Auxiliary equipment fault early warning and health management system and methods for thermal power plants

By constructing an environment-performance benchmark model and an extreme weather scenario database, combined with a game theory decision-making model, the problem of environmental interference and meteorological risk prediction in the management of auxiliary equipment in thermal power plants was solved, achieving accurate fault early warning and optimized operation and maintenance decisions, reducing equipment failure risks and downtime losses.

CN123069009APending Publication Date: 2026-11-10HUANENG ANYUAN POWER GENERATION CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610946912.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-06-29
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

The existing auxiliary equipment failure early warning and health management system for thermal power plants cannot effectively distinguish between external environmental factors and equipment degradation, resulting in inaccurate early warnings, lack of meteorological risk prediction, and inability to coordinate the demands of multiple parties in operation and maintenance decisions, which easily leads to sudden equipment failures and unplanned shutdowns.

Method used

An environment-performance benchmark model is constructed, and multi-dimensional data collected by IoT sensing nodes is standardized, the model is dynamically calibrated, multi-dimensional deviation analysis and fault feature extraction are performed, an energy efficiency health index is generated, an extreme weather scenario database is built, vulnerability scoring is performed, a game-theoretic decision model is constructed to solve for Nash equilibrium, and the operation strategy is optimized.

Benefits of technology

It enables precise differentiation between environmental impacts and equipment performance degradation, quantifies equipment health status, reduces false alarm rate of fault warning, detects hidden faults in advance, avoids equipment failures caused by extreme weather, coordinates equipment safety, operating costs and power grid stability, locks in the optimal maintenance window, and reduces downtime losses and energy consumption increases.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123069009A_ABST
    Figure CN123069009A_ABST
Patent Text Reader

Abstract

This invention provides a fault early warning and health management system and method for auxiliary equipment in thermal power plants, belonging to the field of auxiliary equipment management technology. The system includes: a data acquisition and processing module for collecting multi-dimensional data on the environment and operating conditions to obtain a standardized dataset; a model building and calibration module for constructing an environment-performance benchmark model; a quantitative analysis module for performing multi-dimensional deviation analysis and fault feature extraction to obtain a baseline health value; a risk prediction module for combining energy efficiency and health indices to simulate and calculate vulnerability scores and generate a comprehensive risk value; a decision optimization module for calculating the total benefit of all execution strategies to determine the optimal shutdown window and corresponding execution strategy; and an optimized operation module for performing multi-objective Pareto optimization to select the optimal execution strategy. This invention achieves accurate differentiation between environmental interference and equipment degradation, early warning of hidden faults, and coordinated safety and economy, achieving pre-emptive risk prevention and optimal overall plant benefits.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of auxiliary equipment management technology for thermal power plants, and specifically to a fault early warning and health management system and method for auxiliary equipment in thermal power plants. Background Technology

[0002] Auxiliary equipment in thermal power plants, such as forced draft fans, induced draft fans, cooling towers, and air-cooled islands, are crucial supporting equipment for the stable operation of the units. Their operating status directly affects the overall power generation efficiency, energy consumption level, and production safety of the plant. Currently, the industry has widely deployed auxiliary equipment fault early warning and health management systems, which mainly rely on single operating parameters such as equipment vibration and temperature for status monitoring, and combine historical thresholds to provide simple alarm reminders, becoming a routine means of power plant operation and maintenance management. However, as the operating conditions of thermal power units become increasingly complex, and the interference from differentiated site environments such as coastal areas, high dust levels, and strong winds intensifies, traditional management models are difficult to adapt to the needs of refined operation and maintenance.

[0003] The existing technology for early warning and health management systems for auxiliary equipment in thermal power plants has the following shortcomings: First, the existing technology does not distinguish between performance changes caused by external environmental factors and equipment degradation, and cannot effectively isolate external interference factors such as weather, dust, temperature and humidity. It relies solely on fixed thresholds to judge equipment status, which easily leads to missed or false alarms and makes it difficult to accurately identify early hidden equipment faults. Second, the existing technology only focuses on monitoring the current operating status of the equipment and does not combine weather forecasts to conduct risk simulations. It cannot quantify the load-bearing capacity and failure probability of sub-healthy auxiliary equipment under extreme weather conditions, and lacks the ability to predict risks in advance. When facing severe weather such as typhoons and strong sandstorms, the risk of sudden equipment failures and unplanned unit shutdowns remains high. Third, the operation and maintenance decisions of the existing technology rely heavily on manual experience and only consider equipment safety or maintenance costs. It does not take into account the interests of multiple stakeholders such as equipment management, business accounting, and grid dispatching, which easily leads to problems such as unreasonable maintenance timing, excessive downtime losses, and grid load fluctuations, and fails to achieve globally optimal decision-making. Summary of the Invention

[0004] The purpose of this invention is to provide a fault early warning and health management system and method for thermal power plant auxiliary equipment, in order to solve the technical problems of existing thermal power plant auxiliary equipment management being susceptible to environmental interference leading to inaccurate early warnings, lacking meteorological risk prediction, and being unable to coordinate the demands of multiple parties in operation and maintenance decisions.

[0005] To achieve the above objectives, this invention provides a fault early warning and health management system for auxiliary equipment in thermal power plants. The system includes: a data acquisition and processing module, used to collect multi-dimensional data on the environment and operating conditions through IoT sensing nodes deployed on the auxiliary equipment of the target thermal power plant, and perform standardized preprocessing to obtain a standardized dataset; a model building and calibration module, used to select historical health samples from the historical standardized dataset of the auxiliary equipment of the target thermal power plant, construct an environment-performance benchmark model, and dynamically calibrate the trained environment-performance benchmark model; and a quantitative analysis module, used to calculate the multi-dimensional deviation between the performance benchmark value and the actual operating value of the auxiliary equipment of the target thermal power plant based on the environment-performance benchmark model, and perform multi-dimensional deviation analysis and fault feature extraction. The system generates an energy efficiency health index and determines the type of fault, thus obtaining a baseline health level. A risk prediction module is used to construct and match an extreme weather scenario database based on the baseline health level and meteorological data for a preset future time period. Combined with the energy efficiency health index, it simulates and calculates vulnerability scores and generates a comprehensive risk value. A decision optimization module is used to construct a game-theoretic decision model based on the comprehensive risk value, generate a payoff matrix, configure dynamic weight coefficients, calculate the total payoff of all execution strategies, and perform Nash equilibrium to determine the optimal shutdown window and corresponding execution strategy. An optimized operation module is used to define the safe operation boundary under sub-health conditions based on the optimal shutdown window and multi-dimensional deviations, perform multi-objective Pareto optimization, select the optimal execution strategy, execute it, and monitor the system's operating status.

[0006] Optionally, the step of collecting multi-dimensional data on environment and operating conditions through IoT sensing nodes deployed on auxiliary equipment of the target thermal power plant and performing standardized preprocessing to obtain a standardized dataset includes: building a physical environment sensing network based on the differences in the layout of auxiliary equipment in the target thermal power plant, deploying various monitoring sensors, performing full-domain environmental data transmission and time-series alignment to obtain standardized environmental data; collecting operating condition data of various auxiliary equipment through the monitoring information system of the target thermal power plant, performing noise reduction and anomaly removal processing, and fusing it with the standardized environmental data to obtain an auxiliary equipment data base; based on the auxiliary equipment data base, supplementing the target power grid dispatching constraint data and the target enterprise operation and maintenance resource data, performing full data standardization conversion and storing it in a time-series database to obtain a standardized dataset.

[0007] Optionally, the step of selecting historical healthy samples from the historical standardized dataset of auxiliary equipment of the target thermal power plant, constructing an environment-performance benchmark model, and dynamically calibrating the trained environment-performance benchmark model includes: retrieving historical standardized data for a preset time period from the time-series database, using an intelligent screening program to select high-quality healthy operating samples, and performing normalization processing to obtain a golden dataset; based on the golden dataset, determining the input and output variable system of the model, selecting an ensemble learning algorithm to train the model, and iteratively optimizing the model's accuracy and generalization ability to obtain the trained environment-performance benchmark model; setting a standard sliding time window with a preset time interval, dynamically updating the training samples, performing incremental learning within a preset period, configuring an environmental mutation triggering mechanism, performing forced retraining of the model, and completing the dynamic calibration.

[0008] Optionally, the step of calculating the multidimensional deviation between the performance benchmark value and the actual operating value of the auxiliary equipment of the target thermal power plant based on the environmental-performance benchmark model, performing multidimensional deviation analysis and fault feature extraction, generating an energy efficiency health index and determining the fault type to obtain a health baseline includes: using the received operating parameters of the auxiliary equipment of the target thermal power plant as input parameters, calling the environmental-performance benchmark model, calculating and caching the performance benchmark value, and filtering out effective steady-state data; calculating multiple target deviation indicators based on the actual operating value and performance benchmark value of the auxiliary equipment of the target thermal power plant, identifying typical abnormal operating characteristics, locating hidden faults, and weighting and integrating multiple target deviation indicators into a comprehensive deviation vector; using a preset nonlinear mapping function, combining the comprehensive deviation vector to calculate the energy efficiency health index, classifying the health status of auxiliary equipment according to a preset threshold, analyzing trends to determine the fault type, and obtaining a health baseline.

[0009] Optionally, the step of constructing and matching an extreme weather scenario database based on the baseline health data and meteorological data within a preset future time period, combining the energy efficiency health index, simulating and calculating vulnerability scores, and generating a comprehensive risk value includes: connecting the system to the target meteorological service interface, receiving meteorological data within a preset future time period, calling the system's built-in extreme weather scenario database for matching, and extracting the corresponding scenario impact coefficients; setting simulation target constraints based on the energy efficiency health index, conducting full-component stress simulation calculations in conjunction with meteorological loads, and comparing the simulation calculation results with preset meteorological load thresholds to assess the remaining safety margin; calculating the vulnerability score of auxiliary equipment based on the simulation calculation results, conducting cascading failure impact analysis, and calculating the comprehensive risk value of auxiliary equipment in conjunction with the cascading failure impact coefficient.

[0010] Optionally, the step of constructing a game decision model based on comprehensive risk value, generating a payoff matrix and configuring dynamic weight coefficients, calculating the total payoff of all execution strategies, and solving for Nash equilibrium to determine the optimal shutdown window and corresponding execution strategy includes: determining multiple game participants and work objectives based on comprehensive risk value, vulnerability score, and energy efficiency health index; delineating the optional strategy space for multiple game participants; quantifying data and constructing a payoff matrix; configuring dynamic weight coefficients for the payoff matrix based on the operating conditions of the auxiliary equipment of the target thermal power plant; determining and calculating the payoff calculation formula; traversing all strategy combinations of game participants to calculate the corresponding total payoff; eliminating inferior strategies from the total payoff of all strategy combinations; solving for Nash equilibrium points by category; and determining the time period corresponding to the Nash equilibrium point as the optimal shutdown window.

[0011] Optionally, the step of defining the safe operating boundary under sub-health conditions based on the optimal shutdown window and multidimensional deviation, performing multi-objective Pareto optimization, selecting the optimal execution strategy, and executing and monitoring the system operating status includes: analyzing the dominant causes of failures in the sub-health state of auxiliary equipment based on the energy efficiency health index and multidimensional deviation, performing sensitivity analysis of auxiliary equipment for different failure types, and defining the safe operating boundary; determining the optimization objective within the safe operating boundary, performing multi-objective Pareto optimization calculation using a preset intelligent algorithm, selecting the execution strategy with the highest priority as the optimal execution strategy; sending the optimal operating parameters in the optimal execution strategy to the system for adjustment, initiating closed-loop tracking and monitoring of the system, and configuring an emergency adjustment mechanism.

[0012] Optionally, the fault warning and health management system further includes: a work order generation module, used to generate adaptive maintenance work orders containing preset content based on the optimal execution strategy and the optimal downtime window, perform resource demand prediction and inventory linkage based on the work order content, compile emergency plans in conjunction with meteorological data, and establish an execution feedback channel.

[0013] Optionally, the fault warning and health management system further includes: a feedback evolution module, used to quantify and verify the maintenance effect based on the running data after the execution of adaptive maintenance work orders, and to attribute the root causes of faults and accumulate knowledge in combination with on-site working condition feedback, so as to optimize the game decision model and the environment-performance benchmark model and complete the closed-loop iteration of the system.

[0014] On the other hand, this invention provides a method for fault early warning and health management of auxiliary equipment in thermal power plants. The method includes: collecting multi-dimensional data on environment and operating conditions through IoT sensing nodes deployed on the auxiliary equipment of the target thermal power plant, and performing standardized preprocessing to obtain a standardized dataset; selecting historical health samples from the historical standardized dataset of the auxiliary equipment of the target thermal power plant, constructing an environment-performance benchmark model, and dynamically calibrating the trained environment-performance benchmark model; calculating the multi-dimensional deviation between the performance benchmark value and the actual operating value of the auxiliary equipment of the target thermal power plant based on the environment-performance benchmark model, performing multi-dimensional deviation analysis and fault feature extraction, generating an energy efficiency health index and determining the fault type to obtain a health baseline; constructing and matching an extreme weather scenario database based on the health baseline and meteorological data within a preset future time period, combining the energy efficiency health index to simulate and calculate a vulnerability score, and generating... A comprehensive risk value is generated; a game-theoretic decision-making model is constructed based on the comprehensive risk value, a payoff matrix is ​​generated and dynamic weight coefficients are configured, the total payoff of all execution strategies is calculated, and Nash equilibrium is solved to determine the optimal downtime window and the corresponding execution strategy; based on the optimal downtime window and multidimensional deviation, the safe operating boundary under sub-health conditions is defined, and multi-objective Pareto optimization is performed to select the optimal execution strategy, execute it, and monitor the system's operating status; based on the optimal execution strategy and the optimal downtime window, an adaptive maintenance work order containing preset content is generated, resource demand is predicted and inventory is linked based on the work order content, emergency plans are prepared in conjunction with meteorological data, and an execution feedback channel is established; based on the operating data after the adaptive maintenance work order is executed, the maintenance effect is quantitatively verified, and the root cause of the failure is attributed and knowledge is accumulated in conjunction with on-site operating condition feedback to optimize the game-theoretic decision-making model and the environment-performance benchmark model, completing the system closed-loop iteration.

[0015] Through the above technical solutions, this invention constructs an environment-performance benchmark model and combines it with a sliding time window dynamic calibration mechanism to accurately distinguish between environmental impacts and equipment performance degradation, quantify the true health status of equipment, significantly reduce the false alarm rate and missed alarm rate of fault warnings, and can detect early hidden faults such as blade dust accumulation and component wear in advance. By building an extreme weather scenario database and combining it with the energy efficiency health index to conduct stress simulation and extrapolation, it achieves the goal of quantifying the load-bearing limit, failure probability, and cascading effects of sub-healthy equipment under severe weather conditions, forming risk quantification indicators such as vulnerability scores, and completing the transformation from post-event alarm to pre-event risk prediction, effectively avoiding equipment failures and unplanned unit shutdowns caused by extreme weather. By introducing game theory and solving for Nash equilibrium, combined with Pareto optimization to complete the adjustment of operating parameters with defects, it achieves the goal of coordinating the demands of equipment safety, operating costs, and grid stability, locking in the optimal maintenance window and operating strategy, getting rid of the limitations of human experience-based decision-making, minimizing downtime losses and energy consumption increases while ensuring equipment safety, and maximizing the overall benefits of the entire plant.

[0016] Other features and advantages of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating the fault early warning and health management system for auxiliary equipment in thermal power plants according to the present invention. Figure 2 This is a flowchart illustrating the method for early warning and health management of auxiliary equipment in thermal power plants according to the present invention. Figure 1 ; Figure 3 This is a schematic diagram of the process for obtaining the standardized dataset in this invention; Figure 4 This is a schematic diagram of the process for constructing the environment-performance benchmark model in this invention; Figure 5 This is a schematic diagram of the process for obtaining the baseline health data in this invention; Figure 6 This is a schematic diagram of the process for generating a comprehensive risk value in this invention; Figure 7 This is a flowchart illustrating the process of determining the optimal shutdown window and the corresponding execution strategy in this invention. Figure 8 This is a flowchart illustrating the process of selecting the optimal execution strategy in this invention; Figure 9 This is a flowchart illustrating the method for early warning and health management of auxiliary equipment in thermal power plants according to the present invention. Figure 2 . Detailed Implementation

[0018] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0019] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with the relevant provisions of national laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0020] Please refer to Figure 1This invention provides a fault early warning and health management system for auxiliary equipment in thermal power plants. The system includes: a data acquisition and processing module for collecting multi-dimensional environmental and operational data via IoT sensing nodes deployed on auxiliary equipment in the target thermal power plant, and performing standardized preprocessing to obtain a standardized dataset; a model building and calibration module for selecting historical health samples from the historical standardized dataset of the auxiliary equipment in the target thermal power plant, constructing an environment-performance benchmark model, and dynamically calibrating the trained model; a quantitative analysis module for calculating the multi-dimensional deviation between the performance benchmark value and the actual operating value of the auxiliary equipment in the target thermal power plant based on the environment-performance benchmark model, performing multi-dimensional deviation analysis and fault feature extraction, generating an energy efficiency health index, determining the fault type, and obtaining a health baseline; and a risk prediction module for constructing and matching an extreme weather scenario database based on the health baseline and meteorological data within a preset future time period, and combining the energy efficiency health index to simulate and calculate a vulnerability score. The system generates a comprehensive risk value; a decision optimization module is used to construct a game-theoretic decision model based on the comprehensive risk value, generate a payoff matrix and configure dynamic weight coefficients, calculate the total payoff of all execution strategies, and perform Nash equilibrium to determine the optimal downtime window and corresponding execution strategy; an optimized operation module is used to define the safe operation boundary under sub-health conditions based on the optimal downtime window and multi-dimensional deviation, perform multi-objective Pareto optimization, select the optimal execution strategy, execute and monitor the system operation status; a work order generation module is used to generate adaptive maintenance work orders containing preset content based on the optimal execution strategy and optimal downtime window, perform resource demand prediction and inventory linkage based on the work order content, compile emergency plans in conjunction with meteorological data, and establish an execution feedback channel; a feedback evolution module is used to quantify and verify the maintenance effect based on the operation data after the execution of adaptive maintenance work orders, combine on-site operating condition feedback to attribute the root causes of failures and accumulate knowledge, so as to optimize the game-theoretic decision model and the environment-performance benchmark model and complete the closed-loop iteration of the system.

[0021] In this embodiment of the invention, the data acquisition and processing module can be used to collect multi-dimensional data on the environment and operating conditions through IoT sensing nodes deployed on the auxiliary equipment of the target thermal power plant, and perform standardized preprocessing to obtain a standardized dataset.

[0022] Please refer to Figure 3 In a preferred embodiment of the present invention, multi-dimensional data on the environment and operating conditions are collected by IoT sensing nodes deployed on auxiliary equipment of the target thermal power plant, and standardized preprocessing is performed to obtain a standardized dataset, which may include: Step S110: Based on the differences in the auxiliary equipment layout of the target thermal power plant, build a physical environment sensing network, deploy various monitoring sensors, perform full-domain environmental data transmission and time-series alignment, and obtain standardized environmental data.

[0023] In a preferred embodiment of the invention, an integrated high-precision sensor can be installed on the core meteorological tower of the target thermal power plant area to continuously collect three basic meteorological parameters—atmospheric temperature, relative humidity, and standard atmospheric pressure—around the clock, serving as the basis for environmental analysis. For special operating scenarios such as coastal plants and chemical industrial parks, additional dust concentration detectors, salt spray collection devices, and acid rain / pH online monitoring equipment are installed to focus on capturing the content of corrosive substances such as chloride ions and sulfides in the air. For outdoor auxiliary equipment such as air-cooled islands, cooling towers, and outdoor forced draft fans, dedicated wind speed and direction sensors are added to monitor changes in natural wind load in real time.

[0024] In a preferred embodiment of the present invention, all raw environmental data collected by the front-end sensors are uniformly aggregated and transmitted to the edge gateway device. The gateway performs millisecond-level timestamp verification and timing alignment operations on all accessed environmental data to fundamentally avoid data timing misalignment caused by signal transmission delay and asynchronous clocks of front-end devices, ensuring that the environmental data produced in this step has a unified time reference and obtains standardized environmental data.

[0025] Step S120: Collect operating condition data of various auxiliary machines through the monitoring information system of the target thermal power plant, perform noise reduction and anomaly removal processing, and integrate with standardized environmental data to obtain the auxiliary machine equipment data base.

[0026] In a preferred embodiment of the present invention, the data collection scope can be divided into three major categories of core parameter systems: electrical parameters include the three-phase current of the motor stator, the active power, reactive power, and the operating power factor of the equipment, which can directly reflect the electrical load of the auxiliary equipment and the operating status of the electrical circuit; thermal parameters cover the inlet and outlet pressure of the medium, the operating pressure difference, the real-time flow rate, the opening degree of the regulating valve, and the operating temperature of the medium, which intuitively reflect the medium conveying capacity and thermal operation characteristics of the auxiliary equipment; mechanical parameters cover the bearing temperature of the equipment, the winding temperature of the motor, the lubricating oil temperature, and the radial vibration amplitude of the shaft.

[0027] In a preferred embodiment of the present invention, the system can incorporate a professional digital filtering algorithm and a signal dead zone processing mechanism, which can automatically identify and eliminate abnormal pulse data generated by instantaneous signal jumps and electromagnetic interference, thereby maximizing the authenticity and validity of the original operating condition data. The system binds each operating condition data to corresponding environmental data according to a unified timestamp, realizing a one-to-one correspondence between environmental parameters and equipment parameters. All collected operating condition data and standardized environmental data are deeply integrated to jointly form the auxiliary equipment data base for auxiliary machine health assessment.

[0028] Step S130: Based on the auxiliary equipment data base, supplement the target power grid dispatching constraint data and the target enterprise operation and maintenance resource data, perform full data standardization transformation and store it in the time series database to obtain a standardized dataset.

[0029] In a preferred embodiment of the present invention, the system can capture information such as regional power grid time-of-use pricing, power auxiliary service compensation standards, power grid peak-shaving instructions, and load regulation requirements in real time through the power plant's dedicated dispatch data network. Combined with the operating status of auxiliary equipment, it can accurately calculate the electricity cost and downtime opportunity cost corresponding to different maintenance periods and different operating conditions. The system connects to the Enterprise Resource Planning (ERP) management system to retrieve material management data such as spare parts inventory quantity, material purchase unit price, supplier delivery cycle, and spare parts allocation process in real time. At the same time, it integrates human resource data such as maintenance team personnel scheduling, personnel skill level, number of on-duty personnel, and manpower load. On the edge gateway side, the collected environmental data, equipment operating data, power grid dispatch data, and maintenance resource data are uniformly converted in format and standardized in measurement units. All data are converted to international standard measurement units and written into the time-series database according to two storage frequencies: second-level and minute-level. This completes the digital holographic mapping of the physical entities, external environment, operating conditions, economic dispatch, and maintenance resources corresponding to the auxiliary equipment of the thermal power plant, and finally forms a standardized dataset.

[0030] In this embodiment of the invention, the model building and calibration module can be used to screen historical health samples from the historical standardized dataset of auxiliary equipment of the target thermal power plant, build an environment-performance benchmark model, and dynamically calibrate the trained environment-performance benchmark model.

[0031] Please refer to Figure 4 In a preferred embodiment of the present invention, selecting historical health samples from the historical standardized dataset of auxiliary equipment of the target thermal power plant to construct an environment-performance benchmark model, and dynamically calibrating the trained environment-performance benchmark model, may include: Step S210: Retrieve historical standardized data for a preset time period from the time series database, use an intelligent filtering program to select high-quality healthy operating samples, and perform normalization processing to obtain the golden dataset.

[0032] In a preferred embodiment of the present invention, firstly, the full historical standardized data of the past year stored in the time-series database is directly retrieved, and an intelligent screening program is started. The overall screening work strictly follows the three core principles of steady state, no disturbance, and optimality. Transient fluctuation data generated during the unit start-up and shutdown phase and the phase of large load adjustment are prioritized to be removed. Such data only represents the transitional state of the equipment and cannot reflect the stable operating performance. Secondly, severe disturbance data under sudden operating conditions such as grid faults, equipment tripping, and system anomalies are removed to avoid abnormal operating condition data from polluting the sample database.

[0033] Secondly, after removing all invalid data, the system further filters valid data segments, retaining only high-quality data segments where the equipment operates stably, all operating parameters fluctuate slightly within the rated range, and the comprehensive energy efficiency index ranks in the top 20% percentile of historical data. This part of the data corresponds to the auxiliary machine's fault-free and optimal performance operating state, and is the most ideal sample for building the benchmark model. It can restore the inherent performance of the equipment to the greatest extent and eliminate interference caused by faults and degradation.

[0034] Finally, global normalization is performed on the selected high-quality samples to eliminate data differences caused by different sensor ranges and parameter dimensions, so that all parameters are in a unified calculation dimension. Then, a clustering algorithm is used to intelligently identify and remove discrete outliers to further purify the samples. The final pure dataset is defined as the golden dataset, thus avoiding the problem of the model "learning with defects" from the source.

[0035] Step S220: Based on the Golden Data Set, determine the input and output variable system of the model, select an ensemble learning algorithm to train the model, and iteratively optimize the model's accuracy and generalization ability to obtain the trained environment-performance benchmark model.

[0036] In a preferred embodiment of the present invention, the model input end uses all kinds of collected environmental parameters (e.g., including external factors such as atmospheric temperature, relative humidity, atmospheric pressure, dust concentration, and wind speed), and also accesses the real-time load command of the device as an auxiliary input parameter; the model output end selects two core performance indicators: unit power consumption output flow rate and comprehensive operating efficiency, which have high fault sensitivity and can intuitively reflect the operating status of the device.

[0037] Next, relying on the Golden Data Set, mainstream ensemble learning algorithms such as gradient boosting decision trees and random forests are used to build a nonlinear mapping relationship between environmental parameters, equipment load commands and auxiliary machine core performance indicators. During the training process, the algorithm will autonomously and deeply explore the inherent physical laws between environmental variables and equipment performance. For example, it will analyze the objective physical characteristics such as the decrease in air density caused by the increase in ambient temperature, which in turn causes the effective output of the blower to decrease, so that the model fits the actual operating mechanism of the equipment.

[0038] It should be noted that the training process takes minimizing the residual between the model's predicted values ​​and historical actual operating conditions as its core objective, iteratively optimizing the internal weight parameters of the algorithm, and focusing on strengthening the model's generalization ability to ensure that the model can not only adapt to historical operating conditions that have occurred, but also accurately predict entirely new operating scenarios that have never occurred before.

[0039] Step S230: Set a standard sliding time window with a preset time interval, dynamically update the training samples, perform incremental learning within a preset period, configure an environmental mutation trigger mechanism, perform forced model retraining, and complete dynamic calibration.

[0040] In a preferred embodiment of the present invention, in response to the irreversible performance degradation problems such as auxiliary impeller wear, coating peeling, and component aging, the system uniformly sets 30 days as a standard sliding time window. The system automatically supplements the latest operating data that is determined to be in a healthy state within the past 30 days into the model training set every day, while simultaneously removing the oldest historical data of the past 30 days outside the window, so as to continuously ensure that the training samples always fit the recent real operating state of the equipment and avoid model lag caused by old data. The system performs an incremental learning operation once a week, fine-tuning the internal weight parameters of the model based on the updated sample set. This accurately adapts to the performance degradation caused by the slow aging of the device, solves the problem of static models gradually becoming inaccurate over time, and ensures that the theoretical value output by the model always matches the current hardware status of the device. The system monitors the prediction error of the environment-performance benchmark model in real time. When factors such as seasonal changes and extreme weather cause significant changes in environmental parameters, resulting in the error of multiple consecutive sets of model data exceeding the preset threshold, the system will automatically trigger the model's forced retraining process. After multiple calibrations and optimizations, the finished model is freed from the limitations of static historical data and can output accurate benchmark values ​​stably over a long period of time.

[0041] In this embodiment of the invention, the quantitative analysis module can be used to calculate the multidimensional deviation between the performance benchmark value and the actual operating value of the auxiliary equipment of the target thermal power plant according to the environmental-performance benchmark model, perform multidimensional deviation analysis and fault feature extraction, generate an energy efficiency health index and determine the fault type, and obtain the health baseline.

[0042] Please refer to Figure 5 In a preferred embodiment of the present invention, the multidimensional deviation between the performance benchmark value and the actual operating value of the auxiliary equipment of the target thermal power plant is calculated based on the environmental-performance benchmark model. Multidimensional deviation analysis and fault feature extraction are performed to generate an energy efficiency health index and determine the fault type, thus obtaining a health baseline. This process may include: Step S310: Using the received auxiliary equipment operating parameters of the target thermal power plant as input parameters, call the environment-performance benchmark model, calculate and cache the performance benchmark values, and filter out the valid steady-state data.

[0043] In a preferred embodiment of the present invention, the system receives real-time, continuously collected current full-volume environmental data and equipment load commands, uses these two types of data as fixed input parameters, and instantly invokes the environmental-performance benchmark model that has completed full-dimensional dynamic calibration. The model combines real-time atmospheric temperature, atmospheric pressure, dust concentration, wind speed and other environmental conditions, as well as the current equipment load status, to quickly calculate the theoretical efficiency, theoretical power consumption and other standard performance indicators of the auxiliary machine under the current operating conditions. These values ​​are the ideal operating parameters under the condition of the equipment being in perfect health. The system caches the theoretical benchmark values ​​output by the model locally to ensure data retention. The system performs normalization processing on the actual operating data collected on-site, in the same manner as during the model training phase, unifying data dimensions and calculation standards. Simultaneously, the system automatically identifies transient operating conditions such as sudden changes in grid load and command switching, proactively eliminating such transient fluctuation data and selecting only steady-state data from stable equipment operation to perform comparative analysis between theoretical and actual values.

[0044] Step S320: Based on the actual operating values ​​and performance benchmark values ​​of the auxiliary equipment of the target thermal power plant, calculate multiple target deviation indicators (e.g., power deviation, flow deviation, and efficiency deviation), identify typical abnormal operating characteristics, locate hidden faults, and integrate multiple target deviation indicators into a comprehensive deviation vector with weights.

[0045] In a preferred embodiment of the present invention, the power deviation is the actual operating power of the equipment minus the theoretical power output by the model, the flow deviation is the actual medium flow rate minus the theoretical standard flow rate, and the efficiency deviation is the actual comprehensive operating efficiency of the equipment minus the theoretical efficiency predicted by the model. The system focuses on monitoring the core feature of positive power deviation. When the output flow rate of the equipment remains unchanged but the power consumption increases abnormally, it usually indicates typical hidden faults such as increased internal mechanical friction, flow channel blockage, and dust accumulation on the blades.

[0046] In a preferred embodiment of the present invention, the system assigns differentiated weights to different deviation indicators based on the fluctuation amplitude, rate of change, and impact weight on equipment failure of various deviations, and integrates the scattered deviation data to generate a comprehensive deviation vector, which can be expressed by the following formula. :

[0047] in, Indicates power deviation. Indicates flow deviation. This indicates efficiency deviation. This represents a fixed weighting factor for power deviation. This represents a fixed weighting coefficient for flow deviation. This represents a fixed-weight coefficient for efficiency deviation. It is determined by training with historical fault samples of the unit.

[0048] Step S330: Using a preset nonlinear mapping function and combined with the comprehensive deviation vector, calculate the energy efficiency health index, classify the health status of auxiliary equipment according to the preset threshold, analyze the trend to determine the fault type, and obtain the health baseline.

[0049] In a preferred embodiment of the present invention, the energy efficiency health index can be represented by the following formula. :

[0050] in, 1 indicates complete health, 0 indicates equipment failure. This represents the attenuation adjustment coefficient, which is a calibration constant determined by the unit type and the rated parameters of the auxiliary equipment.

[0051] Next, the system sets four fixed threshold levels to complete the health level classification: when It is judged to be in excellent health; when It is judged to be in good health; when It is determined to be a fault warning state; when This triggers a device fault alarm, clearly classifying the risk level of device operation and making it convenient for maintenance personnel to intuitively identify the status.

[0052] Finally, the system retrieves the historical time-series change curve of the energy efficiency health index and distinguishes fault types by analyzing the changes in the curve slope: a sudden and significant increase in the curve slope indicates an acute, sudden fault; a slow and gradual decline in the curve over a long period indicates a gradual, chronic performance degradation of the equipment. The final generated energy efficiency health index, health level, and fault type determination results together constitute the current health baseline of the auxiliary equipment.

[0053] In one specific embodiment, the blower was operating at 85% load. Based on the current ambient temperature of 34°C and dust concentration of 150 mg / m³, the model calculated a theoretical efficiency of 85.5%. However, actual monitoring showed an efficiency of only 81.2%, resulting in a negative deviation of 4.3%. System analysis revealed a power deviation of +40kW and a flow rate deviation of -5 kg / s. This indicates that the blower was performing unproductive work (e.g., blocked). Based on this, the system calculated an EHI of 0.76, within the warning range. Reviewing historical trends, the index had slowly decreased from 0.92 to 0.76 over the past two weeks. The system determined this to be a chronic blockage fault, rather than a sensor transient, and marked this state for future use.

[0054] In this embodiment of the invention, the risk prediction module can be used to construct and match an extreme weather scenario database based on the baseline health data and meteorological data within a preset future time period, combine the energy efficiency health index, simulate and calculate the vulnerability score, and generate a comprehensive risk value.

[0055] Please refer to Figure 6 In a preferred embodiment of the present invention, based on baseline health data and meteorological data within a preset future time period, an extreme weather scenario database is constructed and matched. Combined with an energy efficiency health index, a vulnerability score is simulated and calculated, and a comprehensive risk value is generated. This may include: Step S410: Connect the system to the target meteorological service interface, receive meteorological data for the future preset time period, call the system's built-in extreme weather scenario library for matching, and extract the corresponding scenario impact coefficient.

[0056] In a preferred embodiment of the present invention, the system connects to a professional meteorological service interface to obtain full-dimensional meteorological information for the next 72 hours, specifically including core data such as regional temperature extremes, instantaneous and average wind speeds, cumulative precipitation, probability of thunderstorms, and salt spray intensity, so as to fully grasp the future meteorological change trend of the factory area and use it as the external boundary condition for risk analysis. The system has a built-in standardized extreme weather scenario database, which pre-enters severe weather scenarios that thermal power plants frequently encounter, such as typhoons, strong sandstorms, extreme cold and freezing, high temperature and humidity, and heavy rainfall. At the same time, it presets corresponding environmental impact coefficients for each type of scenario, including parameters such as wind load increase coefficient, air density correction coefficient, and corrosion acceleration coefficient. The system compares real-time weather forecast data with various features in the scenario database one by one to automatically identify the types of extreme weather that the plant area will face in the future. For a successfully matched meteorological scenario, the system retrieves the corresponding full set of impact coefficients to clarify the extent to which extreme weather changes the equipment's operating load and operating environment. The entire set of meteorological analysis results serves as fixed boundary conditions.

[0057] Step S420: Set simulation target constraints based on the energy efficiency health index, conduct full component stress simulation calculations in conjunction with meteorological loads, and compare the simulation calculation results with the preset meteorological load thresholds to assess the remaining safety margin.

[0058] In a preferred embodiment of the present invention, firstly, the energy efficiency health index is used as the core constraint parameter of the equipment's carrying capacity, and a unified judgment rule is established: when At that time, the load-bearing capacity of core components such as bearings, blades, and motors is 100% of the rated capacity; when the equipment is in a sub-healthy state, When the value is too low, the load-bearing capacity of the component will be reduced proportionally, truly reflecting the actual situation of the reduced load-bearing capacity after the equipment deteriorates.

[0059] Secondly, combining the matched extreme weather scenarios and environmental impact coefficients, external additional loads are superimposed on the simulation model. Focusing on the three core load-bearing components—auxiliary bearings, blades, and drive motors—the system accurately calculates parameters such as mechanical stress, operating temperature, and vibration amplitude under extreme weather conditions, comprehensively simulating the equipment's operating status in harsh environments. Finally, the system compares the simulated actual load data with the actual load data obtained through simulation. By comparing the reduced load-bearing capacity of components, the operational stability of sub-health equipment under severe weather conditions can be determined, and the remaining safety margin of the equipment can be accurately calculated.

[0060] Step S430: Calculate the vulnerability score of auxiliary equipment based on the simulation results, conduct cascading failure impact analysis, and calculate the comprehensive risk value of auxiliary equipment based on the cascading failure impact coefficient.

[0061] In a preferred embodiment of the present invention, the system determines the probability of auxiliary equipment failure under future extreme weather conditions based on equipment load and safety margin data derived from stress simulation. A vulnerability score is generated according to the probability value, and the vulnerability score is directly proportional to the equipment failure probability. A higher score indicates a greater operational risk for a single piece of equipment, directly reflecting the equipment's resilience. The vulnerability score can be expressed by the following formula. :

[0062]

[0063] in, This indicates that the maximum stress of the component was obtained from the extreme weather simulation. Indicates The reduced allowable ultimate stress of the component This indicates the rated load-bearing stress of brand-new equipment; the vulnerability score range is [range missing]. The value ranges from 0 to 1, with higher values ​​indicating higher risks.

[0064] In a preferred embodiment of the present invention, after completing the risk assessment of a single device, the system further deduces the fault propagation path, analyzes whether a sudden failure of a single auxiliary machine will cause the main generator unit to reduce its load and shut down urgently, accurately calculates the load reduction of the unit, and assesses the cascading impact of the fault on the continuity of power supply and operational stability of the regional power grid, quantifying the overall loss caused by the fault. The system multiplies the vulnerability score by the cascading fault impact coefficient to calculate the comprehensive risk value of the equipment. This result not only clarifies the probability of equipment failure and the estimated time of failure, but also quantifies the economic losses and power grid impact caused by the failure, clearly defining the risk level and the urgency of handling.

[0065] In a preferred embodiment of the present invention, the comprehensive risk value can be expressed by the following formula. :

[0066] in, This represents the cascading failure impact coefficient, indicating the rate at which auxiliary equipment failure causes load reduction and grid loss in the generating unit, and is preset based on the generating unit capacity.

[0067] In one specific embodiment, a weather forecast indicates that a typhoon will make landfall in 48 hours, with gusts reaching 30 m / s in the factory area. The system retrieves data from the ventilation fan. The vulnerability score is 0.76 (e.g., with existing blockage). Simulation results show that the fan can withstand a wind speed of 30 m / s under normal conditions, but due to the blockage causing unstable airflow and uneven stress on the blades, the actual load-bearing limit drops to 25 m / s. The system determines a 90% probability of failure during a typhoon. The cascading effects are also assessed: a failure of this blower will lead to a 50% reduction in the load of auxiliary units, resulting in significant power loss. The system generates a vulnerability score of 0.9, marking it as an urgent risk, and recommends that it must be addressed before the typhoon arrives.

[0068] In this embodiment of the invention, the decision optimization module can be used to construct a game decision model based on the comprehensive risk value, generate a payoff matrix and configure dynamic weight coefficients, calculate the total payoff of all execution strategies, and perform Nash equilibrium solution to determine the optimal shutdown window and the corresponding execution strategy.

[0069] Please refer to Figure 7 In a preferred embodiment of the present invention, a game-theoretic decision-making model is constructed based on a comprehensive risk value, a payoff matrix is ​​generated and dynamic weight coefficients are configured, the total payoff of all execution strategies is calculated, and a Nash equilibrium is obtained to determine the optimal shutdown window and the corresponding execution strategy. This may include: Step S510: Based on the comprehensive risk value, vulnerability score and energy efficiency health index, identify multiple game participants and work objectives, delineate the optional strategy space of multiple game participants, quantify the data and construct the payoff matrix.

[0070] In a preferred embodiment of the present invention, the multiple game participants and their objectives may include: the first party is the equipment manager, whose core objective is to ensure the safe operation of the equipment and minimize the probability of failure; the second party is the operator, whose core objective is to control the operating costs of the equipment throughout its entire life cycle and reduce the economic losses caused by maintenance and downtime; and the third party is the power grid dispatcher, whose core objective is to ensure the continuous power supply and stable load of the regional power grid. The objectives of the three parties have a natural conflict of interest, which is also the core premise of game analysis.

[0071] It should be noted that the equipment manager's strategy space is fixed at two categories: immediate maintenance and delayed maintenance; the operator's strategy space is to approve maintenance budgets and reject maintenance budgets; and the grid dispatcher's strategy space is to agree to equipment shutdown for maintenance and refuse equipment shutdown for maintenance. The system converts the energy efficiency health index, vulnerability score, and comprehensive risk value into quantitative benefit values ​​under different strategy combinations, and fully fills the three-party game benefit matrix according to the logic of subject, strategy, and benefit, thus building a complete decision analysis framework.

[0072] Step S520: Configure dynamic weight coefficients for the payoff matrix based on the operating conditions of the auxiliary equipment of the target thermal power plant, determine and calculate the payoff formula, and traverse all strategy combinations of the game participants to calculate the corresponding total payoff.

[0073] In a preferred embodiment of the present invention, the system can automatically adjust the weights of the three parties based on real-time operating conditions and risk levels: when the equipment is determined to be extremely high-risk and its vulnerability score exceeds the standard, the weight of the equipment manager is automatically increased to prioritize equipment safety; when the power grid is in peak electricity consumption and time-of-use electricity prices are in a high range, the weight of the operator is increased to focus on controlling economic losses from downtime; when the regional power grid has insufficient reserve capacity and the power supply pressure is high, the weight of the power grid dispatcher is increased to prioritize power grid stability.

[0074] In a preferred embodiment of the present invention, the total revenue can be expressed by the following formula. :

[0075] in, This indicates the safety benefits of equipment, which are negatively correlated with the probability of failure. This indicates operating cost and revenue, which is negatively correlated with maintenance and downtime losses. This indicates that the revenue from power grid dispatching is negatively correlated with the magnitude of power grid load fluctuations. This indicates the dynamic weight of the equipment manager. Indicates the dynamic weight of the operator. Represents the dynamic weights of the power grid dispatcher, and the constraints are... The weights are adaptively adjusted in real time based on risk, electricity price, and grid load.

[0076] In a preferred embodiment of the present invention, the system automatically traverses all strategy combinations of the three parties, calculates the corresponding total global benefit for each strategy, and fully simulates the decision-making logic of offline collaborative negotiation among multiple departments of an enterprise, replacing manual experience judgment with quantitative data.

[0077] Step S530: Eliminate the inferior strategy from the total return of all strategy combinations, solve for the Nash equilibrium point by category, and determine the time period corresponding to the Nash equilibrium point as the optimal shutdown window.

[0078] In a preferred embodiment of the present invention, the system, based on the calculated total payoff of all strategy combinations, gradually eliminates strategy combinations with low payoff and obvious disadvantages through an iterative algorithm, continuously narrowing the range of selectable strategies, reducing ineffective calculations, focusing on feasible decision-making schemes, and improving solution efficiency. The system can perform calculations in two scenarios: if a pure strategy equilibrium result is obtained after iteration, the corresponding strategy combination is directly locked; if a pure strategy equilibrium is not obtained, a mixed strategy equilibrium is further calculated, and the optimal execution probability corresponding to different maintenance periods is output. Both solution logics strictly follow game theory rules to ensure that the results are objective and rigorous. The time period corresponding to the Nash equilibrium point is defined as the optimal shutdown window recommended by the system. This window simultaneously considers the triple constraints of equipment operation safety, enterprise operating costs, and power grid power supply stability, thereby maximizing global benefits.

[0079] In one specific implementation, under a typhoon warning, the equipment manager advocates immediate maintenance (high benefit), while the operator worries about losing 500,000 yuan in electricity costs due to daytime downtime (low benefit), and the grid dispatcher refuses to shut down due to the evening peak. The system calculates the benefit matrix: if maintenance is forced during the day, the operator's benefit is extremely low; if maintenance is delayed until 2 AM (off-peak period), all three parties benefit more. The system dynamically adjusts the weights: as the typhoon approaches, the equipment's weight spikes. The final Nash equilibrium is: shutdown for maintenance between 2:00 AM and 4:00 AM. At this time, the grid load is low, electricity prices are low, and equipment can be repaired before the typhoon. The system locks this window and notifies all parties.

[0080] In this embodiment of the invention, the optimization operation module can be used to define the safe operation boundary under sub-health conditions based on the optimal shutdown window and multi-dimensional deviation, perform multi-objective Pareto optimization, select the optimal execution strategy, execute and monitor the system operation status.

[0081] Please refer to Figure 8 In a preferred embodiment of the present invention, based on the optimal shutdown window and multidimensional deviation, the safe operating boundary under sub-health conditions is defined, multi-objective Pareto optimization is performed, the optimal execution strategy is selected, and the system operating status is executed and monitored. This may include: Step S610: Based on the energy efficiency health index and multidimensional deviation, analyze the main causes of failures in the sub-health state of auxiliary equipment, conduct sensitivity analysis of auxiliary equipment for different failure types, and delineate safe operation boundaries.

[0082] In a preferred embodiment of the present invention, the extracted multidimensional deviation features and energy efficiency health index are used as the core basis, and combined with the determined waiting time for maintenance, the core deterioration causes of the equipment in a sub-healthy state are comprehensively analyzed. If the diagnosis result is that the efficiency is reduced due to blockage of the flow channel or filter, then blockage is identified as the dominant fault; if it is determined that the mechanical friction is increased due to wear of the bearing or shaft, then component wear is identified as the core cause, and the fault type is accurately distinguished. Then, for different fault types, operating condition sensitivity analysis can be carried out: for blockage faults, the vibration change pattern of the equipment under high speed conditions is identified; for wear faults, the component temperature rise pattern of the equipment under high torque conditions is monitored, so as to fully understand the deterioration rate of the equipment under different operating intensities and identify the shortcomings of the equipment operation. Finally, by combining fault characteristics, energy efficiency health index values, and waiting time, the system accurately delineates the safe operating boundaries of the equipment under its current state, clarifies the hard upper limit thresholds of operating parameters such as speed, load, and frequency, and plots the correlation curve between health status and allowable load of the equipment, clearly marking all feasible operating ranges. The entire set of boundary data serves as a constraint condition.

[0083] Step S620: Within the safe operating boundary, determine the optimization objective, use a preset intelligent algorithm to perform multi-objective Pareto optimization calculation, and select the execution strategy with the highest priority as the optimal execution strategy.

[0084] In a preferred embodiment of the present invention, firstly, within the defined safe operating range, the system sets dual optimization objectives. Objective one is to minimize the equipment wear rate, delay the continuous deterioration of the fault, and ensure that the equipment can operate smoothly until the determined optimal maintenance window. Objective two is to strictly control the additional energy consumption of the equipment and minimize the increase in electricity costs during the faulty operation phase. The two objectives are mutually restrictive and cannot be simultaneously optimized. Therefore, Pareto optimization logic is used to solve the problem. Secondly, the system uses genetic algorithm and multi-objective particle swarm optimization algorithm to carry out multi-objective Pareto optimization calculation within the range of all feasible operating parameters, and solves the complete Pareto optimal solution set. Each set of parameters in the solution set corresponds to an independent operating strategy, including specific parameters such as equipment operating frequency, valve opening, and load set value. Finally, following the core screening principle of "large reduction in wear and small increase in energy consumption", the system selects the parameter combination with the best overall performance from the Pareto optimal solution set, and finds the best balance between equipment safety and economic operation as the optimal execution strategy.

[0085] Step S630: Send the optimal operating parameters from the optimal execution strategy to the system for adjustment, start closed-loop tracking and monitoring of the system, and configure an emergency adjustment mechanism.

[0086] In a preferred embodiment of the present invention, the adjustment sent to the system may include lowering the operating frequency of the fan inverter, reducing the rated speed of the equipment, and fine-tuning the valve opening. After the parameter adjustment is completed, the system starts 24 / 7 closed-loop tracking and monitoring, continuously collecting key data such as real-time energy efficiency health index, equipment energy consumption, bearing temperature, and vibration amplitude, and judging in real time whether the energy efficiency health index has stopped declining rapidly and whether the equipment deterioration trend has been effectively curbed. At the same time, it checks whether the increase in energy consumption is within the preset acceptable range.

[0087] It should be noted that if the total load demand in the plant area surges temporarily, the system will activate an emergency mechanism, which can temporarily exceed the optimized set value to meet production requirements. After the load subsides, the system will automatically restore the optimal operating parameters. The entire dynamic adjustment mechanism ensures that the equipment smoothly transitions to the maintenance window.

[0088] In this embodiment of the invention, the work order generation module can be used to generate an adaptive maintenance work order containing preset content based on the optimal execution strategy and the optimal downtime window, perform resource demand prediction and inventory linkage based on the work order content, compile emergency plans in conjunction with meteorological data, and establish an execution feedback channel.

[0089] In a preferred embodiment of the present invention, the deviation type and root cause of the fault are retrieved from the fault diagnosis, and combined with the recorded equipment deterioration characteristics, and then connected to the system's built-in standardized maintenance strategy library. If the fault is diagnosed as a filter or flow channel blockage, a dust removal or cleaning operation strategy is matched; if the fault is determined to be a bearing or shaft wear, a clearance measurement or component replacement strategy is matched, thus achieving a precise correspondence between the fault and the operation content. Secondly, the system retrieves the vulnerability score output to determine the risk level of on-site operations and adds corresponding safety protection requirements to the work order. For example, in high-risk scenarios, specific clauses such as "two-person on-site supervision", "prepare emergency spare parts in advance" and "prohibit single-person high-altitude operations" are marked to comprehensively protect the personal safety of on-site maintenance personnel and the safety of equipment.

[0090] In a preferred embodiment of the present invention, the system can automatically perform resource statistics based on the various maintenance operations completed after disassembly: for example, matching professional maintenance personnel such as fitters, electricians, and welders according to the work procedures, and calculating the on-the-job hours required for each type of work; at the same time, sorting out the necessary tools for the work, including special tools such as hydraulic pullers, feeler gauges, vibration analyzers, and hoisting equipment; and then, according to the type of failure, compiling the model, specifications, and quantity of spare parts to form a complete resource list. The system automatically connects to the power plant's ERP material management system to query the real-time inventory quantity, storage location, and requisition process of all spare parts in the list. If the inventory quantity of spare parts cannot meet the maintenance needs, the system automatically generates a material transfer application or an emergency procurement application. Finally, based on the determined maintenance window as the final time node, the system works backward to calculate the required time for material transfer, procurement, and personnel scheduling, and reserves additional emergency buffer time to prevent missing the best maintenance window due to insufficient resources.

[0091] In a preferred embodiment of the present invention, a specific emergency plan can be prepared for the maintenance work order based on the predicted extreme weather risks and secondary risks of equipment failure. For example, if typhoons or strong winds occur during the maintenance period, the work order can clearly indicate actionable emergency response rules such as "stop outdoor work immediately, close the equipment windproof cover, and transfer personnel to a safe indoor area" to deal with sudden risks. The system establishes a mobile work feedback channel, allowing on-site maintenance personnel to report work progress in real time, upload on-site photos of equipment, and record newly discovered anomalies during the work through handheld terminals, realizing digital traceability of the entire maintenance process. If the actual severity of the fault reported on-site far exceeds the previous diagnosis or the fault type changes, the system will immediately re-trigger the game decision model, recalculate the payoff matrix and Nash equilibrium based on the new fault state, and dynamically adjust the maintenance plan.

[0092] In this embodiment of the invention, the feedback evolution module can be used to quantify and verify the maintenance effect based on the running data after the execution of the adaptive maintenance work order, and combine the feedback of the on-site working conditions to attribute the root cause of the fault and accumulate knowledge, so as to optimize the game decision model and the environment-performance benchmark model and complete the closed-loop iteration of the system.

[0093] In a preferred embodiment of the present invention, after all maintenance work is completed and the auxiliary equipment is reconnected to the grid and running stably, the system continuously collects all parameters of the equipment's steady-state operation after restarting, ensuring that the parameter categories, collection standards, and data collection and processing are completely consistent. Then, the energy efficiency health index is recalculated and compared horizontally with the recorded energy efficiency health index value before maintenance to accurately calculate the equipment health recovery rate. The system classifies the effectiveness level according to the recovery rate: a recovery rate higher than 95% indicates that the maintenance was completely successful, a recovery rate of 50% to 95% indicates that the maintenance was basically completed, and a recovery rate lower than 50% indicates that the maintenance failed.

[0094] In a preferred embodiment of the present invention, the typical fault identification rules summarized are entered into the global fault feature library to continuously strengthen the automatic fault identification logic of the system. For example, practical rules such as "the overall efficiency of the equipment decreases and the motor current fluctuates irregularly, and the probability of blade dust accumulation fault reaches 80%" are summarized and recorded to continuously enrich the system knowledge base.

[0095] In a preferred embodiment of the present invention, the system feeds back the recorded economic data, such as labor costs, spare parts procurement costs, and power loss during maintenance, as well as the summarized fault characteristic data, to the game-theoretic decision-making model. For maintenance costs and downtime losses corresponding to similar faults, the system dynamically adjusts the revenue weights of the equipment manager, operator, and grid dispatcher, optimizing the decision-making logic for subsequent similar faults. The system also supplements the model training set with the health operation data of the equipment after maintenance as a new golden dataset, fine-tuning the model's internal weight parameters to ensure the environment-performance benchmark model accurately adapts to the new physical state of the equipment after maintenance. Finally, relying on continuous data feedback, sample updates, and parameter corrections, the entire technical solution achieves closed-loop self-optimization, continuously improving the comprehensive capabilities of fault prediction, health assessment, risk evaluation, and decision optimization.

[0096] Please refer to Figure 2 and Figure 9 On the other hand, the present invention also provides a method for early warning and health management of auxiliary equipment in thermal power plants, which may include: Step S100: Collect multi-dimensional data on environment and operating conditions by deploying IoT sensing nodes on the auxiliary equipment of the target thermal power plant, and perform standardized preprocessing to obtain a standardized dataset.

[0097] Step S200: Select historical health samples from the historical standardized dataset of auxiliary equipment of the target thermal power plant, construct an environment-performance benchmark model, and dynamically calibrate the trained environment-performance benchmark model.

[0098] Step S300: Calculate the multidimensional deviation between the performance benchmark value and the actual operating value of the auxiliary equipment of the target thermal power plant based on the environmental-performance benchmark model, perform multidimensional deviation analysis and fault feature extraction, generate the energy efficiency health index and determine the fault type to obtain the health baseline.

[0099] Step S400: Based on the baseline health data and meteorological data for a future preset time period, construct and match an extreme weather scenario database, combine it with the energy efficiency health index, simulate and calculate the vulnerability score, and generate a comprehensive risk value.

[0100] Step S500: Construct a game decision model based on the comprehensive risk value, generate a payoff matrix and configure dynamic weight coefficients, calculate the total payoff of all execution strategies, and perform Nash equilibrium solution to determine the optimal shutdown window and the corresponding execution strategy.

[0101] Step S600: Based on the optimal shutdown window and multidimensional deviation, define the safe operating boundary under sub-health conditions, perform multi-objective Pareto optimization, select the optimal execution strategy, execute and monitor the system's operating status.

[0102] Step S700: Based on the optimal execution strategy and the optimal shutdown window, generate an adaptive maintenance work order containing preset content, perform resource demand forecasting and inventory linkage based on the work order content, compile emergency plans in conjunction with meteorological data, and establish an execution feedback channel.

[0103] Step S800: Based on the operational data after the execution of the adaptive maintenance work order, quantitatively verify the maintenance effect, combine on-site working condition feedback to attribute the root causes of the faults and accumulate knowledge, so as to optimize the game decision model and the environment-performance benchmark model and complete the closed-loop iteration of the system.

[0104] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0105] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0106] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0107] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0108] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0109] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0110] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0111] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0112] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A fault early warning and health management system for auxiliary equipment in thermal power plants, characterized in that, The fault early warning and health management system includes: The data acquisition and processing module is used to collect multi-dimensional data on the environment and operating conditions through IoT sensing nodes deployed on the auxiliary equipment of the target thermal power plant, and to perform standardized preprocessing to obtain a standardized dataset. The model building and calibration module is used to select historical health samples from the historical standardized dataset of auxiliary equipment of the target thermal power plant, build an environment-performance benchmark model, and dynamically calibrate the trained environment-performance benchmark model. The quantitative analysis module is used to calculate the multidimensional deviation between the performance benchmark value and the actual operating value of the auxiliary equipment of the target thermal power plant based on the environmental-performance benchmark model, perform multidimensional deviation analysis and fault feature extraction, generate energy efficiency health index and determine fault type, and obtain health baseline. The risk prediction module is used to construct and match an extreme weather scenario database based on the baseline health data and meteorological data within a preset future time period. Combined with the energy efficiency health index, it simulates and calculates vulnerability scores and generates a comprehensive risk value. The decision optimization module is used to construct a game decision model based on the comprehensive risk value, generate a payoff matrix and configure dynamic weight coefficients, calculate the total payoff of all execution strategies, and perform Nash equilibrium solution to determine the optimal shutdown window and the corresponding execution strategy. The optimization module is used to define the safe operating boundary under sub-health conditions based on the optimal shutdown window and multi-dimensional deviation, perform multi-objective Pareto optimization, select the optimal execution strategy, execute and monitor the system's operating status.

2. The auxiliary equipment fault early warning and health management system for thermal power plants according to claim 1, characterized in that, The process involves using IoT sensing nodes deployed at the auxiliary equipment of the target thermal power plant to collect multi-dimensional data on the environment and operating conditions, and then performing standardized preprocessing to obtain a standardized dataset, including: Based on the differences in the layout of auxiliary equipment in the target thermal power plant, a physical environment sensing network is built, various monitoring sensors are deployed, and environmental data transmission and time-series alignment are carried out across the entire domain to obtain standardized environmental data. The operating condition data of various auxiliary machines are collected through the monitoring information system of the target thermal power plant, noise reduction and anomaly removal are performed, and the data base of auxiliary equipment is obtained by merging with standardized environmental data. Based on the auxiliary equipment data base, supplement the target power grid dispatching constraint data and the target enterprise operation and maintenance resource data, perform full data standardization transformation and store it in the time series database to obtain a standardized dataset.

3. The auxiliary equipment fault early warning and health management system for thermal power plants according to claim 2, characterized in that, The process of selecting historical healthy samples from the historical standardized dataset of auxiliary equipment in the target thermal power plant, constructing an environment-performance benchmark model, and dynamically calibrating the trained environment-performance benchmark model includes: Retrieve historical standardized data for a preset time period from the time series database, use an intelligent filtering program to select high-quality healthy operating samples, and perform normalization processing to obtain the golden dataset. Based on the Golden Data Set, the input and output variable system of the model is determined, an ensemble learning algorithm is selected to train the model, and the model accuracy and generalization ability are iteratively optimized to obtain the trained environment-performance benchmark model. Set a standard sliding time window with a preset time interval, dynamically update training samples, perform incremental learning within a preset period, configure an environmental mutation trigger mechanism, force model retraining, and complete dynamic calibration.

4. The auxiliary equipment fault early warning and health management system for thermal power plants according to claim 1, characterized in that, The process involves calculating the multidimensional deviation between the performance benchmark value and the actual operating value of the auxiliary equipment of the target thermal power plant based on the environmental-performance benchmark model, performing multidimensional deviation analysis and fault feature extraction, generating an energy efficiency health index, determining the fault type, and obtaining the health baseline, including: The received auxiliary equipment operating parameters of the target thermal power plant are used as input parameters. The environment-performance benchmark model is called to calculate and cache the performance benchmark values, and the effective steady-state data is filtered out. Based on the actual operating values ​​and performance benchmark values ​​of the auxiliary equipment of the target thermal power plant, multiple target deviation indicators are calculated, typical abnormal operating characteristics are identified, hidden faults are located, and multiple target deviation indicators are weighted and integrated into a comprehensive deviation vector. The energy efficiency health index is calculated by using a preset nonlinear mapping function and a comprehensive deviation vector. The health status of auxiliary equipment is classified according to a preset threshold, and the failure type is determined by analyzing the trend to obtain the baseline health data.

5. The auxiliary equipment fault early warning and health management system for thermal power plants according to claim 1, characterized in that, The process involves constructing and matching an extreme weather scenario database based on baseline health data and meteorological data for a preset future time period. Combined with an energy efficiency and health index, a vulnerability score is calculated through simulation, and a comprehensive risk value is generated, including: The system connects to the target meteorological service interface, receives meteorological data for a future preset time period, calls the system's built-in extreme weather scenario library for matching, and extracts the corresponding scenario impact coefficients. Based on the energy efficiency and health index, simulation target constraints are set, and stress simulation calculations of all components are carried out in combination with meteorological loads. The simulation calculation results are compared with the preset meteorological load threshold to evaluate the remaining safety margin. Based on the simulation results, the vulnerability score of the auxiliary equipment is calculated, the impact of cascading failures is analyzed, and the comprehensive risk value of the auxiliary equipment is calculated by combining the impact coefficient of cascading failures.

6. The auxiliary equipment fault early warning and health management system for thermal power plants according to claim 1, characterized in that, The process involves constructing a game-theoretic decision-making model based on a comprehensive risk value, generating a payoff matrix and configuring dynamic weight coefficients, calculating the total payoff of all execution strategies, and performing a Nash equilibrium solution to determine the optimal shutdown window and corresponding execution strategy. This includes: Based on the comprehensive risk value, vulnerability score and energy efficiency health index, multiple game participants and work objectives are identified, the optional strategy space of multiple game participants is delineated, data is quantified and a payoff matrix is ​​constructed. Based on the operating conditions of the auxiliary equipment of the target thermal power plant, dynamic weight coefficients are configured for the payoff matrix, the payoff calculation formula is determined and calculated, and the total payoff is calculated by traversing all strategy combinations of the game participants. Disadvantaged strategies are removed from the total return of all strategy combinations. Nash equilibrium points are solved by category, and the time period corresponding to the Nash equilibrium point is determined as the optimal shutdown window.

7. The auxiliary equipment fault early warning and health management system for thermal power plants according to claim 1, characterized in that, The process involves defining the safe operating boundary under sub-health conditions based on the optimal shutdown window and multidimensional deviation, performing multi-objective Pareto optimization, selecting the optimal execution strategy, and executing and monitoring the system's operating status, including: Based on the energy efficiency health index and multidimensional deviation, we analyze the dominant causes of failures in the sub-healthy state of auxiliary equipment, conduct sensitivity analysis of auxiliary equipment for different types of failures, and delineate safe operation boundaries. Within the safe operating boundary, the optimization objective is determined, and a pre-set intelligent algorithm is used to perform multi-objective Pareto optimization calculations to select the execution strategy with the highest priority as the optimal execution strategy. The optimal operating parameters from the optimal execution strategy are sent to the system for adjustment, closed-loop tracking and monitoring are initiated for the system, and an emergency adjustment mechanism is configured.

8. The auxiliary equipment fault early warning and health management system for thermal power plants according to claim 1, characterized in that, The fault early warning and health management system also includes: The work order generation module is used to generate adaptive maintenance work orders containing preset content based on the optimal execution strategy and the optimal downtime window. It performs resource demand forecasting and inventory linkage based on the work order content, compiles emergency plans in conjunction with meteorological data, and establishes an execution feedback channel.

9. The auxiliary equipment fault early warning and health management system for thermal power plants according to claim 8, characterized in that, The fault early warning and health management system also includes: The feedback evolution module is used to quantify and verify the maintenance effect based on the running data after the execution of adaptive maintenance work orders, and to attribute the root causes of failures and accumulate knowledge by combining on-site working condition feedback, so as to optimize the game decision model and the environment-performance benchmark model and complete the closed-loop iteration of the system.

10. A method for early warning and health management of auxiliary equipment faults in thermal power plants, characterized in that, The fault early warning and health management methods include: By deploying IoT sensing nodes on the auxiliary equipment of the target thermal power plant, multi-dimensional data on the environment and operating conditions are collected and standardized preprocessed to obtain a standardized dataset. Historical health samples were selected from the historical standardized dataset of auxiliary equipment in the target thermal power plant to construct an environment-performance benchmark model, and the trained environment-performance benchmark model was dynamically calibrated. Based on the environmental-performance benchmark model, the multidimensional deviation between the performance benchmark value and the actual operating value of the auxiliary equipment of the target thermal power plant is calculated. Multidimensional deviation analysis and fault feature extraction are performed to generate an energy efficiency health index and determine the fault type, thereby obtaining the health baseline. Based on the baseline health data and meteorological data for a future preset time period, an extreme weather scenario database is constructed and matched. Combined with the energy efficiency health index, vulnerability scores are simulated and calculated, and a comprehensive risk value is generated. A game-theoretic decision-making model is constructed based on the comprehensive risk value. A payoff matrix is ​​generated and dynamic weight coefficients are configured. The total payoff of all execution strategies is calculated, and Nash equilibrium is solved to determine the optimal shutdown window and the corresponding execution strategy. Based on the optimal shutdown window and multidimensional deviation, the safe operating boundary under sub-health conditions is defined, multi-objective Pareto optimization is performed, the optimal execution strategy is selected, and the system operation status is executed and monitored. Based on the optimal execution strategy and the optimal downtime window, an adaptive maintenance work order containing preset content is generated. Based on the work order content, resource demand is predicted and inventory is linked. Emergency plans are developed in conjunction with meteorological data, and an execution feedback channel is established. Based on the operational data after the execution of adaptive maintenance work orders, the maintenance effect is quantitatively verified. Combined with on-site operating condition feedback, the root causes of failures are attributed and knowledge is accumulated to optimize the game decision-making model and the environment-performance benchmark model, and complete the closed-loop iteration of the system.