Power terminal equipment data association method based on multiple protocols

Abstract

The invention discloses a multi-protocol-based power terminal equipment data association method, and relates to the technical field of power system data processing, and the method comprises the following steps: S1, multi-source heterogeneous data collection and standardized analysis: obtaining original data of different power terminal equipment through a protocol adapter interface, performing format conversion on the original data based on a preset protocol analysis library; s2, data feature enhancement; s3, performing multi-dimensional intelligent association; s4, association result aggregation and confidence evaluation; s5, performing dynamic learning and adaptive optimization; according to the data association method based on the multi-protocol power terminal equipment, in the operation process, through multi-source data standardization analysis, dynamic feature enhancement, multi-dimensional intelligent association and self-optimization learning, the problems that in the prior art, data formats of multi-protocol equipment are heterogeneous, association dimensions are single, and adaptability is poor are solved; cross-protocol and cross-device data deep association is realized, and the accuracy of state perception and fault diagnosis of the power system is improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of power system data processing, and particularly relates to a method for correlating data of multi-protocol power terminal equipment. BACKGROUND

[0002] There are a large number of terminal equipment (such as smart meters, switch cabinets, transformer monitoring terminals, etc.) in a power system, and the communication protocols adopted by these devices are various (such as Modbus, DL / T645, IEC61850, etc.), which leads to significant differences in data formats and semantic rules, forms a “data island”, and the field of correlating data of multi-protocol power terminal equipment involves data fusion, correlation analysis, storage processing, standardization, security guarantee and multiple technical levels. Effective integration and correlation of these data can provide important support for optimized operation of the power system, realization of intelligent energy management and predictive maintenance, therefore, how to effectively integrate various communication protocol data is crucial to realize intelligentization and efficient operation of the power system. To solve the above defects, the prior art (Chinese patent with publication number CN119788694A and publication date of April 8, 2025) is a method, system and medium for correlating data of multi-protocol power terminal equipment. Compared with the existing terminal equipment anomaly detection method, the application can learn the correlation degree of data of different protocols and quickly obtain a similarity score to achieve the purpose of detecting abnormal data. The method has practical application value in network security monitoring, traffic management and other aspects by helping to identify and classify communication data of different protocols.

[0003] However, the above scheme relies on time stamp or device ID for simple matching in the process of operation, ignores deep correlation such as spatial topology and electrical logic, leads to low correlation accuracy, uses a fixed protocol analysis module, needs to reconstruct the code for new protocols, is difficult to adapt to the rapid iteration of power equipment, the correlation rules are fixed and cannot be dynamically adjusted according to new data, is invalid when the equipment is upgraded or the network topology changes, and the reliability of the correlation result is not quantified, which is difficult to support high-reliability power business.

[0004] Therefore, the application provides a method for correlating data of multi-protocol power terminal equipment to solve the problems in the above method. SUMMARY

[0005] The purpose of the present application is to provide a method for data association of multi-protocol power terminal equipment, to solve the problems of the existing data association method of multi-protocol power terminal equipment in the market, which relies on time stamp or equipment ID for simple matching in the process of operation, ignores spatial topology, electrical logic and other deep associations, resulting in low association accuracy, fixed protocol analysis module, new protocol needs to reconstruct the code, which is difficult to adapt to the rapid iteration of power equipment, the association rule is fixed and cannot be dynamically adjusted according to new data, and the reliability of the association result is not quantified, which is difficult to support high-reliability power business.

[0006] To achieve the above purpose, the present application provides the following technical scheme: a method for data association of multi-protocol power terminal equipment, comprising the following steps: S1. Multi-source heterogeneous data acquisition and standardized analysis: obtaining the original data of different power terminal equipment through a protocol adaptation interface, and performing format conversion on the original data based on a preset protocol analysis library to generate standardized data units containing equipment identification, time stamp, data item, numerical value and data quality label; S2. Data feature enhancement: extending the features of the standardized data units, adding spatial topology features, electrical quantity association features, time series trend features and equipment state features, and constructing an enhanced feature set; S3. Multi-dimensional intelligent association: based on the enhanced feature set, realizing multi-dimensional association through a rule engine, a graph neural network model and a time series association algorithm, and outputting preliminary association results; S4. Association result aggregation and confidence evaluation: clustering and aggregating the preliminary association results, calculating the association confidence through a preset weight model, and screening high-confidence association results; S5. Dynamic learning and adaptive optimization: training the association model based on the high-confidence association results, updating the protocol analysis library and the association rules in real time, and realizing self-optimization of the association method; S6. Application of association results: outputting the high-confidence association results to the power system application layer through a standardized interface to support state evaluation, fault diagnosis and other businesses.

[0007] Preferably, in step S1, the protocol adaptation interface supports Modbus, DL / T645, IEC61850, MQTT and private protocol, the protocol analysis library uses an extensible script engine, and the adaptation of new protocols is realized by adding protocol analysis scripts; Data acquisition: accessing terminal equipment data of different protocols through a protocol adaptation interface (hardware uses an industrial Ethernet gateway, and software integrates a protocol conversion module), including telemetry data (voltage, current, etc.), telematics data (switch status, etc.) and event records; Standardized parsing: Parsing raw data based on extensible protocol parsing library (using Python script engine).

[0008] Preferably, in the step S2, the spatial topology features include device GIS coordinates, power grid topology connection relationship and physical position level; the electrical quantity correlation features include logical relationship of voltage-current-power, phase sequence correspondence relationship; the time series trend features include sliding window mean, volatility and mutation coefficient; and the device state features include health score generated based on data quality label; On the basis of standardized data units, the features are extended in the following ways: Spatial topology features: Combined with the power grid GIS system, the longitude and latitude coordinates of the device, the belonging feeder / substation, and the parent device are added; Electrical quantity correlation features: Derivative features are generated based on circuit theory, "power = voltage x current x power factor" and "three-phase unbalance degree = (maximum phase current - minimum phase current) / average current"; Time series trend features: The mean, variance, slope (reflecting the change trend) and mutation point (current sudden increase ≥ 20% marked as mutation) of the data are calculated through a sliding window (window size can be configured, default 5 minutes); Device state features: Based on the quality label and historical data, a health score (0-100, score ≤ 30 marked as abnormal) is generated through an exponential smoothing algorithm.

[0009] Preferably, in the step S3, the multi-dimensional correlation specifically includes: Rule engine: Hard association is established based on preset power business rules (Ohm's law, three-phase balance condition); Graph neural network model: The device is taken as a node and the feature similarity is taken as an edge weight to construct an association graph, and hidden associations are mined through graph embedding algorithm; Time series correlation algorithm: Dynamic time warping (DTW) is used to calculate the time series similarity of different device data to identify time synchronization correlation.

[0010] Preferably, the multi-dimensional correlation is realized by fusing the rule engine, graph neural network (GNN) and time series algorithm: Rule engine correlation: Preset power business rules: Hard rules: "The phase sequence of smart meters A, B and C in the same area should correspond to A, B and C phases respectively" and "The total power of a transformer should be equal to the sum of the outgoing line powers (error ≤ 5%) "; Soft rules: "After the switch cabinet breaks, the corresponding line current should drop to 0 (delay ≤ 10 seconds)".

[0011] The device data meeting the rules are marked as "rule correlation".

[0012] Preferably, the graph neural network is associated with: Constructing an association graph: nodes are devices, and edge weights are feature similarity (the closer the spatial distance, the higher the electrical quantity correlation, and the greater the weight); Using the GraphSAGE algorithm for graph embedding, hidden associations are mined through node vector similarity (multiple monitoring points on the same line have correlation in current anomalies).

[0013] Preferably, the time sequence association: For the same type of data (current) of different devices, the dynamic time warping (DTW) is used to calculate the time sequence similarity, and the similarity ≥0.85 is marked as "time sequence association"; For event-type data (fault alarm), through time window (default 30 seconds) matching, events in the same window are marked as "time association".

[0014] The output includes the preliminary association results containing the association type (rule / graph model / time sequence).

[0015] Preferably, in the step S4, the input of the weight model includes rule matching degree, feature similarity and historical association accuracy, and the output range is a confidence value of 0-1, and the screening threshold can be dynamically adjusted according to business requirements (the default threshold is 0.7); Association result aggregation and confidence evaluation: Aggregation: The preliminary association results are clustered according to "device group + association type", and the "rule association and time sequence association of transformer T1 and switch cabinet K1" are aggregated as a comprehensive association; Confidence calculation: a weighted summation model is used, and the weight distribution is: rule matching degree (0.4) + feature similarity (0.3) + historical association accuracy (0.3), and the output is a confidence value of 0-1; Screening: high-reliability association results with a confidence value ≥0.7 are retained, and results below the threshold are marked as "to be verified" and need to be confirmed manually.

[0016] Preferably, in the step S5, the dynamic learning is realized through an incremental learning algorithm, when new devices or protocols are added, only the association model needs to be fine-tuned based on new data, without the need to retrain the full data, and the model update time is ≤5 minutes; Dynamic learning and adaptive optimization: Model update: high-confidence association results are used as training samples to update the GNN model and rule weights through an incremental learning algorithm (FTRL) on a regular basis (default daily), to ensure that the model adapts to new data; Protocol expansion: when new protocol devices are accessed, users can upload parsing scripts to the protocol parsing library, and the system automatically loads and generates standardized data without downtime; Abnormal feedback: if the application layer feedback associated result is wrong (fault diagnosis false alarm), the result is added to the negative sample library, and the associated model is optimized in reverse.

[0017] Preferably, the associated result in S6 is applied: State evaluation: based on the associated data of "transformer-switch cabinet-electric meter", the load balance degree of the power distribution substation is comprehensively evaluated; Fault diagnosis: using the associated event chain (switch cabinet opening → current sudden drop → protection action) to locate the fault point, the accuracy is improved to more than 90%.

[0018] Compared with the prior art, the method based on multi-protocol power terminal device data association has the following advantages: 1. Strong protocol adaptability: using an extensible parsing library, supporting quick adaptation of new protocols, and the adaptation time is shortened from 72 hours to 1 hour;

[0019] 2. Comprehensive association dimension: integrating rules, space, time sequence, electrical logic and other multi-dimensional features, the association accuracy is 92%, which is improved by 53% compared with the prior art; 3. Self-optimization capability: through dynamic learning, the model is self-adaptive and updated, and high reliability is maintained when the device is upgraded or the topology changes (adaptation period ≤5 minutes); 4. Strong interpretability: through confidence quantification and association type marking, reliable basis is provided for business decision-making, and manual intervention cost is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 The figure is a schematic diagram of the method based on multi-protocol power terminal device data association. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0022] Embodiment one: please refer to Figure 1 The present application provides the following technical solutions: a method based on multi-protocol power terminal device data association, comprising the following steps: S1. Multi-source heterogeneous data acquisition and standardized parsing: obtaining raw data of different power terminal devices through a protocol adaptation interface, and performing format conversion on the raw data based on a preset protocol parsing library to generate standardized data units containing device identification, timestamp, data item, value and data quality label; S2. Data feature enhancement: Feature expansion is performed on standardized data units, and new spatial topology features, electrical quantity correlation features, time series trend features, and device state features are added to construct an enhanced feature set; S3. Multi-dimensional intelligent correlation: Based on the enhanced feature set, multi-dimensional correlation is achieved through rule engines, graph neural network models, and time series correlation algorithms, and preliminary correlation results are output; S4. Correlation result aggregation and confidence evaluation: The preliminary correlation results are clustered and aggregated, the correlation confidence is calculated through a pre-set weight model, and high-confidence correlation results are selected; S5. Dynamic learning and adaptive optimization: Based on high-confidence correlation results, the correlation model is trained, the protocol analysis library and correlation rules are updated in real time, and the correlation method is self-optimized; S6. Correlation result application: High-confidence correlation results are output to the power system application layer through standardized interfaces to support state assessment, fault diagnosis, and other businesses.

[0023] In step S1, the protocol adaptation interface supports Modbus, DL / T645, IEC61850, MQTT, and private protocols, and the protocol analysis library uses an extensible script engine to adapt to new protocols by adding protocol analysis scripts; Data collection: Terminal device data of different protocols is accessed through the protocol adaptation interface (hardware uses an industrial Ethernet gateway, and software integrates a protocol conversion module), including telemetry data (voltage, current, etc.), telematics data (switch status, etc.), and event records; Standardized analysis: The original data is analyzed based on an extensible protocol analysis library (using a Python script engine).

[0024] In step S2, the spatial topology features include device GIS coordinates, power grid topology connection relationships, and physical location levels; the electrical quantity correlation features include the logical relationship of voltage-current-power, and the phase sequence correspondence; the time series trend features include sliding window mean, volatility, and mutation coefficient; and the device state features include health score generated based on data quality labels; Based on standardized data units, features are expanded in the following ways: Spatial topology features: Combined with the power grid GIS system, the latitude and longitude coordinates of the device, the belonging feeder / substation, and the parent device are added; Electrical quantity correlation features: Derivative features are generated based on circuit theory, "Power = Voltage x Current x Power Factor" and "Three-phase unbalance degree = (Maximum phase current - Minimum phase current) / Average current"; Time series trend features: The mean, variance, slope (reflecting the change trend), and mutation point (current surge ≥20% marked as mutation) of the data are calculated through a sliding window (window size configurable, default 5 minutes). Device state feature: Based on quality labels and historical data, generate health score (0-100, score ≤30 marked as abnormal) through exponential smoothing algorithm.

[0025] In step S3, multi-dimensional association specifically includes: Rule engine: Establish hard association based on preset power business rules (Ohm's law, three-phase balance condition); Graph neural network model: Construct association graph by taking devices as nodes and feature similarity as edge weight, and mine hidden associations through graph embedding algorithm; Time series association algorithm: Calculate time series similarity of different device data using dynamic time warping (DTW) to identify time synchronization association.

[0026] Fusion of rule engine, graph neural network (GNN) and time series algorithm to realize multi-dimensional association: Rule engine association: Pre-set power business rules: Hard rules: "The phase sequence of smart meters A, B, and C in the same area should correspond to A, B, and C phases respectively" "The total power of the transformer should be equal to the sum of each outgoing power (error ≤5%) "; Soft rules: "After the switch cabinet breaks, the corresponding line current should drop to 0 (delay ≤10 seconds)".

[0027] Device data meeting the rules are marked as "rule association".

[0028] Graph neural network association: Construct association graph: Nodes are devices, and edge weights are feature similarity (closer spatial distance, higher electrical quantity correlation, higher weight); Use GraphSAGE algorithm for graph embedding to mine hidden associations through node vector similarity (multiple monitoring points on the same line have correlation when their current is abnormal).

[0029] Time series association: Calculate time series similarity of the same type of data (current) of different devices using dynamic time warping (DTW), and mark as "time series association" if the similarity is ≥0.85; For event-type data (fault alarm), match through time window (default 30 seconds), and mark events within the same window as "time association".

[0030] Output preliminary association results containing association types (rules / graph model / time series).

[0031] In step S4, the input of the weight model includes rule matching degree, feature similarity, and historical association accuracy, and the output is a confidence value ranging from 0 to 1. The screening threshold can be dynamically adjusted according to business needs (default threshold 0.7); Correlation result aggregation and confidence evaluation: Aggregation: Cluster the preliminary correlation results by "device group + correlation type", and aggregate the rule correlation and time sequence correlation between transformer T1 and switch cabinet K1 into a comprehensive correlation. Confidence calculation: Use a weighted summation model, with weight distribution: rule matching degree (0.4) + feature similarity (0.3) + historical correlation accuracy (0.3), output confidence value between 0 and 1. Screening: Keep high-reliability correlation results with confidence ≥0.7, and mark results below the threshold as "to be verified" for manual confirmation.

[0032] In step S5, dynamic learning is achieved through an incremental learning algorithm. When new devices or protocols are added, only the correlation model needs to be fine-tuned based on new data, without the need to retrain the full data set. Model update time ≤5 minutes. Dynamic learning and adaptive optimization: Model update: Regularly (default daily) update the GNN model and rule weights through an incremental learning algorithm (FTRL) using high-confidence correlation results as training samples, ensuring that the model adapts to new data. Protocol expansion: When new protocol devices are connected, users can upload parsing scripts to the protocol parsing library, and the system automatically loads and generates standardized data without downtime. Abnormal feedback: If the application layer feedbacks that the correlation result is incorrect (false alarm of fault diagnosis), the result is added to the negative sample library for reverse optimization of the correlation model.

[0033] Application of correlation results in S6: State evaluation: Based on the correlation data of "transformer-switch cabinet-electric meter", the load balance of the distribution area is comprehensively evaluated. Fault diagnosis: Use the correlation event chain (switch cabinet opening → current sudden drop → protection action) to locate the fault point, with accuracy improved to more than 90%.

[0034] The contents not described in detail in the specification belong to the existing technology known to those skilled in the art.

[0035] Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or make equivalent replacements for part of the technical features, and any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for data association based on multi-protocol power terminal equipment, characterized in that: Includes the following steps: S1. Multi-source heterogeneous data acquisition and standardized parsing: Obtain raw data from different power terminal devices through protocol adaptation interfaces, convert the format of the raw data based on the preset protocol parsing library, and generate standardized data units containing device identifiers, timestamps, data items, values ​​and data quality labels; S2. Data Feature Enhancement: The standardized data units are expanded with new spatial topology features, electrical quantity correlation features, time series trend features, and equipment status features to construct an enhanced feature set; S3. Multi-dimensional intelligent association: Based on the enhanced feature set, multi-dimensional association is achieved through rule engine, graph neural network model and time series association algorithm, and preliminary association results are output; S4. Clustering and Confidence Assessment of Association Results: Cluster the preliminary association results, calculate the association confidence using a preset weight model, and filter out association results with high confidence. S5. Dynamic learning and adaptive optimization: The association model is trained based on high-confidence association results, and the protocol parsing library and association rules are updated in real time to achieve self-optimization of the association method; S6. Application of correlation results: High-confidence correlation results are output to the power system application layer through standardized interfaces to support business functions such as status assessment and fault diagnosis.

2. The method for data association of multi-protocol power terminal equipment according to claim 1, characterized in that: In step S1, the protocol adaptation interface supports Modbus, DL / T645, IEC61850, MQTT and proprietary protocols. The protocol parsing library uses an extensible script engine, and the adaptation to new protocols is achieved by adding new protocol parsing scripts. Data acquisition: Access terminal device data with different protocols through protocol adapter interfaces (hardware adopts industrial-grade Ethernet gateway, software integrates protocol conversion module), including telemetry data (voltage, current, etc.), remote signaling data (switch status, etc.) and event records; Standardized parsing: Parses the raw data based on an extensible protocol parsing library (using a Python script engine).

3. The method for data association of multi-protocol power terminal equipment according to claim 2, characterized in that: In step S2, the spatial topology features include equipment GIS coordinates, power grid topology connection relationships, and physical location hierarchy; Electrical quantity correlation characteristics include the logical relationship between voltage, current, and power, and the phase sequence correspondence; Time-series trend features include sliding window mean, volatility, and mutation coefficient; equipment status features include health scores generated based on data quality labels; Based on standardized data units, features are extended in the following ways: Spatial topology features: Integrate with the power grid GIS system to add the latitude and longitude coordinates of the equipment, its feeder / substation, and parent equipment; Electrical quantity correlation characteristics: Based on circuit theory, derived characteristics are generated, such as "Power = Voltage × Current × Power Factor" and "Three-phase unbalance = (Maximum phase current - Minimum phase current) / Average current". Time series trend characteristics: The mean, variance, slope (reflecting the trend of change), and abrupt change points (current surge ≥20% is marked as abrupt change) of the data are calculated by using a sliding window (window size is configurable, default is 5 minutes). Equipment status characteristics: Based on quality labels and historical data, a health score (0-100, scores ≤30 are marked as abnormal) is generated through an exponential smoothing algorithm.

4. The method for data association of multi-protocol power terminal equipment according to claim 3, characterized in that: In step S3, the multi-dimensional association specifically includes: Rule Engine: Establishes hard associations based on preset power business rules (Ohm's law, three-phase balance conditions); Graph Neural Network Model: Constructs a relational graph by using devices as nodes and feature similarity as edge weights, and mines hidden relationships through graph embedding algorithms; Temporal correlation algorithm: Dynamic Time Warping (DTW) is used to calculate the temporal similarity of data from different devices and identify temporal synchronization correlations.

5. The method for data association of multi-protocol power terminal equipment according to claim 4, characterized in that: The fusion rule engine, graph neural network (GNN), and time-series algorithm achieve multi-dimensional association: Rule engine association: Preset power business rules: Strict rules: "The phase sequence of smart meters A, B, and C in the same transformer area should correspond to phases A, B, and C respectively." "The total power of the transformer should be equal to the sum of the power of each outgoing line (error ≤ 5%)." Soft rule: "After a switchgear tripping event, the corresponding line current should drop to 0 (delay ≤ 10 seconds)"; Device data that meets the rules is marked as "rule association".

6. The method for data association of multi-protocol power terminal equipment according to claim 5, characterized in that: The graph neural network associates features: Construct an association graph: nodes represent devices, and edge weights represent feature similarity (the closer the spatial distance and the higher the electrical correlation, the greater the weight). The GraphSAGE algorithm is used for graph embedding, and hidden correlations are mined by node vector similarity (the current anomalies at multiple monitoring points on the same line segment are correlated).

7. The method for data association of multi-protocol power terminal equipment according to claim 6, characterized in that: The temporal correlation: For the same type of data (current) from different devices, dynamic time warping (DTW) is used to calculate the timing similarity. Data with a similarity of ≥0.85 is marked as "timing-related". For event-based data (fault alarms), matching is performed using a time window (default 30 seconds), and events within the same window are marked as "time-related". The output includes preliminary association results based on association type (rule / graph model / time series).

8. The method for data association of multi-protocol power terminal equipment according to claim 7, characterized in that: In step S4, the input of the weight model includes rule matching degree, feature similarity and historical association accuracy, and the output is a confidence value ranging from 0 to 1. The screening threshold can be dynamically adjusted according to business needs (default threshold 0.7). Association Result Aggregation and Confidence Assessment: Aggregation: The preliminary association results are clustered according to "equipment group + association type", and the "rule association and time sequence association between transformer T1 and switchgear K1" are aggregated into a comprehensive association; Confidence score calculation: A weighted summation model is used, with the weights allocated as follows: rule matching degree (0.4) + feature similarity (0.3) + historical association accuracy (0.3), and the output confidence score value is 0-1; Filtering: Retain high-reliability association results with a confidence level ≥ 0.7, and mark results below the threshold as "to be verified" and require manual confirmation.

9. The method for data association of multi-protocol power terminal equipment according to claim 8, characterized in that: In step S5, the dynamic learning is achieved through an incremental learning algorithm. When a new device or protocol is added, the association model only needs to be fine-tuned based on the new data, without retraining the full dataset. The model update time is ≤5 minutes. Dynamic learning and adaptive optimization: Model update: High-confidence association results are used as training samples regularly (daily by default) to update the GNN model and rule weights through the incremental learning algorithm (FTRL) to ensure that the model adapts to new data; Protocol extension: When connecting new protocol devices, users can upload parsing scripts to the protocol parsing library, and the system will automatically load and generate standardized data without downtime; Anomaly feedback: If the application layer reports an error in the correlation result (false alarm in fault diagnosis), the result is added to the negative sample library to optimize the correlation model in reverse.

10. A method for data association of multi-protocol power terminal equipment according to claim 9, characterized in that: Application of the association results in S6: Status assessment: Based on the correlation data of "transformer-switch cabinet-meter", comprehensively assess the load balance of the distribution substation area; Fault diagnosis: By using the associated event chain (switchgear tripping → current drop → protection action) to locate the fault point, the accuracy rate is improved to over 90%.

Images