph to form the credit rating results of power customers.

2. The knowledge graph-based electricity customer credit rating method according to claim 1, characterized in that, Acquire load-side and power-side operational data from electricity customers, and establish a data mapping relationship between the load side and the power supply side based on the electricity customer's identity identifier, including: Obtain load-side operational data from the electricity marketing system, including electricity metering characteristics and overdue payment risk status; Obtain power source side operation data, including power generation metering characteristics and supply capacity status, from the power dispatching system; By using the unified social credit code in the electricity customer identification, a two-way correlation mapping relationship is established between load-side operation data and power-side operation data; The electricity consumption metering characteristics in the load-side operation data and the power generation metering characteristics in the power supply-side operation data are stored in alignment according to the time dimension. A data mapping index table between load-side metering points and power-side metering points is generated based on a bidirectional correlation mapping relationship.

3. The knowledge graph-based electricity customer credit rating method according to claim 1, characterized in that, A knowledge graph containing load-side metering entities and power-side metering entities is generated based on data mapping relationships. The association structure between load-side metering entities and power-side metering entities is constructed within the knowledge graph, including: The load-side operation data in the data mapping relationship is mapped to the load-side metering entity in the knowledge graph, where the load-side metering entity includes electricity metering characteristics and arrears risk status attributes. The power supply side operation data in the data mapping relationship is mapped to the power supply side metering entity in the knowledge graph, where the power supply side metering entity includes power generation metering characteristics and supply capacity status attributes. Based on the data mapping relationship corresponding to the electricity customer identity, a power generation and consumption association relationship is created between the load-side metering entity and the power-side metering entity; A topological edge structure is formed based on the relationship between power generation and consumption, connecting the load-side metering entity and the power-side metering entity; The knowledge graph storage system persistently stores load-side metering entities, power-side metering entities, and their topological edge structures.

4. The knowledge graph-based method for evaluating the credit rating of electricity customers according to claim 1, characterized in that, Based on the nonlinear segment characteristics of the voltage-reactive power droop curve at the grid connection point, derived from power supply side operation data analysis, a hedging mechanism for the risk of overdue payments is generated, including: Extract the voltage and reactive power droop curve at the grid connection point from the power supply side operating data; Identify the nonlinear characteristic section in the voltage reactive power droop curve. The nonlinear characteristic section is the fluctuation section where the voltage deviates from the rated value. Analyze the voltage fluctuation amplitude and duration characteristics of the nonlinear characteristic segment; The voltage stability margin attenuation factor is calculated based on the product relationship between voltage fluctuation amplitude characteristics and duration characteristics. Generate margin compensation coefficients based on voltage stability margin attenuation factors; A hedging adjustment mechanism is constructed by linking the margin compensation coefficient with the arrears risk status. The hedging adjustment mechanism includes the characteristics of the margin compensation parameter.

5. The knowledge graph-based method for evaluating the credit rating of electricity customers according to claim 1, characterized in that, Extract reactive power compensation behavior characteristics from load-side operational data, simultaneously analyze the mismatch between load impedance characteristics corresponding to load-side metering entities and power supply impedance characteristics corresponding to power supply metering entities, and generate a collaborative risk level through tensor fusion, including: Power factor characteristics and reactive power fluctuation characteristics are extracted from load-side operating data to form reactive power compensation behavior characteristics; Load impedance characteristics are obtained based on the load impedance characteristic data corresponding to the load-side metering entity. The power impedance characteristics are obtained based on the power impedance characteristic data corresponding to the power supply side metering entity. The vector distance between the load impedance characteristics and the power supply impedance characteristics in the power frequency band is calculated as the degree of impedance mismatch. The degree of mismatch between reactive power compensation behavior characteristics and impedance characteristics is constructed as a multidimensional feature tensor. The core risk factors are obtained by performing tensor shrinkage on the multidimensional feature tensor, and then the core risk factors are converted into collaborative risk levels.

6. The knowledge graph-based method for evaluating the credit rating of electricity customers according to claim 1, characterized in that, Adjusting the strength of the hedging mechanism based on the level of collaborative risk includes: Obtain the collaborative risk level; Extracting the characteristics of margin compensation parameters in the hedging adjustment mechanism; By using a pre-defined risk level compensation mapping relationship, the adjustment coefficient for the compensation intensity corresponding to the collaborative risk level is determined; Update the margin compensation parameter characteristics using the compensation intensity adjustment coefficient; The updated margin compensation parameter characteristics are rewritten into the hedging adjustment mechanism to complete the adjustment of the action intensity.

7. The knowledge graph-based electricity customer credit rating method according to claim 6, characterized in that, The preset risk level compensation mapping relationship is implemented in the following way: Establish a mapping relationship table between collaborative risk level and compensation intensity adjustment coefficient, wherein the collaborative risk level is divided into different level intervals according to the degree of risk, and each level interval corresponds to a preset compensation intensity adjustment coefficient value; When obtaining the collaborative risk level, match its corresponding level range in the mapping table and output the corresponding compensation intensity adjustment coefficient.

8. The knowledge graph-based method for evaluating the credit rating of electricity customers according to claim 1, characterized in that, The adjusted hedging and regulation mechanisms and related structures are integrated into the knowledge graph to form the credit rating results for electricity customers, including: Obtain the margin compensation parameter characteristics in the adjusted hedging mechanism; The relationship structure between load-side metering entities and power-side metering entities in the location knowledge graph; Add the margin compensation parameter features as dynamic attributes to the topological edge structure of the associated structure; Credit evaluation feature vectors are constructed based on the topological edge attributes and quantitative entity attributes of the association structure. The credit rating feature vector is converted into the credit rating result of the power customer through the preset credit rating conversion rules. The credit rating results of electricity customers will be persistently stored in the customer credit profile node of the knowledge graph.

9. A knowledge graph-based electricity customer credit rating system, used to implement the knowledge graph-based electricity customer credit rating method according to any one of claims 1-8, characterized in that, include: The mapping establishment module is used to acquire load-side and power-side operation data of power customers and establish a data mapping relationship between the load side and the power supply side based on the power customer's identity identifier. The graph construction module is used to generate a knowledge graph containing load-side metering entities and power-side metering entities based on data mapping relationships, and to construct the association structure between load-side metering entities and power-side metering entities in the knowledge graph. The mechanism generation module is used to analyze the nonlinear segment characteristics of the voltage reactive power droop curve at the grid connection point based on the power supply side operation data, and generate a hedging adjustment mechanism for the risk of overdue payment. The risk level generation module is used to extract reactive power compensation behavior characteristics from load-side operation data, simultaneously analyze the degree of mismatch between the load impedance characteristics corresponding to the load-side metering entity and the power supply impedance characteristics corresponding to the power supply metering entity, and generate a collaborative risk level through tensor fusion. The adjustment and update module is used to adjust the strength of the hedging adjustment mechanism according to the collaborative risk level; The evaluation integration module is used to integrate the adjusted hedging and regulation mechanisms and related structures in the knowledge graph to form the credit rating evaluation results of power customers.

10. A knowledge graph-based power customer credit rating evaluation device, characterized in that, The equipment includes: A processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the knowledge graph-based electricity customer credit rating evaluation method as described in any one of claims 1-8.

Images