Power equipment operation state evaluation and prediction method based on big data

Abstract

The invention relates to the technical field of power equipment state monitoring, in particular to a power equipment operation state evaluation and prediction method based on big data, which comprises a multi-source data acquisition module, a feature engineering processing module, an intelligent evaluation and prediction module and a decision support output module. By setting a multi-source sensing end, when the operation state of the power equipment is evaluated, the definiteness of state evaluation of different types of equipment is ensured by formulating multi-modal data fusion standard parameters and setting different state sensing weights for different types of equipment; and meanwhile, real-time state sensing is performed by fusing electrical parameters, mechanical vibration, thermal distribution and environmental stress data, so that whether a sensing blind area problem caused by a single data dimension occurs in an equipment evaluation process or not can be detected in real time, the comprehensiveness and accuracy of equipment operation state evaluation are ensured, and state sensing fragmentation errors are further reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment status monitoring, and in particular to a method for evaluating and predicting the operating status of power equipment based on big data. Background Art

[0002] Power equipment mainly includes two categories: power generation equipment and power supply equipment. Power generation equipment mainly includes power station boilers, steam turbines, gas turbines, hydro turbines, generators, and transformers. Power supply equipment mainly includes transmission lines, mutual inductors, and contactors of various voltage levels. The stable operation of the power system is related to the production and life of the people, especially industrial and domestic electricity consumption. In the operation process of the entire power system, from power generation to transmission, distribution, and even the final power consumption stage, its reliability has always been a process that must be focused on. The analysis of the operating status of power equipment through real-time monitoring, data analysis and preventive maintenance strategies not only improves the safety, reliability and efficiency of the power system, but also reduces operating costs, which is of great significance to promoting the sustainable and healthy development of the power industry.

[0003] The main defects of existing power equipment operating status assessment and prediction methods include: 1. Single dimension of data collection: The existing system primarily relies on the basic electrical parameters of voltage and current provided by the SCADA system, and on static data on insulation resistance and dielectric loss obtained through regular offline testing. This single-dimensional data collection method cannot fully reflect the true operating status of the equipment. In particular, there are blind spots in the detection of multi-physics field coupling faults such as abnormal mechanical vibration, local overheating, and insulation degradation. Key features such as vibration spectrum changes caused by loose transformer cores and thermal image anomalies caused by poor contact of circuit breaker contacts are not captured.

[0004] 2. Insufficient separation of environmental impact factors: Power equipment actually operates under complex environmental stresses, and its aging rate exhibits nonlinear characteristics. However, existing assessment models generally treat the equipment as a closed system and fail to establish a quantitative correlation model between environmental stress and equipment aging. Studies have shown that in high temperature and high humidity environments, traditional methods can predict the remaining life of transformers by up to 40%, restricting the accuracy of maintenance decisions.

[0005] 3. Lack of dynamic model update mechanism: The current mainstream method uses a static threshold warning mechanism, and its evaluation rules remain fixed once set. This mechanism cannot adapt to the gradual degradation characteristics of equipment as the operating time increases, and it cannot dynamically adjust the evaluation strategy according to real-time operating conditions. As a result, when the equipment enters the accelerated aging stage, the monitoring sensitivity cannot be improved in time, and excessive warnings are generated during the stable operation stage.

[0006] 4. The closed loop of operation and maintenance decision-making has not been formed: Existing technologies typically stop at outputting equipment health scores and failure probabilities, without a mechanism for converting assessment results into maintenance actions. Operations and maintenance personnel still rely on experience to manually formulate maintenance plans, lacking decision-making support for accurately locating the root causes of failures, predicting spare parts demand, and optimizing maintenance windows. This disconnect results in 25% of preventive maintenance resources being wasted, while 30% of sudden failures occur due to untimely warning upgrades.

[0007] Therefore, in response to the above problems, the present invention provides a method for evaluating and predicting the operating status of power equipment based on big data, which can integrate and build an environmental stress-adaptive dynamic prediction architecture for the remaining life, realize real-time model updates under changing operating conditions, and design an intelligent conversion engine from evaluation and prediction results to maintenance strategies, supporting fault root cause location, maintenance resource optimization and risk cost control. Summary of the Invention

[0008] (1) Technical problems solved In response to the deficiencies of the existing technology, the present invention provides a method for evaluating and predicting the operating status of power equipment based on big data, which solves the problems raised in the above background technology.

[0009] (2) Technical solution To achieve the above objectives, the present invention provides the following technical solutions: a method for evaluating and predicting the operating status of power equipment based on big data, the method comprising the following steps: S1. Collect real-time operation data, historical fault data, and environmental parameters of power equipment through a multi-source heterogeneous sensor network to build a multimodal operation database of power equipment; S2. Performing spatiotemporal alignment and missing value repair processing on the multimodal operation data to generate a standardized equipment operation time series data set; S3. Extract multi-scale features of equipment operation data based on deep convolutional neural networks, and use wavelet packet decomposition algorithm to separate steady-state features from transient features to generate equipment operation status feature matrix; S4. Construct an equipment health assessment model that integrates physical mechanisms and data-driven methods, input the characteristic matrix into the assessment model to quantitatively score the equipment health status, and output the equipment health index and fault risk level; S5. Establish an equipment degradation prediction model based on the long short-term memory network of the attention mechanism, dynamically predict the remaining service life of the equipment by combining the environmental stress acceleration factor, and generate an equipment life prediction curve; S6. Use knowledge graph technology to correlate historical equipment maintenance records with real-time evaluation results to generate equipment maintenance strategy optimization plans; S7. Build a digital twin of equipment status through a three-dimensional visualization engine, dynamically map equipment assessment results and predicted trends, and trigger graded warning signals.

[0010] Preferably, the S1 includes the following sub-steps: S11. Collect equipment physical status data through vibration sensors, infrared thermal imagers, and partial discharge detectors deployed on the power equipment itself; S12. Obtain electrical operating parameters of equipment voltage, current, and power factor through the SCADA system; S13. Collect environmental temperature, humidity, pollution level, and space electromagnetic field data through weather stations and IoT terminals; S14. Establish a unique device code and spatiotemporal tag, and associate and store multi-source data according to device ID-timestamp.

[0011] Preferably, said S2 comprises the following sub-steps: S21, using dynamic time warping algorithm to align sensor data streams with different sampling frequencies; S22. Build a missing data repair model based on a generative adversarial network to generate complete device runtime data. S23. Perform dimensionless processing on multidimensional data through Z-score standardization and maximum and minimum normalization.

[0012] Preferably, the S3 includes the following sub-steps: S31. Construct a one-dimensional convolutional neural network to extract the time domain features of electrical parameters; S32, using a two-dimensional convolutional neural network to extract the spatial temperature distribution characteristics of infrared thermal images; S33. Decompose the vibration signal frequency band energy through the wavelet packet energy entropy algorithm to identify the mechanical abnormality characteristics of the equipment.

[0013] Preferably, the S4 includes the following sub-steps: S41. Establish a physical failure model library for equipment, including mathematical representations of three types of failure modes: insulation aging, mechanical wear, and contact degradation; S42, fusing the data-driven features with the physical model output and inputting the results into a gradient boosting decision tree classifier; S43. Generate the equipment health index EHI based on the fuzzy comprehensive evaluation method. The calculation formula is: in is the feature weight, is the normalized feature degradation degree, is the time attenuation coefficient, is the time increment, is the time decay factor.

[0014] Preferably, the S5 includes the following sub-steps: S51. Build the Attention-LSTM network architecture. The input layer is the device feature matrix. The attention layer weight distribution formula is: in is the attention weight at time step t, is the hidden state at time step t, is the context vector.

[0015] S52. Introducing the Arrhenius acceleration model to correct the impact of environmental stress: Among them is is the activation energy, is the Boltzmann constant, is the acceleration factor, is the reference temperature, is the actual operating temperature.

[0016] Preferably, the S6 includes the following sub-steps: S61. Build a knowledge graph of power equipment, with nodes including equipment model, failure mode, and maintenance plan; S62. Calculate the similarity between real-time evaluation results and historical cases based on graph neural networks; S63. Generate a maintenance strategy optimization plan, including spare parts allocation path, maintenance window period and cost budget matrix.

[0017] Preferably, the S7 includes the following sub-steps: S71, using Unity3D engine to build a 3D digital twin of the equipment; S72. Visualize device heatmaps using color mapping algorithms: EHI greater than 80% is displayed in green; 60% less than EHI less than 80% is displayed in yellow; EHI less than 60% is displayed in red; S73, when the predicted RUL is lower than the threshold When the warning is triggered, the fourth-level warning mechanism is triggered.

[0018] Preferably, the environmental stress adaptive prediction further comprises the following steps: S58. Establishing an environmental stress-life decay mapping library: Stores the device life attenuation curve in typical environmental scenarios, including: High temperature and high humidity scenarios: temperature greater than 40°C and humidity greater than 80%; Salt spray corrosion scenario: pollution level is greater than level 3 and humidity is greater than 70%; Low temperature drying scenario: temperature below -20°C and humidity below 30%; S59, real-time matching of current environment and historical scenes: in is the scene similarity, is the environmental parameter of historical scene k.

[0019] S60. Lifespan prediction fusion based on similarity weighting: Among them is Match the lifespan reference value of historical scenarios, is the predicted value corrected for environmental stress.

[0020] Preferably, the maintenance strategy optimization comprises the following steps: S64. Build an equipment maintenance cost-benefit optimization model: Constraints: Among them is For maintenance costs, is the downtime loss, is the failure risk cost, is the spare parts procurement lead time, The duration of maintenance implementation.

[0021] S65. Generate a multi-objective optimization solution set: Plan A: Immediate maintenance, high cost, low risk; Option B: Maintenance during the next scheduled maintenance, medium cost and medium risk; Option C: Delay maintenance until the critical point of predicted lifespan, low cost but high risk; S66. Automatically select the optimal solution based on risk preference: in is the scheme number, is the cost weight, is the risk weight, Score the decision of the j-th maintenance plan, is the failure risk probability, is the normalized cost value.

[0022] (3) Beneficial effects Compared with the existing technology, the present invention provides a method for evaluating and predicting the operating status of power equipment based on big data, which has the following beneficial effects: 1. In the present invention, by setting up a multi-source perception terminal, when conducting the operation status evaluation of power equipment, by formulating multimodal data fusion standard parameters and setting different state perception weights for different equipment types, the clarity of the state evaluation of different types of equipment is guaranteed. At the same time, electrical parameters, mechanical vibration, thermal distribution and environmental stress data are integrated for real-time state perception, which can detect in real time whether there is a perception blind spot problem caused by a single data dimension during the equipment evaluation process, ensure the comprehensiveness and accuracy of the equipment operation status evaluation, and further reduce the error of state perception fragmentation.

[0023] 2. In the present invention, by setting up a dual-drive evaluation end, when modeling the health status of power equipment, by calculating the coupling deviation value of the physical failure mechanism model and the deep learning feature extraction model, it is judged in real time whether the evaluation deviation caused by model splitting occurs in the equipment evaluation process, so that the system can reduce the situation where the equipment health score deviates from the actual state, and when the equipment health model deviates due to insufficient interpretability or adaptability, the feature weights corresponding to the status evaluation standard parameters can be optimized in real time, so that when the evaluation model is abnormal, the parameter contribution can be dynamically corrected, ensuring that the interpretability and adaptability of the equipment health evaluation are both improved.

[0024] 3. In the present invention, by setting up a life prediction terminal, when predicting the remaining life of power equipment, different environmental stresses are automatically quantified and graded, and the equipment degradation deviation value under the coupling of multiple working conditions is detected in real time. According to the acceleration effect of temperature, humidity, and contamination stress, the life prediction curve is dynamically generated, so that the system can realize degradation path tracking under the influence of multiple factors, reduce the risk of prediction failure caused by insufficient environmental adaptability of static models, and further improve the accuracy and environmental adaptability of equipment life prediction.

[0025] 4. In the present invention, by setting up an intelligent decision-making terminal, when formulating the maintenance strategy of the power equipment, by calculating the matching deviation value between the health status and resource scheduling, it is judged in real time whether there is a decision lag problem in the maintenance plan, so that the system can reduce the imbalance between maintenance cost and equipment risk, and when the maintenance resource allocation is abnormal, it can optimize the maintenance priority allocation in real time through the cost-risk decision-making model to ensure the balance between downtime loss, risk loss and maintenance cost, and further eliminate the lag risk caused by traditional manual decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0028] See also Figure 1 The method for evaluating and predicting the operating status of power equipment based on big data comprises the following steps: S1. Collect real-time operation data, historical fault data, and environmental parameters of power equipment through a multi-source heterogeneous sensor network to build a multimodal operation database of power equipment; S2. Perform spatiotemporal alignment and missing value repair on the multimodal operation data to generate a standardized equipment operation time series dataset; S3. Extract multi-scale features of equipment operation data based on deep convolutional neural networks, and use wavelet packet decomposition algorithm to separate steady-state features from transient features to generate equipment operation status feature matrix; S4. Construct an equipment health assessment model that integrates physical mechanisms and data-driven methods, input the characteristic matrix into the assessment model to quantitatively score the equipment health status, and output the equipment health index and fault risk level; S5. Establish an equipment degradation prediction model based on the long short-term memory network of the attention mechanism, dynamically predict the remaining service life of the equipment by combining the environmental stress acceleration factor, and generate an equipment life prediction curve; S6. Use knowledge graph technology to correlate historical equipment maintenance records with real-time evaluation results to generate equipment maintenance strategy optimization plans; S7. Build a digital twin of equipment status through a three-dimensional visualization engine, dynamically map equipment assessment results and predicted trends, and trigger graded warning signals.

[0029] S1 includes the following sub-steps: S11. Collect equipment physical status data through vibration sensors, infrared thermal imagers, and partial discharge detectors deployed on the power equipment itself; S12. Obtain the electrical operating parameters of the equipment, such as voltage, current, and power factor, through the SCADA system; S13. Collect environmental temperature, humidity, pollution level, and space electromagnetic field data through weather stations and IoT terminals; S14. Establish a unique device code and spatiotemporal tag, and associate and store multi-source data according to device ID-timestamp.

[0030] S2 includes the following sub-steps: S21, using dynamic time warping algorithm to align sensor data streams with different sampling frequencies; S22. Build a missing data repair model based on a generative adversarial network to generate complete device runtime data. S23. Perform dimensionless processing on multidimensional data through Z-score standardization and maximum and minimum normalization.

[0031] S3 includes the following sub-steps: S31. Construct a one-dimensional convolutional neural network to extract the time domain features of electrical parameters; S32, using a two-dimensional convolutional neural network to extract the spatial temperature distribution characteristics of infrared thermal images; S33. Decompose the vibration signal frequency band energy through the wavelet packet energy entropy algorithm to identify the mechanical abnormality characteristics of the equipment.

[0032] S4 includes the following sub-steps: S41. Establish a physical failure model library for equipment, including mathematical representations of three types of failure modes: insulation aging, mechanical wear, and contact degradation; S42, fusing the data-driven features with the physical model output and inputting the results into a gradient boosting decision tree classifier; S43. Generate the equipment health index EHI based on the fuzzy comprehensive evaluation method. The calculation formula is: in is the feature weight, is the normalized feature degradation degree, is the time attenuation coefficient, is the time increment, is the time decay factor.

[0033] S5 includes the following sub-steps: S51. Build the Attention-LSTM network architecture. The input layer is the device feature matrix. The attention layer weight distribution formula is: in is the attention weight at time step t, is the hidden state at time step t, is the context vector.

[0034] S52. Introducing the Arrhenius acceleration model to correct the impact of environmental stress: Among them is is the activation energy, is the Boltzmann constant, is the acceleration factor, is the reference temperature, is the actual operating temperature.

[0035] S6 includes the following sub-steps: S61. Build a knowledge graph of power equipment, with nodes including equipment model, failure mode, and maintenance plan; S62. Calculate the similarity between real-time evaluation results and historical cases based on graph neural networks; S63. Generate a maintenance strategy optimization plan, including spare parts allocation path, maintenance window period and cost budget matrix.

[0036] S7 includes the following sub-steps: S71, using Unity3D engine to build a 3D digital twin of the equipment; S72. Visualize device heatmaps using color mapping algorithms: EHI greater than 80% is displayed in green; 60% less than EHI less than 80% is displayed in yellow; EHI less than 60% is displayed in red; S73, when the predicted RUL is lower than the threshold When the warning is triggered, the fourth-level warning mechanism is triggered.

[0037] Environmental stress adaptive prediction also includes the following steps: S58. Establishing an environmental stress-life decay mapping library: Stores the device life attenuation curve in typical environmental scenarios, including: High temperature and high humidity scenarios: temperature greater than 40°C and humidity greater than 80%; Salt spray corrosion scenario: pollution level is greater than level 3 and humidity is greater than 70%; Low temperature drying scenario: temperature below -20°C and humidity below 30%; S59, real-time matching of current environment and historical scenes: in is the scene similarity, is the environmental parameter of historical scene k.

[0038] S60. Lifespan prediction fusion based on similarity weighting: Among them is Match the lifespan reference value of historical scenarios, is the predicted value corrected for environmental stress.

[0039] Maintenance strategy optimization includes the following steps: S64. Build an equipment maintenance cost-benefit optimization model: Constraints: Among them is For maintenance costs, is the downtime loss, is the failure risk cost, is the spare parts procurement lead time, The duration of maintenance implementation.

[0040] S65. Generate a multi-objective optimization solution set: Plan A: Immediate maintenance, high cost, low risk; Option B: Maintenance during the next scheduled maintenance, medium cost and medium risk; Option C: Delay maintenance until the critical point of predicted lifespan, low cost but high risk; S66. Automatically select the optimal solution based on risk preference: in is the scheme number, is the cost weight, is the risk weight.

[0041] include: Multi-source data acquisition module: including vibration monitoring unit, electrical parameter acquisition unit, and environmental perception unit; Feature engineering processing module: equipped with a spatiotemporal alignment processor and a feature extraction accelerator card; Intelligent evaluation and prediction module: deploys the physical model library, GBDT classifier, and Attention-LSTM prediction engine; Decision support output module: integrated knowledge graph database and three-dimensional visualization platform.

[0042] Example 1: Multi-source heterogeneous data collection and preprocessing: Step S1: Full-dimensional data collection: 1. Electrical parameter monitoring: (1) Install a high-frequency current transformer on the transformer high-voltage bushing to capture the partial discharge pulse waveform at a sampling rate of 1000 times per second; (2) Real-time monitoring of the winding hot spot temperature through the optical fiber temperature measurement system; (3) Obtain operating parameters such as load rate and power factor from the SCADA system; 2. Mechanical vibration perception: (1) Arrange an array of 8 triaxial acceleration sensors on the wall of the fuel tank: Position: 4 corresponding areas for core and 4 corresponding areas for winding; Measuring range: ±50g, resolution 0.001m / s²; (2) Synchronously collect vibration spectra under no-load, 50% load, and 100% load conditions; 3. Thermal distribution mapping: (1) Deploy 4 infrared thermal imagers to form a panoramic monitoring network: Wavelength: 12 μm; Thermal sensitivity: 0.03℃@30℃; The full-surface temperature field matrix is ​​generated every 5 minutes with a resolution of 1280 × 960 pixels; 4. Environmental stress monitoring: (1) Install the micro-weather station cluster: Temperature: 40℃; Humidity: 70%RH; Pollution degree: equivalent salt density classification based on image recognition; Electromagnetic field strength monitoring: power frequency field strength meter; Step S2: Data spatiotemporal alignment: 1. Unified time base: (1) Based on the Beidou satellite timing system; (2) Segment aggregation of high-frequency vibration signals: in is the signal mean, is the downsampling segment index, is the original signal sampling point index, is the kth original vibration acceleration sampling value.

[0043] Downsampling to 100Hz to synchronize with electrical parameters; 2. Spatial coordinate mapping: (1) Establish the transformer three-dimensional coordinate system; (2) Convert thermal image pixel coordinates to physical coordinates: Where K is the calibration matrix, which matches the temperature point with the vibration sensor position. is the physical coordinate of the transformer body, is the homogeneous coordinate of the thermal image pixel.

[0044] Example 2: Intelligent collection and fusion of multi-source data: Step S1: Adaptive sensor network deployment: Building an intelligent sensing network at the 500kV transformer site: (1) Electrical monitoring unit: A broadband current sensor is installed on the high-voltage bushing to capture the partial discharge pulse waveform in real time. The sampling rate is dynamically adjusted to 500 Hz under light load and 10 kHz under heavy load. (2) Vibration sensing array: 12 triaxial acceleration sensors are placed at key locations on the fuel tank to form a vibration monitoring grid that automatically identifies the locations of vibration sources in the core and windings. (3) Thermal imaging system: Four infrared thermal imagers are used to build a panoramic scanning network, completing a panoramic mapping of the transformer surface temperature field every 3 minutes; (4) Environmental sensing terminal: A meteorological microstation is installed on the top of the transformer to monitor temperature and humidity changes and salt spray deposition rate in real time; Step S2: Intelligent data fusion engine: Developing Adaptive Data Fusion Systems: (1) Time synchronization mechanism: Based on the PTP precision clock protocol, the timestamp error of each sensor is controlled within 100ns; (2) Spatial registration technology: A unified coordinate system is established through laser positioning to control the position deviation between the thermal image hotspot and the vibration sensor within 5 mm; (3) Data quality monitoring: Real-time detection of signal anomalies, automatically triggering re-collection when the vibration data signal-to-noise ratio is lower than 20dB; (4) Feature-level fusion: extract the electrical pulse rise time, vibration main frequency offset, and thermal gradient change rate features to construct a 33-dimensional fusion feature vector; Example 3: Dynamic tracking and evaluation of degradation process: Step S1: progressive degradation feature extraction: Implement state feature evolution tracking: (1) Insulation aging tracking: Identify the development trajectory of insulation paper electrical trees through changes in partial discharge phase distribution; (2) Mechanical wear monitoring: Analyze the vibration fundamental frequency deviation and third harmonic growth to quantify the degree of mechanical looseness; (3) Contact degradation detection: Combine the hot spot temperature rise rate and contact resistance change to evaluate the degradation state of the conductive circuit; (4) Environmental coupling analysis: Establish a correlation model between salt spray deposition and surface leakage current; Step S2: Dual-drive health assessment system: Deploy an online assessment platform: (1) Physical model library: integrated oil-paper insulation thermal aging model, metal fatigue accumulation model and 12 failure mechanism models (2) Real-time weight adjustment: When the ambient humidity is greater than 80%, the insulation weight is automatically increased by 40%; when the load rate is greater than 90%, the mechanical weight is increased by 30% (3) Health index calculation: Generate a health index curve with confidence interval based on characteristic degradation degree and environmental acceleration factor; (4) Fault pattern recognition: Distinguish the fault types such as insulation aging, mechanical looseness, and poor contact through the decision tree model; Example 3: Predictive maintenance decision optimization: Step S1: Full life cycle cost optimization: Build a maintenance decision model: (1) Maintenance cost modeling: Considering factors such as spare parts prices, labor rates, and special equipment usage fees; (2) Calculation of outage losses: Quantify outage losses based on the location of the equipment in the grid and the importance of the load; (3) Risk cost assessment: Calculate the probability of failure and loss based on the health status of the equipment and the operation mode of the power grid; (4) Opportunity cost analysis: assessing the impact of different maintenance time windows on grid operation; Step S2: Dynamic decision engine: Achieve intelligent decision-making closed loop: (1) Automatic plan generation: Based on the health status prediction, three types of plans are output: immediate maintenance, planned maintenance, and delayed maintenance; (2) Three-dimensional evaluation: Evaluate the solution from the perspectives of safety, economy, and reliability; (3) Risk preference configuration: support nuclear power model, urban power supply model, and rural power grid model; (4) Decision visualization: Display the positions of various options in a three-dimensional coordinate system and automatically recommend the Pareto optimal solution; (5) Execution tracking: Generate electronic work orders containing spare parts list, personnel configuration, and time window.

[0045] Example 4: Intelligent diagnosis and root cause analysis system: Step S1: Multimodal fault feature extraction: Deploy a distributed sensor network in the converter station valve hall: (1) Electrical characteristic layer: The current spikes during the thyristor turn-on and turn-off processes are captured by Rogowski coils. The sampling rate is set to 2 MHz to identify abnormal commutation oscillations. (2) Optical monitoring layer: Use ultraviolet imager to scan the discharge spot on the valve tower surface, and synchronize with infrared thermal imager to record the temperature rise in the corresponding area; (3) Voiceprint recognition layer: directional microphone arrays are installed at the four corners of the valve tower to collect the spectral characteristics of cavitation noise in the cooling water pipe; (4) Chemical sensing layer: Micro pH sensors and conductivity probes are embedded in the circulating cooling water pipes to detect changes in corrosion product concentration in real time; Step S2: Cross-dimensional correlation analysis engine: Build a fault feature association model: (1) Spatiotemporal event alignment: mapping current pulses, UV spots, and acoustic events to a unified time axis to establish millisecond-level event correlation; (2) Causal reasoning network: When the current oscillation amplitude is detected to exceed the threshold, the thermal image changes and acoustic features within 5 ms before and after are automatically retrieved; (3) Root cause location algorithm: When current oscillation is accompanied by noise in a specific frequency band and the local temperature rise is greater than 8°C, it is determined to be a cooling water channel blockage; when accompanied by ultraviolet light flashes, it is determined to be insulation surface creepage; Step S3: Adaptive diagnostic decision: Realize dynamic diagnostic closed loop: (1) Primary diagnosis: Quickly match typical failure modes based on the rule base; (2) Deep analysis: When the rule matching degree is less than 85%, the graph neural network is activated to analyze the feature association path; (3) Confidence assessment: A confidence index is generated simultaneously when the diagnosis result is output. When the index is less than 90%, the expert consultation mechanism is automatically triggered; (4) Knowledge evolution: storing manually confirmed new fault features into the case library and optimizing the diagnostic model weights; It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0046] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating and predicting the operating status of power equipment based on big data, characterized in that: The method comprises the following steps: S1. Collect real-time operation data, historical fault data, and environmental parameters of power equipment through a multi-source heterogeneous sensor network to build a multimodal operation database of power equipment; S2. Performing spatiotemporal alignment and missing value repair processing on the multimodal operation data to generate a standardized equipment operation time series data set; S3. Extract multi-scale features of equipment operation data based on deep convolutional neural networks, and use wavelet packet decomposition algorithm to separate steady-state features from transient features to generate equipment operation status feature matrix; S4. Construct an equipment health assessment model that integrates physical mechanisms and data-driven methods, input the characteristic matrix into the assessment model to quantitatively score the equipment health status, and output the equipment health index and fault risk level; S5. Establish an equipment degradation prediction model based on the long short-term memory network of the attention mechanism, dynamically predict the remaining service life of the equipment by combining the environmental stress acceleration factor, and generate an equipment life prediction curve; S6. Use knowledge graph technology to correlate historical equipment maintenance records with real-time evaluation results to generate equipment maintenance strategy optimization plans; S7. Build a digital twin of equipment status through a three-dimensional visualization engine, dynamically map equipment assessment results and predicted trends, and trigger graded warning signals.

2. The method for evaluating and predicting the operating status of power equipment based on big data according to claim 1, characterized in that: The S1 includes the following sub-steps: S11. Collect equipment physical status data through vibration sensors, infrared thermal imagers, and partial discharge detectors deployed on the power equipment itself; S12. Obtain the electrical operating parameters of the equipment, such as voltage, current, and power factor, through the SCADA system; S13. Collect environmental temperature, humidity, pollution level, and space electromagnetic field data through weather stations and IoT terminals; S14. Establish a unique device code and spatiotemporal tag, and associate and store multi-source data according to device ID-timestamp.

3. The method for evaluating and predicting the operating status of power equipment based on big data according to claim 1, characterized in that: The S2 includes the following sub-steps: S21, using dynamic time warping algorithm to align sensor data streams with different sampling frequencies; S22. Build a missing data repair model based on a generative adversarial network to generate complete device runtime data. S23. Perform dimensionless processing on multidimensional data through Z-score standardization and maximum and minimum normalization.

4. The method for evaluating and predicting the operating status of power equipment based on big data according to claim 1, characterized in that: The S3 includes the following sub-steps: S31. Construct a one-dimensional convolutional neural network to extract the time domain features of electrical parameters; S32, using a two-dimensional convolutional neural network to extract the spatial temperature distribution characteristics of infrared thermal images; S33. Decompose the vibration signal frequency band energy through the wavelet packet energy entropy algorithm to identify the mechanical abnormality characteristics of the equipment.

5. The method for evaluating and predicting the operating status of power equipment based on big data according to claim 1, characterized in that: The S4 includes the following sub-steps: S41. Establish a physical failure model library for equipment, including mathematical representations of three types of failure modes: insulation aging, mechanical wear, and contact degradation; S42, fusing the data-driven features with the physical model output and inputting the results into a gradient boosting decision tree classifier; S43. Generate the equipment health index EHI based on the fuzzy comprehensive evaluation method. The calculation formula is: in is the feature weight, is the normalized feature degradation degree, is the time attenuation coefficient, is the time increment, is the time decay factor.

6. The method for evaluating and predicting the operating status of power equipment based on big data according to claim 1, characterized in that: The S5 comprises the following sub-steps: S51. Build the Attention-LSTM network architecture. The input layer is the device feature matrix. The attention layer weight distribution formula is: in is the attention weight at time step t, is the hidden state at time step t, is the context vector; S52. Introducing the Arrhenius acceleration model to correct the impact of environmental stress: Among them is is the activation energy, is the Boltzmann constant, is the acceleration factor, is the reference temperature, is the actual operating temperature.

7. The method for evaluating and predicting the operating status of power equipment based on big data according to claim 1, characterized in that: The S6 comprises the following sub-steps: S61. Build a knowledge graph of power equipment, with nodes including equipment model, failure mode, and maintenance plan; S62. Calculate the similarity between real-time evaluation results and historical cases based on graph neural networks; S63. Generate a maintenance strategy optimization plan, including spare parts allocation path, maintenance window period and cost budget matrix.

8. The method for evaluating and predicting the operating status of power equipment based on big data according to claim 1, characterized in that: The S7 includes the following sub-steps: S71, using Unity3D engine to build a 3D digital twin of the equipment; S72. Visualize device heatmaps using color mapping algorithms: EHI greater than 80% is displayed in green; 60% less than EHI less than 80% is displayed in yellow; EHI less than 60% is displayed in red; S73, when the predicted RUL is lower than the threshold When the warning is triggered, the fourth-level warning mechanism is triggered.

9. The method for evaluating and predicting the operating status of power equipment based on big data according to claim 1, characterized in that: The environmental stress adaptive prediction further comprises the following steps: S58. Establishing an environmental stress-life decay mapping library: Stores the device life attenuation curve in typical environmental scenarios, including: High temperature and high humidity scenarios: temperature greater than 40°C and humidity greater than 80%; Salt spray corrosion scenario: pollution level is greater than level 3 and humidity is greater than 70%; Low temperature drying scenario: temperature below -20°C and humidity below 30%; S59, real-time matching of current environment and historical scenes: in is the scene similarity, is the environmental parameter of historical scene k; S60. Lifespan prediction fusion based on similarity weighting: Among them is Match the lifespan reference value of historical scenarios, is the predicted value corrected for environmental stress.

10. The method for evaluating and predicting the operating status of power equipment based on big data according to claim 1, characterized in that: The maintenance strategy optimization comprises the following steps: S64. Build an equipment maintenance cost-benefit optimization model: Constraints: Among them is For maintenance costs, is the downtime loss, is the failure risk cost, is the spare parts procurement lead time, Length of time for maintenance implementation; S65. Generate a multi-objective optimization solution set: Plan A: Immediate maintenance, high cost, low risk; Option B: Maintenance during the next scheduled maintenance, medium cost and medium risk; Option C: Delay maintenance until the critical point of predicted lifespan, low cost but high risk; S66. Automatically select the optimal solution based on risk preference: in is the scheme number, is the cost weight, is the risk weight, Score the decision of the j-th maintenance plan, is the failure risk probability, is the normalized cost value.

Images