Intelligent positioning method, system and equipment for measuring point error of plant station gateway and medium

Abstract

The invention discloses a plant station gateway metering point error intelligent positioning method, system and device, and a medium. The method comprises the following steps: constructing a comprehensive spatio-temporal data set of plant station gateway metering; performing preprocessing and multi-dimensional feature engineering on the spatio-temporal data set to generate a structured feature matrix; based on a power grid topology model and the structured feature matrix, primarily screening anomalies through physical constraints, and determining a suspicious region; performing anomaly scoring on the metering points in the suspicious area by using an unsupervised anomaly detection model, performing secondary scoring, and outputting a high-confidence suspicious metering point list; performing root cause reasoning in combination with the structured feature matrix, a power grid topology model and an equipment knowledge graph to obtain a fault equipment identifier, an error type and confidence corresponding to each suspicious metering point; and matching a preset knowledge rule base according to the fault equipment identifier, the error type and the confidence coefficient, and generating an ordered operation and maintenance suggestion. The intelligent level and the active operation and maintenance level of the metering management of the plant station are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of metering point management technology, and in particular to an intelligent positioning method, system, equipment and medium for metering point errors at plant stations. Background Technology

[0002] The accuracy of data from metering points at power plants and substations directly impacts the economic benefits of enterprises and profoundly affects the safe and stable operation of the power grid. This importance is particularly pronounced against the backdrop of continuously improving power system intelligence and the large-scale grid connection of new energy sources such as wind and solar power. However, in actual operation, metering points are often affected by various factors, such as wiring errors, equipment aging, poor secondary circuit contact, electromagnetic interference, or harsh environments like high temperature and humidity, leading to frequent metering errors. These errors are diverse, subtle, and complex in origin. Traditional methods relying on regular manual inspections often fail to detect anomalies in a timely manner, let alone accurately pinpoint the root cause of faults—a problem prevalent even in grassroots maintenance, hindering the effective improvement of the efficiency and accuracy of anomaly diagnosis.

[0003] In recent years, power companies have deployed a large number of smart meters, current and voltage sensors, environmental monitoring devices, and various automated terminals at their power plants, accumulating multi-source heterogeneous data from metering, equipment status, operation and maintenance records, environmental parameters, and even power grid topology. However, facing such high-frequency, high-dimensional, and multi-type massive amounts of information, how to effectively achieve deep integration of cross-source data, extract discriminative features highly correlated with metering anomalies, and further complete efficient anomaly identification, accurate classification, and root cause tracing has become a core bottleneck that urgently needs to be overcome in the current process of promoting the digital and intelligent transformation of metering operation and maintenance. Existing methods often operate independently, lacking unified modeling capabilities, which makes it difficult to support a truly reliable intelligent diagnostic system. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method, system, equipment, and medium for intelligent positioning of metering point errors at plant and station checkpoints. This addresses the problem that existing plant and station checkpoint metering anomaly diagnosis still relies on manual inspections, making it difficult to effectively utilize accumulated multi-source data to achieve automatic error identification, accurate positioning, and root cause analysis, resulting in low diagnostic efficiency and poor accuracy.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for intelligent positioning of metering point errors at plant checkpoints, comprising: Acquire multi-source operational data from all metering points at the plant, integrate equipment status, maintenance records, and environmental information, and construct a comprehensive spatiotemporal dataset. The spatiotemporal dataset is preprocessed and subjected to multidimensional feature engineering to generate a structured feature matrix; Based on the power grid topology model and the structured feature matrix, anomalies are initially screened through physical constraints to identify suspicious areas; An unsupervised anomaly detection model is used to score anomalies in measurement points within suspicious areas, and a second scoring is performed to output a list of high-confidence suspicious measurement points. Based on the list of high-confidence suspicious metering points, combined with the structured feature matrix, power grid topology model and equipment knowledge graph, root cause reasoning is performed to obtain the fault equipment identifier, error type and confidence level corresponding to each suspicious metering point; Based on the faulty device identifier, error type, and confidence level, a preset knowledge rule base is matched to generate ordered operation and maintenance suggestions.

[0007] As a preferred embodiment of the intelligent positioning method for metering point errors at power plants and substations described in this invention, the step of initial screening for anomalies based on the power grid topology model and the structured feature matrix through physical constraints includes: Based on the electrical connection relationships between metering points, busbars, transformers and switchgear in the power plant, a topology model reflecting the physical structure of the power grid is constructed. Key electrical operating parameters of each device are extracted from the structured feature matrix, and the extracted key electrical operating parameters are aligned to the same time section to obtain aligned key electrical operating parameters. Based on the aforementioned topology model and aligned key electrical operating parameters, the power balance state of each node is solved using a power system power flow calculation algorithm, and the theoretical power loss and the actual observed imbalance are calculated. Based on the imbalance, a preliminary anomaly score is generated. Combined with the connection relationships in the topology model, the anomalies of the parent nodes are traced upstream, and the impact of anomaly propagation is analyzed downstream to finally determine the comprehensive suspicious area.

[0008] As a preferred embodiment of the intelligent positioning method for metering point errors at plant checkpoints described in this invention, the unsupervised anomaly detection model is an autoencoder, and the autoencoder training and scoring process includes: The autoencoder is trained using historical data from normal measurement points; Input the multi-time feature fragments of the measurement points in the suspicious area into the trained autoencoder to obtain the reconstructed output; The weighted reconstruction error is calculated based on the original features and the reconstructed features, and used as a single anomaly score. The maximum value of a single anomaly score for the same measurement point across multiple time windows is taken as the final anomaly score.

[0009] The beneficial effect of this preferred technical solution is that by training the autoencoder using only normal data and combining it with a maximum reconstruction error scoring mechanism with multiple time windows, the detection sensitivity and location reliability of implicit measurement errors are effectively improved.

[0010] As a preferred embodiment of the intelligent positioning method for metering point errors at plant checkpoints described in this invention, the step of performing root cause reasoning includes: High-confidence suspicious measurement points are mapped to associated device entities in the knowledge graph to obtain a set of candidate faulty devices; Based on the set of candidate faulty devices, the corresponding standard fault feature vector is retrieved, and a matching value is calculated with the observed abnormal feature vector. Based on the matching values, the observed likelihood is determined, and combined with the prior probability of historical equipment failures, the posterior confidence of various failures is calculated using a Bayesian algorithm. Output the fault type with the highest posterior confidence, the corresponding device identifier, and the inference path from the metering point to the corresponding device.

[0011] As a preferred embodiment of the intelligent positioning method for metering point errors at plant checkpoints described in this invention, the preprocessing of the spatiotemporal dataset includes: Linear interpolation is used to complete short-term missing data, and spatial nearest neighbor weighted average is used to complete long-term missing data, resulting in a complete data sequence with continuous time. Based on the complete data sequence, the outlier values ​​are corrected using the Hamper filter algorithm to obtain clean data after anomaly cleaning. Based on the clean data, the multi-source data is resampled according to a unified time zone to obtain time-aligned multi-source data. Based on the time-aligned multi-source data, a dynamic time warping algorithm is used to synchronize asynchronous time series with phase shifts, ultimately obtaining a spatiotemporally consistent comprehensive spatiotemporal dataset.

[0012] The beneficial effects of this preferred technical solution are that, through phased missing data completion, anomaly correction, time alignment and sequence synchronization processing, it effectively improves the integrity, consistency and time alignment accuracy of multi-source heterogeneous measurement data, laying a reliable data foundation for subsequent high-precision anomaly detection and root cause inference.

[0013] As a preferred embodiment of the intelligent positioning method for metering point errors at plant and station checkpoints described in this invention, the features extracted by the feature engineering include statistical features, frequency domain features, spatial correlation features, and environmental features. The statistical characteristics include mean, standard deviation, skewness, and kurtosis; The frequency domain characteristics include the energy ratio of the fundamental wave to each harmonic; The spatial correlation features include the spatial autocorrelation coefficient between measurement points; The environmental characteristics include temperature difference, humidity difference, and environmental fluctuation values.

[0014] As a preferred embodiment of the intelligent positioning method for metering point errors at plant checkpoints described in this invention, the step of acquiring multi-source operational data of all metering points at plant checkpoints includes: Real-time data from checkpoint metering points is obtained through power automation communication methods; Receive device status sensor data through IoT sensing communication; Environmental temperature and humidity data are acquired through industrial environmental monitoring communication methods; Import historical operation and maintenance records from the operation and maintenance management system by extracting and converting data from the database.

[0015] Secondly, the present invention provides an intelligent positioning system for metering point errors at plant checkpoints, comprising: The multi-source data fusion module is used to acquire multi-source operational data from all metering points at the plant, integrate equipment status, operation and maintenance records and environmental information, and construct a comprehensive spatiotemporal dataset. The feature engineering and structuring module is used to preprocess the spatiotemporal dataset and perform multidimensional feature engineering to generate a structured feature matrix. The physical constraint screening module is used to identify suspicious areas based on the power grid topology model and the structured feature matrix by screening anomalies through physical constraints. The unsupervised anomaly assessment module is used to score the anomalies of measurement points in suspicious areas using an unsupervised anomaly detection model, perform secondary scoring, and output a list of high-confidence suspicious measurement points. The knowledge-driven root cause reasoning module is used to perform root cause reasoning based on the list of high-confidence suspicious metering points, combined with the structured feature matrix, power grid topology model and equipment knowledge graph, to obtain the fault equipment identifier, error type and confidence level corresponding to each suspicious metering point; The rule matching decision support module is used to match the faulty device identifier, error type and confidence level with a preset knowledge rule base to generate ordered operation and maintenance suggestions.

[0016] Thirdly, the present invention provides an electronic device, comprising: Memory, used to store programs; A processor is configured to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the intelligent positioning method for metering point errors at plant checkpoints.

[0017] Fourthly, the present invention provides a computer-readable storage medium, comprising: when the program is executed by a processor, the steps of implementing the intelligent positioning method for metering point errors at plant and station checkpoints.

[0018] The beneficial effects of this invention are as follows: This invention utilizes multi-source heterogeneous data fusion and spatiotemporal alignment techniques to construct a high-quality, comprehensive spatiotemporal dataset covering operational data, equipment status, maintenance records, and environmental information, significantly improving the completeness and temporal consistency of input data. Through a two-level anomaly detection technique combining physical constraint initial screening and unsupervised anomaly evaluation, it efficiently and accurately identifies high-confidence suspicious points from all metering points, effectively reducing false alarm rates and focusing on latent errors. By integrating knowledge-driven root cause reasoning techniques based on power grid topology, structured features, and equipment knowledge graphs, it achieves interpretable localization from anomaly metering points to specific faulty equipment, error types, and confidence levels, breaking through the traditional diagnostic bottleneck of "knowing the anomaly but not the cause." Finally, through decision generation techniques based on fault diagnosis results and rule base matching, it automatically generates prioritized, ordered maintenance suggestions, promoting the transformation of plant metering management from passive response to proactive intelligent maintenance. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a basic flowchart illustrating an intelligent positioning method for metering point errors at plant checkpoints, provided as an embodiment of the present invention. Detailed Implementation

[0020] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0021] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for intelligent positioning of metering point errors at plant checkpoints is provided, comprising: S100: Acquire multi-source operational data from all metering points at the plant, integrate equipment status, maintenance records and environmental information to construct a comprehensive spatiotemporal dataset; S200: Preprocess the spatiotemporal dataset and perform multidimensional feature engineering to generate a structured feature matrix; S300: Based on the power grid topology model and the structured feature matrix, anomalies are initially screened through physical constraints to determine suspicious areas; S400: Use an unsupervised anomaly detection model to score anomalies in measurement points within suspicious areas, perform secondary scoring, and output a list of high-confidence suspicious measurement points. S500: Based on the list of high-confidence suspicious metering points, combined with the structured feature matrix, power grid topology model and equipment knowledge graph, root cause reasoning is performed to obtain the fault equipment identifier, error type and confidence level corresponding to each suspicious metering point; S600: Based on the faulty device identifier, error type, and confidence level, match the preset knowledge rule base to generate ordered operation and maintenance suggestions.

[0022] It should be noted that existing methods for diagnosing metering anomalies at plant and substation checkpoints face a series of challenges during operation: these include prominent data silos, with metering data, equipment status, maintenance records, and environmental information scattered across different systems, lacking effective integration and making it difficult to comprehensively characterize the causes of anomalies; anomaly detection relies on manual experience or simple threshold rules, resulting in insufficient sensitivity to latent and gradual errors (such as slow CT polarity drift and zero drift of electricity meters), leading to a high false alarm rate; the lack of physical mechanism constraints makes purely data-driven models susceptible to noise interference, resulting in frequent false alarms, and an inability to distinguish between metering device malfunctions and grid operation fluctuations; furthermore, weak location capabilities, making it difficult to trace specific faulty equipment and error types even when anomalies are detected, still requiring extensive manual investigation; and a disconnect between diagnostic results and maintenance, lacking a closed-loop mechanism from "anomaly identification" to "handling recommendations," making it difficult to support proactive maintenance decisions, resulting in low overall efficiency and insufficient intelligence.

[0023] Therefore, the existing diagnostic methods for metering anomalies at plant and station checkpoints still rely on manual inspections, making it difficult to effectively utilize accumulated multi-source data to achieve automatic error identification, accurate location, and root cause analysis, resulting in low diagnostic efficiency and poor accuracy. Through steps S100-S600, an intelligent diagnostic method integrating multi-source data, physical constraints, and knowledge graphs is used to achieve high-precision detection of metering errors at plant and station checkpoints, interpretable root cause location, and proactive maintenance suggestion generation, significantly improving the accuracy, efficiency, and intelligence level of metering anomaly diagnosis.

[0024] Example 2, this is an embodiment of the present invention, which provides an intelligent positioning method for metering point errors at plant checkpoints based on the previous embodiment, including: In this embodiment of the application, step S100, which involves acquiring multi-source operational data from all metering points at the plant's access points, includes: Real-time data from the metering points at the control points is obtained through power automation communication methods (corresponding to IEC 61850). Receive device status sensor data via IoT sensor communication (corresponding to MQTT); Environmental temperature and humidity data are acquired through industrial environmental monitoring communication methods (corresponding to Modbus TCP). Import historical operation and maintenance records from the operation and maintenance management system by using a database extraction and conversion loading method (corresponding to SQL ETL).

[0025] In an optional implementation, the communication protocol and access method in the multi-source data fusion in step S100 can also be that each data source (metering point, equipment sensor, environmental monitoring unit and operation and maintenance system) uniformly publishes the operating data, status information, environmental parameters and historical records to the message middleware (such as Kafka), and the spatiotemporal data fusion engine subscribes to the corresponding topics and aggregates them into the data lake, thereby constructing a comprehensive spatiotemporal dataset.

[0026] In an optional implementation, the communication protocol and access method in the multi-source data fusion in step S100 can also be deployed on the plant side with an edge computing gateway that supports multi-protocol parsing, to uniformly aggregate metering, equipment status, environment and operation and maintenance data and complete the initial fusion, and then upload the structured data stream to the cloud spatiotemporal data fusion engine through a secure encrypted channel to build a comprehensive spatiotemporal dataset.

[0027] In this embodiment of the application, step S200 involves preprocessing the spatiotemporal dataset, including: Linear interpolation is used to complete short-term missing data, and spatial nearest neighbor weighted average is used to complete long-term missing data, resulting in a complete data sequence with continuous time. Based on the complete data sequence, the Hamper filter algorithm is used to correct outliers, resulting in clean data after anomaly cleaning. Based on clean data, multi-source data is resampled according to a unified time zone to obtain time-aligned multi-source data. Based on time-aligned multi-source data, a dynamic time warping algorithm is used to synchronize asynchronous time series with phase shifts, ultimately obtaining a spatiotemporally consistent comprehensive spatiotemporal dataset.

[0028] In this embodiment of the application, the outlier correction algorithm in step S200 includes using the Hampel filter method to identify and correct outliers based on the median and median absolute deviation (MAD) within the sliding window.

[0029] In an optional implementation, the outlier correction algorithm in step S200 can also use a sliding window to calculate the local mean and standard deviation of the complete data sequence that is continuous in time. Data points that deviate from the window mean ± 3 times the standard deviation are judged as outliers and corrected by linear interpolation or mean within the window, thereby obtaining clean data after cleaning.

[0030] In an optional implementation, the outlier correction algorithm in step S200 can further divide the time-continuous complete data sequence into multi-dimensional feature samples according to a sliding window, input them into an isolated forest or LOF model to calculate outlier scores, and remove or interpolate sample points whose outlier scores exceed an adaptive threshold, thereby obtaining clean data after cleaning.

[0031] In this embodiment of the application, the features extracted by feature engineering in step S200 include statistical features, frequency domain features, spatial correlation features, and environmental features; the statistical features include mean, standard deviation, skewness, and kurtosis; Frequency domain characteristics include the energy ratio of the fundamental wave to each harmonic; spatial correlation characteristics include the spatial autocorrelation coefficient between measurement points; environmental characteristics include temperature difference, humidity difference, and environmental fluctuation values.

[0032] In the embodiments of this application, the frequency domain features are obtained by fast Fourier transform to obtain the fundamental wave and each harmonic component, and the harmonic energy ratio is calculated; the spatial correlation features are obtained by calculating the Pearson correlation coefficient or spatial autocorrelation coefficient between time series at different measurement points.

[0033] In this embodiment, the frequency domain features are obtained by obtaining each frequency component through Fast Fourier Transform. And calculate the harmonic energy ratio : in, It is the fundamental wave energy; Let K be the energy of the kth harmonic; K is the total order of harmonics. In this embodiment, the spatial correlation feature is expressed as the Pearson correlation coefficient, which is used to calculate the correlation between any two measurement points. and Spatial autocorrelation coefficient : in, , They are the measurement points and exist The observed value at time; , Its time average.

[0034] In this embodiment, the multidimensional features extracted from all devices at different time periods are concatenated into a structured feature matrix. : in, Let M be the feature value of the Nth device, where M is the feature dimension of a single device.

[0035] In this embodiment of the application, step S300, based on the power grid topology model and the structured feature matrix, performs preliminary screening of anomalies through physical constraints, including: Based on the electrical connection relationships between metering points, busbars, transformers and switchgear in the power plant, a topology model reflecting the physical structure of the power grid is constructed. Key electrical operating parameters of each device are extracted from the structured feature matrix, and the extracted key electrical operating parameters are aligned to the same time section to obtain aligned key electrical operating parameters. Based on the aforementioned topology model and aligned key electrical operating parameters, the power balance state of each node is solved using a power system power flow calculation algorithm, and the theoretical power loss and the actual observed imbalance are calculated. Based on the imbalance, a preliminary anomaly score is generated. Combined with the connection relationships in the topology model, the anomalies of the parent nodes are traced upstream, and the impact of anomaly propagation is analyzed downstream to finally determine the comprehensive suspicious area.

[0036] In this embodiment of the application, the unsupervised anomaly detection model in step S400 is an autoencoder, and the autoencoder training and scoring process includes: The autoencoder is trained using historical data from normal measurement points; Input the multi-time feature fragments of the measurement points in the suspicious area into the trained autoencoder to obtain the reconstructed output; The weighted reconstruction error is calculated based on the original features and the reconstructed features, and used as a single anomaly score. The maximum value of a single anomaly score for the same measurement point across multiple time windows is taken as the final anomaly score.

[0037] In this embodiment of the application, the unsupervised anomaly detection model in step S400 adopts a deep neural network autoencoder (DNN Autoencoder), which is trained only with normal data and scores the error through weighted reconstruction.

[0038] In an optional implementation, in step S400, the unsupervised anomaly detection model can also train the VAE using only historical data of normal measurement points to learn the potential probability distribution of normal patterns. Multi-time feature fragments of measurement points in the suspicious area are input into the trained VAE. The weighted anomaly score is calculated jointly by the reconstruction error and the KL divergence of the latent space, and the final anomaly confidence is determined based on the maximum score of multiple time windows.

[0039] In an optional implementation, the unsupervised anomaly detection model in step S400 can also construct a feature distribution model based on historical data of normal measurement points, input multi-time feature fragments of suspicious measurement points into an isolated forest or LOF model, calculate their anomaly scores as a single score, and take the maximum value of the anomaly scores of the same measurement point under multiple time windows to form a final anomaly score to output a list of high-confidence suspicious measurement points.

[0040] In this embodiment of the application, the mean and standard deviation are calculated based on the final abnormal score distribution of all measurement points, and the abnormal threshold is adaptively determined by adding multiple times the standard deviation to the mean. At the same time, the abnormal scores of each measurement point are converted into standardized confidence scores for likelihood calculation in subsequent root cause inference.

[0041] In the embodiments of this application, the weighted reconstruction error in, For the total number of features, For the first Preset weights for dimensional features and These are the original value and the reconstructed value, respectively.

[0042] In this embodiment, the autoencoder uses weighted mean square error as the loss function: in, This represents the number of samples.

[0043] In this embodiment of the application, the maximum value of a single score for the same measurement point across multiple time windows is taken as the node anomaly score, as follows: in, Assess the anomalies of nodes; Reconstruction error score for the same node under different time windows; T is the number of time windows; Based on the score distribution of all nodes, the threshold is adaptively set: in, The mean; For all rating standard deviations; This is the default value; The scores for each node are then standardized into confidence scores: in, For the first Confidence scores for each sample; In this embodiment of the application, root cause reasoning is performed in step S500, including: High-confidence suspicious measurement points are mapped to associated device entities in the knowledge graph to obtain a set of candidate faulty devices; Based on the set of candidate faulty devices, the corresponding standard fault feature vector is retrieved, and the matching value is calculated with the observed anomaly feature vector. The observed likelihood is determined based on the matching value, and the posterior confidence of various types of faults is calculated by Bayesian algorithm in combination with the prior probability of equipment historical faults. Output the fault type with the highest posterior confidence, the corresponding device identifier, and the inference path from the metering point to the corresponding device.

[0044] In this embodiment of the application, the knowledge representation and matching mechanism in step S500 root cause reasoning includes constructing a device knowledge graph (in the form of triplet RDF), fusing prior probabilities and observational likelihoods with Bayesian formulas, and outputting posterior confidence.

[0045] In an optional implementation, the knowledge representation and matching mechanism in root cause reasoning in step S500 can also construct a fuzzy rule base based on power regulations and expert experience. The observed abnormal features of high-confidence suspicious metering points are fuzzified and matched with preset rules (such as "significant decrease in current and large shift in phase angle → reversed CT polarity"). The matching degree of each fault type is output through the fuzzy inference engine, and the fault equipment, error type and reasoning path with the highest confidence are selected by combining topological association.

[0046] In an optional implementation, the knowledge representation and matching mechanism in root cause reasoning in step S500 can also construct a heterogeneous graph based on the power grid topology and equipment attributes, take the abnormal features of high-confidence suspicious metering points as the initial input of the nodes, perform message passing and aggregation through graph neural networks (such as GAT or RGCN), directly output the probability distribution of fault types of each equipment node, and select the fault type with the highest probability, the corresponding equipment identifier and the implicit reasoning path as the root cause diagnosis result.

[0047] In this embodiment of the application, the observed anomaly feature vector is calculated. Compared with standard fault feature vector Cosine similarity is used as the matching value: Furthermore, this is combined with the prior probability of historical equipment failures. Observational Likelihood The posterior confidence level was calculated using Bayes' theorem: in, For all available fault types.

[0048] In this embodiment, the equipment knowledge graph is constructed based on plant equipment ledgers, defect records from the past five years, real-time monitoring data, and power industry standard documents. The entities in the knowledge graph include metering points, energy meters, current transformers, voltage transformers, transformers, and busbars. Relationship types include physical connection relationships, membership relationships, metering assignment relationships, and fault mode associations. Attributes include equipment model, commissioning time, typical error type, historical fault probability, and maintenance cycle. The knowledge graph is stored in triplet form, examples of which include: <metering point A, monitors, transformer X>, <transformer X, has_fault, "secondary side short circuit">.

[0049] In this embodiment of the application, the orderly operation and maintenance suggestion calculates the optimization score based on the fault confidence and the preset severity score, and outputs the optimization score in descending order.

[0050] Example 3 is an embodiment of the present invention. This embodiment differs from the first embodiment in that it provides an intelligent positioning system for metering point errors at plant checkpoints.

[0051] It should be noted that the technical solution of the intelligent positioning system for metering point error at the plant and station checkpoint is based on the same concept as the technical solution of the intelligent positioning method for metering point error at the plant and station checkpoint described above. For details not described in detail in the technical solution of the intelligent positioning system for metering point error at the plant and station checkpoint in this embodiment, please refer to the description of the technical solution of the intelligent positioning method for metering point error at the plant and station checkpoint described above.

[0052] This embodiment provides an intelligent positioning system for metering point errors at plant checkpoints, comprising: The multi-source data fusion module is used to acquire multi-source operational data from all metering points at the plant, integrate equipment status, operation and maintenance records and environmental information, and construct a comprehensive spatiotemporal dataset. The feature engineering and structuring module is used to preprocess the spatiotemporal dataset and perform multidimensional feature engineering to generate a structured feature matrix. The physical constraint screening module is used to identify suspicious areas based on the power grid topology model and the structured feature matrix by screening anomalies through physical constraints. The unsupervised anomaly assessment module is used to score the anomalies of measurement points in suspicious areas using an unsupervised anomaly detection model, perform secondary scoring, and output a list of high-confidence suspicious measurement points. The knowledge-driven root cause reasoning module is used to perform root cause reasoning based on the list of high-confidence suspicious metering points, combined with the structured feature matrix, power grid topology model and equipment knowledge graph, to obtain the fault equipment identifier, error type and confidence level corresponding to each suspicious metering point; The rule matching decision support module is used to match the faulty device identifier, error type and confidence level with a preset knowledge rule base to generate ordered operation and maintenance suggestions.

[0053] This embodiment also provides an electronic device applicable to a method for intelligent positioning of metering point errors at plant checkpoints, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement a method for intelligent positioning of metering point errors at plant checkpoints, as proposed in the above embodiments.

[0054] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements a method for intelligent positioning of metering point errors at plant checkpoints as proposed in the above embodiments.

[0055] The storage medium proposed in this embodiment belongs to the same inventive concept as the intelligent positioning method for metering point error at a plant station proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0056] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0057] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligently locating errors at metering points in plant and station checkpoints, characterized in that, include: Acquire multi-source operational data from all metering points at the plant, integrate equipment status, maintenance records, and environmental information, and construct a comprehensive spatiotemporal dataset. The spatiotemporal dataset is preprocessed and subjected to multidimensional feature engineering to generate a structured feature matrix; Based on the power grid topology model and the structured feature matrix, anomalies are initially screened through physical constraints to identify suspicious areas; An unsupervised anomaly detection model is used to score anomalies in measurement points within suspicious areas, and a second scoring is performed to output a list of high-confidence suspicious measurement points. Based on the list of high-confidence suspicious metering points, combined with the structured feature matrix, power grid topology model and equipment knowledge graph, root cause reasoning is performed to obtain the fault equipment identifier, error type and confidence level corresponding to each suspicious metering point; Based on the faulty device identifier, error type, and confidence level, a preset knowledge rule base is matched to generate ordered operation and maintenance suggestions.

2. The intelligent positioning method for metering point errors at plant checkpoints as described in claim 1, characterized in that: The initial screening of anomalies based on the power grid topology model and the structured feature matrix through physical constraints includes: Based on the electrical connection relationships between metering points, busbars, transformers and switchgear in the power plant, a topology model reflecting the physical structure of the power grid is constructed. Key electrical operating parameters of each device are extracted from the structured feature matrix, and the extracted key electrical operating parameters are aligned to the same time section to obtain aligned key electrical operating parameters. Based on the aforementioned topology model and aligned key electrical operating parameters, the power balance state of each node is solved using a power system power flow calculation algorithm, and the theoretical power loss and the actual observed imbalance are calculated. Based on the imbalance, a preliminary anomaly score is generated. Combined with the connection relationships in the topology model, the anomalies of the parent nodes are traced upstream, and the impact of anomaly propagation is analyzed downstream to finally determine the comprehensive suspicious area.

3. The intelligent positioning method for metering point errors at plant checkpoints as described in claim 1 or 2, characterized in that: The unsupervised anomaly detection model is an autoencoder, and the autoencoder training and scoring process includes: The autoencoder is trained using historical data from normal measurement points; Input the multi-time feature fragments of the measurement points in the suspicious area into the trained autoencoder to obtain the reconstructed output; The weighted reconstruction error is calculated based on the original features and the reconstructed features, and used as a single anomaly score. The maximum value of a single anomaly score for the same measurement point across multiple time windows is taken as the final anomaly score.

4. The intelligent positioning method for metering point errors at plant checkpoints as described in claim 3, characterized in that: The root cause reasoning includes: High-confidence suspicious measurement points are mapped to associated device entities in the knowledge graph to obtain a set of candidate faulty devices; Based on the set of candidate faulty devices, the corresponding standard fault feature vector is retrieved, and a matching value is calculated with the observed abnormal feature vector. Based on the matching values, the observed likelihood is determined, and combined with the prior probability of historical equipment failures, the posterior confidence of various failures is calculated using a Bayesian algorithm. Output the fault type with the highest posterior confidence, the corresponding device identifier, and the inference path from the metering point to the corresponding device.

5. The intelligent positioning method for metering point errors at plant checkpoints as described in claim 4, characterized in that: The preprocessing of the spatiotemporal dataset includes: Linear interpolation is used to complete short-term missing data, and spatial nearest neighbor weighted average is used to complete long-term missing data, resulting in a complete data sequence with continuous time. Based on the complete data sequence, the outlier values ​​are corrected using the Hamper filter algorithm to obtain clean data after anomaly cleaning. Based on the clean data, the multi-source data is resampled according to a unified time zone to obtain time-aligned multi-source data. Based on the time-aligned multi-source data, a dynamic time warping algorithm is used to synchronize asynchronous time series with phase shifts, ultimately obtaining a spatiotemporally consistent comprehensive spatiotemporal dataset.

6. The intelligent positioning method for metering point errors at plant checkpoints as described in claim 5, characterized in that: The features extracted by the feature engineering include statistical features, frequency domain features, spatial correlation features, and environmental features; The statistical characteristics include mean, standard deviation, skewness, and kurtosis; The frequency domain characteristics include the energy ratio of the fundamental wave to each harmonic; The spatial correlation features include the spatial autocorrelation coefficient between measurement points; The environmental characteristics include temperature difference, humidity difference, and environmental fluctuation values.

7. The intelligent positioning method for metering point errors at plant checkpoints as described in claim 6, characterized in that: The acquisition of multi-source operational data from all metering points at the plant and substation includes: Real-time data from checkpoint metering points is obtained through power automation communication methods; Receive device status sensor data through IoT sensing communication; Environmental temperature and humidity data are acquired through industrial environmental monitoring communication methods; Import historical operation and maintenance records from the operation and maintenance management system by extracting and converting data from the database.

8. A smart positioning system for metering point errors at plant checkpoints, using the method described in any one of claims 1-7, characterized in that, include: The multi-source data fusion module is used to acquire multi-source operational data from all metering points at the plant, integrate equipment status, operation and maintenance records and environmental information, and construct a comprehensive spatiotemporal dataset. The feature engineering and structuring module is used to preprocess the spatiotemporal dataset and perform multidimensional feature engineering to generate a structured feature matrix. The physical constraint screening module is used to identify suspicious areas based on the power grid topology model and the structured feature matrix by screening anomalies through physical constraints. The unsupervised anomaly assessment module is used to score the anomalies of measurement points in suspicious areas using an unsupervised anomaly detection model, perform secondary scoring, and output a list of high-confidence suspicious measurement points. The knowledge-driven root cause reasoning module is used to perform root cause reasoning based on the list of high-confidence suspicious metering points, combined with the structured feature matrix, power grid topology model and equipment knowledge graph, to obtain the fault equipment identifier, error type and confidence level corresponding to each suspicious metering point; The rule matching decision support module is used to match the faulty device identifier, error type and confidence level with a preset knowledge rule base to generate ordered operation and maintenance suggestions.

9. An electronic device, characterized in that, include: Memory, used to store programs; A processor for loading the program to perform the steps of the method as claimed in any one of claims 1-7.

10. A computer-readable storage medium storing a program, characterized in that, When the program is executed by a processor, it implements the steps of the method as described in any one of claims 1-7.

Images