Substation site table review method based on multi-dimensional feature fusion and template reasoning machine

Abstract

This invention discloses a substation site table review method based on multidimensional feature fusion and template inference engine. The review process is as follows: importing and preprocessing substation site table data to output a standardized substation site table record set; extracting multidimensional feature vectors from the record set; automatically generating a standard template substation site table based on the substation's structural parameters using a template inference engine; calculating a comprehensive similarity between the multidimensional feature vectors extracted from the standardized field signal record set and the template signals of the standard template substation site table; comparing the comprehensive similarity with a set threshold; and performing the substation site table review based on the comparison result. By combining a template inference engine with a multidimensional feature fusion algorithm, the entire process from substation site table import, signal feature recognition, automatic template matching, and intelligent review is automated.

Description

Technical Field

[0001] This invention relates to the fields of power system control automation and data standardization, and in particular to a method for reviewing substation tables based on multi-dimensional feature fusion and template inference engine. Background Technology

[0002] Substation site meters, as the foundational data for power dispatching and equipment monitoring, encompass all signals required for grid security monitoring. Their accuracy and standardization are crucial for ensuring the safe and stable operation of the grid. Currently available substation site meter management systems can be considered the closest existing engineering technologies to this invention. They possess basic functions such as site meter import, rule review, and report export, but lack extended functions such as automatic template generation and expansion capabilities, multi-dimensional semantic matching and signal merging recognition algorithms, and automated reporting and closed-loop archiving systems. Based on these shortcomings, this invention proposes an overall technical architecture of "template inference engine + multi-dimensional feature fusion algorithm + RPA archiving," achieving a qualitative leap in site meter review from "static rule matching" to "semantic understanding and intelligent reasoning."

[0003] With the accelerated construction of new power systems, the power grid scale continues to expand, and the types of connected equipment are becoming increasingly complex. The traditional method of compiling and reviewing monitoring information point tables (hereinafter referred to as "point tables") at power plants and substations is no longer sufficient to meet the higher requirements of modern power grids for automation, standardization, and lean management. In recent years, although the industry has successively issued a series of guiding documents, due to multiple constraints in technical tools, management mechanisms, and implementation processes, the deep implementation of these standards at the grassroots level still faces several practical problems. 1. The workload is enormous. Currently, most power supply companies still employ a sequential work model of "manual compilation of each item—offline review and approval—manual archiving." This model faces three major challenges: First, compilation efficiency is low, requiring repeated communication between dispatching, operation and maintenance, and substations; a single remote signaling definition often requires multiple revisions from draft to finalization. Second, verification costs are high; manual verification of signal names, range coefficients, and alarm levels is prone to omissions or mismatches due to visual fatigue or experience differences. Third, it involves repetitive and labor-intensive work; template reusability is poor across different projects and intervals, requiring repeated compilation of signals for the same type of circuit breaker and disconnector at different sites. This high-intensity, low-reusability manpower input model is increasingly contradictory to the exponential growth of power grid equipment.

[0004] 2. Inconsistent implementation of standards Although standardized point table templates have been published, clearly defining elements such as signal description, information classification, and categorization coding, significant inconsistencies remain in standard implementation during actual engineering projects due to the lack of an automated verification mechanism embedded in the process. On the one hand, equipment naming conventions are difficult to unify, with design institutes, integrators, and maintenance teams using different methods to describe the same type of signal. On the other hand, alarm level classification is highly arbitrary, with some emergency alarm signals being mistakenly classified as general alarms, or a large number of low-frequency secondary signals crowding out high-level alarm channels, directly weakening the sensitivity of control personnel to abnormal grid conditions and their ability to prioritize responses. Currently, the significant differences in equipment naming and the lack of an automated verification mechanism easily lead to problems such as non-standard naming, signal omissions, and incorrect alarm levels.

[0005] 3. Difficulty in version updates As the standard version is updated, old point tables cannot be quickly and batch-corrected, and templates are difficult to maintain and iterate.

[0006] 4. Report archiving is not intelligent. The review results and scheduled inspection reports need to be entered and archived manually, which is inefficient and makes it easy to lose the review process.

[0007] In summary, power grid companies urgently need an intelligent tool that can automatically generate, review, and report outputs, thereby automating the entire process from data import and standardization comparison to report archiving. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the prior art and provide a substation site list review method based on multi-dimensional feature fusion and template inference engine, realizing full automation of the process from site list import → model parsing → intelligent review.

[0009] The method of this invention integrates template generation technology, rule reasoning engine, and signal semantic recognition algorithm, which can automatically identify device type, match signal specifications, and detect anomalies, thereby significantly improving the intelligence level of point table creation and review.

[0010] This invention can be used for the standardized review of monitoring signal point tables in the integrated control system of power grid companies. This invention also involves interdisciplinary fields such as data standardization, knowledge rule reasoning, natural language semantic analysis, and process automation, and is an important foundational supporting technology in the context of the digital transformation of smart grids.

[0011] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A substation site table review method based on multidimensional feature fusion and template inference engine is proposed. The specific process of the substation site table review method is as follows: importing and preprocessing the site table data to output a standardized site table record set; extracting multidimensional feature vectors from the record set; automatically generating a standard template site table based on the structural parameters of the substation using a template inference engine; calculating the comprehensive similarity between the multidimensional feature vectors extracted from the standardized field signal record set and the template signals of the standard template site table; comparing the comprehensive similarity with a set threshold, and implementing the substation site table review based on the comparison result.

[0012] Furthermore, the point table data preprocessing includes: uniform formatting, synonym replacement, and numbering normalization.

[0013] Furthermore, the standardized point table record set The attributes include: full signal name, device number, signal type, alarm level, and device model.

[0014] Furthermore, the extraction of multidimensional feature vectors from the record set includes: extracting device numbers using regular expressions, parsing voltage levels from names, classifying by keywords, generating signal name embedding vectors through semantic models, and matching the model library of measurement and control or protection devices.

[0015] Furthermore, the template inference engine automatically generates a standard template point table based on the structural parameters of the substation. The generation method is as follows: enter: , Where V represents the voltage level, L represents the wiring method, and M represents the device model; Formula generation: , in, This represents the collection of all equipment under this station. The signal specification rules corresponding to this device; Output: Standard template point list T={t1,t2,…,t n Each template point table contains standard naming, signal type, alarm level, and device association information.

[0016] Furthermore, based on the on-site signal D i Template signal T of the standard template point table j The formula for calculating the overall similarity is: , Among them, S dev For device number matching degree, S sem S represents semantic cosine similarity. type For type consistency, S volt For voltage level consistency, S modelTo ensure consistency in device model; The result of comparing the comprehensive similarity with the set threshold is as follows: S≥ Determined to be standard data; ≤S< The data was deemed to be suspected of being non-compliant. S< The data was determined to be abnormal / missing. in, and The threshold value is used.

[0017] Furthermore, the substation meter review method also includes merging and identifying field signals and standard template meter signals, and semantic clustering.

[0018] Furthermore, the merging and semantic clustering of the field signal and the standard template point table signal specifically involves: Field signal D i Simultaneously with template signals from multiple standard template point tables satisfy ,in If the threshold is D, then the D i Signals identified as merged are tagged with signal merging problem type and named according to semantic clustering rules, which are as follows: .

[0019] Furthermore, the substation meter review method also includes on-site signal alarm classification verification.

[0020] Furthermore, the on-site signal alarm classification verification specifically involves checking whether the on-site signal classification conforms to the specifications through the alarm classification library and marking it accordingly.

[0021] The beneficial effects of this invention are as follows: This invention aims to achieve substation site meter review through automation and intelligent means. By combining a template inference engine and a multi-dimensional feature fusion algorithm, the entire process from site meter import, signal feature recognition, automatic template matching, and intelligent review is automated. Specifically: 1. Significantly improve the automation and accuracy of checklist review. (1) Automated generation and verification The system automatically generates standard point tables that conform to the specifications of China Southern Power Grid based on template inference engine. It can realize automatic inference and expansion of templates for different voltage levels, wiring methods and equipment models, effectively replacing the work of manually creating and maintaining templates.

[0022] (2) High-precision review algorithm By employing a multi-dimensional feature fusion similarity algorithm, which comprehensively considers five dimensions—device number, voltage level, signal type, semantic features, and device model—the algorithm effectively avoids the misjudgments and omissions caused by traditional regular expression matching. Verification has shown that the accuracy of automated review has increased from approximately 80% to 95% compared to manual methods, while the average review time has been reduced from 2 hours to less than 10 minutes.

[0023] 2. Enhance naming consistency and signal integrity identification (1) Synonym recognition regardless of order The system can automatically identify equivalent names such as "circuit breaker closing signal" and "closed circuit breaker", avoiding false alarms caused by differences in sequence or terminology.

[0024] (2) Intelligent recognition of signal merging By using a semantic clustering model, the system can automatically separate and identify composite signals such as "motor / heating power supply failure", effectively supporting accurate comparison of on-site signal merging.

[0025] (3) Automatic signal loss detection The system automatically scans for differences between the standard template and the set of field signals, accurately identifying missing signals, redundant signals, and duplicate definitions to ensure the integrity of the monitoring point table. Attached Figure Description

[0026] Figure 1 This is a flowchart of the substation site list review method based on multi-dimensional feature fusion and template inference engine of the present invention. Detailed Implementation

[0027] The present invention will be further described in detail below with reference to the embodiments.

[0028] Flowchart of the substation site list review method based on multi-dimensional feature fusion and template inference engine, as follows: Figure 1 As shown. The specific steps are as follows: (1) Import and preprocessing of point tables The system supports importing dot table files in formats such as Excel, CSV, and XML. During the import phase, the following processing is performed: Uniform formatting: Remove spaces, special characters, and newline characters; Synonym replacement: Standardize the terms "switch closing" and "circuit breaker closing" to reflect the same meaning; Numbering standardization: For example, “#1”, “1#”, and “No.1” are all standardized to “#1”.

[0029] Output: Standardized point record set Each signal contains attribute fields: .

[0030] (2) Multidimensional feature point extraction To achieve intelligent matching and review, the system extracts multi-dimensional feature vectors from each signal: , in: .

[0031] (3) Template inference engine automatically generates standard point table The template inference engine derives a standard template point table based on the substation's structural parameters: enter: in: V: Voltage level (e.g., 220kV) L: Wiring method (e.g., double busbar / 3 / 2 wiring). M: Device model; Reasoning logic: , in, This represents the collection of all equipment under this station. This refers to the signal specification rules corresponding to this device.

[0032] Output: Standard template point table Each template signal includes a standard name, signal type, alarm level, and device association information.

[0033] (4) Multidimensional feature matching and normative review For each actual signal With template signal Calculate the overall similarity: , in: ; Decision logic: .

[0034] (5) Signal merging and recognition and semantic clustering If a certain field signal D_i is simultaneously associated with multiple template signals satisfy If the signal is identified as a "merging signal", the system output should be named as follows: , And label the problem type as "signal merging".

[0035] For example, if the field signal "Circuit breaker power failure" corresponds to "motor power failure" and "heating power failure" in the template, it is recommended to name it "motor / heating power failure".

[0036] (6) Signal alarm classification verification The system uses the alarm classification library (Accident / Abnormal / General) to check whether the field signal classification conforms to the specifications: if the template signal is "Accident" but the field is "Abnormal", it marks "Classification Error"; if the field classification is missing, it is recommended to complete it.

[0037] The above content is only used to illustrate the technical solution of the present invention. Simple modifications or equivalent substitutions made by those skilled in the art to the technical solution of the present invention do not depart from the essence and scope of the technical solution of the present invention.

Claims

1. A method for reviewing substation site lists based on multi-dimensional feature fusion and template inference engine, characterized in that: The specific process of the substation meter review method is as follows: Importing and preprocessing point table data, outputting a standardized point table record set; the data preprocessing involves adjusting the data format. Extract multidimensional feature vectors from the record set; The template inference engine automatically generates a standard template point table based on the structural parameters of the substation. The comprehensive similarity is calculated based on the multidimensional feature vectors extracted from the standardized record set of field signals and the template signals of the standard template point table; The overall similarity is compared with a set threshold, and the substation site list is reviewed based on the comparison result.

2. The substation site list review method based on multi-dimensional feature fusion and template inference engine according to claim 1, characterized in that: The point table data preprocessing includes: uniform formatting, synonym replacement, and numbering normalization.

3. The substation site list review method based on multi-dimensional feature fusion and template inference engine according to claim 1, characterized in that: The standardized point table record set The attributes include: full signal name, device number, signal type, alarm level, and device model.

4. The substation site list review method based on multi-dimensional feature fusion and template inference engine according to claim 1, characterized in that: The extraction of multidimensional feature vectors from the record set includes: extracting device numbers using regular expressions, parsing voltage levels from names, classifying by keywords, generating signal name embedding vectors through semantic models, and matching the model library of measurement and control or protection devices.

5. The substation site list review method based on multi-dimensional feature fusion and template inference engine according to claim 1, characterized in that: The template inference engine automatically generates a standard template point table based on the structural parameters of the substation. The generation method is as follows: enter: , Where V represents the voltage level, L represents the wiring method, and M represents the device model; Formula generation: , in, This represents the collection of all equipment under this station. The signal specification rules corresponding to this device; Output: Standard template point list T={t1,t2,…,t n Each template point table contains standard naming, signal type, alarm level, and device association information.

6. The substation site list review method based on multi-dimensional feature fusion and template inference engine according to claim 1, characterized in that: The result of comparing the comprehensive similarity with the set threshold is as follows: S≥ Determined to be standard data; ≤S< The data was deemed to be suspected of being non-compliant. S< The data was determined to be abnormal / missing. Where S represents the overall similarity. and The threshold value is used.

7. The substation site list review method based on multi-dimensional feature fusion and template inference engine according to claim 1, characterized in that: The substation meter review method also includes merging and recognizing field signals and standard template meter signals, and semantic clustering.

8. The substation table review method based on multi-dimensional feature fusion and template inference engine according to claim 7, characterized in that: The specific process of merging and identifying the field signal and the standard template point table signal, and performing semantic clustering, is as follows: Field signal D i Simultaneously with template signals from multiple standard template point tables satisfy ,in If the threshold is D, then the D i Signals identified as merged are tagged with signal merging problem type and named according to semantic clustering rules, which are as follows: 。 9. The substation site list review method based on multi-dimensional feature fusion and template inference engine according to claim 1, characterized in that: The substation meter review method also includes on-site signal alarm classification verification.

10. The substation table review method based on multi-dimensional feature fusion and template inference engine according to claim 9, characterized in that: The on-site signal alarm classification verification specifically involves checking whether the on-site signal classification conforms to the specifications through the alarm classification library and marking it accordingly.

Images