Integrated analysis method for digital twinning mechanism model of power transmission and transformation equipment

Abstract

The invention discloses an integrated analysis method for a digital twinning mechanism model of power transmission and transformation equipment, which comprises the following steps of: 1, constructing an electrical parameter-equipment state-environment information-image information four-dimensional data acquisition system, and realizing comprehensive perception on the physical state of the equipment and the surrounding environment; step 2, carrying out standardization processing on dimension isomerism, value domain difference, space-time asynchronization and abnormal interference of the multi-source data; 3, integrating the multi-dimensional data features by adopting a hierarchical fusion strategy, and generating a precise fusion feature set and a decision result; 4, constructing an intelligent diagnosis and evaluation rule based on the fused data, and realizing data-driven decision reasoning; the technical problems of incomplete data coverage dimension, low analysis precision, high missed diagnosis rate and misdiagnosis rate, untimely early warning, no collaborative architecture support and the like in the existing power transmission and transformation equipment monitoring and analysis method are solved.

Description

An integrated analysis method for digital twin mechanism models of power transmission and transformation equipment Technical Field

[0001] This invention belongs to the field of power transmission and transformation technology, and in particular relates to an integrated analysis method for a digital twin mechanism model of power transmission and transformation equipment. Background Technology

[0002] As a critical component of the power system, the stable operation of power transmission and transformation equipment is crucial to the reliability and security of the entire power grid. With the development of power technology, massive amounts of multi-source data are available, covering equipment operating parameters, monitoring images, environmental data, etc., making it difficult to deeply extract information value using traditional single-dimensional analysis. Existing methods for analyzing digital twin mechanism models of power transmission and transformation equipment have the following drawbacks:

[0003] 1. Relying solely on a single data source such as electrical parameters without integrating equipment status, environmental factors, and image features results in incomplete data coverage and an inability to provide comprehensive support for equipment status analysis;

[0004] 2. Multi-source data suffers from problems such as heterogeneous dimensions, differences in value ranges, spatiotemporal asynchrony, and abnormal interference. The lack of standardized processing procedures leads to low analytical accuracy and poor reliability of results when directly fused.

[0005] 3. Relying on manual inspection and experience-based judgment, it is highly subjective, with a high rate of missed diagnoses and misdiagnoses. It cannot detect fault signs in advance, has low warning timeliness, and is difficult to avoid the risk of fault escalation.

[0006] 4. Data collection, processing, analysis, and decision-making operate independently without the support of a collaborative architecture, resulting in high reliance on manual labor, low efficiency, and blind spots in inspection coverage. Summary of the Invention

[0007] The technical problem to be solved by this invention is to provide an integrated analysis method for digital twin mechanism models of power transmission and transformation equipment, so as to solve the technical problems in existing power transmission and transformation equipment monitoring and analysis methods, such as incomplete data coverage dimensions, low analysis accuracy, high rate of missed diagnosis and false diagnosis, untimely early warning and lack of collaborative architecture support.

[0008] Technical solution of the present invention:

[0009] An integrated analysis method for a digital twin mechanism model of power transmission and transformation equipment, the method comprising:

[0010] Step 1: Construct a four-dimensional data acquisition system of "electrical parameters - equipment status - environmental information - image information" to achieve comprehensive perception of the physical status of the equipment and its surrounding environment;

[0011] Step 2: Standardize the data from multiple sources to address dimensional heterogeneity, value range differences, spatiotemporal asynchrony, and abnormal interference.

[0012] Step 3: Adopt a hierarchical fusion strategy to integrate multi-dimensional data features and generate a precise fusion feature set and decision results;

[0013] Step 4: Construct intelligent diagnosis and evaluation rules based on the fused data to achieve data-driven decision reasoning.

[0014] The method further includes:

[0015] Step 5: Adopt a layered distributed architecture. The bottom data acquisition layer interfaces with sensors and monitoring equipment; the middle data processing and fusion layer performs preprocessing, fusion, and rough set analysis; the top human-computer interaction layer displays results and receives instructions. Each layer communicates and operates collaboratively according to the interface specifications.

[0016] Electrical parameter acquisition includes collecting voltage, current, power, and frequency data, which are then aggregated to the monitoring center via a communication network to construct a time-series dataset of electrical parameters. , for The measured values ​​of each electrical quantity at each time point; i is the sequence number of the acquisition time, and n is the total number of acquisitions;

[0017] Equipment condition monitoring and data acquisition includes: installing vibration, oil temperature, or partial discharge sensors on key parts of transformers and circuit breakers to monitor their mechanical and insulation conditions; and obtaining vibration amplitude and frequency datasets. , for Vibration characteristic vector at any given time; j is the sequence number of the acquisition time, ranging from 1 to m; partial discharge monitoring obtains discharge quantity and frequency data. ; This represents the partial discharge monitoring sample at the k-th acquisition time; k represents the sequence number of the acquisition time.

[0018] Environmental data collection includes: installing temperature, humidity, light, and wind speed sensors in the substation to collect environmental data. , for The environmental parameter value at time, r represents the number of data samples in the environmental dataset, and s is the sample number;

[0019] Image data acquisition includes: acquiring images of the equipment's appearance and connection points; extracting defects and loose connections in the equipment's appearance through image recognition; and denoting the image dataset as follows. , This is the annotation for the p-th frame image and its corresponding features; p is the sequence number of the image sample, ranging from 1 to q, and q is the total number of image samples.

[0020] Methods for generating accurate fusion feature sets and decision results include:

[0021] Step 3.1: Weighted fusion is applied to the preprocessed and normalized electrical parameters, condition monitoring data, and environmental data; the fusion formula is:

[0022] ;

[0023] For fusion value, For the i-th type of source data value, The data reliability weights are used; N represents the number of categories in all multi-source data, and the fused dataset is obtained by embedding environmental data. ;

[0024] Step 3.2: Extract features from the image dataset, reduce dimensionality using PCA, and the covariance matrix is:

[0025] ;

[0026] C is the covariance matrix, used to measure the correlation between different features; m is the total number of data samples. Let be the feature vector of the i-th sample; This is the feature mean vector of all samples; The matrix is ​​transposed; based on the obtained eigenvalues, a fused feature set is generated using fuzzy logic fusion.

[0027] Step 3.3: For equipment fault diagnosis and condition assessment tasks, use DS evidence theory to fuse multi-source decision results to obtain the final decision result.

[0028] The method of fusing multi-source decision results using DS evidence theory includes: using DS evidence theory, treating numerical fusion results and feature fusion results as two independent evidence sources, setting a basic probability allocation function, and fusing them through evidence combination rules to obtain the final decision fusion result, so as to generate the final decision result.

[0029] Methods for implementing data-driven decision reasoning include:

[0030] Step 4.1: Construct a decision table: Using the fused features as conditional attributes and the device operating status as the decision attribute, construct a decision table. Organize the logical relationships of the data; U is the sample set, C is the condition attribute set, D is the decision attribute set, V is the attribute range, and f is the information function;

[0031] Attribute reduction: Constructing the difference matrix using the difference matrix method Based on this, the core attributes and reduced set can be obtained, or a heuristic algorithm can be used to remove redundant attributes and simplify the decision table to obtain the simplest rule set.

[0032] Rule extraction and reasoning: Extracting rules from the reduced decision table, let the rule be... If the attribute values ​​of the new sample match, the corresponding decision is inferred based on the new data matching rules to infer the state, thereby realizing intelligent diagnosis and evaluation based on rough sets; R is the core decision rule; the core condition attribute retained after reduction corresponds to the specific threshold or value of the condition attribute; d is the decision attribute, which is the type of equipment state to be judged in the end.

[0033] The beneficial effects of this invention are:

[0034] This invention collects four-dimensional core data—electrical parameters, equipment status, environmental information, and image information—and combines a hierarchical fusion strategy with rough set theory to optimize reasoning. This improves the fault diagnosis accuracy of the test group, significantly outperforming the control group which relies on traditional single-data monitoring and manual judgment. It effectively reduces the risk of missed or misdiagnosed faults due to missing data and improves the identification accuracy of five typical faults (winding overheating, circuit breaker failure to operate, etc.).

[0035] This invention reduces data redundancy by 67% through Min-Max normalization calculation, spatiotemporal alignment and PCA dimensionality reduction, and fuzzy logic fusion processing. The effective information ratio of the fused feature set is increased to over 90%, and the state assessment accuracy of the test group reaches 94.1%, which is 31.2% higher than the accuracy of the control group (62.9%). The consistency between the assessment results and the actual operating state of the equipment is significantly improved.

[0036] This invention uses decision-level fusion and rule-based reasoning based on DS evidence theory to capture potential signs of equipment failure in advance. The average early warning time in the test group was 72 hours, while traditional methods do not have an effective early warning function. This provides a sufficient window period for preventive maintenance of equipment, reduces the risk of failure escalation by more than 90%, and shortens the average downtime from the traditional 24 hours to 4 hours, reducing downtime losses by 83.3%.

[0037] This invention achieves full-process automation through a layered distributed architecture, encompassing comprehensive acquisition of multi-source heterogeneous data, multi-source data preprocessing, construction of multi-source information fusion models, rough set theory optimization and reasoning, and model integration and systems engineering. Data processing efficiency is improved by 95.8% compared to traditional manual analysis. While traditional manual analysis of one set of equipment data takes 2 hours, this system automatically processes it in just 5 minutes. Simultaneously, it reduces the frequency of manual inspections; traditionally, inspections are conducted weekly, while this method involves monthly drone inspections combined with fixed monitoring and automatic monitoring. The annual inspection manpower cost for a single 220kV substation is reduced by 80%, and the inspection coverage rate is increased from the traditional 75% to 100%.

[0038] This invention provides an integrated analysis method that covers the entire process of "comprehensive acquisition of multi-source heterogeneous data - multi-source data preprocessing - construction of multi-source information fusion model - rough set theory optimization reasoning - model integration and system engineering implementation", which is adapted to the digital twin's need for digital mapping and intelligent analysis of the physical state of equipment.

[0039] It solves the technical problems in existing power transmission and transformation equipment monitoring and analysis methods, such as incomplete data coverage dimensions, low analysis accuracy, high rate of missed diagnosis and false diagnosis, untimely early warning, and lack of collaborative architecture support. Attached Figure Description

[0040] Figure 1 is a schematic diagram of the process of this invention. Detailed Implementation

[0041] An integrated analysis method for a digital twin mechanism model of power transmission and transformation equipment includes:

[0042] Step 1: Comprehensive Collection of Multi-Source Heterogeneous Data

[0043] Construct a four-dimensional data acquisition system encompassing "electrical parameters, equipment status, environmental information, and image information" to achieve comprehensive perception of the equipment's physical state and surrounding environment.

[0044] Electrical parameter acquisition: Using smart meters, sensors, etc., electrical parameters such as voltage, current, power, and frequency are collected at high frequency and aggregated to the monitoring center via a communication network to construct a time series dataset of electrical parameters. , for The measured values ​​of each electrical quantity at each time point; i is the sequence number of the acquisition time, and n is the total number of acquisitions, meaning that electrical parameter data were acquired at a total of n times.

[0045] Equipment condition monitoring: Vibration, oil temperature, and partial discharge sensors are installed in critical parts of equipment such as transformers and circuit breakers to monitor their mechanical and insulation conditions. For example, vibration monitoring of transformers yields vibration amplitude and frequency datasets. , for Vibration feature vector at each time point; j is the sequence number of the acquisition time, ranging from 1 to m, indicating that monitoring data were collected at a total of m time points.

[0046] Partial discharge monitoring obtains discharge quantity and frequency data. ; This represents the partial discharge monitoring sample at the k-th acquisition time; k represents the sequence number of the acquisition time, ranging from 1 to K, where K is the total number of partial discharge data acquisitions.

[0047] Environmental data collection: Sensors for temperature, humidity, light intensity, and wind speed are installed in the substation to collect environmental data. , for The environmental parameter value at any time indicates that environmental factors affect the device's heat dissipation, insulation, and other performance characteristics; r represents the number of data samples in this environmental dataset, and s is the sample number, ranging from 1 to r.

[0048] Image data acquisition: Images of the equipment's appearance and connections are periodically collected using drones and surveillance cameras. Image recognition technology is then used to extract features such as defects in the equipment's appearance and loose connections. The image dataset is denoted as... , This section defines the p-th frame image and its corresponding feature annotations. p is the sequence number of the image sample, ranging from 1 to q, where q is the total number of image samples.

[0049] Step 2: Multi-source data preprocessing

[0050] Standardization is implemented to address issues such as heterogeneous dimensions, differences in value ranges, spatiotemporal asynchrony, and abnormal interference in multi-source data.

[0051] Abnormal data processing: Identify, correct, or remove abnormal data according to set thresholds and rules. For electrical parameters, if the current suddenly changes beyond the normal range, it is corrected by combining adjacent data; in image data, frames that are blurred or severely occluded are discarded after image quality assessment to ensure data quality and improve fusion accuracy;

[0052] Data normalization: Normalization calculations are performed using Min-Max. , Here, x is the normalized value, and x is the original value. as well as For the maximum and minimum values ​​of the data; or the Z-Score normalization formula. Standardizing the units of measurement facilitates integrated calculations and resolves the issues of heterogeneous units and differing value ranges in multi-source data. The mean of all samples from the multi-source data. The standard deviation of all samples in the multi-source data;

[0053] Spatiotemporal alignment processing: Based on the timestamp difference method, time synchronization of time series data is achieved; spatial location is completed through image and geographic information registration to ensure the spatiotemporal consistency of multi-source data.

[0054] Step 3: Construction of Multi-Source Information Fusion Model

[0055] A hierarchical fusion strategy is adopted to integrate multi-dimensional data features and generate accurate fusion feature sets and decision results:

[0056] First-level fusion (numerical data fusion): The weighted fusion method is used to calculate the preprocessed and normalized electrical parameters, condition monitoring data, and environmental data. The formula is as follows:

[0057]

[0058] in For fusion value, For the i-th type of source data value, is the data reliability weight. N represents the number of categories of all multi-source data; the environmental data is embedded as a supplement in the present invention to obtain a fused dataset , mining the underlying associations of multi-sources to provide a basis for upper-layer analysis; the environmental data is specifically embedded in the present invention, and the model can combine background information such as temperature, humidity, and air pressure to distinguish between "equipment itself failures" and "temporary anomalies induced by the environment", avoiding misjudgments and unnecessary power outage repairs; based on the联动 analysis of environmental data and equipment operation data, early warning of failures and dynamic adjustment of operation and maintenance strategies can be achieved. For example, if it is predicted that the temperature will continue to be high in the next few days, the transformer load can be reduced in advance and the cooling device can be started to prevent overheating failures; in coastal areas with high salt fog, the insulator cleaning cycle can be optimized according to humidity and salt fog concentration data to reduce the risk of flashover due to pollution.

[0059] Second-level fusion (feature-based data fusion): After extracting features from the image dataset, PCA is used for dimensionality reduction, and the covariance matrix is as follows:

[0060]

[0061] where C is the covariance matrix, used to measure the correlation between different features (such as the correlation degree between the two features of "current" and "power");

[0062] m is the total number of data samples (for example, if data at 1000 moments is collected, m = 1000);

[0063] is the feature vector of the i-th sample (for example, a value containing multiple dimensions such as current, vibration amplitude, temperature, and image rust area);

[0064] is the mean feature vector of all samples, used to eliminate the data offset;

[0065] is the matrix transpose, which is a mathematical operation for covariance calculation to ensure the dimensional matching of matrix operations.

[0066] According to the obtained eigenvalues, and then based on fuzzy logic fusion (the triangular fuzzy membership function depicts the feature membership relationship, and the fusion rule is determined according to expert experience and data distribution), a fused feature set is generated to condense key features and reduce redundancy; through three steps of preprocessing - weighted fusion - optimized dimensionality reduction in the present invention, an accurate and concise unified dataset is obtained.

[0067] The third level of fusion (decision-level fusion): For equipment fault diagnosis and condition assessment tasks, DS evidence theory is used to fuse multi-source decision results. Using DS evidence theory, numerical fusion results and feature fusion results are treated as two independent evidence sources. A basic probability allocation function is set, and the final decision fusion result is obtained through evidence combination rules to fuse multi-source decision results and generate the final decision result.

[0068] Set up the recognition framework Θ is the general term for the recognition frame. For specific elements within the framework, namely individual device status or fault type, each source of evidence Assigning basic probability values , integration rules ,

[0069] This indicates that the intersection of the judgment results from multiple sources of evidence is exactly one. ;

[0070] To perform a product operation on the probabilities of evidence sources that satisfy the intersection condition;

[0071] This is a normalization factor to eliminate the impact of conflicting evidence. K is the conflict factor, which measures the conflict of evidence. Decisions are made based on the fusion results, integrating multiple factors to improve accuracy.

[0072] This invention assigns weights to each type of data based on the reliability of the data source and its impact on the device status. The preprocessed multi-source data is integrated using a weighted fusion formula to obtain preliminary fused data. The covariance matrix of the fused dataset is calculated, and the high-dimensional fused data is mapped to the principal component space. Redundant features are removed, and after dimensionality reduction and fuzzy fusion, a simplified and accurate fused dataset is obtained.

[0073] Step 4: Rough Set Theory Optimization Reasoning

[0074] Based on the fused data, intelligent diagnostic and evaluation rules are constructed to achieve data-driven decision reasoning:

[0075] Constructing a decision table: Using the fused features as conditional attributes and the equipment operating status (normal, fault, and fault type) as decision attributes, construct a decision table. Organize the logical relationships of the data; U is the sample set, C is the condition attribute set, D is the decision attribute set, V is the attribute range, and f is the information function.

[0076] Attribute reduction: Constructing the difference matrix using the difference matrix method. Based on this, the core attributes and reduced set can be obtained, or a heuristic algorithm can be used to remove redundant attributes, simplify the decision table, and obtain the simplest rule set, such as... , To simplify subsets, aiding in rapid decision-making and reducing computational load.

[0077] This represents the element in the i-th row and j-th column of the difference matrix;

[0078] C represents the set of conditional attributes, which correspond to the features in the fused data of power transmission and transformation equipment, such as voltage, current, vibration amplitude, partial discharge, etc.

[0079] 'a' represents a specific conditional attribute in C, such as the feature "vibration amplitude".

[0080] U is the universe of discourse, which is the sample set of power transmission and transformation equipment, such as the operating data sample of 100 transformers;

[0081] |U| represents the total number of samples in the universe of discourse U;

[0082] and Let be the i-th and j-th samples in the domain U, such as the complete monitoring data of the 1st and 2nd transformers.

[0083] Rule extraction and reasoning: Extract rules from the reduced decision table, in the form "IF (condition) THEN (decision)", and infer the state based on the new data and matching the rules. Let the rule be... If the attribute values ​​of new samples match, a corresponding decision is made, thus realizing intelligent diagnosis and evaluation based on rough sets.

[0084] R represents the core decision rule; the core conditional attributes retained after reduction correspond to specific thresholds or values ​​of the conditional attributes.

[0085] d represents the decision attribute, which is the type of equipment status to be determined, such as "equipment operating status" or "fault type".

[0086] Step 5: Model Integration and System Engineering Implementation

[0087] System architecture design: A layered distributed architecture is adopted. The bottom data acquisition layer interfaces with sensors and monitoring equipment; the middle data processing and fusion layer performs preprocessing, fusion, and rough set analysis; the top human-computer interaction layer displays results and receives instructions. Each layer communicates and operates collaboratively according to the interface specifications.

[0088] System engineering implementation: Implement algorithms using Python combined with libraries such as NumPy and Scikit-learn; store data in the database; develop interactive interfaces using a web framework; optimize code structure and algorithm efficiency; ensure stable and reliable system operation through testing and debugging; and embed visualization modules to intuitively display device status and analysis results.

[0089] Performance testing and evaluation: Multiple substation equipment were selected and a test group was set up. The test group used this method and a control group. The control group used the traditional manual inspection + single data monitoring method to collect long-term operation data. The evaluation was based on manual inspection records and fault statistics. The performance of the method was comprehensively evaluated through quantitative indicators such as fault diagnosis accuracy, status assessment accuracy, and early warning time.

[0090] The present invention will be further described below:

[0091] An integrated analysis method for a digital twin mechanism model of power transmission and transformation equipment includes:

[0092] Step 1: Comprehensive Collection of Multi-Source Heterogeneous Data

[0093] Electrical parameter acquisition: High-frequency smart meters are installed on the high-voltage and low-voltage sides of the 220kV transformer in the substation to collect voltage, current, power, and frequency parameters and transmit them to the monitoring center;

[0094] Equipment condition monitoring: Vibration sensors are deployed on the top of the transformer tank, oil temperature sensors (measurement range -40℃~125℃) are deployed on the windings, and partial discharge sensors are deployed in the circuit breaker arc-extinguishing chamber, with a sampling frequency of 5Hz;

[0095] Environmental information collection: Three sets of temperature and humidity sensors, light sensors, and wind speed sensors are evenly deployed in the outdoor area of ​​the substation, with a sampling frequency of 1Hz;

[0096] Image data acquisition: Drones are used to inspect the substation equipment once a week, taking images of the equipment appearance, wiring terminals, and other parts; fixed monitoring cameras are installed at key connection points of the equipment, and image data is collected once a day.

[0097] Step 2: Multi-source data preprocessing

[0098] Abnormal data processing: The threshold for current mutation is set to ±30% of the normal range. When a current mutation exceeding the threshold is detected, the moving average method of 5 adjacent data points is used for correction. Image data is evaluated using sharpness and occlusion rate. Frames with sharpness below 0.6 and occlusion rate above 30% are discarded.

[0099] Data normalization: The Min-Max normalization formula is used to map all numerical data to the [0,1] interval;

[0100] Spatiotemporal alignment processing: Based on the timestamps of electrical parameters, the time deviation of vibration and oil temperature data is corrected by timestamp interpolation; by matching the geographic information collected by GPS with the pixel coordinates of the image, the spatial registration of the image data and the equipment components is achieved, with a positioning accuracy of ≤10cm.

[0101] Step 3: Construction of Multi-Source Information Fusion Model

[0102] Numerical data weighted fusion: Based on experience, weights are set for electrical parameters, condition monitoring data, and environmental data (voltage is set to 0.3, current to 0.25, vibration to 0.15, oil temperature to 0.15, temperature and humidity to 0.1, and wind speed to 0.05). The weighted fusion method is used to calculate and fuse the data to obtain the numerical fusion result.

[0103] Feature-based data fusion: PCA is used to reduce the image data and retain the principal components with a cumulative contribution rate of ≥90%. A triangular fuzzy membership function is constructed to characterize the membership relationship of defect features. The fusion rules are determined based on expert experience and data distribution to obtain the feature fusion results.

[0104] Decision-level fusion: Using DS evidence theory, numerical fusion results and feature fusion results are treated as two independent evidence sources. A basic probability allocation function is set, and the final decision fusion result is obtained through evidence combination rules.

[0105] Step 4: Rough Set Theory Optimization Reasoning

[0106] Decision table construction: 1000 sets of equipment operation samples were selected, including 600 normal samples and 400 fault samples. The condition attributes are 12 fused feature parameters, and the decision attributes are 6 states such as "normal", "winding overheating" and "circuit breaker failure to operate".

[0107] Attribute reduction: The importance of conditional attributes is calculated using the difference matrix method, and three redundant conditional attributes are removed to obtain the simplest decision table with nine conditional attributes.

[0108] Rule extraction and reasoning: 85 reasoning rules are extracted from the simplest decision table, such as "IF (oil temperature > 85℃ and partial discharge > 100pC), THEN (winding overheating fault)". After new data is acquired, the rules are matched by a fuzzy matching algorithm to output the equipment status decision.

[0109] Step 5: Model Integration and System Engineering Implementation

[0110] System architecture design and system engineering implementation: Python 3.9 is used, combined with NumPy 1.24 and Scikit-learn 1.2 to implement the fusion algorithm and rough set inference, MySQL 8.0 is used to store data, a web interactive interface is developed based on the Django framework, and ECharts is embedded to visualize the device status;

[0111] Performance testing and evaluation: Two 220kV substations were selected as test objects. Substation A was the test group (using this method), and substation B was the control group (traditional manual inspection + single electrical parameter monitoring). The test period was one year. The fault diagnosis accuracy of the test group reached 92.3%, which was 28.5% higher than that of the control group. The condition assessment accuracy reached 94.1%, which was 31.2% higher than that of the control group. The average early warning time was 72 hours. The control group had no effective early warning function, which verified the superiority of this method.

Claims

1. An integrated analysis method for a digital twin mechanism model of power transmission and transformation equipment, characterized in that: The method includes: Step 1, constructing a four-dimensional data acquisition system of "electrical parameters - equipment status - environmental information - image information" to achieve comprehensive perception of the physical status of the equipment and its surrounding environment; Step 2, standardizing the data from multiple sources to address dimensional heterogeneity, value range differences, spatiotemporal asynchrony, and abnormal interference; Step 3, adopting a hierarchical fusion strategy to integrate multi-dimensional data features and generate a precise fusion feature set and decision results; Step 4, constructing intelligent diagnosis and evaluation rules based on the fused data to achieve data-driven decision reasoning.

2. The integrated analysis method for a digital twin mechanism model of power transmission and transformation equipment according to claim 1, characterized in that: The method further includes: Step 5, adopting a layered distributed architecture, with the bottom data acquisition layer connecting to sensors and monitoring equipment; the intermediate data processing and fusion layer performing preprocessing, fusion and rough set analysis; and the top human-computer interaction layer displaying results and receiving instructions. Each layer communicates and operates collaboratively according to the interface specifications.

3. The integrated analysis method for a digital twin mechanism model of power transmission and transformation equipment according to claim 1, characterized in that: Electrical parameter acquisition includes collecting voltage, current, power, and frequency data, which are then aggregated to the monitoring center via a communication network to construct a time-series dataset of electrical parameters. ， for The measured values ​​of various electrical quantities at each time point; i is the sequence number of the acquisition time, and n is the total number of acquisitions; equipment condition monitoring acquisition includes: installing vibration, oil temperature, or partial discharge sensors on key parts of transformers and circuit breakers to monitor mechanical and insulation conditions; obtaining vibration amplitude and frequency datasets. ， for Vibration characteristic vector at any given time; j is the sequence number of the acquisition time, ranging from 1 to m; partial discharge monitoring obtains discharge quantity and frequency data. ； This represents the partial discharge monitoring sample at the k-th acquisition time; k represents the sequence number of the acquisition time; environmental information acquisition includes: deploying temperature, humidity, light, and wind speed sensors in the substation to collect environmental data. ， for The environmental parameter value at any given time, r represents the number of data samples in the environmental dataset, and s is the sample index; image data acquisition includes: acquiring images of the equipment's appearance and connection points, extracting defects and loose connections in the equipment's appearance through image recognition, and the image dataset is denoted as... ， This is the annotation for the p-th frame image and its corresponding features; p is the sequence number of the image sample, ranging from 1 to q, and q is the total number of image samples.

4. The integrated analysis method for a digital twin mechanism model of power transmission and transformation equipment according to claim 1, characterized in that: The method for generating accurate fused feature sets and decision results includes: Step 3.1, applying weighted fusion to preprocessed and normalized electrical parameters, condition monitoring data, and environmental data; the fusion formula is: ； For fusion value, For the i-th type of source data value, The data reliability weights are used; N represents the number of categories in all multi-source data, and the fused dataset is obtained by embedding environmental data. Step 3.2: Extract features from the image dataset, reduce dimensionality using PCA, and the covariance matrix is: C is the covariance matrix, used to measure the correlation between different features; m is the total number of data samples. Let be the feature vector of the i-th sample; This is the feature mean vector of all samples; The matrix is ​​transposed; based on the obtained eigenvalues, a fusion feature set is generated based on fuzzy logic fusion; step 3.3: for equipment fault diagnosis and condition assessment tasks, the multi-source decision results are fused using DS evidence theory to obtain the final decision result.

5. The integrated analysis method for a digital twin mechanism model of power transmission and transformation equipment according to claim 1, characterized in that: The method of fusing multi-source decision results using DS evidence theory includes: using DS evidence theory, treating numerical fusion results and feature fusion results as two independent evidence sources, setting a basic probability allocation function, and fusing them through evidence combination rules to obtain the final decision fusion result, so as to generate the final decision result.

6. The integrated analysis method for a digital twin mechanism model of power transmission and transformation equipment according to claim 1, characterized in that: The method for implementing data-driven decision reasoning includes: Step 4.1, constructing a decision table: using the fused features as conditional attributes and the device operating status as the decision attribute, construct a decision table. First, clarify the logical relationships between the data; U is the sample set, C is the condition attribute set, D is the decision attribute set, V is the attribute range, and f is the information function; attribute reduction: construct the difference matrix using the difference matrix method. Based on this, the core attributes and reduced set are obtained, or redundant attributes are removed using heuristic algorithms to simplify the decision table and obtain the simplest rule set; Rule extraction and reasoning: Rules are extracted from the reduced decision table, let the rule be... If the attribute values ​​of the new sample match, the corresponding decision is inferred based on the new data matching rules to infer the state, thereby realizing intelligent diagnosis and evaluation based on rough sets; R is the core decision rule; the core condition attribute retained after reduction corresponds to the specific threshold or value of the condition attribute; d is the decision attribute, which is the type of equipment state to be judged in the end.

Images