Lamination structure matching method and system for GIM model and equipment ledger, equipment and medium

Abstract

The invention relates to the technical field of digital conversion of a power system, and discloses a layered structure matching method, system and equipment for a GIM model and an equipment ledger, and a medium, and the method comprises the steps: obtaining first data containing ledger data and grid information model data under target equipment, and carrying out the preprocessing of the first data; establishing a first matching algorithm based on the preprocessed first data; establishing a model training set according to a first matching algorithm result, and training a second matching adjustment model by using the model training set; and carrying out layered structure matching according to the output of the trained second matching adjustment model. A multi-dimensional set and a dynamic verification process are adopted, and accurate matching of the complex relation between the GIM model and the equipment ledger is achieved. The space, electrical, structure and time sequence dimensions respectively ensure physical correspondence of equipment, capture an electrical connection relationship, reveal an internal structure and an assembly relationship, record an operation time sequence and state change, and comprehensively use to improve matching accuracy and comprehensiveness.

Description

Technical Field

[0001] This invention relates to the field of power system digital conversion technology, and in particular to a method, system, equipment and medium for matching the layered structure of GIM model and equipment ledger. Background Technology

[0002] To accelerate the construction of a smart and safe power grid and achieve a high degree of integration of "power flow, information flow, and business flow," information technology is the core foundation. GIM models have become a key technology for the informatization of power transmission and transformation engineering, but several problems exist. First, regarding the standards system, although a preliminary system has been established, compatibility is insufficient, industry coverage is incomplete, interface specifications with other systems within the State Grid are not fully unified, and the modeling specifications for secondary equipment are not detailed enough. Second, regarding transmission efficiency and real-time performance, large model transmission faces bottlenecks, has poor adaptability to weak networks, and poses security risks. Third, regarding the integration of multi-source heterogeneous data, integration is difficult, the efficiency of full lifecycle data management, storage, and retrieval is low, and information mining technology has not fully released the value of GIM data. Fourth, regarding matching with power transmission and transformation equipment ledger data, complete matching is difficult, and real-time performance is not high.

[0003] The GIM model is of great significance to the informatization, security, and intelligentization of the State Grid. Future development can focus on four aspects: enriching standards and specifications, improving transmission efficiency, matching multi-source data, and AI-driven fusion analysis. Against the backdrop of building a new power system and promoting digital transformation, digital twin technology for power transmission and transformation equipment is entering a period of rapid development, placing higher demands on accurate perception and intelligent control of equipment status.

[0004] In data computation, raw data is prone to loss and corruption. This invention proposes a layered matching algorithm for GIM models and equipment ledgers of digital twins for power transmission and transformation equipment. It comprehensively considers equipment geographic information, operating status, and historical records to establish correspondences, collects and processes equipment data to ensure information integrity and accuracy, and employs a multi-dimensional matching algorithm to optimize the matching degree. Simultaneously, machine learning technology is introduced to improve matching efficiency and accuracy, and the algorithm includes a real-time data verification module to correct matching deviations. This algorithm can provide accurate data support for power transmission and transformation equipment, and has significant research value and application prospects. Summary of the Invention

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] Therefore, this invention provides a method, system, device, and medium for matching the layered structure of GIM models and equipment ledgers, which can solve the problems of insufficient compatibility, incomplete matching, and low real-time performance when matching GIM models and equipment ledgers, improve matching efficiency and accuracy, and ensure the integrity and accuracy of information.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] In a first aspect, the present invention provides a method for matching the hierarchical structure of a GIM model and an equipment ledger, comprising:

[0009] Acquire the first data under the target device and preprocess the first data under the target device. The first data includes ledger data and grid information model data (GIM data).

[0010] The first data under the target device includes static data, three-dimensional geometric and spatial data, operational status data, and business-related data;

[0011] A first matching algorithm is established based on the first data under the preprocessed target device.

[0012] The first matching algorithm is used to perform multi-dimensional matching on the preprocessed ledger data and grid information model data;

[0013] A model training set is established based on the matching results of the first matching algorithm, and the second matching adjustment model is trained using the model training set.

[0014] Based on the second matching after training, adjust the model's output to perform layered structure matching.

[0015] As a preferred embodiment of the layered structure matching method for GIM model and equipment ledger described in this invention, the first matching algorithm includes:

[0016] A multi-dimensional set is preset, which includes several dimensions for measuring ledger data and grid information model data;

[0017] Establish matching logic for several dimensions based on the aforementioned multi-dimensional set;

[0018] The matching logic of the aforementioned dimensions is combined to form the first matching algorithm.

[0019] This preferred solution comprehensively considers the various relationships between ledger data and grid information model data, improving the accuracy and efficiency of matching. By pre-setting a multi-dimensional set, this invention can formulate corresponding matching standards for different types of ledger data and grid information model data, thereby ensuring the rationality and reliability of the matching. Simultaneously, by establishing dimensional matching logic based on the multi-dimensional set, this invention can perform more detailed comparisons and analyses of the data, further improving the accuracy of matching. Combining several dimensional matching logics to form the first matching algorithm comprehensively considers the relationships between various dimensions, yielding more accurate matching results.

[0020] As a preferred embodiment of the layered structure matching method for GIM model and equipment ledger described in this invention, the step of establishing a model training set based on the matching results of the first matching algorithm includes:

[0021] Establish matching criteria for different dimensions;

[0022] The matching degree is obtained based on the matching degree judgment criteria of the first matching algorithm.

[0023] Matching results that meet the matching degree judgment criteria are taken as positive samples, and matching results that do not meet the matching degree judgment criteria are taken as negative samples.

[0024] The positive and negative samples are combined to form the model training set.

[0025] As a preferred embodiment of the layered structure matching method for GIM model and equipment ledger described in this invention, the step of adjusting the model output based on the second matching after training to perform layered structure matching includes:

[0026] A preset matching verification mechanism is provided, which is used to verify whether the matching deviation of the first data after performing layer structure matching meets the preset matching deviation requirements.

[0027] The matching and verification mechanism includes several layers of dynamic verification processes.

[0028] As a preferred embodiment of the layered structure matching method for GIM model and equipment ledger described in this invention, the multi-dimensional set includes at least one or more of the following:

[0029] Spatial dimension, electrical dimension, structural dimension, and temporal dimension;

[0030] The spatial dimension is used to describe the positional relationship of the target devices in physical space, including the relative positions and layout between devices;

[0031] The electrical dimension is used to describe the electrical connection relationship or the transmission path of electrical parameters between target devices;

[0032] The structural dimension is used to describe the composition structure of the target device and the assembly relationships between its components;

[0033] The time sequence dimension is used to describe the time sequence and state changes of the target device during operation.

[0034] As a preferred embodiment of the layered structure matching method for GIM model and equipment ledger described in this invention, the plurality of dynamic verification processes include a primary verification process, an intermediate verification process, and a high-level verification process.

[0035] The initial verification process includes intercepting abnormal data based on device physical thresholds;

[0036] The intermediate verification process includes using a sliding window combined with the 3σ principle to detect transient anomalies and using the median value within the window for temporary correction.

[0037] As a preferred embodiment of the layered structure matching method for GIM model and equipment ledger described in this invention, the advanced verification process includes calculating the cosine similarity between the real-time data of the target equipment and the historical normal pattern, and initiating a hybrid correction strategy for data with a similarity lower than 0.8.

[0038] Secondly, this invention provides a layered structure matching system for GIM models and equipment ledgers, comprising four modules:

[0039] The data acquisition and processing module is used to acquire first data under the target device and preprocess the first data under the target device. The first data includes ledger data and grid information model data.

[0040] The first data under the target device includes static data, three-dimensional geometric and spatial data, operational status data, and business-related data;

[0041] The algorithm establishment module is used to establish a first matching algorithm based on the first data under the preprocessed target device;

[0042] The first matching algorithm is used to perform multi-dimensional matching on the preprocessed ledger data and grid information model data;

[0043] The model building module is used to build a model training set based on the matching results of the first matching algorithm, and to train the second matching adjustment model using the model training set;

[0044] The matching module is used to adjust the model's output based on the second matching after training to perform layered structure matching.

[0045] Thirdly, the present invention provides an electronic device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0046] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0047] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention proposes a method for matching the layered structure of GIM models and equipment ledgers. It acquires first data under the target equipment and preprocesses this first data, which includes ledger data and grid information model data. A first matching algorithm is established based on the preprocessed first data under the target equipment. A model training set is established based on the matching results of the first matching algorithm, and a second matching adjustment model is trained using the model training set. Layered structure matching is performed based on the output of the trained second matching adjustment model. By employing the above-mentioned multi-dimensional set and dynamic verification process, this invention can achieve accurate matching of complex relationships between GIM models and equipment ledgers. Specifically, the spatial dimension ensures accurate correspondence of equipment in physical space, the electrical dimension captures the electrical connection relationships between equipment, the structural dimension reveals the internal structure and component assembly relationships of the equipment, and the temporal dimension records the time sequence and state changes of the equipment during operation. The comprehensive application of these dimensions greatly improves the accuracy and comprehensiveness of the matching.

[0048] In the dynamic verification process, the primary verification process quickly intercepts abnormal data using physical thresholds of the device. The intermediate verification process further detects transient anomalies by using a sliding window combined with the 3σ principle and uses the median within the window for temporary correction to ensure data stability and reliability. The advanced verification process calculates the cosine similarity between real-time data and historical normal patterns, and initiates a hybrid correction strategy for data with low similarity, further improving the accuracy and robustness of the matching.

[0049] In summary, this invention not only solves the problems of insufficient compatibility, incomplete matching, and low real-time performance in matching GIM models with equipment ledgers in existing technologies, but also significantly improves the efficiency and accuracy of matching by introducing multi-dimensional matching algorithms and dynamic verification processes, ensuring the integrity and accuracy of information. This is of great significance for accelerating the construction of smart and safe power grids and achieving a high degree of integration of "power flow, information flow, and business flow". Attached Figure Description

[0050] 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.

[0051] Figure 1 This is a flowchart illustrating a method for matching the layered structure of a GIM model and an equipment ledger, as provided in an embodiment of the present invention.

[0052] Figure 2This is an internal structure diagram of an electronic device for a layered structure matching method of GIM model and equipment ledger provided in an embodiment of the present invention. Detailed Implementation

[0053] 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.

[0054] This application provides a method that can effectively solve the problems mentioned above. The following will describe in detail how to implement the layered structure matching method of the GIM model and the equipment ledger with multiple embodiments.

[0055] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a method for matching the layered structure of a GIM model and an equipment ledger, including:

[0056] Existing technologies have several problems, such as inaccurate matching between GIM models and equipment ledgers, and insufficient real-time performance. These problems seriously affect the progress of power grid informatization.

[0057] Figure 1 A flowchart illustrating a method for matching the hierarchical structure of a GIM model and an equipment ledger is shown, including:

[0058] S101, acquire the first data under the target device, and preprocess the first data under the target device. The first data includes ledger data and GIM data, wherein:

[0059] It should be noted that if you want to achieve the layered structure matching between the GIM model and the equipment ledger of a certain power equipment, you must obtain the relevant data of the GIM model and the equipment ledger. Therefore, it is necessary to collect the ledger data and GIM data of the power equipment.

[0060] In some specific implementations, ledger data is the core business data for power grid companies to manage the entire lifecycle of transmission and transformation equipment. It records the technical parameters, operation and maintenance records, and asset information of the equipment from commissioning to decommissioning. Ledger data typically includes basic information such as the equipment's name, model, manufacturer, installation location, and date of commissioning. At the same time, ledger data provides accurate static attributes and dynamic management data support for the GIM model.

[0061] In some specific implementations, it is also necessary to obtain the device's mesh information model data, i.e., GIM data. This type of data typically includes the device's three-dimensional geometric information, spatial location information, operating status information, and electrical connection information with other devices. This data forms the basis for subsequent matching algorithms.

[0062] It should be noted that the present invention achieves subsequent structural matching by acquiring the first data.

[0063] In this embodiment of the invention, the first data under the target device includes static data, three-dimensional geometric and spatial data, operating status data, and business-related data.

[0064] It should be noted that the data contained in the first set of data can comprehensively reflect the various attributes and states of the target device, providing a rich information foundation for subsequent matching algorithms.

[0065] In some specific implementations, static data records basic information about the equipment, such as name, model, electrical parameters, and physical characteristics, and serves as an important identifier for the equipment.

[0066] In some specific implementations, three-dimensional geometric and spatial data describes the location and layout of the equipment in physical space, encompassing three-dimensional geometric models, geospatial coordinates, installation orientation parameters, and electrical connection topology between devices, which helps to achieve accurate reproduction of the equipment in a virtual environment.

[0067] In some specific implementations, the operating status data reflects the real-time operating status of the equipment, which is crucial for monitoring the health status of the equipment and predicting potential failures. It also includes historical operating records, which are of great reference value for subsequent fault handling.

[0068] In some specific implementations, business-related data reveals the connections between equipment and other business processes, helping to understand the role and value of equipment from a more macro perspective.

[0069] In this embodiment of the invention, a multi-source data acquisition network including GIM, PMS ledger data, etc., is established to obtain the first data of the equipment, including multi-dimensional data of the entire life cycle of the power transmission and transformation equipment, which can be divided into the following categories, as shown in Table 1.

[0070] Table 1. Multidimensional Data Classification of Data Input Module

[0071]

[0072] In this invention, static data includes unique equipment identification information (ID, name, model), manufacturer information, electrical parameters (voltage level, rated capacity, etc.), physical characteristics (material, structural dimensions), and management information (asset code, commissioning time, operation and maintenance unit), etc. This is the most fundamental data source for power transmission and transformation equipment.

[0073] In this embodiment of the invention, the three-dimensional geometric and spatial data encompasses the three-dimensional geometric model of the device (overall and component modeling), geospatial coordinates (latitude and longitude, altitude), installation orientation parameters, and electrical connection topology between devices, providing support for visualization and spatial analysis.

[0074] In this embodiment of the invention, the operating status data is required to include both real-time operating parameters (electrical quantity measurements, switch status, protection signals) collected by the SCADA system and historical operating records (load change curves, fault event records, and performance comparison data before and after maintenance).

[0075] In this embodiment of the invention, the business-related data includes full-process business data, spanning all stages of the equipment's lifecycle: technical drawings and simulation analysis reports in the design phase; process standards and material lists in the construction phase; inspection records, defect ledgers, and test reports in the operation and maintenance phase; and life assessment and disposal records in the decommissioning phase. This organic integration of structured and unstructured data provides comprehensive data support for the digital management of the power grid.

[0076] It should be noted that PMS (Production Management System) ledger data is the core business data for power grid companies to manage transmission and transformation equipment throughout its entire lifecycle. It records the technical parameters, operation and maintenance records, and asset information of the equipment from commissioning to decommissioning. Essentially, it is a "digital archive" of power equipment, providing accurate static attributes and dynamic management data support for the GIM model.

[0077] In specific GIM model and equipment ledger layered structure matching systems, the data input module is the foundation for feature extraction, matching algorithms, and real-time verification, directly impacting the accuracy and reliability of the digital twin system. It requires the collection and processing of structural and operational data from various types of equipment to ensure the integrity and accuracy of information at different levels.

[0078] Specific implementation methods include: obtaining equipment spatial coordinates through the power grid GIS platform API; collecting equipment status such as current, voltage, and power from substation RTU / IED equipment through the OPC UA / Modbus protocol; and obtaining information such as equipment model, commissioning time, maintenance records, and technical parameters through PMS ledger data (production management system data).

[0079] It's important to note that after obtaining the initial data, some preprocessing operations are necessary to ensure the efficient operation of subsequent matching algorithms. Preprocessing steps include data cleaning (removing duplicate, erroneous, or missing data); data normalization (converting data of different magnitudes to the same scale for easier algorithm processing); and data transformation (converting the raw data into a format suitable for algorithm input). Preprocessed data improves its quality and usability, laying a solid foundation for subsequent multi-dimensional matching and dynamic verification processes. During preprocessing, special attention must be paid to protecting data privacy and security, ensuring that sensitive information is not leaked.

[0080] In some specific implementations, preprocessing can be automated by setting a series of rules and algorithms. Specific methods include: using regular expressions to remove invalid characters, using interpolation to fill missing values, and using Z-score normalization to normalize the data. These preprocessing steps can significantly improve data quality, reduce the impact of noise and outliers on matching results, and thus ensure the accuracy and stability of subsequent algorithms. Furthermore, the timeliness and completeness of the data must be considered during preprocessing to ensure that the data used accurately reflects the current state and historical changes of the device, providing strong support for accurate matching.

[0081] In other specific implementations, preprocessing can further improve data quality and usability by introducing machine learning algorithms to automatically identify and correct outliers and missing values. For example, clustering algorithms can be used to identify and group similar data points, making it easier to identify outliers; or decision tree algorithms can be used to predict and fill in missing values, ensuring data integrity. The application of these advanced preprocessing techniques can further enhance the accuracy and reliability of the data, providing a more solid foundation for subsequent multi-dimensional matching and dynamic verification processes.

[0082] In this embodiment of the invention, preprocessing includes data cleaning, unified data standardization, and time-series data processing, specifically:

[0083] Taking the matching of a 500kV transformer GIM model with its ledger as an example. Input data includes...

[0084] 1) GIM model: 3D mesh in .obj format

[0085] 2) PMS ledger: Maintenance records in JSON format

[0086] 3) SCADA: OPC UA real-time data stream, etc.

[0087] The preprocessing module standardizes and unifies the above input data:

[0088] 1) Data cleaning: including missing value handling, outlier detection, and deduplication.

[0089] 2) Unified data standards: Since the data formats of the GIM model and the equipment ledger records are different, it is necessary to convert the GIS coordinate system WGS84 to the CGCS2000 standard and convert the local time of each system in the equipment ledger to UTC+8 timestamp (accurate to milliseconds).

[0090] 3) Time series data processing: Time series data consists of a series of observation results obtained from the observed objects in chronological order. Since there is a gap between the GIM model and the equipment ledger time series data benchmark and sampling frequency, it is required to unify the time benchmark and standardize the sampling frequency, and use dynamic alignment technology to complete the dynamic calibration of time series data.

[0091] By preprocessing the input data, the preprocessing module improves the accuracy of subsequent matching algorithms by more than 40% through standardized pipeline operations, while reducing the need for manual intervention in anomaly handling by 70%.

[0092] It's important to note that acquiring and preprocessing the initial data from the target device significantly improves data quality, laying a solid foundation for subsequent data matching and analysis. Missing value handling ensures data integrity, preventing analytical biases caused by missing data; outlier detection and correction effectively eliminate erroneous data, preventing it from misleading matching results; and deleting duplicate data avoids redundant information dragging down algorithm efficiency. This series of preprocessing steps works together to make the data more accurate and reliable, thereby improving the accuracy and efficiency of the matching algorithm.

[0093] S102, establish a first matching algorithm based on the preprocessed first data of the target device, wherein:

[0094] It should be noted that after obtaining the preprocessed data, it is necessary to analyze the data to obtain the final structure matching result. This can be achieved by designing some matching algorithms or matching logic.

[0095] In this embodiment of the invention, a first matching algorithm is designed, which is used to perform multi-dimensional matching on the preprocessed ledger data and grid information model data;

[0096] In some specific implementations, the first matching algorithm can utilize methods based on multi-dimensional feature extraction and machine learning. By deeply analyzing the preprocessed data, key features reflecting equipment characteristics and status, such as geometry, spatial location, and operating parameters, are extracted. Then, a matching model is constructed using these features, and the model is trained and optimized using machine learning algorithms to accurately identify and match the correspondence between the GIM model and the equipment ledger. This method not only improves matching accuracy but also enhances the algorithm's adaptability and robustness, enabling it to meet matching needs across different equipment and scenarios.

[0097] In other specific implementations, the first matching algorithm can also utilize image recognition and natural language processing technologies to further enhance the intelligence of the matching process. Through image recognition technology, the system can automatically identify and compare the 3D geometric information in the GIM model with images or video data in the equipment register, thereby more accurately locating the equipment and confirming its identity. Simultaneously, natural language processing technology allows the system to parse and understand the textual descriptions in the equipment register, such as equipment name, model, and function, further improving the accuracy and flexibility of the matching. The integrated application of these advanced technologies not only improves matching efficiency but also significantly enhances the automation and intelligence of the system.

[0098] In this embodiment of the invention, the first matching algorithm includes:

[0099] A multi-dimensional set is pre-defined, which includes several dimensions for measuring ledger data and grid information model data; several dimension matching logics are established based on the multi-dimensional set.

[0100] The first matching algorithm is formed by combining several dimensions of matching logic.

[0101] In embodiments of the present invention, multi-dimensional sets include at least one or more of the following:

[0102] Spatial dimension, electrical dimension, structural dimension, and temporal dimension;

[0103] Spatial dimension is used to describe the positional relationship of target devices in physical space, including the relative positions and layout of devices;

[0104] The electrical dimension is used to describe the electrical connection relationships between target devices or the transmission path of electrical parameters;

[0105] The structural dimension is used to describe the composition structure of the target device and the assembly relationships between its components;

[0106] The time sequence dimension is used to describe the time sequence and state changes of the target device during operation.

[0107] Specifically, the first matching algorithm in this invention is a three-dimensional solution strategy for the complex correspondence between GIM models and equipment ledgers. Its core lies in comprehensively determining the matching relationship between entities through multiple complementary observation perspectives. These multiple dimensions are reflected in spatial, electrical, structural, and temporal dimensions. Given the multidimensionality and complexity of power transmission and transformation system data, optimizing data relationships at different levels further complicates the matching process. Therefore, feature matrices can be constructed based on four different dimensions, as shown in Table 2. Different weighting coefficients are used for different dimensions according to actual needs to achieve multi-dimensional matching between GIM models and equipment ledgers.

[0108] Table 2 Construction of the feature matrix

[0109]

[0110] Furthermore, suppose the multidimensional data of the GIM model constitutes a matrix A = x ij The multidimensional data in the equipment ledger constitutes a matrix B=y ij .

[0111] In some specific implementation examples, the L1 norm, also known as Manhattan distance, is proposed to measure the distance between two matrices. It has the characteristics of considering the absolute difference of each dimension, being unaffected by scale, and being suitable for continuous numerical data. It is suitable for comparing the similarity of numerical data, such as geographic coordinates and time series.

[0112] The mathematical formula for Manhattan distance can be expressed as:

[0113]

[0114] In other specific implementation examples, the L2 norm, also known as Euclidean distance, was proposed. It is defined as the square root of the sum of the squares of each dimension of a vector. It is suitable for comparing the similarity of numerical data, such as vectors and time series.

[0115] The mathematical formula for Euclidean distance can be expressed as:

[0116]

[0117] In the multi-dimensional matching of GIM models and equipment ledgers for power transmission and transformation equipment, the two algorithms each have their applicable scenarios:

[0118] The L1 norm is suitable for matching discrete values ​​of equipment status (such as switch opening / closing status 0 / 1) and for rapid initial screening when data noise exists (such as SCADA transient interference). The L2 norm is suitable for precise spatial coordinate matching (such as GIS and GIM model alignment) and continuous parameter similarity calculation (such as transformer oil temperature curves).

[0119] Of course, there are more than just L1 and L2 matrix norms. Choosing the appropriate matching algorithm based on the characteristics of data in different dimensions is beneficial for more accurately describing the degree of matching between the two.

[0120] It should be noted that establishing the first matching algorithm based on the preprocessed target device's initial data significantly improves the accuracy and efficiency of the matching. Preprocessing ensures the integrity and accuracy of the input data, providing a high-quality data foundation for subsequent algorithms. The design of the first matching algorithm, through multi-dimensional feature extraction and machine learning methods, comprehensively considers multiple aspects such as the device's geometry, spatial location, and operating status, thereby achieving a precise solution to the complex correspondence between the GIM model and the device ledger. This method not only improves the accuracy of the matching but also enhances the algorithm's adaptability and robustness, providing strong data support for subsequent device management and maintenance.

[0121] S103, Establish a model training set based on the matching results of the first matching algorithm, and train the second matching algorithm to adjust the model using the model training set, wherein:

[0122] It should be noted that after determining the first matching algorithm, it is necessary to combine the algorithms to achieve the final automatic structure recognition operation. Currently, this is done by designing models, such as deep learning models, support vector machine models, or decision tree models. In this embodiment of the invention, in order to improve the accuracy and generalization ability of the matching, a second matching adjustment model is adopted. This model is further optimized and adjusted based on the matching results of the first matching algorithm.

[0123] In some specific implementations, when the second matching adjustment model is designed using deep learning, the specific steps can be as follows:

[0124] Step 1.1: Collect the matching result data of the first matching algorithm, including samples of correct and incorrect matches, to build the training dataset and validation dataset required for the deep learning model.

[0125] Step 1.2: Preprocess the training dataset, such as normalization and feature selection, to improve the training efficiency and performance of the deep learning model.

[0126] Step 1.3: Select a suitable deep learning network architecture, such as a convolutional neural network (CNN), a recurrent neural network (RNN), or their variants, and customize the design according to the complexity of the problem and the characteristics of the data.

[0127] Step 1.4: Train the deep learning model using the preprocessed training dataset, and continuously adjust the network weights through the backpropagation algorithm until the model reaches the expected accuracy or converges on the validation dataset.

[0128] In some other specific implementations, when the second matching adjustment model is designed using a decision tree model, the specific steps can be as follows:

[0129] Step 2.1: Collect the matching result data of the first matching algorithm. This data also includes samples of correct and incorrect matches, which are used to build the training and test datasets required for the decision tree model.

[0130] Step 2.2: Preprocess the collected data, identify and handle missing values, and fill in the missing information using statistical methods or domain knowledge; at the same time, detect and handle outliers to ensure data consistency and accuracy; finally, remove duplicate data to avoid overfitting during model training.

[0131] Step 2.3: Select a suitable decision tree algorithm, such as ID3, C4.5, or CART, and customize the design based on data characteristics and problem requirements. Construct a decision tree model by recursively selecting the optimal feature to split nodes.

[0132] Step 2.4: Train the decision tree model using the preprocessed training dataset, employing techniques such as pruning to prevent overfitting and improve the model's generalization ability. Finally, evaluate the model's performance on the test dataset to ensure its accuracy and stability in real-world applications.

[0133] In this embodiment of the invention, establishing a model training set based on the matching results of the first matching algorithm includes:

[0134] Establish matching criteria for different dimensions;

[0135] The matching degree is obtained based on the matching degree judgment criteria of the first matching algorithm.

[0136] Matching results that meet the matching degree judgment criteria are taken as positive samples, and matching results that do not meet the matching degree judgment criteria are taken as negative samples.

[0137] Positive and negative samples are combined to form the model training set.

[0138] In this invention, machine learning technology is introduced to automatically identify potential inconsistencies and redundant information by learning historical data features, thereby improving the efficiency and accuracy of matching.

[0139] Step 3.1: Perform data preprocessing, including data cleaning, feature extraction, and normalization, converting the data into a format that DNN can process. This addresses compatibility issues with multi-source heterogeneous data, eliminates noise and outlier interference, constructs standardized feature representations, and ensures data security and compliance.

[0140] Data cleaning refers to removing obvious noise, outliers, and missing values. For GIM model and equipment ledger data, electrical parameter filtering is used, and sliding window midpoint filtering is used to eliminate SCADA transient interference. For missing values, an intelligent imputation method is used, as detailed in Table 3.

[0141] Table 3. Intelligent Data Imputation Method

[0142]

[0143] Feature extraction refers to the process of extracting feature curves or other feature information from time series or other forms of data and using them as input to a DNN model.

[0144] In this invention, feature extraction needs to include multiple dimensions, including spatial features, electrical features, temporal features, and structural features. Normalization is used to ensure that the numerical range of the data is consistent. Data is typically normalized or standardized so that the values ​​of all data points are within the same range (usually 0 to 1). The normalization formula is:

[0145]

[0146] Step 3.2: Constructing the DNN model requires dividing the DNN according to the position of different layers. This invention designs a deep neural network model for power equipment anomaly detection, whose network structure includes an input layer, multiple hidden layers, and an output layer. The number of nodes in the input layer strictly corresponds to the dimensions of the preprocessed features, receiving 28-dimensional standardized input data, including spatial coordinates (6-dimensional), electrical parameters (12-dimensional), and temporal features (10-dimensional). The hidden layers adopt a three-layer progressive design, containing 64, 32, and 16 neurons respectively, using LeakyReLU, Swish, and GELU activation functions sequentially to adapt to the nonlinear characteristics of power equipment data. LeakyReLU retains a small gradient (α = 0.1) in the negative value range to handle potentially negative impedance features. A dynamic Dropout layer is inserted after each hidden layer, with its dropout rate adaptively adjusted according to the standard deviation of the input data (range 0.1-0.5), significantly improving the model's robustness in noisy environments. The output layer adopts a multi-task design, simultaneously outputting anomaly probability (sigmoid activation) and fault type classification (5-class softmax output), and solves the sample imbalance problem through a weighted loss function.

[0147] Step 3.3: For multi-dimensional datasets, they can be divided into training set, validation set and test set according to 70%, 20% and 10% respectively. After the model is built, the model is trained using historical data. The purpose is to learn the mapping relationship between input data and target output, so as to achieve intelligent correction of abnormal data or missing values.

[0148] It should be noted that, for regression tasks, the mean squared error (MSE) is used as the loss function to quantify the deviation between the predicted and actual values. The calculation formula is as follows:

[0149]

[0150] Among them, y i For the true value, Here, n represents the model's predicted value, and n is the number of samples. During training, the dataset is divided into a training set and a validation set: the training set is used to optimize model parameters (such as weights and biases) through backpropagation, while the validation set is used to monitor the model's generalization ability and prevent overfitting. By iteratively optimizing and minimizing the MSE, the model gradually improves its ability to correct for data anomalies, ultimately achieving stable performance on the test set.

[0151] Step 3.4: After model training is complete, it can be deployed to the data correction process. First, anomalies in the current data are identified based on the model's output anomaly probability threshold and its deviation from historical data distribution, and the similarity of their feature curves is analyzed. The detected anomaly data, after standardization preprocessing, is input into the trained DNN model. The model extracts deep features through an encoder-decoder structure and generates correction values ​​based on the learned normal operating mode of the device. These correction values ​​gradually replace the original anomaly data, preserving the original data trend while eliminating abrupt noise. The entire process is logged in real-time through a monitoring system to ensure traceability.

[0152] It should be noted that building a model training set based on the matching results of the first matching algorithm, and then training the second matching adjustment model using this training set, can further refine the matching logic and improve the accuracy and reliability of the matching. Through the initial matching of the first matching algorithm, this invention has obtained a large amount of matching result data, which includes both correctly matched and incorrectly matched samples. Using this data to build a model training set allows the second matching adjustment model to learn more matching features and patterns, thereby better adjusting and optimizing the matching results. This method not only improves matching accuracy but also reduces manual intervention and increases matching efficiency.

[0153] S104, adjust the model's output based on the second matching after training to perform layered structure matching, where:

[0154] In this embodiment of the invention, performing layered structure matching based on the output of the model adjusted according to the second matching after training includes:

[0155] A preset matching verification mechanism is used to verify whether the matching deviation of the first data after hierarchical matching meets the preset matching deviation requirements. Algorithms that can be used include simulated annealing and cluster analysis.

[0156] In some specific implementations, the matching verification mechanism can be finely tuned using the Simulated Annealing algorithm. Simulated Annealing is a probabilistic optimization algorithm inspired by the annealing process in physics. During the matching process, the algorithm initially accepts a relatively poor match, and then accepts an even worse match with a certain probability. This probability decreases as the temperature gradually decreases until it eventually converges to a better match. This method can escape local optima and increase the likelihood of finding the global optimum.

[0157] Specifically, the application of simulated annealing algorithm in layered structure matching can include the following steps:

[0158] Step 1: Initialize the matching results and temperature parameters. Use the matching results of the first matching algorithm as the initial matching results and set a relatively high initial temperature.

[0159] Step 2: Based on the current temperature, randomly select a matching pair for adjustment and calculate the change in matching deviation before and after the adjustment.

[0160] Step 3: If the adjusted matching deviation decreases, or if the adjusted matching deviation increases but meets a certain probability acceptance condition (this probability is related to the current temperature and the amount of change in the matching deviation), then the adjusted matching result is accepted.

[0161] Step 4: Reduce the temperature parameter and repeat steps 2 and 3 until the temperature drops to the preset termination temperature or the matching result converges.

[0162] By fine-tuning the simulated annealing algorithm, matching bias can be further reduced, improving matching accuracy and reliability. Furthermore, because the simulated annealing algorithm has the ability to escape local optima, it can, to some extent, avoid mismatches during the matching process.

[0163] In some specific implementations, the matching verification mechanism can also use cluster analysis, such as K-means and DBSCAN, to group and filter a large number of matching results. The K-means algorithm can divide the matching results into K clusters, where matching results within each cluster have high similarity, while matching results between different clusters show significant differences. This method can quickly identify outliers and irregularities in the matching results, facilitating subsequent manual review. The DBSCAN algorithm is a density-based clustering method that can discover clusters of arbitrary shapes and identify noise points, making it suitable for processing matching result data with complex distribution characteristics. Through cluster analysis, we can perform preliminary classification and filtering of the matching results, providing a foundation for subsequent fine-tuning.

[0164] In summary, this invention proposes a method for matching the layered structure of GIM models and equipment ledgers. It acquires first data for the target equipment and preprocesses this first data, which includes ledger data and grid information model data. A first matching algorithm is established based on the preprocessed first data. A model training set is built based on the matching results of the first matching algorithm, and a second matching adjustment model is trained using this training set. Layered structure matching is then performed based on the output of the trained second matching adjustment model. By employing the aforementioned multi-dimensional set and dynamic verification process, this invention can achieve accurate matching of the complex relationships between GIM models and equipment ledgers. Specifically, the spatial dimension ensures accurate correspondence of equipment in physical space, the electrical dimension captures the electrical connections between equipment, the structural dimension reveals the internal structure and component assembly relationships of the equipment, and the temporal dimension records the time sequence and state changes of the equipment during operation. The comprehensive application of these dimensions greatly improves the accuracy and comprehensiveness of the matching.

[0165] Example 2, in a preferred embodiment, the matching verification mechanism includes several layers of dynamic verification process, including a primary verification process, an intermediate verification process, and a high-level verification process;

[0166] The initial verification process includes intercepting abnormal data based on device physical thresholds;

[0167] The intermediate verification process includes using a sliding window combined with the 3σ principle to detect transient anomalies and using the median value within the window for temporary correction.

[0168] Specifically, in order to promptly correct matching deviations caused by equipment failures or data transmission errors, this patent proposes a data verification module, which can be divided into a three-stage dynamic verification process: primary, intermediate, and advanced stages.

[0169] First, a primary verification is performed to quickly intercept obviously abnormal data based on the device's physical thresholds (such as voltage over-limit ±10%, coordinate offset tolerance ±5 meters, and other hard rules).

[0170] Secondly, intermediate verification is performed using a sliding window (60 seconds) statistical method. Transient anomalies are detected by the 3σ principle and temporary corrections are made using the median within the window.

[0171] Finally, a high-level validated DNN model (inputting 28-dimensional device features) is used to calculate the cosine similarity between real-time data and historical normal patterns. For data with a similarity of less than 0.8, a hybrid correction strategy is initiated.

[0172] The 3σ principle is the core statistical method for real-time data verification. A 60-second sliding window is established for the operating parameters of each device (such as voltage and temperature), and the mean (μ) and standard deviation (σ) of the data within the window are calculated in real time. The formulas for calculating the mean and standard deviation are shown below.

[0173]

[0174] It should be noted that the 3σ principle is a commonly used statistical rule. This principle is based on the characteristics of the normal distribution. In a normal distribution, approximately 99.7% of the data falls within the range of μ ± 3σ. By applying the 3σ principle, analysts can effectively identify and handle outliers in the dataset, thereby improving the accuracy and reliability of data analysis.

[0175] Example 3: This example also provides a layered structure matching system for GIM models and equipment ledgers, comprising four modules:

[0176] Module 1: Data Acquisition and Processing Module, used to acquire the first data from the target device and preprocess the first data from the target device. The first data includes ledger data and grid information model data.

[0177] The first data under the target device includes static data, three-dimensional geometric and spatial data, operational status data, and business-related data;

[0178] Module 2: Algorithm Establishment Module, used to establish the first matching algorithm based on the first data under the preprocessed target device;

[0179] The first matching algorithm is used to perform multi-dimensional matching on the preprocessed ledger data and grid information model data;

[0180] Module 3: Model Building Module, used to build a model training set based on the matching results of the first matching algorithm, and to train the second matching algorithm to adjust the model using the model training set;

[0181] Module 4: Matching Module, used to adjust the model's output based on the second matching after training to perform layered structure matching.

[0182] The above-mentioned unit modules can be embedded in the processor of the electronic device in hardware form or independent of it, or they can be stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0183] Specifically, this embodiment also provides an electronic device, which can be a terminal, and its internal structure diagram can be as follows. Figure 2As shown, the electronic device includes a processor, memory, communication interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a layered structure matching method for GIM models and equipment ledgers. The display screen can be an LCD screen or an e-ink display screen. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the device's casing, or an external keyboard, touchpad, or mouse.

[0184] This embodiment also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it performs the following steps:

[0185] Step 1: Obtain the first data under the target device and preprocess the first data under the target device. The first data includes ledger data and grid information model data.

[0186] The first data under the target device includes static data, three-dimensional geometric and spatial data, operational status data, and business-related data;

[0187] Step 2: Establish a first matching algorithm based on the first data from the preprocessed target device;

[0188] The first matching algorithm is used to perform multi-dimensional matching on the preprocessed ledger data and grid information model data;

[0189] Step 3: Establish a model training set based on the matching results of the first matching algorithm, and train the second matching algorithm to adjust the model using the model training set;

[0190] Step 4: Adjust the model output based on the second matching after training to perform layered structure matching.

[0191] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended 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.

[0192] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented using various computer languages.

[0193] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0194] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0195] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0196] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0197] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for matching the layered structure of a GIM model and an equipment ledger, characterized in that, include: Acquire first data from the target device and preprocess the first data from the target device. The first data includes ledger data and grid information model data. The first data under the target device includes static data, three-dimensional geometric and spatial data, operational status data, and business-related data; A first matching algorithm is established based on the first data under the preprocessed target device. The first matching algorithm is used to perform multi-dimensional matching on the preprocessed ledger data and grid information model data; A model training set is established based on the matching results of the first matching algorithm, and the second matching adjustment model is trained using the model training set. Based on the second matching after training, adjust the model's output to perform layered structure matching.

2. The method for matching the layered structure of a GIM model and an equipment ledger as described in claim 1, characterized in that, The first matching algorithm includes: A multi-dimensional set is preset, which includes several dimensions for measuring ledger data and grid information model data; Establish matching logic for several dimensions based on the aforementioned multi-dimensional set; The matching logic of the aforementioned dimensions is combined to form the first matching algorithm.

3. The method for matching the layered structure of a GIM model and an equipment ledger as described in claim 2, characterized in that, The step of building a model training set based on the matching results of the first matching algorithm includes: Establish matching criteria for different dimensions; The matching degree is obtained based on the matching degree judgment criteria of the first matching algorithm. Matching results that meet the matching degree judgment criteria are taken as positive samples, and matching results that do not meet the matching degree judgment criteria are taken as negative samples. The positive and negative samples are combined to form the model training set.

4. The method for matching the layered structure of a GIM model and an equipment ledger as described in claim 3, characterized in that, The step of adjusting the model's output based on the second matching after training to perform layered structure matching includes: A preset matching verification mechanism is provided, which is used to verify whether the matching deviation of the first data after performing layer structure matching meets the preset matching deviation requirements. The matching and verification mechanism includes several layers of dynamic verification processes.

5. The method for matching the layered structure of a GIM model and an equipment ledger as described in claim 4, characterized in that, The multidimensional set includes at least one or more of the following: Spatial dimension, electrical dimension, structural dimension, and temporal dimension; The spatial dimension is used to describe the positional relationship of the target devices in physical space, including the relative positions and layout between devices; The electrical dimension is used to describe the electrical connection relationship or the transmission path of electrical parameters between target devices; The structural dimension is used to describe the composition structure of the target device and the assembly relationships between its components; The time sequence dimension is used to describe the time sequence and state changes of the target device during operation.

6. The method for matching the layered structure of a GIM model and an equipment ledger as described in claim 5, characterized in that, The aforementioned multi-layered dynamic verification process includes a primary verification process, an intermediate verification process, and a high-level verification process. The initial verification process includes intercepting abnormal data based on device physical thresholds; The intermediate verification process includes using a sliding window combined with the 3σ principle to detect transient anomalies and using the median value within the window for temporary correction.

7. The method for matching the layered structure of a GIM model and an equipment ledger as described in claim 6, characterized in that, The advanced verification process includes calculating the cosine similarity between the real-time data of the target device and the historical normal pattern, and initiating a hybrid correction strategy for data with a similarity lower than 0.

8.

8. A layered structure matching system for GIM models and equipment ledgers, using the method described in any one of claims 1 to 7, characterized in that, include: The data acquisition and processing module is used to acquire first data under the target device and preprocess the first data under the target device. The first data includes ledger data and grid information model data. The first data under the target device includes static data, three-dimensional geometric and spatial data, operational status data, and business-related data; The algorithm establishment module is used to establish a first matching algorithm based on the first data under the preprocessed target device; The first matching algorithm is used to perform multi-dimensional matching on the preprocessed ledger data and grid information model data; The model building module is used to build a model training set based on the matching results of the first matching algorithm, and to train the second matching adjustment model using the model training set; The matching module is used to adjust the model's output based on the second matching after training to perform layered structure matching.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the layered structure matching method for GIM model and equipment ledger as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the layered structure matching method for GIM model and equipment ledger as described in any one of claims 1 to 7.

Images