Boiler fan performance monitoring and fault diagnosis method and system based on data driving

Abstract

The invention belongs to the technical field of heat energy power system operation optimization, and particularly relates to a boiler fan performance monitoring and fault diagnosis method and system based on data driving. Real-time operation data of a boiler fan are collected to form an input observation vector, a dynamic historical memory matrix is constructed according to a power generation power interval, the abnormal weight of each parameter is calculated, operation simulation is carried out to obtain simulation data, and then the average deviation degree and an alarm threshold value are calculated. Early warning is carried out by comparing the real-time comprehensive deviation degree with an alarm threshold value, and abnormal parameters are positioned and regulated according to the contribution degree of each parameter deviation to the comprehensive deviation degree. The system comprises a data acquisition module, a simulation module, a fault diagnosis module and a control execution module. The problems that in the prior art, real-time simulation precision of the boiler fan is insufficient, fault early warning and root judgment defects exist and the like are solved, the monitoring precision and the fault response speed under the variable load working condition are improved, closed-loop management from early warning to regulation and control is achieved, and reliability and safety of operation of the boiler fan are enhanced.

Description

Technical Field

[0001] This invention belongs to the field of thermal power system operation optimization technology, specifically relating to a data-driven method and system for monitoring the performance and diagnosing faults of boiler fans. Background Technology

[0002] Since the beginning of the 21st century, the world's energy structure has been continuously transforming towards a cleaner, lower-carbon, more efficient, and diversified direction. However, due to the strong time-varying characteristics of renewable energy sources such as wind and solar power, my country faces difficulties in absorbing renewable energy power generation, resulting in serious problems of wind and solar curtailment. Improving the operational flexibility of conventional thermal power systems to provide absorption services for renewable energy power generation is an important technological direction. Thermal power will shift from a primary energy source to a basic energy source, and the load variation range and frequency of thermal power units will gradually increase. Therefore, thermal power systems will be operating under transient conditions of frequent load variations for a long period of time.

[0003] As a core subsystem ensuring stable, safe, and economical combustion in the furnace, the reliability and stability of boiler fans are becoming increasingly important. The boiler fans used in many in-service coal-fired power generating units generally include primary air fans, forced draft fans, and induced draft fans. The operational stability and safety of the core equipment of the boiler fans are central to the operational reliability of coal-fired power generating units.

[0004] To ensure the safe operation of boiler fans during unit operation, a reliable boiler fan simulation model needs to be established to achieve online performance monitoring. Simultaneously, a boiler fan fault diagnosis model needs to be established to diagnose faults in key parameters of the boiler fan, enabling early warning of faults and determination of their root causes. This further protects the various components of the boiler fan and provides operation and maintenance recommendations. However, due to the complexity of boiler fan modeling and the diversity of faults, existing technologies lack sufficient real-time simulation accuracy for boiler fans and have shortcomings in early warning of system faults and determination of their root causes.

[0005] In the invention patent with publication number CN117195623A, a method for analyzing boiler tube wall temperature based on finite element method is proposed. The method mainly includes: acquiring boiler data; establishing a boiler tube wall thermal stress model based on the boiler data; obtaining a boiler tube wall stress field model based on the boiler tube wall thermal stress model; calculating the boiler tube wall temperature and pressure parameters under different operating conditions using finite element analysis based on the calculated tube wall temperature and pressure parameters; and establishing a corresponding data model based on the calculated tube wall temperature and pressure parameters; comparing the real-time operating temperature and pressure parameters of the boiler with the corresponding parameters in the data model; and providing an early warning of the boiler tube wall temperature based on the absolute value of the temperature difference. This invention improves the accuracy of boiler tube wall temperature analysis and reduces the incidence of boiler accidents. However, this invention mainly focuses on boiler tube wall temperature analysis, establishing a thermal stress model using the finite element method, but it does not involve boiler fans. Furthermore, finite element analysis requires multiple iterations to calculate the temperature and stress fields under different operating conditions, resulting in a large computational load, which is difficult to meet the needs of real-time online monitoring of fans. Under transient conditions of varying load, the response speed of the traditional finite element method cannot match the characteristics of rapid parameter changes; moreover, the early warning mechanism mentioned in the proposed scheme is based only on the absolute value of the pipe wall temperature difference and does not integrate multi-dimensional parameters such as fan vibration and bearing temperature, so it cannot fully reflect fan failures.

[0006] In the invention patent with publication number CN119914876A, a method and system for detecting and warning of boiler wear and explosion prevention is proposed. The method includes: deploying sensors to collect boiler operating data; calculating the temperature field inside the boiler using a thermal-fluid coupling equation; calculating the flow velocity field inside the boiler using fluid dynamics equations; constructing a damage degree equation based on the temperature field, flow velocity field, and stress field distribution of boiler components in the working environment to assess the damage status of boiler components; and determining the damage risk level of boiler components based on the calculation results of the damage degree equation, executing risk warning operations, and adjusting the boiler's operating parameters. The boiler wear and explosion prevention detection and warning method provided by this invention comprehensively considers the multi-physics field effects inside the boiler, enabling more accurate assessment of the damage state and potential risks of boiler components. By employing the finite element method to finely solve the temperature field and flow velocity field, combined with the stress field distribution, the damage degree of the boiler is dynamically assessed, reflecting the health status of boiler components in real time. This invention calculates the temperature and velocity fields using thermal-fluid coupling equations and fluid dynamics equations. However, it primarily focuses on the overall boiler structure and does not model the aerodynamic characteristics of the fan, nor does it consider the impact of fan blade wear on airflow distribution. Therefore, it cannot be directly used for fan fault diagnosis. Its damage degree equation integrates stress, temperature, and velocity, but it does not set corresponding parameters for fan-specific faults. This can lead to missed warnings in fan fault diagnosis and early warning systems. While it employs adaptive mesh refinement to improve accuracy, the computational load of the refined mesh increases significantly, making real-time parameter updates impossible during transient load changes in the fan, resulting in warning delays.

[0007] In summary, neither of the two existing technologies mentioned above achieves multi-dimensional real-time fault monitoring for the transient and transient characteristics of boiler fans. Summary of the Invention

[0008] This invention provides a data-driven method and system for monitoring the performance and diagnosing faults of boiler fans, in order to solve the technical problems of insufficient real-time simulation accuracy of boiler fans in the prior art, and the deficiencies in early warning of faults and determination of fault root causes within the system.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: A data-driven method for monitoring and diagnosing the performance and faults of boiler fans includes the following steps: Collect real-time operating data of boiler fans to form input observation vectors, obtain historical operating data of boiler fans, and establish a dynamic historical memory matrix; Based on the dynamic historical memory matrix, calculate the abnormal weights of each parameter of the boiler fan; perform boiler fan operation simulation to obtain simulation operation data of each parameter of the boiler fan; based on the abnormal weights of each parameter of the boiler fan and the simulation operation data, calculate the comprehensive deviation and average deviation between the historical operation data and the simulation operation data of the boiler fan; calculate the alarm threshold based on the average deviation. Based on the input observation vector, calculate the real-time comprehensive deviation. Based on the real-time comprehensive deviation and the alarm threshold, issue an early warning. Based on the real-time comprehensive deviation, calculate the contribution of each parameter deviation to the comprehensive deviation. The abnormal parameters in the boiler fan are adjusted according to the contribution of each parameter deviation to the overall deviation.

[0010] The establishment of the dynamic historical memory matrix is ​​specifically as follows: construct a historical dataset T from the historical operating data of the boiler fan, calculate the Euclidean distance between the data in the historical dataset T and each input observation vector, finally select the data in the historical dataset T that has the closest Euclidean distance to the real-time input observation vector, and construct a dynamic historical memory matrix based on the selected data.

[0011] The process of constructing a historical dataset T from the historical operating data of the boiler fan is as follows: the historical operating data is divided into three power ranges according to the power generation capacity of the coal-fired boiler unit, and 1000 sets of historical data are selected at equal intervals from different power ranges to form a dataset T.

[0012] The correlation coefficients between the variables are calculated using the acquired real-time operating data of the boiler fan. The calculation formula is as follows:

[0013] in, for One of the variables in the constantly collected data of the boiler fan. , for Another variable in the constantly collected data of the boiler fan. , For variables The sample mean, For variables The sample mean.

[0014] The calculation of the anomaly weights of each parameter of the boiler fan based on the dynamic historical memory matrix is ​​as follows:

[0015] in, Assigning coefficients to abnormal weights. This refers to the number of times a certain parameter is an anomaly within a historical period. It is the sum of the number of anomalies for all parameters within a historical period.

[0016] The average deviation between historical operating data and simulated operating data of the boiler fan is calculated. A comprehensive deviation calculation is then performed based on the combined deviation between the historical and simulated operating data of the boiler fan. The method for calculating the comprehensive deviation between the historical and simulated operating data of the boiler fan is as follows:

[0017]

[0018] in, The abnormal weights for each parameter of the boiler fan. The relative error between historical operating data and simulation values ​​for each parameter. These are the simulated values ​​of the parameters. Historical running data for parameters.

[0019] The average deviation is calculated based on the combined deviation of historical operating data and simulated operating data of the boiler fan. The calculation method is as follows:

[0020] in, The amount of data for comprehensive deviation The combined deviation between the simulated value of parameter i and historical operating data.

[0021] The calculation of the alarm threshold based on the average deviation specifically involves: setting an alarm threshold coefficient, and calculating the alarm threshold based on the alarm threshold coefficient and the average deviation.

[0022] In the formula: This refers to the alarm threshold coefficient of the fault diagnosis model. This represents the average deviation.

[0023] The process involves comparing the real-time overall deviation with the alarm threshold. Based on the comparison result, a corresponding early warning is triggered. Furthermore, the contribution of each parameter deviation to the overall deviation is calculated based on the real-time overall deviation. Specifically, when the real-time overall deviation is lower than the alarm threshold, a fault alarm is triggered, and fault root cause determination begins. The contribution of each parameter deviation to the overall deviation is calculated accordingly. The calculation is as follows:

[0024] in, The abnormal weights for each parameter of the boiler fan. This represents the relative error between historical operating data and simulation values ​​for each parameter.

[0025] A data-driven boiler fan performance monitoring and fault diagnosis system includes a data acquisition module, a simulation module, a fault diagnosis module, and a control execution module. The data acquisition module is used to collect real-time operating data of the boiler fan, form an input observation vector, obtain historical operating data of the boiler fan, and establish a dynamic historical memory matrix. The simulation module is used to calculate the abnormal weights of each parameter of the boiler fan based on the dynamic historical memory matrix; to perform boiler fan operation simulation to obtain simulation operation data of each parameter of the boiler fan; to calculate the comprehensive deviation and average deviation between the historical operation data and the simulation operation data of the boiler fan based on the abnormal weights of each parameter of the boiler fan and the simulation operation data; and to calculate the alarm threshold based on the average deviation. The fault diagnosis module is used to calculate the real-time comprehensive deviation based on the input observation vector, issue an early warning based on the real-time comprehensive deviation and the alarm threshold, and calculate the contribution of each parameter deviation to the comprehensive deviation based on the real-time comprehensive deviation. The control execution module is used to adjust abnormal parameters in the boiler fan based on the contribution of each parameter deviation to the overall deviation.

[0026] Compared with the prior art, the present invention has the following beneficial effects: This invention utilizes a data-driven dynamic memory matrix to construct historical data across power generation ranges, accurately matching varying load conditions and significantly improving the simulation accuracy of boiler fans under different loads, thus solving the problem of insufficient real-time simulation accuracy in existing technologies. By employing variable correlation analysis to eliminate redundant parameters and assigning abnormal weights based on abnormal frequencies, it enhances sensitivity to key fault parameters, making fault identification more accurate. Through comprehensive deviation and contribution analysis, it overcomes the limitations of traditional single-indicator early warning systems, enabling precise location of abnormal parameters and rapid determination of fault root causes. The control execution module adjusts abnormal parameters in conjunction with their contribution, forming a closed-loop management system from monitoring and diagnosis to control, shortening fault handling time, reducing maintenance costs, and effectively improving the reliability and safety of boiler fan operation, providing strong support for the stable operation of thermal power systems.

[0027] Historical data is partitioned by power generation capacity, with 1000 data sets selected at equal intervals in each interval. A dynamic historical memory matrix is ​​constructed by filtering the data most similar to real-time observations through the Euclidean distance between the centralized historical data and each input observation vector. Through power partitioning, historical operating conditions can be accurately matched, significantly improving simulation accuracy. The dynamic historical memory matrix filters only similar historical data, reducing redundant calculations, shortening model response time, and meeting real-time monitoring requirements.

[0028] The correlation coefficients between variables are calculated from the acquired data, and redundant variables with high correlation are eliminated to reduce the impact of duplicate parameters on fault diagnosis. This allows wind turbine performance monitoring to focus more on key variables and improves the accuracy of fault identification.

[0029] Based on the dynamic historical memory matrix, the abnormal weights of each parameter of the boiler fan are calculated. That is, the abnormal weights are assigned according to the historical abnormal frequency of the parameters. Parameters with frequent abnormalities are given higher abnormal weights. Parameters that have frequently been abnormal in history are given more attention, and potential faults are warned in advance. This avoids the neglect of key parameters by traditional abnormal weighting methods and improves the fault detection coverage.

[0030] By integrating multiple parameter errors through comprehensive deviation and quantifying the impact of each parameter on the deviation through contribution metric, the limitations of traditional single-indicator early warning are overcome. This allows for direct identification of specific abnormal parameters, and the alarm threshold is automatically adjusted according to operating conditions, avoiding false alarms or missed alarms caused by fixed thresholds and improving the reliability of early warning.

[0031] Based on the contribution of each parameter deviation to the overall deviation, abnormal parameters in the boiler fan are adjusted to form a closed loop from fault detection to automatic control, shortening the time for manual intervention and reducing the risk of equipment damage; after accurately locating the fault source, blind repairs are avoided, the number of unplanned shutdowns is reduced, and maintenance resources are saved. Attached Figure Description

[0032] Figure 1 : Schematic diagram of the dynamic historical memory matrix construction process; Figure 2 A schematic diagram of a data-driven method for monitoring and diagnosing the performance of boiler fans; Figure 3 : A schematic diagram of a data-driven boiler fan performance monitoring and fault diagnosis system module. Detailed Implementation

[0033] To further understand the content of this invention, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments are merely illustrative and not limiting of the invention.

[0034] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0035] Example 1 like Figure 2 As shown in the figure, this embodiment proposes a data-driven method for monitoring and diagnosing the performance of boiler fans, including the following steps: Collect real-time operating data of boiler fans to form input observation vectors, obtain historical operating data of boiler fans, and establish a dynamic historical memory matrix; Based on the dynamic historical memory matrix, calculate the abnormal weights of each parameter of the boiler fan; perform boiler fan operation simulation to obtain simulation operation data of each parameter of the boiler fan; based on the abnormal weights of each parameter of the boiler fan and the simulation operation data, calculate the comprehensive deviation and average deviation between the historical operation data and the simulation operation data of the boiler fan; calculate the alarm threshold based on the average deviation. Based on the input observation vector, calculate the real-time comprehensive deviation. Based on the real-time comprehensive deviation and the alarm threshold, issue an early warning. Based on the real-time comprehensive deviation, calculate the contribution of each parameter deviation to the comprehensive deviation. The abnormal parameters in the boiler fan are adjusted according to the contribution of each parameter deviation to the overall deviation.

[0036] Based on the above steps, this embodiment simulates the operation of the boiler fan to obtain simulated operation data. Using historical and simulated operation data, alarm thresholds for the boiler fan's operating parameters are calculated. Early fault warnings are then issued based on these alarm thresholds. Furthermore, the contribution of parameter deviations to the overall deviation is obtained based on the combined deviation of the historical and simulated operation data. The root cause of boiler fan failures is determined based on the contribution of each parameter. This embodiment provides a detailed explanation of the above method, and its specific implementation method is as follows: The operating data of the boiler fan is collected in real time from the distributed control system (DCS) of the thermal power plant, and the historical operating data of the boiler fan is obtained through the distributed control system (DCS) of the thermal power plant. The types of operating data include fan bearing temperature, fan motor front and rear bearing temperature, fan bearing X-axis vibration, fan bearing Y-axis vibration, fan lubricating oil pressure, fan motor lubricating oil pressure, fan lubricating oil temperature, fan blade opening, fan current, fan outlet air pressure and air volume, and fan outlet air temperature.

[0037] The real-time and historical operating data of the boiler fan exhibit strong correlation due to the strong interrelationships among the various variables. This embodiment calculates the correlation coefficient between the parameters in the boiler fan operating data to remove non-linearly correlated variables and reduce data interference. The calculation formula is as follows:

[0038] in, for One of the variables in the constantly collected data of the boiler fan. , for Another variable in the constantly collected data of the boiler fan. , For variables The sample mean, For variables The sample mean. When two variables and If the correlation coefficient is greater than 0.8, the two variables can be considered to be strongly correlated, and they can be removed to significantly reduce the number of variables.

[0039] Furthermore, the data after correlation coefficient calculation undergoes data preprocessing to remove outliers and noisy data, and is subjected to data standardization and variable outlier weight analysis. The specific data preprocessing is as follows: Real-time and historical operating data of boiler fans were analyzed to screen for missing, outlier, and erroneous values. Missing values ​​were filled, outliers were corrected, and erroneous values ​​were removed. A sliding window method was used to detect outliers, and missing values ​​were filled using linear interpolation or the mean of adjacent data. Time series data were smoothed using a Savitzky-Golay filter to eliminate high-frequency noise. Because the variables in the historical operating data of boiler fans have different dimensions and significant differences in absolute values, data standardization was necessary. The z-score standardization method was used, and the z-score standardization formula is as follows:

[0040] In the formula: x is a variable; The average value of x; Let x be the standard deviation.

[0041] Based on the preprocessed data, the real-time operating data of the boiler fan is organized into an input observation vector. This serves as the core input for subsequent dynamic memory matrix matching and fault diagnosis. A dynamic historical memory matrix is ​​established based on the historical operating data of the boiler blower. The establishment of a dynamic historical memory matrix like Figure 1 As shown, based on the power generation capacity of the coal-fired boiler unit, the historical data is divided into three parts at equal intervals. Within each part, 1000 data points are selected at equal intervals to form data sets for the three power ranges, resulting in a historical data set T. Then, the historical data set T is used in conjunction with the input observation vector. Calculate the Euclidean distance between vectors in the historical data set T and the input observation vector X, sort the vectors in T according to the Euclidean distance, and select the n vectors with the smallest distance in the historical data set T to form a dynamic historical memory matrix. This method is used to construct a dynamic historical memory matrix. This not only shortens the calculation time for fault diagnosis, but also improves the calculation accuracy of fault diagnosis.

[0042] Based on the dynamic historical memory matrix constructed above Considering the abnormal frequency of various parameters of the actual boiler fan, higher abnormal weights should be set for parameters with high abnormal frequencies. The abnormal weights of each parameter of the boiler fan are calculated based on the actual number of abnormalities. The calculation formula for the abnormal weights of each parameter of the boiler fan is as follows:

[0043] in, Assigning coefficients to abnormal weights. This refers to the number of times a certain parameter is an anomaly within a historical period. It is the sum of the number of anomalies for all parameters within a historical period.

[0044] Based on historical operating data of the boiler fan, a boiler fan simulation model is established to simulate boiler fan operation and obtain simulation data during the operation process. According to the abnormal weights of various boiler fan parameters and the simulation data, a boiler fan fault diagnosis model is established to calculate indicators such as comprehensive deviation, average deviation, and alarm thresholds, and to locate the parameters where faults occur. First, the average deviation between the historical operating data and the simulated operating data of the boiler fan is calculated. Then, a comprehensive deviation calculation is performed based on the combined deviation between the historical operating data and the simulated operating data of the boiler fan. The method for calculating the comprehensive deviation between the historical operating data and the simulated operating data of the boiler fan is as follows:

[0045]

[0046] in, The abnormal weights for each parameter of the boiler fan. The relative error between historical operating data and simulation values ​​for each parameter. These are the simulated values ​​of the parameters. Historical running data for parameters.

[0047] Subsequently, the average deviation is calculated based on the combined deviation between historical operating data and simulated operating data of the boiler fan. The calculation method is as follows:

[0048] in, The amount of data for comprehensive deviation The combined deviation between the simulated value of parameter i and historical operating data.

[0049] Set an alarm threshold coefficient, and calculate the alarm threshold based on the alarm threshold coefficient and the average deviation:

[0050] In the formula: This refers to the alarm threshold coefficient of the fault diagnosis model. This represents the average deviation.

[0051] Finally, based on the input observation vector of the real-time operating data of the boiler fan... The system calculates the real-time comprehensive deviation between the boiler fan's real-time operating data and the simulated operating data. This real-time comprehensive deviation is compared to an alarm threshold. When the real-time comprehensive deviation falls below the alarm threshold, a fault alarm is triggered, and the tracking of abnormal thermal parameters begins, initiating fault root cause determination. This fault root cause determination is primarily based on two dimensions: the absolute value of each parameter's deviation's contribution to the comprehensive deviation and its growth rate. If a parameter's deviation's contribution to the comprehensive deviation increases rapidly and exceeds 0.95 at the time of alarm issuance, that parameter is identified as abnormal. The contribution of each parameter's deviation to the comprehensive deviation is then considered. The calculation is as follows:

[0052] in, The abnormal weights for each parameter of the boiler fan. This represents the relative error between historical operating data and simulated values ​​for each parameter. The contribution of each parameter's deviation to the overall deviation is monitored. The size of the parameter can be used to locate the fault.

[0053] The boiler fan fault diagnosis model feeds back the specific parameter data of the fault to the distributed control system (DCS) of the thermal power plant, actively adjusts the control system corresponding to the abnormal parameters, and thus realizes fault diagnosis and closed-loop control of the boiler fan. At the same time, it transmits the fault information to the unit operation and management personnel, provides the fault cause analysis results, and provides the unit operation and management personnel with suggested equipment operation and maintenance solutions.

[0054] Example 2 This implementation is based on the data-driven boiler fan performance testing and fault diagnosis method proposed in Example 1. It relies on data from the distributed control system of a thermal power plant and is composed of a data acquisition module, a simulation module, a fault diagnosis module, and a control execution module working together. Figure 3 The functions and interaction logic of each module are as follows: The data acquisition module is used to collect real-time operating data of the boiler fan, form an input observation vector, obtain historical operating data of the boiler fan, and establish a dynamic historical memory matrix. The simulation module is used to calculate the abnormal weights of each parameter of the boiler fan based on the dynamic historical memory matrix; to perform boiler fan operation simulation to obtain simulation operation data of each parameter of the boiler fan; to calculate the average deviation between the historical operation data and the simulation operation data of the boiler fan based on the abnormal weights of each parameter of the boiler fan and the simulation operation data; and to calculate the alarm threshold based on the average deviation. The fault diagnosis module is used to calculate the real-time comprehensive deviation based on the input observation vector, issue an early warning based on the real-time comprehensive deviation and the alarm threshold, and calculate the contribution of each parameter deviation to the comprehensive deviation. The control execution module is used to adjust abnormal parameters in the boiler fan based on the contribution of each parameter deviation to the overall deviation.

[0055] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can be appropriately combined to form other embodiments that can be understood by those skilled in the art. The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A data-driven method for monitoring and diagnosing the performance and faults of boiler fans, characterized in that, Includes the following steps: Collect real-time operating data of boiler fans to form input observation vectors, obtain historical operating data of boiler fans, and establish a dynamic historical memory matrix; Based on the dynamic historical memory matrix, calculate the abnormal weights of each parameter of the boiler fan; Perform boiler fan operation simulation to obtain simulation operation data of various parameters of the boiler fan; calculate the average deviation between historical operation data and simulation operation data of the boiler fan based on the abnormal weight of each parameter and the simulation operation data; calculate the alarm threshold based on the average deviation. Based on the input observation vector, calculate the real-time comprehensive deviation. Based on the real-time comprehensive deviation and the alarm threshold, issue an early warning and calculate the contribution of each parameter deviation to the comprehensive deviation. Based on the contribution of each parameter deviation to the overall deviation, abnormal parameters in the boiler fan are adjusted.

2. The data-driven method for monitoring and diagnosing the performance of boiler fans according to claim 1, characterized in that, The establishment of the dynamic historical memory matrix is ​​specifically as follows: construct a historical dataset T from the historical operating data of the boiler fan, calculate the Euclidean distance between the data in the historical dataset T and each input observation vector, finally select the data in the historical dataset T that has the closest Euclidean distance to the real-time input observation vector, and construct a dynamic historical memory matrix based on the selected data.

3. The data-driven method for monitoring and diagnosing the performance of boiler fans according to claim 2, characterized in that, The process of constructing a historical dataset T from the historical operating data of the boiler fan is as follows: the historical operating data is divided into three power ranges according to the power generation capacity of the coal-fired boiler unit, and 1000 sets of historical data are selected at equal intervals from different power ranges to form a dataset T.

4. The data-driven method for monitoring and diagnosing the performance of boiler fans according to claim 1, characterized in that, The correlation coefficients between the variables are calculated using the acquired real-time operating data of the boiler fan. The calculation formula is as follows: in, for One of the variables in the constantly collected data of the boiler fan. , for Another variable in the constantly collected data of the boiler fan. , For variables The sample mean, For variables The sample mean.

5. The data-driven method for monitoring and diagnosing the performance of boiler fans according to claim 1, characterized in that, The calculation of the anomaly weights of each parameter of the boiler fan based on the dynamic historical memory matrix is ​​as follows: in, Assigning coefficients to abnormal weights. This refers to the number of times a certain parameter is an anomaly within a historical period. It is the sum of the number of anomalies for all parameters within a historical period.

6. The data-driven method for monitoring and diagnosing the performance of boiler fans according to claim 1, characterized in that, The average deviation between historical operating data and simulated operating data of the boiler fan is calculated. A comprehensive deviation calculation is then performed based on the combined deviation between the historical and simulated operating data of the boiler fan. The method for calculating the comprehensive deviation between the historical and simulated operating data of the boiler fan is as follows: in, The abnormal weights for each parameter of the boiler fan. The relative error between historical operating data and simulation values ​​for each parameter. These are the simulated values ​​of the parameters. Historical running data for parameters.

7. The data-driven method for monitoring and diagnosing the performance of boiler fans according to claim 6, characterized in that, The average deviation is calculated based on the combined deviation of historical operating data and simulated operating data of the boiler fan. The calculation method is as follows: in, The amount of data for comprehensive deviation The combined deviation between the simulated value of parameter i and historical operating data.

8. The data-driven method for monitoring and diagnosing the performance of boiler fans according to claim 7, characterized in that, The calculation of the alarm threshold based on the average deviation specifically involves: setting an alarm threshold coefficient, and calculating the alarm threshold based on the alarm threshold coefficient and the average deviation. In the formula: This refers to the alarm threshold coefficient for the fault diagnosis model. This represents the average deviation.

9. The data-driven method for monitoring and diagnosing the performance of boiler fans according to claim 1, characterized in that, The process involves comparing the real-time overall deviation with the alarm threshold. Based on the comparison result, a corresponding early warning is triggered. Furthermore, the contribution of each parameter deviation to the overall deviation is calculated based on the real-time overall deviation. Specifically, when the real-time overall deviation is lower than the alarm threshold, a fault alarm is triggered, and fault root cause determination begins. The contribution of each parameter deviation to the overall deviation is calculated accordingly. The calculation is as follows: in, The abnormal weights for each parameter of the boiler fan. This represents the relative error between the historical operating data and the simulation values ​​of each parameter.

10. A data-driven boiler fan performance monitoring and fault diagnosis system, based on the data-driven boiler fan performance monitoring and fault diagnosis method according to any one of claims 1 to 9, characterized in that, It includes a data acquisition module, a simulation module, a fault diagnosis module, and a control execution module; The data acquisition module is used to collect real-time operating data of the boiler fan, form an input observation vector, obtain historical operating data of the boiler fan, and establish a dynamic historical memory matrix. The simulation module is used to calculate the abnormal weights of each parameter of the boiler fan based on the dynamic historical memory matrix. Perform boiler fan operation simulation to obtain simulation operation data of various parameters of the boiler fan; calculate the average deviation between historical operation data and simulation operation data of the boiler fan based on the abnormal weight of each parameter and the simulation operation data; calculate the alarm threshold based on the average deviation. The fault diagnosis module is used to calculate the real-time comprehensive deviation based on the input observation vector, issue an early warning based on the real-time comprehensive deviation and the alarm threshold, and calculate the contribution of each parameter deviation to the comprehensive deviation. The control execution module is used to adjust abnormal parameters in the boiler fan based on the contribution of each parameter deviation to the overall deviation.

Images