A coal-fired cogeneration unit fault early warning system

Abstract

The present application belongs to the field of coal-fired units, and discloses a coal-fired cogeneration unit fault early warning system. The present application collects the operation parameters of the coal-fired cogeneration unit under different operation conditions, forms unit time sequence data capable of reflecting the change rule of the unit operation state, and constructs an adjacency matrix based on the unit time sequence data, so as to convert the dispersed operation parameters into graph structure data containing thermodynamic constraints, so that the fault early warning process can not only analyze the change of a single operation parameter, but also comprehensively judge in combination with the correlation and thermodynamic connection between the devices of the unit, thereby improving the completeness and accuracy of the unit state representation. Through data cleaning and data arrangement on the graph structure data, noise data and error data can be eliminated, the interference of abnormal collected data on the analysis result is reduced, and a reliable data basis is provided for subsequent model analysis.

Description

Technical Field

[0001] This invention belongs to the field of coal-fired power units, specifically relating to a fault early warning system for coal-fired cogeneration units. Background Technology

[0002] Coal-fired combined heat and power (CHP) units are energy conversion devices that use coal as fuel to generate heat through combustion and simultaneously output electrical and thermal energy. They are characterized by high energy utilization efficiency and stable power supply, and are widely used in industrial production, district heating, and regional energy supply. A typical coal-fired CHP unit includes a boiler, turbine, generator, heating network, and corresponding auxiliary equipment. These devices are interconnected, have numerous operating parameters, and their operating status is affected by factors such as fuel quality, load changes, environmental conditions, and equipment performance degradation.

[0003] During the operation of coal-fired combined heat and power (CHP) units, parameters such as temperature, pressure, flow rate, current, vibration, and load reflect the operating status of the unit and its equipment. When equipment experiences wear, blockage, leakage, overheating, or control abnormalities, these operating parameters typically change to varying degrees. Therefore, real-time monitoring of unit operating data through distributed control systems, sensors, and data acquisition devices, followed by data cleaning, analysis, and comparison, provides a data foundation for equipment status assessment and fault early warning.

[0004] Existing fault warning systems typically rely on manually setting upper and lower limits for operating parameters, issuing alarms when monitored parameters exceed these thresholds. Some systems can also combine historical operating data with regression analysis, time series analysis, or machine learning models to predict equipment operating trends and generate corresponding warning information based on the degree of anomaly. However, coal-fired combined heat and power (CHP) units experience significant fluctuations in their normal operating parameters under various conditions, including start-up and shutdown, load increases and decreases, heating regulation, and changes in coal quality. Fixed thresholds are insufficient to adapt to the full range of operating characteristics of the unit, easily leading to false alarms or missed alarms.

[0005] Furthermore, existing fault early warning systems typically focus on independent analysis of single parameters or individual devices, failing to fully consider the thermodynamic relationships between boilers, turbines, generators, and heating systems. They also struggle to effectively integrate real-time operating data with historical normal data, fault data, and physical models. When equipment operating conditions have not yet exceeded alarm thresholds, but multiple related parameters have already shown coordinated changes, existing systems struggle to identify potential faults in a timely manner, resulting in delayed warnings. Maintenance typically requires further development of the fault or equipment shutdown before repairs can be carried out.

[0006] Therefore, existing fault early warning technologies for coal-fired cogeneration units have problems such as relying on manual setting of early warning thresholds, difficulty in adapting to changes in all operating conditions, insufficient utilization of related parameters, and lag in the identification of potential faults, which can easily lead to untimely fault detection and increase unit downtime for maintenance. Summary of the Invention

[0007] The purpose of this invention is to overcome the problem that existing fault early warning methods for coal-fired cogeneration units are difficult to adapt to changes in all operating conditions and the correlation characteristics between equipment, and to provide a fault early warning system for coal-fired cogeneration units.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a fault early warning method for coal-fired cogeneration units, comprising the following steps: The operating parameters of coal-fired cogeneration units under different operating conditions are collected to form unit time-series data. Based on the unit time-series data, an adjacency matrix is ​​constructed, and the unit time-series data is converted into graph structure data containing thermodynamic constraints. Data cleaning and organization are performed on the graph structure data to obtain the unit data to be analyzed. The data of the unit to be analyzed is input into multiple pre-established fault early warning models, and the analysis results of each fault early warning model are obtained. The target fault early warning model is determined based on the analysis results of each fault early warning model on historical unit data. The target fault early warning model is used to analyze the data of the unit to be analyzed, and the analysis results are compared with the data corresponding to the normal operating status of the unit to obtain the fault diagnosis results. The system determines whether to output a fault alarm based on the relationship between the fault diagnosis results and the warning threshold. When the fault diagnosis results do not reach the warning threshold, the current unit data is fed back to the historical unit database.

[0009] A further improvement of this invention lies in the following method for collecting operating parameters of coal-fired cogeneration units under different operating conditions to form unit time-series data, constructing an adjacency matrix based on the unit time-series data, and converting the unit time-series data into graph-structured data containing thermodynamic constraints: Full-condition data is collected from the distributed control system of the coal-fired cogeneration unit, including temperature, pressure, flow rate, and current. Based on the physical connection relationships and thermodynamic correlation relationships between the various equipment in a coal-fired cogeneration unit, determine the nodes in the adjacency matrix and the connection relationships between the nodes; The unit time series data are mapped to the corresponding nodes, and the adjacency matrix is ​​used as the association constraint between nodes to form a graph structure data containing thermodynamic constraints.

[0010] A further improvement of this invention lies in the following specific method for cleaning and organizing graph-structured data to obtain the data of the unit to be analyzed: Identify and remove noisy and erroneous data from graph-structured data; Based on the node connection relationships represented by the adjacency matrix, the unit time series data from different sensors are correlated and integrated; The cleaned and integrated graph-structured data is converted into continuous variable data to form the unit data to be analyzed.

[0011] A further improvement of this invention lies in that the data of the unit to be analyzed is input into multiple pre-established fault early warning models, and the analysis results of each fault early warning model are obtained. The specific method for determining the target fault early warning model based on the analysis results of each fault early warning model on historical unit data is as follows: Multiple fault early warning models were established using linear regression and ridge regression methods, respectively. Historical unit data are input into each fault early warning model to obtain the historical analysis results corresponding to each fault early warning model. The historical analysis results are compared with the actual operating status corresponding to the historical unit data. Based on the comparison results, a fault early warning model that matches the operating characteristics of the coal-fired cogeneration unit is determined, and this fault early warning model is identified as the target fault early warning model.

[0012] A further improvement of this invention lies in using a target fault early warning model to analyze the data of the unit to be analyzed, and comparing the analysis results with the data corresponding to the normal operating state of the unit to obtain the fault diagnosis results. The specific method is as follows: A mathematical model for unit operation was established based on the physical characteristics of coal-fired cogeneration units. Input the data of the unit to be analyzed into the target fault early warning model to obtain the model output data of the coal-fired cogeneration unit. The model output data and the data of the unit to be analyzed are compared with the data corresponding to the normal operating state of the unit. The fault diagnosis results of the coal-fired cogeneration unit are determined based on the deviation generated by the comparison.

[0013] A further improvement of this invention lies in the following method for determining whether to output fault alarm information based on the relationship between fault diagnosis results and warning thresholds: Compare the fault diagnosis results with the preset warning thresholds; When the fault diagnosis result reaches the warning threshold, output the fault alarm information and record the fault occurrence time, fault location, fault cause, fault handling process and fault handling result; When the fault diagnosis result does not reach the warning threshold, the current unit data is stored in the historical unit database and used as the historical unit data when determining the target fault warning model.

[0014] A further improvement of the present invention is that a fault warning report is generated based on the data of the unit to be analyzed, the fault diagnosis results, and the fault handling results corresponding to the fault alarm information. The fault warning report displays the changing trend of the unit operating parameters in different time periods in at least one of the forms of charts, curves, and heat maps.

[0015] Secondly, the present invention provides a fault early warning system for coal-fired cogeneration units, comprising: The data acquisition module is used to collect operating parameters of coal-fired cogeneration units under different operating conditions, form unit time-series data, construct an adjacency matrix based on the unit time-series data, and convert the unit time-series data into graph structure data containing thermodynamic constraints. The data processing module is used to clean and organize the graph-structured data to obtain the data of the unit to be analyzed. The modeling and comparison module is used to input the data of the unit to be analyzed into multiple pre-established fault warning models, obtain the analysis results of each fault warning model, and determine the target fault warning model based on the analysis results of each fault warning model on historical unit data. The fault diagnosis module is used to analyze the data of the unit to be analyzed using the target fault early warning model, compare the analysis results with the data corresponding to the normal operating status of the unit, and obtain the fault diagnosis results. The early warning processing module is used to determine whether to output fault alarm information based on the relationship between the fault diagnosis result and the early warning threshold. When the fault diagnosis result does not reach the early warning threshold, the current unit data is fed back to the historical unit database.

[0016] A further improvement of this invention is that the data acquisition module is communicatively connected to the distributed control system of the coal-fired cogeneration unit, and is used to collect full-condition data including temperature, pressure, flow rate and current. Based on the physical connection relationship and thermodynamic correlation relationship between the various devices of the coal-fired cogeneration unit, the nodes in the adjacency matrix and the connection relationship between the nodes are determined, and the unit time series data is mapped to the corresponding nodes. The data processing module is used to identify and remove noisy and erroneous data in the graph structure data, and to associate and integrate the unit time series data from different sensors according to the node connection relationship represented by the adjacency matrix to form the unit data to be analyzed.

[0017] A further improvement of the present invention is that the modeling and comparison module includes multiple fault early warning models established using linear regression and ridge regression methods, which are used to compare the historical analysis results corresponding to each fault early warning model with the actual operating status corresponding to the historical unit data, and determine the target fault early warning model based on the comparison results. The fault diagnosis module is used to input the data of the unit to be analyzed into the target fault early warning model, obtain the model output data, and determine the fault diagnosis result based on the deviation between the model output data and the data of the unit to be analyzed relative to the corresponding data of the unit's normal operating state. The early warning processing module is used to output fault alarm information when the fault diagnosis result reaches the early warning threshold, and to store the current unit data in the historical unit database when the fault diagnosis result does not reach the early warning threshold, so as to serve as the historical unit data for subsequent determination of the target fault early warning model.

[0018] Compared with the prior art, the present invention has the following beneficial effects: Based on the full-condition data acquisition, graph structure conversion, multi-model comparison, fault diagnosis, and historical data feedback process defined in claim 1, it can be summarized as follows.

[0019] This invention collects operating parameters of coal-fired cogeneration units under different operating conditions, forming time-series data that reflects the changing patterns of the unit's operating status. Based on this time-series data, an adjacency matrix is ​​constructed, converting the dispersed operating parameters into graph-structured data containing thermodynamic constraints. This allows the fault early warning process to not only analyze changes in individual operating parameters but also to make comprehensive judgments based on the relationships and thermodynamic connections between various equipment within the unit, thereby improving the completeness and accuracy of the unit's status characterization. Data cleaning and processing of the graph-structured data removes noisy and erroneous data, reducing the interference of abnormally collected data on the analysis results and providing a reliable data foundation for subsequent model analysis. By inputting the unit data to be analyzed into multiple fault early warning models and determining the target fault early warning model based on the analysis results of each model on historical unit data, a model with a high degree of adaptability can be selected according to the actual operating characteristics of the unit, avoiding early warning bias caused by using a single fixed model. By comparing the analysis results of the target fault early warning model with data corresponding to the unit's normal operating state, abnormal changes in the unit's operating status can be identified in a timely manner, and fault diagnosis results can be obtained, improving the timeliness of potential fault detection. When the fault diagnosis result does not reach the warning threshold, the current unit data is fed back to the historical unit database, which continuously enriches the data foundation required for model comparison and fault analysis, enabling the fault warning process to adapt to changes in unit operating conditions, thereby reducing false alarms and missed alarms, shortening fault response and downtime maintenance time, and ensuring the safe and stable operation of coal-fired cogeneration units. Attached Figure Description

[0020] Figure 1 This is a flowchart of the present invention; Figure 2 This is a system diagram of the present invention. Detailed Implementation

[0021] 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.

[0022] Example 1: like Figure 1 As shown, the fault early warning method for coal-fired cogeneration units of the present invention includes the following steps: Operating parameters of coal-fired cogeneration units under different operating conditions are collected to form unit time-series data. An adjacency matrix is ​​constructed based on the unit time-series data, and the unit time-series data is converted into graph structure data containing thermodynamic constraints.

[0023] Specifically, the data acquisition module is communicatively connected to the distributed control system of the coal-fired cogeneration unit. Through the distributed control system and sensors deployed on various devices of the coal-fired cogeneration unit, the operating status of the unit is monitored in real time. The collected operating parameters include temperature, pressure, flow rate, and current, and may also include other parameters reflecting the operating status of the boiler, turbine, generator, heating equipment, and auxiliary equipment.

[0024] Different operating conditions include normal unit operation, start-up, shutdown, load increase, load decrease, heating load adjustment, and operation when fuel quality changes. By collecting full-condition data under different operating conditions, we avoid relying solely on a single stable condition to build a fault early warning model, and ensure that the obtained historical unit data can reflect the changing patterns of the operating status of coal-fired cogeneration units over time and under different operating conditions.

[0025] At sampling time The various operating parameters of a coal-fired cogeneration unit are represented as a time-series data vector of the unit:

[0026] in, Indicates the sampling time The unit time-series data vector, Indicates the first Each node at the sampling time The corresponding operating parameters This indicates the number of nodes in the graph structure data.

[0027] The nodes and edges in the graph structure data are determined based on the physical connections, working fluid flow relationships, energy transfer relationships, and thermodynamic correlations among the various equipment in a coal-fired combined heat and power (CHP) unit. Nodes can correspond to boilers, steam turbines, generators, heat exchangers, heating equipment, pumps, fans, pipelines, or corresponding measuring points, while edges represent the physical connections or thermodynamic correlations between nodes.

[0028] The adjacency matrix is ​​denoted as:

[0029] Elements in the adjacency matrix It can be determined according to the following formula: ,node With nodes There are physical or thermodynamic connections between them; ,node With nodes There is no physical connection or thermodynamic relationship between them; in, Represents a node With nodes The correlation weight between them. The correlation weight can be determined based on the connection status between the equipment, the direction of working fluid flow, the heat transfer relationship, and the degree of correlation of the time series data of the corresponding units.

[0030] Thermodynamic constraints in coal-fired combined heat and power (CHP) units include at least mass conservation constraints and energy conservation constraints. In one optional implementation, the mass conservation constraint for the equipment or equipment region corresponding to a node is expressed as:

[0031] The energy conservation constraint is expressed as:

[0032] in, This represents the mass flow rate flowing into the corresponding node. This represents the mass flow rate exiting the corresponding node. Indicates the mass of the working fluid within the node. This indicates the energy of the corresponding node. This indicates the output of the energy of the corresponding node. This represents the stored energy within a node.

[0033] when and When the fluctuations are within the normal range, it indicates that the corresponding node satisfies the preset physical conservation relationship; when or When the fluctuations deviate from the normal range, it indicates that the corresponding node or the equipment connected to it may be malfunctioning. The aforementioned physical conservation relationship is used to constrain the connections and data associations between nodes, so that the resulting graph structure data not only contains the temporal variation characteristics of each measuring point, but also reflects the physical association characteristics between the equipment of the coal-fired cogeneration unit.

[0034] The unit time-series data are mapped to corresponding nodes, and the adjacency matrix is ​​used as the association constraint between nodes to form a graph structure data:

[0035] in, Represents a set of nodes. Denotes the set of edges. Represents the adjacency matrix. Indicates the sampling time of each node The corresponding unit timing data.

[0036] Data cleaning and organization are performed on the graph-structured data to obtain the data of the units to be analyzed.

[0037] The data processing module receives the graph structure data generated by the data acquisition module, performs integrity checks, rationality checks, and continuity checks on the node data in the graph structure data, and identifies and removes noise data and erroneous data caused by sensor failure, communication interruption, signal interference, or sampling abnormalities.

[0038] Data exceeding the reasonable range of corresponding operating parameters is marked as erroneous and removed; data exhibiting frequent fluctuations within a short period and inconsistent with the actual operating patterns of the equipment is marked as noisy and cleaned. During data cleaning, the connections between nodes are preserved to prevent disruption of the thermodynamic relationships represented by the graph structure data when cleaning individual node data.

[0039] Data from different sensors is correlated and integrated based on the node connection relationships represented by the adjacency matrix. For example, when analyzing steam transport status, steam temperature, steam pressure, and steam flow rate are integrated according to the connection relationships between corresponding pipelines and equipment; when analyzing the operating status of pumps or fans, equipment current, inlet and outlet pressure, flow rate, and equipment temperature are correlated. By integrating data from different sensors, the utilization efficiency of unit data is improved, and the operating status of coal-fired cogeneration units is comprehensively reflected from the coordinated changes of multiple operating parameters.

[0040] The cleaned and integrated graph structure data is converted into continuous variable data to form the unit data to be analyzed. The unit data to be analyzed includes both data on the changes of various operating parameters over time and node association data determined by the adjacency matrix and thermodynamic constraints.

[0041] The data processing module can also convert complex raw data into charts and reports, allowing operators and managers to intuitively view the operating status and fault warning information of coal-fired cogeneration units. By synchronizing the data corresponding to the normal operating status with the data of the unit currently being analyzed, abnormal changes can be quickly identified and responded to.

[0042] The data of the unit to be analyzed is input into multiple pre-established fault early warning models, and the analysis results of each fault early warning model are obtained. The target fault early warning model is determined based on the analysis results of each fault early warning model on historical unit data.

[0043] Multiple fault early warning models are pre-established in the modeling and comparison module. In this embodiment, the multiple fault early warning models include a fault early warning model established using the linear regression method and a fault early warning model established using the ridge regression method. Different models are used to model and analyze continuous variables to predict the operating status and remaining life of coal-fired cogeneration units.

[0044] The linear regression model can be expressed as:

[0045] in, This indicates that the linear regression model at the sampling time... The output prediction results, Represents the intercept term. Indicates the first The regression coefficients corresponding to each operating parameter Indicates the sampling time The One running parameter, This indicates the number of parameters used to run the input model.

[0046] Ridge regression models reduce the impact of correlations between different operating parameters on model stability by adding a regularization term to the regression loss function. The parameters of a ridge regression model can be determined using the following formula:

[0047] in, This indicates the actual operating status corresponding to the historical unit data. This represents the parameter estimation results of the ridge regression model. Represents the regularization coefficient. Greater than or equal to 0.

[0048] Historical unit data stored in the historical unit database is input into each fault early warning model to obtain the historical analysis results corresponding to each fault early warning model. The historical unit data includes historical normal operation data, historical abnormal operation data, historical fault data, and current unit data that has not reached the early warning threshold for coal-fired cogeneration units.

[0049] The historical analysis results of each fault early warning model are compared with the actual operating status corresponding to historical unit data to evaluate the performance of different fault early warning models. In one optional implementation, the first... The evaluation error of each fault early warning model is expressed as:

[0050] in, Indicates the first The evaluation error of each fault early warning model This indicates the number of historical samples used for model evaluation. This indicates the actual operating status corresponding to the historical samples. Indicates the first Analysis results of a fault early warning model.

[0051] Based on the evaluation error, model output stability, and adaptability of each fault warning model to the corresponding operating conditions, a fault warning model that matches the operating characteristics of the coal-fired cogeneration unit is determined, and this fault warning model is identified as the target fault warning model.

[0052] For example, when the linear relationship between various operating parameters in historical unit data is relatively stable, a linear regression model can be selected as the target fault early warning model; when there is a strong correlation between multiple operating parameters in historical unit data, a ridge regression model can be selected as the target fault early warning model. By comparing the performance of multiple models on historical unit data, a model suitable for the current unit and current operating conditions can be selected and applied to real-time monitoring and fault early warning.

[0053] The target fault early warning model is used to analyze the data of the unit to be analyzed. The analysis results are compared with the data corresponding to the normal operating status of the unit to obtain the fault diagnosis results.

[0054] Fault diagnosis adopts a data model-driven approach, establishing a mathematical model of unit operation based on the physical characteristics of coal-fired cogeneration units, comparing real-time monitoring data with the model output data of the target fault early warning model, and combining condition monitoring technology to identify operational deviations.

[0055] Input the data of the unit to be analyzed into the target fault early warning model to obtain the model output data of the coal-fired cogeneration unit. For the first The deviation between the real-time monitoring data and the model output data of each operating parameter is expressed as:

[0056] in, Indicates the first Each operating parameter at the sampling time deviation, Indicates the first Real-time monitoring data of each operating parameter This represents the corresponding predicted data output by the target fault early warning model.

[0057] Furthermore, fault diagnosis indicators can be obtained based on the deviations of multiple operating parameters:

[0058] in, Indicates the sampling time The fault diagnosis results Indicates the first The weights corresponding to each running parameter, and the sum of all weights is 1. Indicates the first The standardized coefficients of each operating parameter.

[0059] Weight The weighting can be determined based on the degree of impact of the corresponding operating parameters on the safe operation of the coal-fired cogeneration unit. Parameters that directly reflect the safety status of the boiler, turbine, generator, or heating equipment can be assigned a higher weighting; parameters that auxiliaryly characterize the operating status can be assigned a lower weighting.

[0060] The model output data and the data of the unit under analysis are compared with the data corresponding to the normal operating state of the unit. The fault diagnosis result is determined based on the deviation generated by the comparison. The fault diagnosis result is used to characterize the degree to which the current unit data deviates from the normal operating state and can be used to determine the severity of potential faults.

[0061] Analyzing the operational data of coal-fired cogeneration units provides a comprehensive understanding of the equipment's health status, enabling maintenance personnel to develop more accurate maintenance plans and implement preventative and planned maintenance. Simultaneously, analyzing energy efficiency-related operating parameters identifies inefficient processes and resource waste in the operation of coal-fired cogeneration units, providing a basis for optimizing operational processes and improving energy utilization efficiency.

[0062] The decision to output a fault alarm is based on the relationship between the fault diagnosis results and the warning threshold. When the fault diagnosis results do not reach the warning threshold, the current unit data is fed back to the historical unit database.

[0063] The early warning processing module will process the fault diagnosis results. Compared with the pre-set warning threshold The comparison and early warning judgment process is represented as follows: when When the fault alarm information is displayed, output fault alarm information. when At this time, no fault alarm information will be output, and the current unit data will be fed back to the historical unit database.

[0064] When the fault diagnosis result reaches the warning threshold, it indicates that one or more operating parameters of the coal-fired cogeneration unit have deviated from the normal operating state. The warning processing module controls the alarm to output fault alarm information so that the operator can quickly identify potential faults.

[0065] Fault alarm information may include alarm time, alarm device, alarm location, abnormal operating parameters, fault diagnosis results, and warning threshold. After receiving the fault alarm information, operators will perform maintenance and management of the coal-fired cogeneration unit and record in detail the fault occurrence time, fault location, fault cause, fault handling process, and fault handling results.

[0066] During the maintenance and management process, corresponding maintenance plans are formulated based on the fault diagnosis results and the actual condition of the equipment. Through timely response, analysis and handling, the safe and stable operation of coal-fired cogeneration units is ensured, downtime and maintenance costs caused by faults are reduced, and a data foundation is provided for subsequent technical improvements and management optimization.

[0067] When the fault diagnosis result does not reach the warning threshold, the current unit data is fed back to the modeling comparison module through Unit Data 8 and stored in the historical unit database. The current unit data serves as the historical unit data used for subsequent model training, model comparison, and selection of the target fault warning model.

[0068] The historical unit database categorizes and stores normal operation data, abnormal operation data, fault data, fault handling data, and model analysis results to efficiently organize and retrieve data, and to ensure the availability and integrity of historical unit data. The modeling and comparison module utilizes the historical unit data and fault early warning models in the database for subsequent analysis, enabling the fault early warning models to be continuously evaluated as unit operation data increases, thereby improving the applicability of fault early warnings.

[0069] After alarm and maintenance processing is completed, a fault early warning report is generated based on the unit data to be analyzed, fault diagnosis results, and fault handling results corresponding to the fault alarm information. The fault early warning report can be presented in at least one form, such as charts, curves, and heat maps.

[0070] Curves are used to display the changing trends of temperature, pressure, flow rate, current, and fault diagnosis results over different time periods; heat maps are used to display the degree of abnormality of different equipment, different measuring points, or different operating parameters; charts are used to display the comparison results between current unit data and historical data, optimal operating state data, and target data.

[0071] Example 2: See Figure 2 The present invention also provides a fault early warning system for coal-fired cogeneration units, including a data acquisition module, a data processing module, a modeling and comparison module, a fault diagnosis module, and an early warning processing module.

[0072] The data acquisition module is used to collect operating parameters of coal-fired cogeneration units under different operating conditions, form unit time-series data, construct an adjacency matrix based on the unit time-series data, and convert the unit time-series data into graph structure data containing thermodynamic constraints.

[0073] The data processing module is used to clean and organize the graph-structured data to obtain the unit data to be analyzed. It integrates data from different sensors, transforming complex raw data into continuous variable data, charts, and reports that are easy to analyze.

[0074] The modeling and comparison module is used to input the data of the unit to be analyzed into multiple pre-established fault warning models, obtain the analysis results of each fault warning model, and determine the target fault warning model based on the analysis results of each fault warning model on historical unit data.

[0075] The fault diagnosis module is used to analyze the data of the unit to be analyzed using the target fault early warning model, compare the analysis results with the data corresponding to the normal operating status of the unit, and obtain the fault diagnosis results.

[0076] The early warning processing module determines whether to output a fault alarm based on the relationship between the fault diagnosis results and the early warning threshold. When the fault diagnosis results reach the early warning threshold, the early warning processing module outputs a fault alarm, records equipment maintenance information through maintenance and management, and finally generates a fault early warning report. When the fault diagnosis results do not reach the early warning threshold, the early warning processing module feeds back the current unit data to the modeling comparison through unit data and stores it in the historical unit database.

[0077] In actual operation, this invention collects operational data from coal-fired cogeneration units in real time via network connection and transmits the collected data to the modeling and comparison stage. Data processing and analysis clean, organize, integrate, and analyze the real-time collected data. The modeling and comparison stage determines the target fault warning model based on the performance of multiple fault warning models on historical unit data.

[0078] Fault diagnosis compares the analysis results of the target fault early warning model with the data corresponding to the unit's normal operating status to obtain the fault diagnosis result. The "whether it exceeds the early warning mechanism" compares the fault diagnosis result with the early warning threshold.

[0079] When the fault diagnosis result reaches the warning threshold, an alarm is output and fault alarm information is generated. The operator performs maintenance and management of the unit and records the fault occurrence time, fault location, fault cause, fault handling process and fault handling result. Subsequently, a fault warning report is generated by the report generation department.

[0080] When the fault diagnosis result does not reach the warning threshold, the unit data 8 transmits the current unit data to the modeling comparison 2 and forms a historical unit database, so that the current unit data can be called for comparison during subsequent data processing, model comparison and fault analysis to identify potential faults.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for early warning of faults in a coal-fired combined heat and power (CHP) unit, characterized in that, Includes the following steps: The operating parameters of coal-fired cogeneration units under different operating conditions are collected to form unit time-series data. Based on the unit time-series data, an adjacency matrix is ​​constructed, and the unit time-series data is converted into graph structure data containing thermodynamic constraints. Data cleaning and organization are performed on the graph structure data to obtain the unit data to be analyzed. The data of the unit to be analyzed is input into multiple pre-established fault early warning models, and the analysis results of each fault early warning model are obtained. The target fault early warning model is determined based on the analysis results of each fault early warning model on historical unit data. The target fault early warning model is used to analyze the data of the unit to be analyzed, and the analysis results are compared with the data corresponding to the normal operating status of the unit to obtain the fault diagnosis results. The system determines whether to output a fault alarm based on the relationship between the fault diagnosis results and the warning threshold. When the fault diagnosis results do not reach the warning threshold, the current unit data is fed back to the historical unit database.

2. The fault early warning method for a coal-fired cogeneration unit according to claim 1, characterized in that, The specific method for collecting operating parameters of coal-fired cogeneration units under different operating conditions to form unit time-series data, constructing an adjacency matrix based on the unit time-series data, and converting the unit time-series data into graph structure data containing thermodynamic constraints is as follows: Full-condition data is collected from the distributed control system of the coal-fired cogeneration unit, including temperature, pressure, flow rate, and current. Based on the physical connection relationships and thermodynamic correlation relationships between the various equipment in a coal-fired cogeneration unit, determine the nodes in the adjacency matrix and the connection relationships between the nodes; The unit time series data are mapped to the corresponding nodes, and the adjacency matrix is ​​used as the association constraint between nodes to form a graph structure data containing thermodynamic constraints.

3. The fault early warning method for a coal-fired cogeneration unit according to claim 1, characterized in that, The specific methods for cleaning and organizing graph-structured data to obtain the unit data to be analyzed are as follows: Identify and remove noisy and erroneous data from graph-structured data; Based on the node connection relationships represented by the adjacency matrix, the unit time series data from different sensors are correlated and integrated; The cleaned and integrated graph-structured data is converted into continuous variable data to form the unit data to be analyzed.

4. The fault early warning method for a coal-fired cogeneration unit according to claim 1, characterized in that, The data of the unit to be analyzed is input into multiple pre-established fault early warning models, and the analysis results of each fault early warning model are obtained. The specific method for determining the target fault early warning model based on the analysis results of each fault early warning model on historical unit data is as follows: Multiple fault early warning models were established using linear regression and ridge regression methods, respectively. Historical unit data are input into each fault early warning model to obtain the historical analysis results corresponding to each fault early warning model. The historical analysis results are compared with the actual operating status corresponding to the historical unit data. Based on the comparison results, a fault early warning model that matches the operating characteristics of the coal-fired cogeneration unit is determined, and this fault early warning model is identified as the target fault early warning model.

5. A fault early warning method for a coal-fired cogeneration unit according to claim 1, characterized in that, The specific method for obtaining fault diagnosis results by analyzing the data of the unit under analysis using the target fault early warning model and comparing the analysis results with the data corresponding to the normal operating state of the unit is as follows: A mathematical model for unit operation was established based on the physical characteristics of coal-fired cogeneration units. Input the data of the unit to be analyzed into the target fault early warning model to obtain the model output data of the coal-fired cogeneration unit. The model output data and the data of the unit to be analyzed are compared with the data corresponding to the normal operating state of the unit. The fault diagnosis results of the coal-fired cogeneration unit are determined based on the deviation generated by the comparison.

6. The fault early warning method for a coal-fired cogeneration unit according to claim 1, characterized in that, The specific method for determining whether to output fault alarm information based on the relationship between fault diagnosis results and early warning thresholds is as follows: Compare the fault diagnosis results with the preset warning thresholds; When the fault diagnosis result reaches the warning threshold, output the fault alarm information and record the fault occurrence time, fault location, fault cause, fault handling process and fault handling result; When the fault diagnosis result does not reach the warning threshold, the current unit data is stored in the historical unit database and used as the historical unit data when determining the target fault warning model.

7. A fault early warning method for a coal-fired cogeneration unit according to claim 6, characterized in that, A fault warning report is generated based on the unit data to be analyzed, the fault diagnosis results, and the fault handling results corresponding to the fault alarm information. The fault warning report displays the changing trend of unit operating parameters in different time periods using at least one of the following forms: charts, curves, and heat maps.

8. A fault early warning system for a coal-fired cogeneration unit, characterized in that, include: The data acquisition module is used to collect operating parameters of coal-fired cogeneration units under different operating conditions, form unit time-series data, construct an adjacency matrix based on the unit time-series data, and convert the unit time-series data into graph structure data containing thermodynamic constraints. The data processing module is used to clean and organize the graph-structured data to obtain the data of the unit to be analyzed. The modeling and comparison module is used to input the data of the unit to be analyzed into multiple pre-established fault warning models, obtain the analysis results of each fault warning model, and determine the target fault warning model based on the analysis results of each fault warning model on historical unit data. The fault diagnosis module is used to analyze the data of the unit to be analyzed using the target fault early warning model, compare the analysis results with the data corresponding to the normal operating status of the unit, and obtain the fault diagnosis results. The early warning processing module is used to determine whether to output fault alarm information based on the relationship between the fault diagnosis result and the early warning threshold. When the fault diagnosis result does not reach the early warning threshold, the current unit data is fed back to the historical unit database.

9. A fault early warning system for a coal-fired cogeneration unit according to claim 8, characterized in that, The data acquisition module communicates with the distributed control system of the coal-fired cogeneration unit to collect full-condition data including temperature, pressure, flow rate and current. Based on the physical connection relationship and thermodynamic correlation relationship between the equipment of the coal-fired cogeneration unit, it determines the nodes in the adjacency matrix and the connection relationship between the nodes, and maps the unit time series data to the corresponding nodes. The data processing module is used to identify and remove noisy and erroneous data in the graph structure data, and to associate and integrate the unit time series data from different sensors according to the node connection relationship represented by the adjacency matrix to form the unit data to be analyzed.

10. A fault early warning system for a coal-fired cogeneration unit according to claim 8, characterized in that, The modeling and comparison module includes multiple fault early warning models established using linear regression and ridge regression methods. These models are used to compare the historical analysis results corresponding to each fault early warning model with the actual operating status corresponding to historical unit data, and to determine the target fault early warning model based on the comparison results. The fault diagnosis module is used to input the data of the unit to be analyzed into the target fault early warning model, obtain the model output data, and determine the fault diagnosis result based on the deviation between the model output data and the data of the unit to be analyzed relative to the corresponding data of the unit's normal operating state. The early warning processing module is used to output fault alarm information when the fault diagnosis result reaches the early warning threshold, and to store the current unit data in the historical unit database when the fault diagnosis result does not reach the early warning threshold, so as to serve as the historical unit data for subsequent determination of the target fault early warning model.

Images