A digital intelligent integrated platform terminal display operation and maintenance decision-making method and system

Abstract

The present invention discloses a digital intelligent integrated platform terminal display operation and maintenance decision-making method and system, belonging to the field of GIS equipment partial discharge data monitoring and processing. The method includes: S100, real-time collection of partial discharge signals, environmental data, and equipment operating status data; S200, identifying the discharge type using a discharge type recognition model, and generating an environmental or equipment operating status correlation analysis through gradient backpropagation and physical rule matching; S300, generating a feature contribution report based on SHAP+LIME, verifying the model decision based on a physical rule library, and dynamically adjusting the confidence level; S400, matching operation and maintenance strategies based on a knowledge graph; S500, displaying a 3D spectrum graph, heat map, and natural language report of the operation and maintenance strategy on a visual interface; and S600, updating model parameters through incremental learning. This invention can lower the threshold for operation and maintenance.

Description

Technical Field

[0001] The present invention relates to the field of GIS partial discharge monitoring data processing, and in particular to a digital intelligent integrated platform terminal display operation and maintenance decision-making method and system. Background Art

[0002] Partial discharge (PD) is a pulsed discharge that produces a series of physical phenomena and chemical changes, including light, sound, electrical, and mechanical vibrations, within and around electrical equipment. These physical and chemical changes can provide detection signals for monitoring the insulation condition within the equipment. When insulation defects occur within high-voltage electrical equipment, they are accompanied by the generation of PD signals. By detecting and analyzing PD signals, it is possible to determine whether insulation defects exist within the equipment and prevent potential accidents from escalating.

[0003] Gas-insulated switchgear (GIS) are widely used in power systems. Their complex internal structure makes fault repair difficult and can easily lead to significant losses. GIS equipment failures can be caused by a variety of factors, primarily insulation defects accompanied by partial discharge (PD). Analyzing the severity of defects based on PD signals, predicting their development trends, providing effective early warnings, and formulating appropriate maintenance strategies are crucial for ensuring safe and reliable equipment operation.

[0004] Existing partial discharge monitoring systems have the following problems: classification models are poorly interpretable, making it difficult for operations and maintenance personnel to trust automated decisions; maintenance strategies rely on static rule bases and cannot dynamically adapt to complex working conditions; and there is a lack of a model optimization mechanism driven by user feedback. Summary of the Invention

[0005] This invention aims to address at least one of the technical problems existing in the prior art. To this end, it proposes a digital intelligent integrated platform terminal display operation and maintenance decision-making method and system. This method lowers technical barriers, accelerates fault location and response, exposes potential model flaws, promotes algorithm and data iteration, and establishes a closed "monitoring-interpretation-decision-making" loop to improve power grid reliability.

[0006] On the one hand, an embodiment of the present invention provides a digital intelligent integrated platform terminal display operation and maintenance decision-making method, comprising the following steps: S100, collecting partial discharge signals in real time through ultra-high frequency sensors and ultrasonic sensors, and obtaining environmental data and equipment operation status data; the environmental data include temperature and humidity; the equipment operation status data include load rate, operating time and voltage; S200, inputting the partial discharge signal into the CNN+Transfomer network to extract time-frequency domain features, inputting the environmental data and equipment operation status data into the fully connected network encoding auxiliary features, and dynamically weighting the time-frequency domain features and auxiliary features through the attention mechanism to obtain a fused feature vector; outputting the discharge type probability distribution based on the fused feature vector, and generating the environment or equipment operation status data based on the auxiliary feature vector and the discharge type classification result through gradient back propagation and physical rule matching. State association analysis; S300, generate feature contribution report based on SHAP+LIME, obtain key features according to the feature contribution report, match physical rules of the physical rule base based on key features, verify model decisions according to physical rules, and determine dynamic adjustment of confidence or output of environmental risk prompts based on the model decision verification results; S400, obtain discharge type, discharge type confidence, environmental data and equipment operation status data, and match operation and maintenance strategies based on knowledge graphs; S500, generate 3D spectrum graphs and heat maps based on partial discharge signals, generate natural language reports based on operation and maintenance strategies, and display 3D spectrum graphs, heat maps and natural language reports on a visual interface; S600, record user correction decisions and operation and maintenance strategy execution results, update model parameters through incremental learning, dynamically adjust knowledge graph node weights, and optimize physical rule base logic.

[0007] The embodiments of the present invention achieve at least the following beneficial effects: The embodiments of the present invention improve adaptability to complex working conditions by adaptively adjusting the weights of environmental / status data; explainability-driven strategy generation: quantification of environmental contributions and linkage with the knowledge graph ensure transparent and reliable decision-making; closed-loop feedback optimization: user ratings drive synchronous updates of the knowledge graph and model parameters. SHAP+LIME is deeply integrated with domain rules, achieving a transition from "black box prediction" to "white box decision-making"; dynamic updates and multimodal interaction of the knowledge graph improve the speed and accuracy of operation and maintenance response. It lowers the technical threshold and accelerates fault location and response; exposes potential defects in the model and promotes algorithm and data iteration; and builds a "monitoring-interpretation-decision-making" closed loop to improve power grid reliability.

[0008] According to some embodiments of the present invention, step S200 includes: S210, preprocessing the collected partial discharge signal, including filtering and noise reduction; and dynamically adjusting the signal gain according to temperature data.

[0009] According to some embodiments of the present invention, step S200 includes: S220, extracting the spectral spatial features of the partial discharge signal based on CNN, capturing the time domain sequence dependency based on Transformer, and obtaining the time-frequency domain features of the partial discharge signal; S230, encoding the environmental features and the equipment operation status features based on the fully connected network to obtain auxiliary features; S240, generating the partial discharge signal weight and the environmental / state weight based on the fully connected layer, weightedly fusing the time-frequency domain features and the auxiliary features of the partial discharge signal to obtain a fused feature vector; and outputting the discharge type probability distribution based on the fused feature vector.

[0010] According to some embodiments of the present invention, step S200 includes: S250, calculating the gradient of environmental characteristics or equipment operating status characteristics on the classification results through back propagation, and quantifying the influence weight of environmental characteristics or equipment operating status characteristics; S260, comparing the quantified results with preset domain knowledge, verifying the rationality of the discharge type identification model, and obtaining the environment or equipment operating status association analysis results.

[0011] According to some embodiments of the present invention, step S300 includes: S310, performing feature contribution analysis on the discharge type classification results, obtaining a SHAP global interpretation by calculating the Shapley value of each feature of the partial discharge signal, and obtaining a LIME local interpretation by generating a single sample heat map and identifying key discriminant features.

[0012] According to some embodiments of the present invention, step S300 includes: S320, constructing a physical rule library; S330, when environmental data or equipment operating status data exceeds a preset threshold, triggering physical rule verification; determining whether environmental anomalies or equipment anomalies may cause feature distortion based on the environmental data or equipment operating status data; if so, adjusting the confidence level and generating an environmental / equipment anomaly alarm; S340, when the confidence level of the discharge type classification result is lower than the set threshold, triggering physical rule verification; obtaining key features based on the feature contribution of the discharge type classification result, matching the physical rules in the physical rule library based on the key features, and determining whether the partial discharge features on which the discharge type identification model depends comply with the physical rules; when the discharge type identification model depends on conflicting features, triggering a rule conflict alarm.

[0013] According to some embodiments of the present invention, the S400 includes: S410, constructing a knowledge graph; S420, matching a preset operation and maintenance strategy according to an input discharge type; S430, dynamically adjusting the operation and maintenance strategy according to the equipment operation status data; S440, calculating the risk level according to the confidence level, environmental data and equipment operation status data to obtain the operation and maintenance priority; S450, outputting an operation and maintenance text report according to the operation and maintenance strategy and operation and maintenance priority.

[0014] According to some embodiments of the present invention, S600 includes: S610, using newly labeled data to fine-tune the classification model and optimize the attention weight distribution; S620, updating the knowledge graph node weights based on the operation and maintenance personnel's scoring of the operation and maintenance strategy.

[0015] According to some embodiments of the present invention, step S500 includes: S510, generating AR positioning guidance according to the operation and maintenance strategy.

[0016] Another embodiment of the present invention provides a digital intelligent integrated platform terminal display operation and maintenance decision system, including: a data acquisition module, which is used to collect partial discharge signals in real time through ultra-high frequency sensors and ultrasonic sensors, and obtain environmental data and equipment operation status data; the environmental data includes temperature and humidity; the equipment operation status data includes load rate, operating time and voltage; a partial discharge identification module, which is used to input partial discharge signals into a CNN+Transfomer network to extract time-frequency domain features, input environmental data and equipment operation status data into a fully connected network to encode auxiliary features, and dynamically weight the time-frequency domain features and auxiliary features through an attention mechanism to obtain a fused feature vector; output a discharge type probability distribution based on the fused feature vector, and generate an environment or equipment operation status correlation analysis based on the auxiliary feature vector and the discharge type classification result through gradient back propagation and physical rule matching. Analysis; an interpretability module, which is used to generate a feature contribution report based on SHAP+LIME, obtain key features based on the feature contribution report, match the physical rules of the physical rule base based on the key features, verify the model decision according to the physical rules, and determine the dynamic adjustment of confidence or output of environmental risk prompts based on the model decision verification results; an operation and maintenance strategy module, which is used to obtain the discharge type, discharge type confidence, environmental data and equipment operating status data, and match the operation and maintenance strategy based on the knowledge graph; a visualization module, which is used to generate 3D spectrum graphs and heat maps based on partial discharge signals, generate natural language reports based on the operation and maintenance strategy, and display the 3D spectrum graphs, heat maps and natural language reports on the visualization interface; a feedback optimization module, which is used to record user correction decisions and operation and maintenance strategy execution results, update model parameters through incremental learning, dynamically adjust the knowledge graph node weights, and optimize the physical rule base logic.

[0017] The embodiments of the present invention achieve at least the following beneficial effects: The embodiments of the present invention improve adaptability to complex working conditions by adaptively adjusting the weights of environmental / status data; explainability-driven strategy generation: quantification of environmental contributions and linkage with the knowledge graph ensure transparent and reliable decision-making; closed-loop feedback optimization: user ratings drive synchronous updates of the knowledge graph and model parameters. SHAP+LIME is deeply integrated with domain rules, achieving a transition from "black box prediction" to "white box decision-making"; dynamic updates and multimodal interaction of the knowledge graph improve the speed and accuracy of operation and maintenance response. It lowers the technical threshold and accelerates fault location and response; exposes potential defects in the model and promotes algorithm and data iteration; and builds a "monitoring-interpretation-decision-making" closed loop to improve power grid reliability.

[0018] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments with reference to the accompanying drawings, in which:

[0020] Figure 1 Schematic diagram of a method according to an embodiment of the present invention;

[0021] Figure 2 Schematic diagram of the structure of the system according to an embodiment of the present invention.

[0022] Reference numerals:

[0023] Data acquisition module 100, partial discharge identification module 200, explainability module 300, operation and maintenance strategy module 400, visualization module 500, feedback optimization module 600. DETAILED DESCRIPTION

[0024] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0025] In the description of the present invention, "several" means one or more, "more" means two or more, "greater than," "less than," and "exceed" are understood to exclude the number itself, while "above," "below," and "within" are understood to include the number itself. The use of "first" and "second" in the description is solely for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance, implicitly specifying the number of the indicated technical features, or implicitly specifying the order of the indicated technical features.

[0026] Under normal circumstances, the insulation performance of electrical equipment operating in the power grid can withstand the operating voltage. However, due to the influence of the manufacturing and installation process, bubbles, impurities, cracks, etc. may exist inside the insulation of electrical equipment. Under the action of high-voltage alternating electric fields, the internal or surface electric field strength will be too high, and periodic partial discharge will occur within the insulation. Since the discharge amount is very small, it will not cause the entire circuit to break down in a short time, but it can cause the insulation performance to degrade and even lose the withstand voltage performance. As the number of high-voltage power equipment increases and the voltage level increases, these problems accumulate and seriously threaten the safe operation of the power system. Partial discharge detection is an effective means of verifying the insulation condition of equipment and is also a hot topic in current application. Common methods in partial discharge detection include ultra-high frequency detection, ultrasound, high-frequency current and ground wave detection equipment. These devices have solved a large number of potential accidents for maintenance personnel.

[0027] Partial discharge signals appear as single or continuous electrical pulses. The pulse waveform not only contains information about the discharge mechanism of the defect, but also the severity of the defect. Therefore, one of the main tasks of partial discharge detection is to identify the defect type through the partial discharge pulse waveform, thereby diagnosing the severity of the defect and formulating reasonable equipment maintenance measures and strategies.

[0028] Ultra High Frequency (UHF) sensors operate in the 300MHz to 3GHz frequency range. They feature high frequency, wide bandwidth, and strong penetration, making them widely used in communications, industrial monitoring, healthcare, the Internet of Things, and other fields. Partial discharge (PD) generates broadband electromagnetic waves. Different types of PD experience different electrical breakdown processes, generating pulse currents of varying amplitudes and steepness, and consequently, electromagnetic transients and waves with varying frequency components. The sensor utilizes a Hilbert fractal antenna, a non-frequency-variable antenna whose electrical performance is independent of frequency. It offers advantages such as wide frequency bandwidth, circular polarization, compact size, high efficiency, and flush-mountability. The amplifier utilizes a low-noise, high-gain (40dB) UHF signal.

[0029] Ultrasonic partial discharge sensor: Sound waves propagate primarily as longitudinal waves in liquid and gaseous media. Longitudinal waves transmit pressure primarily through collisions between molecules whose vibration directions are parallel to the wave propagation direction. Sound waves propagate primarily as transverse and longitudinal waves in solid media. When propagating as transverse waves, the vibration direction of particles is perpendicular to the wave propagation direction. This requires sufficient attraction between particles so that the vibration of a particle can drive the vibration of neighboring particles. Therefore, transverse waves only occur in highly concentrated liquids or solids. Transverse waves only emerge when longitudinal waves penetrate a gas or liquid medium and reach a metal casing, where they continue to propagate through the metal medium.

[0030] The digital intelligent integrated platform terminal display operation and maintenance decision-making method of the embodiment of the present invention assists technical personnel in making operation and maintenance decisions through explainability-driven decision support in order to lower the technical threshold. Explainability-driven decision support is a technology that assists users in making decisions by revealing the internal decision logic of the machine learning model and converting complex model predictions into human-understandable evidence. Its core lies in: Transparent model decision: Through visualization or numerical methods, the key features, data areas or logical paths that the model relies on during classification are displayed. Causal association analysis: Explain how the model associates input features with output categories, such as which signal features (such as spectral peaks, phase distribution) play a decisive role in identifying "metal tip discharge". Trust building and risk control: Enhance user trust in the model through explainability and help discover potential deviations or errors in the model, such as locating deficiencies in feature extraction when misjudging "air gap discharge" as "free particle discharge."

[0031] Reference Figure 1 The method of the embodiment of the present invention can implement a digital intelligent integrated platform terminal display operation and maintenance decision-making method, including the following steps:

[0032] S100, collect partial discharge signals in real time through ultra-high frequency sensors and ultrasonic sensors, and obtain environmental data and equipment operating status data; environmental data includes temperature and humidity; equipment operating status data includes load rate, operating time and voltage.

[0033] In this embodiment, environmental data includes temperature and humidity. Temperature is used for signal correction. High temperatures can cause sensor drift, necessitating dynamic signal gain adjustment. Discharge type correction is also performed: the probability of air-gap discharge increases in high-temperature environments. Implementation involves incorporating a temperature compensation algorithm (such as a linear regression model) into the signal preprocessing stage. Temperature is then input into the classification model as a feature to calculate discharge type probability.

[0034] This embodiment uses humidity to determine interference suppression and sensor protection. Interference suppression involves distinguishing between environmental interference and actual discharge, as high humidity can easily trigger surface discharge. Sensor protection includes triggering an alert when humidity exceeds a limit. Implementation measures include building a humidity-interference association rule base (e.g., "When humidity exceeds 90%, block low-frequency signals") and dynamically adjusting alarm thresholds during operation and maintenance decision-making.

[0035] In this embodiment, the device operating status includes load factor, operating time, and voltage fluctuation. The load factor is used to assess discharge severity: a minor discharge under high load can escalate into a serious fault. Implementation methods include combining the load factor with the discharge magnitude to calculate a composite risk index (e.g., risk = discharge amount × load factor) and triggering graded alarms during language decision-making.

[0036] In this example, operating time is used for predictive maintenance: Aging equipment can exacerbate insulation defects. By associating the "operating time - failure probability" relationship in the knowledge graph, maintenance cycle recommendations are optimized.

[0037] In this embodiment, voltage fluctuations are used to identify transient interference: sudden voltage changes may trigger false short-term discharge alarms. The implementation method includes marking the voltage change time window in the preprocessing stage and filtering abnormal signals within this period.

[0038] S200: Input the partial discharge signal into a CNN+Transformer network to extract time-frequency domain features. Input environmental data and equipment operating status data into a fully connected network to encode auxiliary features. Dynamically weight the time-frequency domain features and auxiliary features using an attention mechanism to generate a fused feature vector. Output a discharge type probability distribution based on the fused feature vector. Based on the auxiliary feature vector and the discharge type classification results, generate an environmental or equipment operating status correlation analysis using gradient backpropagation and physical rule matching. This embodiment uses the attention mechanism to adaptively adjust the weights of environmental / status features, improving robustness under complex operating conditions.

[0039] The discharge type identification model in this embodiment uses ultra-high frequency / ultrasonic partial discharge signals as primary input features, and environmental data (temperature, humidity) and device status data (load rate, voltage) as auxiliary feature vectors. These two features are weighted and fused using a multimodal fusion network. The multimodal fusion network includes a signal processing branch that extracts time-frequency domain features using a CNN+Transformer model, and an environmental / device status data processing branch that encodes auxiliary features using a fully connected network (FCN). The fusion layer dynamically weights and fuses these two features using an attention mechanism. The model outputs a probability distribution of the discharge type (e.g., "Metal tip discharge: 89%") and an environmental correlation analysis (e.g., "High temperature causes a 5% decrease in confidence level").

[0040] S300. Generate a feature contribution report based on SHAP+LIME, obtain key features based on the feature contribution report, match physical rules of the physical rule library based on the key features, verify the model decision based on the physical rules, and determine whether to dynamically adjust the confidence or output environmental risk warnings based on the model decision verification results.

[0041] S400: Obtain discharge type, discharge type confidence, environmental data, and equipment operating status data, and match operation and maintenance strategies based on the knowledge graph.

[0042] S500: Generate a 3D spectrum graph and a thermal map based on the partial discharge signal, generate a natural language report based on the operation and maintenance strategy, and display the 3D spectrum graph, thermal map and natural language report on a visual interface.

[0043] S600: Record user correction decisions and operation and maintenance strategy execution results, update model parameters through incremental learning, dynamically adjust knowledge graph node weights, and optimize the physical rule base logic.

[0044] In some embodiments, step S200 includes: S210, pre-processing the collected partial discharge signal, including filtering and noise reduction; and dynamically adjusting the signal gain according to temperature data.

[0045] In some embodiments, step S200 includes:

[0046] S220. Extract the spectral spatial features of the partial discharge signal based on CNN, capture the time domain sequence dependency based on Transformer, and obtain the time-frequency domain features of the partial discharge signal.

[0047] S230: Obtain auxiliary features based on the fully connected network coding environment features and the device operation status features.

[0048] S240 , generating a partial discharge signal weight and an environment / state weight based on the fully connected layer, weightedly fusing the time-frequency domain features and auxiliary features of the partial discharge signal to obtain a fused feature vector; and outputting a discharge type probability distribution based on the fused feature vector.

[0049] In some embodiments, step S200 includes:

[0050] S250 , calculating the gradient of the environmental feature or the device operating status feature on the classification result through back propagation, and quantifying the influence weight of the environmental feature or the device operating status feature.

[0051] S260: Compare the quantified results with the preset domain knowledge to verify the rationality of the discharge type identification model and obtain the correlation analysis results of the environment or equipment operation status.

[0052] The working principle of environmental / device status data in this embodiment includes: first, contribution quantification is performed. Backpropagation is used to calculate the gradient of environmental features (such as temperature and humidity) on the classification results, quantifying their influence weights. For example, a temperature gradient value of +0.15 indicates that the probability of being classified as "air-gap discharge" increases by 15% for every 1°C increase in temperature. Attention weights are used to reflect the relative importance of environmental data in the fusion process. Physical rule verification is then performed: the quantified results are compared with domain knowledge (e.g., "When humidity exceeds 90%, the probability of surface discharge increases") to verify the rationality of the model. If the model relies on features that conflict with the rules (e.g., excessively high weights for high frequencies at low temperatures), an alarm is triggered and manual review is recommended. Confidence adjustment: The classification confidence is dynamically adjusted based on the environmental contribution. For example, an original confidence of 92% decreases by 5% due to high-temperature compensation, resulting in an output of "87% (high-temperature impact)." When the environmental risk (e.g., humidity > 95%) exceeds a threshold, a "high risk of environmental interference" alert is added.

[0053] In some embodiments, step S300 includes: S310, performing feature contribution analysis on the discharge type classification results, obtaining a SHAP global explanation by calculating the Shapley value of each feature of the partial discharge signal, and obtaining a LIME local explanation by generating a single sample heat map and identifying key discriminant features.

[0054] The SHAP+LIME collaborative interpretation mechanism of this embodiment includes: feature contribution calculation:

[0055] SHAP global analysis: Based on the Shapley value in game theory, it calculates the marginal contribution of each feature in all possible subsets and outputs the global feature importance (such as the main frequency of the spectrum and the dispersion of the discharge amplitude).

[0056] LIME local interpretation: Generate a perturbation dataset for a single sample, train a local linear model, and extract key features (such as the phase stability index of a certain discharge).

[0057] Visual output: Use SHAP's summary_plot to display global feature contributions, and LIME to generate heat maps or text reports (such as "1200MHz band contribution +35%").

[0058] In some embodiments, step S300 includes:

[0059] S320: Build a physical rule library.

[0060] S330. When the environmental data or the device operating status data exceeds the preset threshold, the physical rule verification is triggered; based on the environmental data or the device operating status data, it is determined whether the environmental anomaly or the device anomaly may cause feature distortion; if so, the confidence level is adjusted and an environmental / device anomaly alarm is generated.

[0061] S340. When the confidence level of the discharge type classification result is lower than the set threshold, physical rule verification is triggered; key features are obtained based on the feature contribution of the discharge type classification result, and physical rules in the physical rule library are matched based on the key features to determine whether the partial discharge features that the discharge type identification model relies on comply with the physical rules; when the discharge type identification model relies on conflicting features, a rule conflict alarm is triggered.

[0062] The physical rule base verification in this embodiment includes: Rule base construction: Domain knowledge (such as "polarity effect index > 0.7 → metal tip discharge") is encoded into logical rules and stored in a structured database. Dynamic verification logic: When SHAP / LIME identifies key features (such as high-frequency band energy), the system automatically matches physical rules (such as "high-frequency band is susceptible to communication interference"), and triggers an alarm if there is a conflict; Example: If the model determines that it is "air gap discharge", but the SHAP value shows dependence on atypical features (such as ambient temperature), it is marked as "manual review required" in combination with the rule base. Feedback optimization: The user's correction results (such as marking the misjudgment case as "surface discharge") are transmitted back to the rule base to update the rule weights or add new constraints.

[0063] The specific implementation steps include: Data input: The partial discharge signal is preprocessed (filtering and noise reduction) and then input into a hybrid model (such as CNN + Transformer). Model prediction: Outputs the discharge type and confidence level (such as "metal tip discharge, 92%"). Explanation generation: SHAP calculates global feature importance, and LIME generates single sample explanations; key features are extracted (such as "uniformity of discharge times"). Rule matching: Features are matched with a physical rule library (such as "uniform discharge times → metal tip"). If there is a conflict in rules (such as feature dependence on the interference frequency band), dynamic threshold adjustment is triggered. Decision output: Generates a report containing explanations and rule verification (such as "High risk: Repair recommended within 24 hours").

[0064] In some embodiments, S400 includes:

[0065] S410. Build a knowledge graph.

[0066] S420: Match a preset operation and maintenance strategy according to the input discharge type.

[0067] S430. Dynamically adjust the operation and maintenance strategy according to the equipment operation status data.

[0068] S440: Calculate the risk level based on the confidence level, environmental data, and equipment operating status data to obtain the operation and maintenance priority. This embodiment can achieve more accurate risk grading by integrating the load rate and discharge amount.

[0069] S450: Output an operation and maintenance text report according to the operation and maintenance strategy and operation and maintenance priority.

[0070] The knowledge graph construction in this embodiment includes entity and relationship definition: Entities are defined: equipment type (GIS cabinet), failure mode (metal discharge), and maintenance action (power outage maintenance). Relationships are established: equipment-fault association (GIS cabinet → metal discharge) and failure-maintenance mapping (metal discharge → replacement of insulation basin). Data sources are integrated, including structured data (historical work orders and equipment parameter tables) and unstructured data (operation and maintenance logs and expert experience documents) (entity relationships are extracted using natural language processing).

[0071] The operation and maintenance strategy matching process of this embodiment includes the following: Fault type input: The model outputs the discharge type (such as "suspended potential body discharge") and confidence level. Graph query: Query the associated maintenance strategies in the knowledge graph (such as "suspended discharge → check flange pouring mouth"); Combine historical data to filter strategies with high success rates (such as "The success rate of using strategy A for similar faults in the past three months has reached 90%"). Priority sorting: Generate a priority list based on strategy effectiveness, implementation cost, and equipment status data; Example: High-risk faults automatically trigger work orders to be dispatched to the operation and maintenance system 91214. Multimodal output: Text report: "Repair is recommended within 7 days, priority: high"; AR visualization: Overlay fault location guidance on the GIS cabinet entity.

[0072] In some embodiments, S600 includes:

[0073] S610. Use the newly labeled data to fine-tune the classification model and optimize the attention weight distribution.

[0074] S620. Update the knowledge graph node weights based on the operation and maintenance personnel's scores for the operation and maintenance strategies.

[0075] Reference Figure 2 The embodiment of the present invention proposes to implement a digital intelligent integrated platform terminal display operation and maintenance decision system, including:

[0076] The data acquisition module 100 is used to collect partial discharge signals in real time through ultra-high frequency sensors and ultrasonic sensors, and obtain environmental data and equipment operating status data; environmental data includes temperature and humidity; equipment operating status data includes load rate, operating time and voltage.

[0077] The partial discharge identification module 200 is used to input the partial discharge signal into the CNN+Transformer network to extract time-frequency domain features, input the environmental data and equipment operation status data into the fully connected network to encode auxiliary features, and dynamically weight the time-frequency domain features and auxiliary features through the attention mechanism to obtain a fused feature vector; output the discharge type probability distribution based on the fused feature vector, and generate the environment or equipment operation status association analysis based on the auxiliary feature vector and the discharge type classification result through gradient backpropagation and physical rule matching.

[0078] The explainability module 300 is used to generate a feature contribution report based on SHAP+LIME, obtain key features according to the feature contribution report, match physical rules of the physical rule library based on the key features, verify the model decision according to the physical rules, and determine to dynamically adjust the confidence or output environmental risk prompts based on the model decision verification results.

[0079] The operation and maintenance strategy module 400 is used to obtain the discharge type, discharge type confidence, environmental data and equipment operation status data, and match the operation and maintenance strategy based on the knowledge graph.

[0080] The visualization module 500 is used to generate a 3D spectrum diagram and a thermal map based on the partial discharge signal, generate a natural language report based on the operation and maintenance strategy, and display the 3D spectrum diagram, thermal map and natural language report on a visualization interface.

[0081] Feedback optimization module 600 is used to record user correction decisions and operation and maintenance strategy execution results, update model parameters through incremental learning, dynamically adjust knowledge graph node weights, and optimize the physical rule base logic.

[0082] Although specific embodiments are described herein, those skilled in the art will recognize that many other modifications or alternative embodiments are also within the scope of the present disclosure. For example, any of the functions and / or processing capabilities described in conjunction with a particular device or component may be performed by any other device or component. In addition, although various exemplary implementations and architectures have been described in accordance with the embodiments of the present disclosure, those skilled in the art will recognize that many other modifications to the exemplary implementations and architectures described herein are also within the scope of the present disclosure.

[0083] Some aspects of the present disclosure have been described above with reference to the block diagrams and flow charts of the systems, methods, systems and / or computer program products according to the exemplary embodiments. It should be understood that the combination of one or more blocks in the block diagram and the flow chart and the blocks in the block diagram and the flow chart can be realized by executing computer executable program instructions respectively. Equally, according to some embodiments, some blocks in the block diagram and the flow chart may not need to be executed in the order shown, or may not need to be executed in full. In addition, additional components and / or operations beyond those components and / or operations shown in the blocks in the block diagram and the flow chart may be present in certain embodiments.

[0084] Therefore, the blocks in the block diagrams and flow charts support combinations of means for performing the specified functions, combinations of elements or steps for performing the specified functions, and program instruction means for performing the specified functions. It should also be understood that each block in the block diagrams and flow charts, and combinations of blocks in the block diagrams and flow charts, can be implemented by a dedicated hardware computer system that performs the specific functions, elements, or steps, or a combination of dedicated hardware and computer instructions.

[0085] The program modules, applications, etc. described herein may include one or more software components, including, for example, software objects, methods, data structures, etc. Each such software component may include computer-executable instructions that, in response to execution, cause at least a portion of the functionality described herein (e.g., one or more operations of the illustrative methods described herein) to be performed.

[0086] Software component can be encoded with any one in various programming languages.A kind of exemplary programming language can be low-level programming language, such as the assembly language associated with specific hardware architecture and / or operating system platform.Comprise that the software component of assembly language instruction may need to be converted to executable machine code by assembler before being executed by hardware architecture and / or platform.Another exemplary programming language can be a more advanced programming language, and it can be transplanted across multiple architectures.Comprise that the software component of more advanced programming language may need to be converted to intermediate representation by interpreter or compiler before execution.Other examples of programming language include but are not limited to macro language, shell or command language, job control language, script language, database query or search language or report writing language.In one or more exemplary embodiments, the software component that comprises the instruction of one in the above-mentioned programming language example can be directly executed by operating system or other software component, without first being converted into another form.

[0087] Software components can be stored as files or other data storage structures. Software components of similar types or related functions can be stored together, such as in a specific directory, folder, or library. Software components can be static (e.g., preset or fixed) or dynamic (e.g., created or modified at execution time).

[0088] The embodiments of the present invention are described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Various changes can be made within the knowledge of ordinary technicians in the relevant technical field without departing from the scope of the present invention.

Claims

1. A digital intelligent integrated platform terminal display operation and maintenance decision-making method, characterized in that: The following steps are involved: S100, collecting partial discharge signals in real time through a UHF sensor and an ultrasonic sensor, and obtaining environmental data and device operating status data; the environmental data includes temperature and humidity; the device operating status includes data load rate, operating time, and voltage; S200, identifying the discharge type through a discharge type identification model, including: inputting the partial discharge signal into a CNN+Transformer network to extract time-frequency domain features, inputting environmental data and equipment operating status data into a fully connected network to encode auxiliary features, and dynamically weighting the time-frequency domain features and auxiliary features through an attention mechanism to obtain a fused feature vector; outputting a discharge type probability distribution based on the fused feature vector, and generating an environment or equipment operating status association analysis based on the auxiliary feature vector and the discharge type classification result through gradient backpropagation and physical rule matching; S300: Generate a feature contribution report based on SHAP+LIME, obtain key features based on the feature contribution report, match physical rules of the physical rule library based on the key features, verify the model decision based on the physical rules, and determine dynamically adjust the confidence level or output environmental risk warnings based on the model decision verification results; S400: Obtain discharge type, discharge type confidence, environmental data, and equipment operating status data, and match operation and maintenance strategies based on the knowledge graph; S500: Generate 3D spectrum and heat map based on partial discharge signal, generate natural language report based on operation and maintenance strategy, and display 3D spectrum, heat map and natural language report on the visual interface; S600: Record user correction decisions and operation and maintenance strategy execution results, update model parameters through incremental learning, dynamically adjust knowledge graph node weights, and optimize the physical rule base logic.

2. The digital intelligent integrated platform terminal display operation and maintenance decision-making method according to claim 1 is characterized in that: The step S200 includes: S210 , pre-processing the collected partial discharge signal, including filtering and noise reduction; and dynamically adjusting the signal gain according to the temperature data.

3. The digital intelligent integrated platform terminal display operation and maintenance decision-making method according to claim 1 is characterized in that: The step S200 includes: S220, extracting the spectral spatial features of the partial discharge signal based on CNN, capturing the time domain sequence dependency based on Transformer, and obtaining the time-frequency domain features of the partial discharge signal; S230, obtaining auxiliary features based on the fully connected network coding environment features and the device operation status features; S240 , generating a partial discharge signal weight and an environment / state weight based on the fully connected layer, weightedly fusing the time-frequency domain features and auxiliary features of the partial discharge signal to obtain a fused feature vector; and outputting a discharge type probability distribution based on the fused feature vector.

4. The digital intelligent integrated platform terminal display operation and maintenance decision-making method according to claim 1 is characterized in that: The step S200 includes: S250. Calculate the gradient of the environmental feature or the device operating status feature on the classification result through back propagation, and quantify the influence weight of the environmental feature or the device operating status feature; S260: Compare the quantified results with the preset domain knowledge to verify the rationality of the discharge type identification model and obtain the correlation analysis results of the environment or equipment operation status.

5. The digital intelligent integrated platform terminal display operation and maintenance decision-making method according to claim 1 is characterized in that: The step S300 includes: S310. Perform feature contribution analysis on the discharge type classification results, calculate the Shapley value of each feature of the partial discharge signal to obtain a SHAP global explanation, and generate a single-sample heat map to identify key discriminant features to obtain a LIME local explanation.

6. The digital intelligent integrated platform terminal display operation and maintenance decision-making method according to claim 1, characterized in that: The step S300 includes: S320, constructing a physical rule base, including encoding domain knowledge into logical rules, storing them in a structured database, and obtaining a physical rule base; S330: When the environmental data or the device operating status data exceeds a preset threshold, a physical rule verification is triggered; based on the environmental data or the device operating status data, it is determined whether the environmental anomaly or the device anomaly may cause feature distortion; if so, the confidence level is adjusted and an environmental / device anomaly alarm is generated; S340. When the confidence level of the discharge type classification result is lower than the set threshold, physical rule verification is triggered; key features are obtained based on the feature contribution of the discharge type classification result, and physical rules in the physical rule library are matched based on the key features to determine whether the partial discharge features that the discharge type identification model relies on comply with the physical rules; when the discharge type identification model relies on conflicting features, a rule conflict alarm is triggered.

7. The digital intelligent integrated platform terminal display operation and maintenance decision-making method according to claim 1 is characterized in that: The S400 includes: S410: Obtain historical work orders, equipment parameter tables, operation and maintenance logs, and expert experience documents to build a knowledge graph, including defining equipment types, failure modes, and operation and maintenance actions as entities and establishing relationships between entities. S420: Matching associated operation and maintenance strategies in the knowledge graph according to the input discharge type, and screening operation and maintenance strategies with high success rates based on historical data; S430, dynamically adjusting the operation and maintenance strategy based on the equipment operation status data; S440. Calculate the risk level based on the confidence level, environmental data, and equipment operating status data to obtain the operation and maintenance priority; S450: Output an operation and maintenance text report according to the operation and maintenance strategy and operation and maintenance priority.

8. The digital intelligent integrated platform terminal display operation and maintenance decision-making method according to claim 1 is characterized in that: The S600 includes: S610: Use the newly labeled data to fine-tune the classification model and optimize the attention weight distribution; S620. Update the knowledge graph node weights based on the operation and maintenance personnel's scores for the operation and maintenance strategies.

9. The digital intelligent integrated platform terminal display operation and maintenance decision-making method according to claim 1, characterized in that: The step S500 includes: S510: Generate AR positioning guidance according to the operation and maintenance strategy.

10. A digital intelligent integrated platform terminal display operation and maintenance decision system, used to execute the method according to any one of claims 1 to 9, characterized in that: include: A data acquisition module is used to collect partial discharge signals in real time through ultra-high frequency sensors and ultrasonic sensors, and obtain environmental data and equipment operating status data; the environmental data includes temperature and humidity; the equipment operating status data includes load rate, operating time and voltage; The PD recognition module is used to input the PD signal into the CNN+Transformer network to extract the time-frequency domain features, input the environmental data and equipment operation status data into the fully connected network to encode the auxiliary features, and dynamically weight the time-frequency domain features and auxiliary features through the attention mechanism to obtain the fused feature vector; The fusion feature vector outputs the probability distribution of the discharge type. Based on the auxiliary feature vector and the discharge type classification results, the environment or equipment operation status association analysis is generated through gradient back propagation and physical rule matching. The explainability module is used to generate a feature contribution report based on SHAP+LIME, obtain key features based on the feature contribution report, match the physical rules of the physical rule library based on the key features, verify the model decision based on the physical rules, and determine the dynamic adjustment of confidence or output of environmental risk warnings based on the model decision verification results; The operation and maintenance strategy module is used to obtain discharge type, discharge type confidence, environmental data, and equipment operating status data, and match operation and maintenance strategies based on the knowledge graph; The visualization module is used to generate 3D spectrum graphs and thermal maps based on partial discharge signals, and natural language reports based on operation and maintenance strategies. The 3D spectrum graphs, thermal maps, and natural language reports will be displayed on the visualization interface. The feedback optimization module is used to record user correction decisions and operation and maintenance strategy execution results, update model parameters through incremental learning, dynamically adjust knowledge graph node weights, and optimize the physical rule base logic.

Images