A thermal power plant inspection data driven equipment health assessment and prediction method

Abstract

This invention discloses a data-driven equipment health assessment and prediction method for thermal power plants, belonging to the field of equipment health assessment and prediction technology. The method includes: performing dynamic boundary pre-classification of standard operating data based on operating conditions to obtain equipment health assessment matching rules; using Granger causality tests to analyze the multi-dimensional correlation between inspection analysis data and standard operating data; constructing a three-dimensional simulation model of the equipment and performing multi-physics simulation deduction combined with real-time equipment inspection data; combining the correlation mapping relationship between thermal power plant inspection simulation data and equipment operating status for mapping inference to obtain the predicted equipment operating status; and matching the predicted equipment operating status with the equipment health assessment matching rules to obtain the equipment health assessment result. This invention can quantitatively distinguish the true impact of inspection parameters on equipment operating status, avoid false correlations caused by simple correlations, and significantly reduce the probability of misjudgment under varying operating conditions.

Description

Technical Field

[0001] This invention relates to the field of equipment health assessment and prediction technology, and more specifically, to a data-driven method for equipment health assessment and prediction in thermal power plants. Background Technology

[0002] Inspection data from thermal power plants encompasses multi-source time-series signals continuously collected during coal mill operation, including vibration, bearing temperature, inlet and outlet differential pressure, motor current, and primary air velocity. This data contains early signs of typical faults such as liner wear, roller fatigue, and stone / coal blockage, but is also heavily influenced by load fluctuations, coal quality changes, and start-up / shutdown operations, exhibiting characteristics such as high noise, strong nonlinearity, and complex state coupling. For the coal mill, a core piece of equipment in boiler fuel preparation, sudden failures can easily lead to combustion instability or even unplanned unit shutdowns. Therefore, there is an urgent need to develop a method for equipment health assessment and prediction driven by coal mill inspection data, providing crucial support for the safe and efficient operation of thermal power plants.

[0003] However, existing equipment health assessment and prediction methods driven by inspection data in thermal power plants generally rely on manual experience to set fixed thresholds, which are difficult to adapt to frequently changing operating conditions such as load and coal type, resulting in high false alarm or missed alarm rates; modeling based solely on statistical correlation between parameters cannot distinguish between causal and co-occurring relationships, and is easily affected by redundant or interfering parameters, producing false warnings; it relies too much on surface observation data; and the results are uninterpretable, making it difficult to gain the trust of operators and unable to support accurate maintenance decisions.

[0004] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0005] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a data-driven equipment health assessment and prediction method for thermal power plants. It has the advantages of accurate health status assessment and interpretable status prediction, thereby solving the problems of lagging traditional manual inspections, misjudgment due to fixed thresholds, redundancy of multiple parameters, and difficulty in monitoring internal status.

[0006] (II) Technical Solution To achieve the aforementioned advantages / objectives of accurate health status assessment and interpretable status prediction, the specific technical solution adopted by this invention is as follows: A data-driven method for equipment health assessment and prediction in thermal power plants includes: Historical inspection data and operating status data of coal mills in thermal power plants within the same time period are obtained. Effect measurement and data filtering are performed on the historical inspection data and operating status data to obtain inspection analysis data and operating standard data. Based on the operating conditions, dynamic boundary pre-classification is performed on the standard operating data to obtain equipment health assessment matching rules; By using Granger causality tests to analyze the multi-dimensional correlation between inspection analysis data and standard operating data, the correlation mapping relationship of equipment operating status is obtained; A three-dimensional simulation model of the equipment is constructed, and multi-physics simulation is performed in combination with real-time equipment inspection data to obtain thermal power plant inspection simulation data. The correlation between thermal power plant inspection simulation data and equipment operating status is combined to perform mapping reasoning to obtain the predicted operating status of the equipment. The predicted operating status of the equipment is matched with the equipment health assessment matching rules to obtain the equipment health assessment result.

[0007] Furthermore, effect measurement and data filtering were performed on historical inspection data and operational status data to obtain inspection analysis data and operational standard data, including: Denoising and outlier processing are performed on historical inspection data and operational status data to obtain cleaned historical inspection data and operational status data. The dynamic gray entropy correlation analysis method was used to perform effect quantification analysis on the historical inspection data and operation status data after cleaning, and the comprehensive contribution effect degree was obtained. Based on the comprehensive contribution effect degree and the preset effect degree threshold, the historical inspection data and the cleaned operation status data are filtered to obtain the filtered inspection data and operation status data. The filtered inspection data and operation status data are then standardized to obtain inspection analysis data and operation standard data.

[0008] Furthermore, dynamic gray entropy correlation analysis was used to perform effect quantification analysis on the cleaned historical inspection data and operational status data, resulting in the comprehensive contribution effect degree including: The transfer entropy is used to assess the information flow and dynamic causal influence intensity of the cleaned historical inspection data and operational status data, and the transfer entropy assessment results from inspection data to operational status data are obtained. The results of the transfer entropy assessment are sorted and filtered to obtain several key inspection parameters. Grey relational analysis is then performed on these key inspection parameters and the cleaned operation status data to obtain the grey relational degree of these key inspection parameters. The entropy weight method is used to normalize and weight the results of the transmission entropy assessment and the grey relational degree to generate the comprehensive contribution effect of each inspection parameter to the fluctuation of the operating status.

[0009] Furthermore, based on operating conditions, dynamic boundary pre-classification is performed on the standard operating data to obtain equipment health assessment matching rules, including: The health status of the standard operating data is grouped and statistically analyzed according to the operating condition category to obtain the distribution characteristics of the health status under each operating condition. Based on the health status distribution characteristics under various operating conditions, an adaptive boundary division strategy is used to dynamically determine the judgment boundary under different operating conditions, and obtain the health status classification threshold under each operating condition. Each operating condition category is structurally bound to its corresponding health status grading threshold to generate equipment health assessment matching rules.

[0010] Furthermore, by utilizing Granger causality tests to analyze the multi-dimensional correlation between inspection data and operational standard data, the correlation mapping relationship of equipment operating status was obtained, including: The inspection analysis data and the operation standard data are timestamped to obtain the aligned inspection analysis data and operation standard data. The aligned inspection analysis data and the standard operation data are merged to construct a multivariate time joint sequence. The multivariate time joint sequence is then subjected to stationarity testing and differencing to obtain a stationary joint sequence. The optimal lag order of the stationary joint sequence is determined based on the information criterion. Based on the optimal lag order, a sample set is constructed for the stationary joint sequence to obtain a two-way lag comparison sample set. Based on the two-way lag comparison sample set, a two-way Granger causality test and mapping relationship are constructed to obtain the correlation mapping relationship of equipment operating status.

[0011] Furthermore, based on the optimal lag order, a sample set is constructed for the stationary joint sequence to obtain a two-way lag comparison sample set. Based on this sample set, a two-way Granger causality test and mapping relationship construction are performed to obtain the equipment operating status correlation mapping relationship, including: The stationary joint sequence is segmented according to the optimal lag order, and the segmented results are constructed to obtain a two-way lag comparison sample group. Based on the two-way lag comparison sample groups, the statistical significance of the two groups of inspection parameters affecting the operating status and the operating status affecting the inspection parameters were calculated to obtain the two-way Granger causality determination results. The strength of causal relationships is determined based on the results of bidirectional Granger causality determination, and feature mapping relationships are constructed based on the strength of causal relationships to obtain the device operating status association mapping relationship.

[0012] Furthermore, based on the two-way lag comparison sample groups, the statistical significance of the impact of inspection parameters on operating status and the impact of operating status on inspection parameters were calculated separately. The results of the two-way Granger causality determination are as follows: Based on the bidirectional lag comparison sample group, the lag sequence of the inspection parameters is used to fit the predicted operating status, and the lag sequence of the operating status is used to fit the predicted inspection parameters. The sum of squared residuals of the two sets of fittings is calculated to obtain the error results with and without constraints. Based on the error results, a variance ratio statistic is constructed and compared with the critical value. Based on the comparison results, the corresponding significance probability value is calculated. The significance probability value is used to determine whether the two groups of predicted contributions are statistically significant. Based on the statistical significance judgment results, it is determined whether there is Granger causality between the inspection parameters and the operating status, and between the operating status and the inspection parameters, respectively, and the result of the two-way Granger causality judgment is obtained.

[0013] Furthermore, a three-dimensional simulation model of the equipment is constructed, and multiphysics simulation is performed in conjunction with real-time equipment inspection data to obtain thermal power plant inspection simulation data. The correlation between the thermal power plant inspection simulation data and the equipment operating status is then combined to perform mapping inference, resulting in the predicted equipment operating status, including: A three-dimensional solid model of the coal mill is constructed based on its geometric dimensions, assembly structure, and material properties. The three-dimensional solid model is then meshed and multiphysics parameters are assigned to obtain a three-dimensional simulation model of the equipment. Real-time equipment inspection data is input into the equipment's three-dimensional simulation model to perform multi-physics field simulation deduction that couples temperature field, stress field, and flow field, thereby obtaining thermal power plant inspection simulation data. The simulated inspection data of thermal power plants is substituted into the equipment operation status correlation mapping relationship, and feature mapping reasoning calculation is performed according to causal weights to obtain the predicted equipment operation status.

[0014] Furthermore, real-time equipment inspection data is input into the equipment's three-dimensional simulation model for multi-physics simulation deduction involving coupled temperature, stress, and flow fields, resulting in thermal power plant inspection simulation data including: Real-time equipment inspection data is converted into boundary conditions and excitation parameters required for simulation, and the boundary conditions and excitation parameters are applied to the corresponding positions of the equipment 3D simulation model to obtain the equipment 3D simulation model after the working condition configuration. Based on the 3D simulation model of the equipment after the working condition configuration, the temperature distribution is iteratively solved in the simulation platform by combining the temperature parameters in the real-time equipment inspection data to obtain the temperature field simulation results. The temperature field simulation results are used as thermal loads and applied to the three-dimensional simulation model of the equipment after the working condition is configured to calculate the structural stress and deformation caused by temperature changes, and the stress field simulation results are obtained. Using temperature gradient and structural deformation from the simulation results of temperature field and stress field as constraints, the fluid flow state is iteratively calculated using the three-dimensional simulation model of the equipment after the working condition is configured, and cross-checked and corrected with the simulation results of temperature field and stress field until the changes of each physical field parameter meet the convergence condition, thus obtaining the multi-physics field coupled simulation results. Feature data consistent with the actual inspection dimensions are extracted from the multiphysics coupling simulation results, and the extracted feature data results are normalized to obtain thermal power plant inspection simulation data.

[0015] Furthermore, the predicted operating status of the equipment is matched with the equipment health assessment matching rules to obtain the equipment health assessment results, including: The predicted operating status of the equipment is linearly normalized according to the standard operating data range to obtain a standardized set of predicted status indicators. Based on the indicator definitions in the equipment health assessment matching rules, corresponding state parameters are selected from the standardized set of predicted state indicators to obtain the set of predicted state indicators to be matched. The set of predicted state indicators to be matched is compared with the equipment health assessment matching rules on an indicator-by-indicator basis and interval matching to obtain the results of indicator matching degree and rule compliance judgment. The results of the indicator matching degree and rule compliance judgment are comprehensively weighted and scored, and the comprehensive weighted score results are classified and judged according to the preset health level classification standard to obtain the equipment health assessment results of the coal mill.

[0016] (III) Beneficial Effects Compared with existing technologies, this invention provides a data-driven method for equipment health assessment and prediction in thermal power plants, which has the following advantages: (1) This invention combines equipment operation status correlation mapping with multi-physics simulation to achieve accurate assessment and advanced prediction of the health status of coal mills. It effectively solves problems such as the lag of traditional manual inspection, misjudgment of fixed thresholds, redundancy of multiple parameters and difficulty in monitoring internal status. Through equipment operation status correlation mapping, it can quantitatively distinguish the degree of real causal influence of inspection parameters on equipment operation status, avoid false correlation caused by simple correlation, automatically lock core influencing factors such as vibration, temperature and pressure difference, greatly reduce the probability of misjudgment under variable working conditions, and make the status prediction results more in line with the actual operating mechanism of the equipment, with stronger interpretability and engineering credibility.

[0017] (2) This invention determines the health assessment matching rules, abandons the traditional fixed threshold judgment method, and adaptively divides the health, sub-health and abnormal level thresholds according to different working conditions such as low load, medium load and high load. It effectively solves the problem of large judgment error and high false alarm and false alarm rate under variable working conditions, greatly improves the adaptability, accuracy and scenario matching of health assessment rules, and makes the equipment health status judgment more in line with the actual operating rules, avoiding the assessment distortion caused by load fluctuation.

[0018] (3) This invention constructs a quantitative correlation between inspection data and operating status by mapping the equipment operating status, and clarifies the influence intensity of parameters and the direction of information transmission through bidirectional causal analysis. It overcomes the shortcomings of traditional correlation analysis, which cannot distinguish causality and is difficult to determine the dominant factors. It can accurately identify key influencing factors and significantly improve the reliability and interpretability of status prediction.

[0019] (4) This invention transforms real-time inspection data into simulated boundary conditions and excitation loads by coupling the equipment three-dimensional simulation model with multi-physics field, and realizes sequential coupling iterative calculation of temperature field, stress field and flow field. It can obtain internal state parameters that are difficult to measure directly without disassembling the equipment and without affecting normal production, and generate high-fidelity inspection simulation data. It solves the problems of limited on-site detection methods, difficulty in obtaining key parameters and inability to show potential faults in advance. At the same time, it combines the correlation mapping relationship to realize the quantitative reasoning of future operating status, realizes the transformation from post-event analysis to pre-event prediction, and effectively improves the equipment fault early warning capability. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of a data-driven equipment health assessment and prediction method for thermal power plants, according to an embodiment of the present invention. Detailed Implementation

[0022] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0023] According to an embodiment of the present invention, a method for equipment health assessment and prediction driven by inspection data in thermal power plants is provided.

[0024] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, the equipment health assessment and prediction method driven by thermal power plant inspection data according to an embodiment of the present invention includes: Historical inspection data and operating status data of coal mills in thermal power plants within the same time period are obtained. Effect measurement and data filtering are performed on the historical inspection data and operating status data to obtain inspection analysis data and operating standard data.

[0025] Specifically, effect measurement and data filtering are performed on historical inspection data and operational status data to obtain inspection analysis data and operational standard data, including: Denoising and outlier processing are performed on historical inspection data and operational status data to obtain cleaned historical inspection data and operational status data. The dynamic gray entropy correlation analysis method was used to perform effect quantification analysis on the cleaned historical inspection data and operation status data to obtain the comprehensive contribution effect degree.

[0026] Specifically, dynamic gray entropy correlation analysis was used to perform effect quantification analysis on the cleaned historical inspection data and operational status data, resulting in a comprehensive contribution effect degree including: The transfer entropy is used to assess the information flow and dynamic causal influence intensity of the cleaned historical inspection data and operational status data, and the transfer entropy assessment results from inspection data to operational status data are obtained. The results of the transfer entropy assessment are sorted and filtered to obtain several key inspection parameters. Grey relational analysis is then performed on these key inspection parameters and the cleaned operation status data to obtain the grey relational degree of these key inspection parameters. The entropy weight method is used to normalize and weight the results of the transmission entropy assessment and the grey relational degree to generate the comprehensive contribution effect of each inspection parameter to the fluctuation of the operating status.

[0027] Based on the comprehensive contribution effect degree and the preset effect degree threshold, the historical inspection data and the cleaned operation status data are filtered to obtain the filtered inspection data and operation status data. The filtered inspection data and operation status data are then standardized to obtain inspection analysis data and operation standard data.

[0028] Specifically, historical inspection data and operational status data of medium-speed coal mills in thermal power plants are collected simultaneously within the same time period. The historical inspection data includes six parameters sampled every hour: mill roller vibration amplitude, bearing temperature, lubricating oil temperature, inlet and outlet pressure difference, casing vibration, and sealing air pressure. The operational status data includes five core operational parameters: coal feed rate, primary air volume, motor current, outlet temperature, and milling pressure. All data are uniformly represented using the YYYY-MM-DD standard. HH:MM:SS timestamp alignment, the original data is a time series sample of 30 consecutive days. Data cleaning was performed on the original data, using a mean filter with a sliding window of 5 to remove random noise. Outliers were identified and removed using the 3σ criterion; data exceeding the mean ± 3 standard deviations were identified as outliers and imputed using linear interpolation. Simultaneously, the physical rationality of the data was verified, removing false data such as vibration amplitude greater than 25 mm / s, temperature exceeding the -20℃~300℃ range, and negative wind pressure, which do not conform to the equipment's mechanism. This yielded the cleaned, valid historical inspection data and operational status data. The information transmission strength from each inspection parameter to the operational status data was calculated using transfer entropy. Based on the entropy transfer assessment results, four key inspection parameters—vibration amplitude, bearing temperature, inlet and outlet pressure difference, and lubricating oil temperature—were selected by sorting the values ​​from high to low. Then, a grey relational analysis was performed on these four key parameters and the operating status data to calculate the grey relational degree of each parameter. The entropy transfer value and grey relational degree were normalized and weighted using the entropy weight method to calculate the comprehensive contribution effect of the four key inspection parameters. The effect threshold was set to 0.20, retaining all four high-contribution parameters and removing redundant parameters. The selected data underwent min-max standardization, mapping all values ​​to the [0,1] interval to generate inspection analysis data and operating standard data.

[0029] Based on the operating conditions, dynamic boundary pre-classification is performed on the standard operating data to obtain equipment health assessment matching rules.

[0030] Specifically, based on operating conditions, dynamic boundary pre-classification is performed on standard operating data to obtain equipment health assessment matching rules, including: The health status of the standard operating data is grouped and statistically analyzed according to the operating condition category to obtain the distribution characteristics of the health status under each operating condition. Based on the health status distribution characteristics under various operating conditions, an adaptive boundary division strategy is used to dynamically determine the judgment boundary under different operating conditions, and obtain the health status classification threshold under each operating condition. Each operating condition category is structurally bound to its corresponding health status grading threshold to generate equipment health assessment matching rules.

[0031] Specifically, based on operating conditions, the standard operating data is dynamically pre-classified by boundary. Operating conditions are divided into three groups according to coal feed rate: low load (≤30t / h), medium load (30~50t / h), and high load (>50t / h). The parameter distribution characteristics under healthy conditions are statistically analyzed for each group. Adaptive boundary division is used to determine the three-level judgment thresholds for healthy, sub-healthy, and abnormal conditions under each operating condition. For example, under high load, the vibration amplitude is healthy (<0.40), sub-healthy (0.40~0.60), and abnormal (>0.60); under medium load, healthy (<0.35), sub-healthy (0.35~0.55), and abnormal (>0.55); under low load, healthy (<0.30), sub-healthy (0.30~0.50), and abnormal (>0.50). Operating conditions are then bound to thresholds to form equipment health assessment matching rules.

[0032] By using Granger causality tests to analyze the multi-dimensional correlation between inspection data and standard operating data, the correlation mapping relationship of equipment operating status is obtained.

[0033] Specifically, by using Granger causality tests to analyze the multi-dimensional correlation between inspection data and operational standard data, the correlation mapping relationship of equipment operating status is obtained, including: The inspection analysis data and the operation standard data are timestamped to obtain the aligned inspection analysis data and operation standard data. The aligned inspection analysis data and the standard operation data are merged to construct a multivariate time joint sequence. The multivariate time joint sequence is then subjected to stationarity testing and differencing to obtain a stationary joint sequence. The optimal lag order of the stationary joint sequence is determined based on the information criterion. Based on the optimal lag order, a sample set is constructed for the stationary joint sequence to obtain a two-way lag comparison sample set. Based on the two-way lag comparison sample set, a two-way Granger causality test and mapping relationship are constructed to obtain the correlation mapping relationship of equipment operating status.

[0034] Specifically, a sample set is constructed for the stationary joint sequence based on the optimal lag order to obtain a two-way lag comparison sample set. Based on this sample set, a two-way Granger causality test and mapping relationship construction are performed to obtain the equipment operating status correlation mapping relationship, including: The stationary joint sequence is segmented according to the optimal lag order, and the segmented results are constructed to obtain a two-way lag comparison sample group. Based on the two-way lag comparison sample groups, the statistical significance of the two groups of inspection parameters affecting the operating status and the operating status affecting the inspection parameters were calculated to obtain the two-way Granger causality determination results.

[0035] Specifically, based on the two-way lag comparison sample groups, the statistical significance of the impact of inspection parameters on operating status and the impact of operating status on inspection parameters were calculated separately. The results of the two-way Granger causality determination are as follows: Based on the bidirectional lag comparison sample group, the lag sequence of the inspection parameters is used to fit the predicted operating status, and the lag sequence of the operating status is used to fit the predicted inspection parameters. The sum of squared residuals of the two sets of fittings is calculated to obtain the error results with and without constraints. Based on the error results, a variance ratio statistic is constructed and compared with the critical value. Based on the comparison results, the corresponding significance probability value is calculated. The significance probability value is used to determine whether the two groups of predicted contributions are statistically significant. Based on the statistical significance judgment results, it is determined whether there is Granger causality between the inspection parameters and the operating status, and between the operating status and the inspection parameters, respectively, and the result of the two-way Granger causality judgment is obtained.

[0036] The strength of causal relationships is determined based on the results of bidirectional Granger causality determination, and feature mapping relationships are constructed based on the strength of causal relationships to obtain the device operating status association mapping relationship.

[0037] Specifically, the strength of the causal relationship between each inspection parameter and the operating status is ranked based on the variance ratio statistic and significance probability value in the two-way Granger causality determination results, and a feature association list is output in order of causal strength. Each inspection parameter is assigned a corresponding weight according to the causal relationship strength in the feature association list, redundant parameters without significant causal relationship are eliminated, and a weighted core feature combination is output. Based on the weighted core feature combination, a feature mapping relationship from the inspection parameters to the equipment operating status is constructed by weighted fusion, forming an equipment operating status association mapping relationship.

[0038] Specifically, the Granger causality test is used to construct the correlation mapping relationship. The inspection analysis data and the operating standard data are precisely aligned with timestamps to construct a multivariate time joint series. The stationarity is verified by the ADF test and first-order differencing is performed. The optimal lag order is determined to be 3 using the AIC criterion. A two-way lag comparison sample group is constructed according to this order. The lag sequence of inspection parameters is used to fit the operating state, and the lag sequence of operating state is used to fit the inspection parameters. The sum of squared residuals of the two sets of fittings is calculated. The variance ratio statistic is constructed and the significance probability value is calculated. A p-value < 0.05 is used to determine the existence of a significant Granger causality relationship. The causal correlation strength is determined according to the size of the statistic and weights are assigned. The highest weight is 0.28 for vibration amplitude and the lowest is 0.22 for lubricating oil temperature. Redundant dimensions without significant correlation are removed to construct the correlation mapping relationship of equipment operating status.

[0039] A three-dimensional simulation model of the equipment is constructed, and multi-physics simulation is performed in combination with real-time equipment inspection data to obtain thermal power plant inspection simulation data. The correlation between thermal power plant inspection simulation data and equipment operating status is combined to perform mapping reasoning to obtain the predicted operating status of the equipment.

[0040] Specifically, a 3D simulation model of the equipment is constructed, and multiphysics simulation is performed in conjunction with real-time equipment inspection data to obtain thermal power plant inspection simulation data. The correlation between the thermal power plant inspection simulation data and the equipment operating status is then combined to perform mapping inference, resulting in the predicted equipment operating status, including: A three-dimensional solid model of the coal mill is constructed based on its geometric dimensions, assembly structure, and material properties. The three-dimensional solid model is then meshed and multiphysics parameters are assigned to obtain a three-dimensional simulation model of the equipment. Real-time equipment inspection data is input into the equipment's three-dimensional simulation model to perform multi-physics field simulation deduction that couples temperature field, stress field, and flow field, thus obtaining thermal power plant inspection simulation data.

[0041] Specifically, real-time equipment inspection data is input into the equipment's three-dimensional simulation model to perform multi-physics simulation deduction coupling temperature field, stress field, and flow field, resulting in thermal power plant inspection simulation data including: Real-time equipment inspection data is converted into boundary conditions and excitation parameters required for simulation, and the boundary conditions and excitation parameters are applied to the corresponding positions of the equipment 3D simulation model to obtain the equipment 3D simulation model after the working condition configuration. Based on the 3D simulation model of the equipment after the working condition configuration, the temperature distribution is iteratively solved in the simulation platform by combining the temperature parameters in the real-time equipment inspection data to obtain the temperature field simulation results. The temperature field simulation results are used as thermal loads and applied to the three-dimensional simulation model of the equipment after the working condition is configured to calculate the structural stress and deformation caused by temperature changes, and the stress field simulation results are obtained. Using temperature gradient and structural deformation from the simulation results of temperature field and stress field as constraints, the fluid flow state is iteratively calculated using the three-dimensional simulation model of the equipment after the working condition is configured, and cross-checked and corrected with the simulation results of temperature field and stress field until the changes of each physical field parameter meet the convergence condition, thus obtaining the multi-physics field coupled simulation results. Feature data consistent with the actual inspection dimensions are extracted from the multiphysics coupling simulation results, and the extracted feature data results are normalized to obtain thermal power plant inspection simulation data.

[0042] The simulated inspection data of thermal power plants is substituted into the equipment operation status correlation mapping relationship, and feature mapping reasoning calculation is performed according to causal weights to obtain the predicted equipment operation status.

[0043] Specifically, a three-dimensional solid model is established based on the actual geometric dimensions, assembly structure, and material properties of steel and wear-resistant alloys of the coal mill. Key components such as grinding rollers, bearings, and cylinders are meshed, and multi-physics parameters such as thermal conductivity, elastic modulus, and fluid density are assigned. Real-time inspection data is transformed into simulation boundary conditions and excitation parameters such as temperature boundary, pressure load, and flow constraint, which are applied to the corresponding positions of the simulation model. Sequential coupling iterative calculations of multi-physics are carried out. First, the temperature field distribution is solved based on real-time temperature parameters. Then, the temperature field is used as a thermal load to calculate structural stress and deformation. Finally, the flow field parameters are solved with temperature gradient and structural deformation as constraints. The three fields of data are mutually transmitted and corrected until convergence. Four characteristic data of vibration, temperature, pressure difference, and oil temperature are extracted from the simulation results and standardized to generate thermal power plant inspection simulation data consistent with the dimensions of real inspections. This simulation data is substituted into the equipment operation status correlation mapping relationship, and the mapping inference is completed by weighted calculation according to causal weights. The predicted equipment operation status is output.

[0044] Specifically, the constructed 3D simulation model of the equipment adopts a multi-physics coupling architecture, consisting of a geometric entity modeling module, a mesh generation module, a boundary load loading module, a physics field solution module, and a post-processing output module connected in series. The modules are coupled and transferred sequentially through data interfaces to achieve the sequential transmission of temperature, stress, and flow fields. The data for the model comes from actual engineering drawings of the coal mill, component material mechanical parameters, and real-time inspection data. Preprocessing includes geometric model simplification, chamfering and deburring, omission of non-critical structures, and mesh quality optimization. Mesh orthogonality and torsion rate are used as convergence criteria, and the optimization strategy is adaptive mesh refinement. Key parameters are set as follows: mesh element size 2mm-10mm, temperature field solution iteration steps 50, and stress field convergence tolerance 1×10⁻⁶. -6 The k-ε model was selected as the turbulence model for the flow field. The parameters were determined based on the structural dimensions of the coal mill and the accuracy requirements of multi-physics coupling.

[0045] The predicted operating status of the equipment is matched with the equipment health assessment matching rules to obtain the equipment health assessment result.

[0046] Specifically, the predicted operating status of the equipment is matched with the equipment health assessment matching rules to obtain the equipment health assessment results, including: The predicted operating status of the equipment is linearly normalized according to the standard operating data range to obtain a standardized set of predicted status indicators. Based on the indicator definitions in the equipment health assessment matching rules, corresponding state parameters are selected from the standardized set of predicted state indicators to obtain the set of predicted state indicators to be matched. The set of predicted state indicators to be matched is compared with the equipment health assessment matching rules on an indicator-by-indicator basis and interval matching to obtain the results of indicator matching degree and rule compliance judgment. The results of the indicator matching degree and rule compliance judgment are comprehensively weighted and scored, and the comprehensive weighted score results are classified and judged according to the preset health level classification standard to obtain the equipment health assessment results of the coal mill.

[0047] Specifically, based on the current actual operating conditions of the equipment, the corresponding graded threshold range is located. Each key indicator is compared with the healthy, sub-healthy, and abnormal thresholds in the matching rules, and the preliminary comparison results between the indicators and the threshold ranges are output. Based on the preliminary comparison results, the specific range in which the indicator currently falls within the healthy, sub-healthy, or abnormal range is determined, and the range positioning result is output. Based on the range positioning result, that is, based on the relative position of the indicator within the threshold range, a quantitative score is calculated using a linear mapping method to obtain the matching degree of each individual indicator. The matching degree value is normalized, and the set of matching degrees of each individual indicator is output. Combined with the preset health judgment rules, a comprehensive compliance check is performed on all indicators to determine whether the overall condition meets the conditions for the healthy, sub-healthy, or abnormal level. That is, after obtaining the matching degree of each indicator, the matching status of all indicators is reviewed and comprehensively judged according to the preset health level judgment rules. The overall predicted state is checked to see if it meets all or the core conditions of a certain health level, avoiding misjudgment caused by the abnormality of a single indicator. Thus, the overall health, sub-healthy, or abnormal level of the equipment is finally determined, and the complete indicator matching degree and rule compliance judgment result is output.

[0048] Specifically, a fixed causal weighting is used, with the grinding roller vibration amplitude at 0.27, bearing temperature at 0.25, inlet and outlet pressure difference at 0.24, and lubricating oil temperature at 0.22. The matching degree of each indicator is weighted and summed to obtain a comprehensive weighted score. The comprehensive weighted score is compared with the preset health level classification standard, where a comprehensive weighted score ≥0.80 is considered healthy, 0.60~0.80 is considered sub-healthy, and <0.60 is considered abnormal, thus determining the health level corresponding to the comprehensive score. Using the health level determination result as input, and combining rule compliance to complete the verification and confirmation, the equipment health assessment result of the coal mill is directly output.

[0049] According to another embodiment of the present invention, a data-driven equipment health assessment and prediction system for thermal power plant inspections is provided, the system comprising: The data preprocessing module is used to acquire historical inspection data and operating status data of coal mills in thermal power plants within the same time period, and to perform effect measurement and data filtering on the historical inspection data and operating status data to obtain inspection analysis data and operating standard data. The health assessment matching rule determination module is used to perform dynamic boundary pre-classification of operating standard data based on operating conditions to obtain equipment health assessment matching rules. The equipment operation status correlation mapping module is used to analyze the multi-dimensional correlation between inspection analysis data and operation standard data using Granger causality test to obtain the equipment operation status correlation mapping relationship. The equipment operation status prediction module is used to build a three-dimensional simulation model of the equipment and perform multi-physics simulation deduction by combining real-time equipment inspection data to obtain thermal power plant inspection simulation data. The thermal power plant inspection simulation data is combined with the correlation mapping relationship between the thermal power plant inspection simulation data and the equipment operation status to perform mapping reasoning and obtain the predicted equipment operation status. The equipment health assessment module is used to match the predicted operating status of equipment with the equipment health assessment matching rules to obtain the equipment health assessment results.

[0050] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "setting," "connection," "fixing," "screw connection," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal connection of two components or the interaction between two components. Unless otherwise explicitly limited, those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0051] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A data-driven method for equipment health assessment and prediction in thermal power plants, characterized in that, include: Historical inspection data and operating status data of coal mills in thermal power plants within the same time period are obtained. Effect measurement and data filtering are performed on the historical inspection data and operating status data to obtain inspection analysis data and operating standard data. Based on the operating conditions, dynamic boundary pre-classification is performed on the standard operating data to obtain equipment health assessment matching rules; By using Granger causality tests to analyze the multi-dimensional correlation between inspection analysis data and standard operating data, the correlation mapping relationship of equipment operating status is obtained; A three-dimensional simulation model of the equipment is constructed, and multi-physics simulation is performed in combination with real-time equipment inspection data to obtain thermal power plant inspection simulation data. The correlation between thermal power plant inspection simulation data and equipment operating status is combined to perform mapping reasoning to obtain the predicted operating status of the equipment. The predicted operating status of the equipment is matched with the equipment health assessment matching rules to obtain the equipment health assessment result.

2. The method for equipment health assessment and prediction driven by inspection data in a thermal power plant according to claim 1, characterized in that, The process of measuring the effect and filtering the historical inspection data and operational status data to obtain inspection analysis data and operational standard data includes: Denoising and outlier processing are performed on historical inspection data and operational status data to obtain cleaned historical inspection data and operational status data. The dynamic gray entropy correlation analysis method was used to perform effect quantification analysis on the historical inspection data and operation status data after cleaning, and the comprehensive contribution effect degree was obtained. Based on the comprehensive contribution effect degree and the preset effect degree threshold, the historical inspection data and the cleaned operation status data are filtered to obtain the filtered inspection data and operation status data. The filtered inspection data and operation status data are then standardized to obtain inspection analysis data and operation standard data.

3. The method for equipment health assessment and prediction driven by inspection data in a thermal power plant according to claim 2, characterized in that, The dynamic gray entropy correlation analysis method is used to perform effect quantification analysis on the historical inspection data and operation status data after cleaning, and the comprehensive contribution effect degree is obtained, including: The transfer entropy is used to assess the information flow and dynamic causal influence intensity of the cleaned historical inspection data and operational status data, and the transfer entropy assessment results from inspection data to operational status data are obtained. The results of the transfer entropy assessment are sorted and filtered to obtain several key inspection parameters. Grey relational analysis is then performed on these key inspection parameters and the cleaned operation status data to obtain the grey relational degree of these key inspection parameters. The entropy weight method is used to normalize and weight the results of the transmission entropy assessment and the grey relational degree to generate the comprehensive contribution effect of each inspection parameter to the fluctuation of the operating status.

4. The method for equipment health assessment and prediction driven by inspection data in a thermal power plant according to claim 1, characterized in that, The dynamic boundary pre-classification of operating standard data based on operating conditions to obtain equipment health assessment matching rules includes: The health status of the standard operating data is grouped and statistically analyzed according to the operating condition category to obtain the distribution characteristics of the health status under each operating condition. Based on the health status distribution characteristics under various operating conditions, an adaptive boundary division strategy is used to dynamically determine the judgment boundary under different operating conditions, and obtain the health status classification threshold under each operating condition. Each operating condition category is structurally bound to its corresponding health status grading threshold to generate equipment health assessment matching rules.

5. The method for equipment health assessment and prediction driven by inspection data in a thermal power plant according to claim 1, characterized in that, The method of using Granger causality tests to analyze the multi-dimensional correlation between inspection data and operational standard data to obtain the correlation mapping relationship of equipment operating status includes: The inspection analysis data and the operation standard data are timestamped to obtain the aligned inspection analysis data and operation standard data. The aligned inspection analysis data and the standard operation data are merged to construct a multivariate time joint sequence. The multivariate time joint sequence is then subjected to stationarity testing and differencing to obtain a stationary joint sequence. The optimal lag order of the stationary joint sequence is determined based on the information criterion. Based on the optimal lag order, a sample set is constructed for the stationary joint sequence to obtain a two-way lag comparison sample set. Based on the two-way lag comparison sample set, a two-way Granger causality test and mapping relationship are constructed to obtain the correlation mapping relationship of equipment operating status.

6. The method for equipment health assessment and prediction driven by inspection data in a thermal power plant according to claim 5, characterized in that, The process involves constructing a sample set based on the optimal lag order for the stationary joint sequence, obtaining a two-way lag comparison sample set, and then performing a two-way Granger causality test and mapping relationship construction based on the two-way lag comparison sample set to obtain the equipment operating status correlation mapping relationship, including: The stationary joint sequence is segmented according to the optimal lag order, and the segmented results are constructed to obtain a two-way lag comparison sample group. Based on the two-way lag comparison sample groups, the statistical significance of the two groups of inspection parameters affecting the operating status and the operating status affecting the inspection parameters were calculated to obtain the two-way Granger causality determination results. The strength of causal relationships is determined based on the results of bidirectional Granger causality determination, and feature mapping relationships are constructed based on the strength of causal relationships to obtain the device operating status association mapping relationship.

7. The method for equipment health assessment and prediction driven by inspection data in a thermal power plant according to claim 6, characterized in that, The calculation of the two sets of statistical significance of the influence of inspection parameters on operating status and the influence of operating status on inspection parameters based on the two-way lag comparison sample groups, and the resulting determination of the two-way Granger causality relationship, include: Based on the bidirectional lag comparison sample group, the lag sequence of the inspection parameters is used to fit the predicted operating status, and the lag sequence of the operating status is used to fit the predicted inspection parameters. The sum of squared residuals of the two sets of fittings is calculated to obtain the error results with and without constraints. Based on the error results, a variance ratio statistic is constructed and compared with the critical value. The corresponding significance probability value is calculated based on the comparison results. The significance probability value is used to determine whether the two groups of predicted contributions are statistically significant. Based on the statistical significance judgment results, it is determined whether there is Granger causality between the inspection parameters and the operating status, and between the operating status and the inspection parameters, respectively, and the result of the two-way Granger causality judgment is obtained.

8. The method for equipment health assessment and prediction driven by inspection data in a thermal power plant according to claim 1, characterized in that, The construction of a three-dimensional simulation model of the equipment, combined with real-time equipment inspection data, is used for multi-physics simulation deduction to obtain thermal power plant inspection simulation data. The correlation between the thermal power plant inspection simulation data and the equipment operating status is then used for mapping reasoning to obtain the predicted equipment operating status, including: A three-dimensional solid model of the coal mill is constructed based on its geometric dimensions, assembly structure, and material properties. The three-dimensional solid model is then meshed and multiphysics parameters are assigned to obtain a three-dimensional simulation model of the equipment. Real-time equipment inspection data is input into the equipment's three-dimensional simulation model to perform multi-physics field simulation deduction that couples temperature field, stress field, and flow field, thereby obtaining thermal power plant inspection simulation data. The simulated inspection data of thermal power plants is substituted into the equipment operation status correlation mapping relationship, and feature mapping reasoning calculation is performed according to causal weights to obtain the predicted equipment operation status.

9. A method for equipment health assessment and prediction driven by inspection data in a thermal power plant according to claim 8, characterized in that, The process of inputting real-time equipment inspection data into the equipment's three-dimensional simulation model for multi-physics simulation deduction involving coupled temperature, stress, and flow fields, to obtain thermal power plant inspection simulation data includes: Real-time equipment inspection data is converted into boundary conditions and excitation parameters required for simulation, and the boundary conditions and excitation parameters are applied to the corresponding positions of the equipment 3D simulation model to obtain the equipment 3D simulation model after the working condition configuration. Based on the 3D simulation model of the equipment after the working condition configuration, the temperature distribution is iteratively solved in the simulation platform by combining the temperature parameters in the real-time equipment inspection data to obtain the temperature field simulation results. The temperature field simulation results are used as thermal loads and applied to the three-dimensional simulation model of the equipment after the working condition is configured to calculate the structural stress and deformation caused by temperature changes, and the stress field simulation results are obtained. Using temperature gradient and structural deformation from the simulation results of temperature field and stress field as constraints, the fluid flow state is iteratively calculated using the three-dimensional simulation model of the equipment after the working condition is configured, and cross-checked and corrected with the simulation results of temperature field and stress field until the changes of each physical field parameter meet the convergence condition, thus obtaining the multi-physics field coupled simulation results. Feature data consistent with the actual inspection dimensions are extracted from the multiphysics coupling simulation results, and the extracted feature data results are normalized to obtain thermal power plant inspection simulation data.

10. The method for equipment health assessment and prediction driven by inspection data in a thermal power plant according to claim 1, characterized in that, The process of matching the predicted equipment operating status with the equipment health assessment matching rules to obtain the equipment health assessment result includes: The predicted operating status of the equipment is linearly normalized according to the standard operating data range to obtain a standardized set of predicted status indicators. Based on the indicator definitions in the equipment health assessment matching rules, corresponding state parameters are selected from the standardized set of predicted state indicators to obtain the set of predicted state indicators to be matched. The set of predicted state indicators to be matched is compared with the equipment health assessment matching rules on an indicator-by-indicator basis and interval matching to obtain the results of indicator matching degree and rule compliance judgment. The results of the indicator matching degree and rule compliance judgment are comprehensively weighted and scored, and the comprehensive weighted score results are classified and judged according to the preset health level classification standard to obtain the equipment health assessment results of the coal mill.

Images