Data forwarding and model mapping matching method based on CAG

Abstract

The invention relates to a CAG-based data forwarding and model mapping matching method, which comprises the following steps of: 1, collecting multi-source heterogeneous electric power CAG data in real time, and carrying out standardization processing on the collected data; step 2, based on the standardized CAG data obtained in the step 1, performing feature extraction and fault matching by using a dynamic model; and step 3, completing CAG data forwarding and model mapping matching processes based on the extracted features and fault matching results obtained in the step 2, and further realizing credible data sharing and collaborative decision-making. According to the invention, the data processing efficiency and accuracy can be obviously improved.

Description

Technical Field

[0001] This invention belongs to the field of power equipment condition monitoring technology, and relates to a data forwarding and model mapping matching method, especially a CAG-based data forwarding and model mapping matching method. Background Technology

[0002] Existing technologies for monitoring the status of power equipment in power systems mainly rely on "rule-based experience judgment", "statistical analysis of a single data source" or "centralized database storage". Existing solutions typically involve manual inspections to obtain voltage, current and switch status, followed by maintenance personnel manually formulating handling plans based on the analysis results.

[0003] Through comparative analysis, existing technologies have at least the following shortcomings that urgently need to be addressed: Defect 1: Low data processing efficiency and insufficient real-time performance Problems: Traditional solutions rely on manual or low-frequency sampling (e.g., once per hour), which cannot capture transient anomalies in the power system (e.g., millisecond-level voltage dips, microsecond-level arc flashovers); data analysis uses offline batch processing mode (e.g., daily summary and analysis), resulting in delayed fault detection.

[0004] Defect 2: Poor model generalization ability and low accuracy. Performance: Existing algorithms are mostly trained on fixed rules based on historical data (such as the "summer load peak formula of a certain region"), but do not consider dynamic factors such as the access of new energy sources (such as the fluctuation of photovoltaic output) and changes in user behavior (such as the randomness of electric vehicle charging); single models (such as using only neural networks) are sensitive to small sample abnormal data, with a false alarm rate as high as 15%-20%.

[0005] Defect 3: Weak system collaboration and severe data silos. Manifestations: Different business systems (such as dispatch automation systems, marketing management systems, and equipment monitoring systems) use independent databases, and the data formats are not uniform (such as a mixture of text, XML, and binary data). Cross-departmental data sharing requires manual conversion. Trust technologies such as blockchain are not applied, and the risk of data tampering is high (such as maintenance records being modified by humans to evade responsibility).

[0006] Defect 4: Incomplete matching logic, low target achievement rate Problem: In existing technical solutions, there are gaps in the main process of "data acquisition → feature extraction → result output"—for example, the raw data collected is not standardized and cleaned (e.g., the units are not uniform), and it is directly input into the analysis model, resulting in feature distortion; there is a lack of feedback mechanism between sub-steps (e.g., the model output error is not corrected in reverse to correct the acquisition parameters), and ultimately it cannot accurately match the core requirements of "quickly locating faults + reducing false alarms".

[0007] To solve the above problems, it is urgent to develop a method based on CAG data forwarding and model mapping matching.

[0008] A search revealed no prior art patents that are identical or similar to this invention. Summary of the Invention

[0009] To address the shortcomings of existing technologies, this invention proposes a CAG data forwarding and model mapping matching method. Based on big data processing and real-time analysis, this method significantly improves data processing efficiency and accuracy through an efficient CAG data forwarding mechanism and a precise model mapping matching algorithm.

[0010] The above-mentioned objective of this invention is achieved through the following technical solution: A method for CAG-based data forwarding and model mapping matching includes the following steps: Step 1: Collect multi-source heterogeneous power CAG data in real time and standardize the collected data; Step 2: Based on the standardized CAG data obtained in Step 1, feature extraction and fault matching are performed using a dynamic model; Step 3: Based on the extracted features and fault matching results obtained in Step 2, complete the CAG data forwarding and model mapping matching process, thereby realizing trusted data sharing and collaborative decision-making.

[0011] Furthermore, the specific steps of step 1 include: Step 1.1: Deploy the distributed sensor network. High-precision sensors are deployed on transmission lines, substations, and user sides to collect parameters such as voltage, current, temperature, and switch status, with the sampling frequency dynamically adjusted according to the scenario.

[0012] Step 1.2: Edge Data Preprocessing By deploying edge computing terminals in substations, the collected raw data is denoised, normalized, and dimensionally unified to generate standardized data packets.

[0013] Step 1.3: Data Upload and Timestamp Marking Standardized data is uploaded to the cloud server in real time via a 5G / fiber optic private network, with each data packet appended with a BeiDou satellite time stamp.

[0014] Furthermore, the specific method for step 2 is as follows: Step 2.1: Perform multimodal feature fusion Extract the time-domain features, FFT spectrum energy distribution, and wavelet coefficients of the standardized data obtained in step 1, and combine them with physical mechanism features to construct a "physical + data" dual-driven feature set.

[0015] Step 2.2: Build, train, and update the adaptive model; A hybrid model combining a physical model and a lightweight neural network is adopted. The physical model provides basic constraints, while the neural network learns the nonlinear mapping relationship of historical faults. Through online incremental learning, the established adaptive model can adapt to dynamic scenarios such as new energy fluctuations and seasonal changes.

[0016] Step 2.3: Accurate Matching of Fault Types Input real-time features into the trained adaptive model, and output the failure probability and type. When the probability is ≥90%, an alert is triggered; if the probability is <90% but the features deviate from the normal range, it is marked as "suspicious anomaly" and pushed to the manual review module.

[0017] Furthermore, the specific steps of step 3 include: Step 3.1: Deploy the blockchain network module Based on consortium blockchain technology, a blockchain network is built that includes "dispatch center nodes", "operations and maintenance department nodes" and "equipment manufacturer nodes", with each node agreeing on data access permissions through smart contracts; Step 3.2: Data Uploading and Tamper-Proof Storage The collected raw data, feature extraction results, and model output conclusions are all encrypted using a hash algorithm before being uploaded to the blockchain. Each data block contains the hash value of the preceding block, and any tampering will result in a mismatch in subsequent hash values.

[0018] Step 3.3: Cross-departmental collaborative handling When a high-risk failure is predicted, the system automatically broadcasts an early warning message via the blockchain, and each node receives the message synchronously and triggers its local contingency plan.

[0019] The advantages and beneficial effects of this invention are as follows: 1. This invention proposes a CAG data forwarding and model mapping matching method. By collecting and preprocessing power CAG data in a standardized format, it improves data quality and data processing efficiency, and achieves efficient data transmission and utilization.

[0020] 2. This invention proposes a model mapping and matching method for CAG data, which integrates multimodal features and constructs a hybrid model to achieve online incremental learning of data and accurate matching of fault types.

[0021] 3. The CAG-based data forwarding and model mapping matching method of the present invention uses blockchain to achieve data tamper-proofing and data access read / write rule agreement, thereby enhancing data credibility and collaborative decision-making capabilities. Attached Figure Description

[0022] Figure 1 This is a flowchart of the processing of the present invention; Figure 2 This is a schematic diagram of the structure of Embodiment 2 of the present invention. Detailed Implementation

[0023] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings: A method for CAG-based data forwarding and model mapping matching includes the following steps: Step 1: Collect multi-source heterogeneous power data in real time and standardize the collected data; The specific steps of step 1 include: Step 1.1: Deploy the distributed sensor network. High-precision sensors are deployed on transmission lines, substations, and user sides to collect parameters such as voltage, current, temperature, and switch status at high frequencies, with the sampling frequency dynamically adjusted according to the scenario.

[0024] Step 1.2: Edge Data Preprocessing By deploying edge computing terminals in substations, the collected raw data is denoised, normalized, and dimensionally unified to generate standardized data packets.

[0025] Step 1.3: Data Upload and Timestamp Marking Standardized data is uploaded to the cloud server in real time via a 5G / fiber optic private network, with each data packet appended with a BeiDou satellite time stamp.

[0026] The working principle of step 1 is as follows: Step 1: Real-time acquisition and standardization of multi-source heterogeneous power data Step 1.1: Deployment of Distributed Sensor Networks High-precision sensors (such as fiber optic current transformers and MEMS voltage sensors) are deployed on transmission lines, substations, and user sides to collect parameters such as voltage (35KV-1000KV), current (0-500A), temperature (-40℃~200℃), and switch status (0 / 1). The sampling frequency is dynamically adjusted according to the scenario (1kHz in high-fault areas and 100Hz in normal areas).

[0027] Step 1.2: Edge Data Preprocessing By deploying edge computing terminals (equipped with ARM processors) in substations, the raw data is denoised (wavelet transform to filter out high-frequency interference), normalized (Min-Max standardized to the [0,1] interval), and dimension-unified (e.g., the current unit is unified to "A"), generating standardized data packets (format: JSON / Protobuf).

[0028] Step 1.3: Data Upload and Timestamp Marking Standardized data is uploaded to the cloud server in real time via a 5G / fiber optic private network. Each data packet is appended with a BeiDou satellite time stamp (accuracy ±10ns) to ensure time alignment of multi-source data.

[0029] Step 2: Based on the standardized CAG data obtained in Step 1, feature extraction and fault matching are performed using a dynamic model; The specific method for step 2 is as follows: Step 2.1: Perform multimodal feature fusion Extract the time-domain features, FFT spectrum energy distribution, and wavelet coefficients of the standardized data obtained in step 1, and combine them with physical mechanism features to construct a "physical + data" dual-driven feature set.

[0030] Step 2.2: Build, train, and update the adaptive model; A hybrid model combining a physical model and a lightweight neural network is adopted. The physical model provides basic constraints, while the neural network learns the nonlinear mapping relationship of historical faults. Through online incremental learning, the established adaptive model can adapt to dynamic scenarios such as new energy fluctuations and seasonal changes.

[0031] Step 2.3: Accurate Matching of Fault Types Input real-time features into the trained adaptive model, and output the failure probability and type. When the probability is ≥90%, an alert is triggered; if the probability is <90% but the features deviate from the normal range, it is marked as "suspicious anomaly" and pushed to the manual review module.

[0032] The working principle of step 2 is as follows: Step 2: Feature extraction and fault matching based on dynamic models Step 2.1: Multimodal Feature Fusion We extract the time-domain features, FFT spectral energy distribution, and wavelet coefficients of standardized data, and combine them with physical mechanism features to construct a "physical + data" dual-driven feature set.

[0033] Step 2.2: Adaptive Model Training and Update A hybrid model of "physical model + lightweight neural network" is adopted: the physical model provides basic constraints, and the neural network learns the nonlinear mapping relationship of historical faults; through online incremental learning (automatically fine-tuning the model parameters for every 100 new data points), the established adaptive model can adapt to dynamic scenarios such as new energy fluctuations and seasonal changes.

[0034] Step 2.3: Accurate Matching of Fault Types Input real-time features into the trained model and output the fault probability (0-1) and type (such as "single-phase grounding", "phase-to-phase short circuit", "overload"). When the probability is ≥90%, an early warning is triggered; if the probability is <90% but the feature deviates from the normal range (such as a sudden increase in current of 50%), it is marked as "suspicious abnormality" and pushed to the manual review module.

[0035] Step 3: Based on the extracted features and fault matching results obtained in Step 2, complete the CAG data forwarding and model mapping matching process, thereby realizing trusted data sharing and collaborative decision-making.

[0036] The specific steps of step 3 include: Step 3.1: Deploy the blockchain network module Based on consortium blockchain technology, a blockchain network is built that includes "dispatch center nodes", "operations and maintenance department nodes" and "equipment manufacturer nodes", with each node agreeing on data access permissions through smart contracts; Step 3.2: Data Uploading and Tamper-Proof Storage The collected raw data, feature extraction results, and model output conclusions are all encrypted using a hash algorithm before being uploaded to the blockchain. Each data block contains the hash value of the preceding block, and any tampering will result in a mismatch in subsequent hash values.

[0037] Step 3.3: Cross-departmental collaborative handling When a high-risk failure is predicted, the system automatically broadcasts an early warning message via the blockchain, and each node receives the message synchronously and triggers its local contingency plan.

[0038] The working principle of step 3 is as follows: Step 3: Blockchain-enabled Trusted Data Sharing and Collaborative Decision-Making Step 3.1: Deployment of Blockchain Network Module Based on consortium blockchain technology (such as Hyperledger Fabric), a blockchain network is built that includes "schedule center nodes", "operations and maintenance department nodes" and "equipment manufacturer nodes". Each node agrees on data access permissions through smart contracts (e.g., the dispatch center can read and write all data, while the operations and maintenance department can only read and write data within its jurisdiction).

[0039] Step 3.2: Data Uploading and Tamper-Proof Storage The collected raw data, feature extraction results, and model output conclusions are all encrypted using a hash algorithm (SHA-256) and then uploaded to the blockchain. Each data block contains the hash value of the preceding block (forming a chain structure). Any tampering will result in a mismatch of subsequent hash values ​​(traceable back to the original operator).

[0040] Step 3.3: Cross-departmental collaborative handling When a high-risk fault is predicted, the system automatically broadcasts early warning information (including fault location, probability, and suggested measures) through the blockchain. Each node receives the information synchronously and triggers its local contingency plan (such as the dispatch center adjusting the power flow distribution, and maintenance personnel receiving work orders and navigating to the site). The response time is ≤1 minute.

[0041] Example 1 like Figure 1 The diagram shown is a flowchart of the processing of this invention. From data acquisition, preprocessing, model matching to result output, after fault warning / suspicious anomaly review, the fault loop is completed through collaborative handling, and finally, the entire process of data acquisition is monitored and the fault is handled in real time, completing the closed-loop matching of the entire process of "data acquisition → preprocessing → model matching → result output".

[0042] Example 2 like Figure 2 As shown in the table below, power data involves diverse equipment, inconsistent data index accuracy, and varying collection frequencies, lacking a unified standard model. Based on the PMS3.0 overall architecture, a new CAG service is built in Zone IV to aggregate data sent from gateways such as CAC at various substations, parse it, and forward it to the power grid resource business platform and the unstructured data platform. Through an efficient CAG data forwarding mechanism and a precise model mapping matching algorithm, data processing efficiency and accuracy can be significantly improved.

[0043]

[0044] System input data status.

[0045]

[0046] System output data.

[0047]

[0048] The innovation of this invention lies in: To address the shortcomings of existing technologies, this invention proposes a method based on CAG data forwarding and model mapping matching, aiming to achieve: ① Real-time acquisition of high-precision power data (sampling frequency ≥ 1kHz); ② By using dynamic model adaptive correction (combining physical mechanisms and machine learning), the accuracy of fault prediction is improved to over 95%; ③ Utilize blockchain networks to ensure trusted data sharing, with a response delay of ≤1 minute; ④ Complete the closed-loop matching process of "data acquisition → preprocessing → model matching → result output".

[0049] The CAG-based data forwarding and model mapping matching method invented in this invention can significantly improve data transmission efficiency and matching accuracy, and is suitable for big data processing and real-time analysis scenarios. Through an efficient CAG data forwarding mechanism and a precise model mapping matching algorithm, it can effectively reduce data transmission latency and matching errors, thereby improving data processing efficiency and accuracy.

[0050]

[0051] It should be emphasized that the embodiments described in this invention are illustrative rather than limiting. Therefore, this invention includes, but is not limited to, the embodiments described in the specific implementation. Any other implementations derived by those skilled in the art based on the technical solutions of this invention are also within the scope of protection of this invention.

Claims

1. A method for CAG data forwarding and model mapping matching, characterized in that: Includes the following steps: Step 1: Collect multi-source heterogeneous power CAG data in real time and standardize the collected data; Step 2: Based on the standardized CAG data obtained in Step 1, feature extraction and fault matching are performed using a dynamic model; Step 3: Based on the extracted features and fault matching results obtained in Step 2, complete the CAG data forwarding and model mapping matching process, thereby realizing trusted data sharing and collaborative decision-making.

2. The method for CAG-based data forwarding and model mapping matching according to claim 1, characterized in that: The specific steps of step 1 include: Step 1.1: Deploy the distributed sensor network. High-precision sensors are deployed on transmission lines, substations, and user sides to collect parameters such as voltage, current, temperature, and switch status, with the sampling frequency dynamically adjusted according to the scenario. Step 1.2: Edge Data Preprocessing By deploying edge computing terminals in substations, the collected raw data is denoised, normalized, and dimensionally unified to generate standardized data packets. Step 1.3: Data Upload and Timestamp Marking Standardized data is uploaded to the cloud server in real time via a 5G / fiber optic private network, with each data packet appended with a BeiDou satellite time stamp.

3. The method for CAG-based data forwarding and model mapping matching according to claim 1, characterized in that: The specific method for step 2 is as follows: Step 2.1: Perform multimodal feature fusion Extract the time-domain features, FFT spectral energy distribution, and wavelet coefficients of the standardized data obtained in step 1, and combine them with physical mechanism features to construct a "physical + data" dual-driven feature set; Step 2.2: Build, train, and update the adaptive model; A hybrid model of "physical model + lightweight neural network" is adopted. The physical model provides basic constraints, and the neural network learns the nonlinear mapping relationship of historical faults. Through online incremental learning, the established adaptive model can adapt to dynamic scenarios such as new energy fluctuations and seasonal changes. Step 2.3: Accurate Matching of Fault Types Input real-time features into the trained adaptive model, output the fault probability and type, and trigger an early warning when the probability is ≥90%. If the probability is less than 90% but the feature deviates from the normal range, it is marked as "suspicious anomaly" and pushed to the manual review module.

4. The method for CAG-based data forwarding and model mapping matching according to claim 1, characterized in that: The specific steps of step 3 include: Step 3.1: Deploy the blockchain network module Based on consortium blockchain technology, a blockchain network is built that includes "dispatch center nodes", "operations and maintenance department nodes" and "equipment manufacturer nodes", with each node agreeing on data access permissions through smart contracts; Step 3.2: Data Uploading and Tamper-Proof Storage The collected raw data, feature extraction results, and model output conclusions are all encrypted using a hash algorithm before being uploaded to the blockchain. Each data block contains the hash value of the preceding block, and any tampering will result in a mismatch in subsequent hash values. Step 3.3: Cross-departmental collaborative handling When a high-risk failure is predicted, the system automatically broadcasts an early warning message via the blockchain, and each node receives the message synchronously and triggers its local contingency plan.

Images