Thermal power plant immersion type equipment state monitoring method and system based on AI

Abstract

The invention discloses a thermal power plant immersive equipment state monitoring method and system based on AI, and relates to the technical field of equipment monitoring, and the method comprises the steps: building a digital twinborn model of a thermal power plant after a monitoring sensor is deployed on equipment of the thermal power plant; executing attraction clustering according to an equipment topological relation in the digital twin model, and establishing a multi-level clustering result; reading a monitoring data set of the monitoring sensor; activating an independent anomaly recognition network, and establishing an independent anomaly recognition result; establishing equipment criticality according to the independent anomaly recognition result; establishing equipment abnormity joint early warning; and carrying out immersive visual presentation. The technical problems that in the prior art, the operation state of the thermal power plant equipment is not comprehensively and accurately monitored, the equipment management efficiency is low, and the reliability is difficult to guarantee are solved, and the technical effects that the operation state of the thermal power plant equipment is comprehensively and accurately monitored, and the efficiency and the reliability of equipment management are effectively improved are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment monitoring, and in particular to an AI-based immersive equipment status monitoring method and system for a thermal power plant. Background Art

[0002] As a key place for power production, the safe and stable operation of thermal power plant equipment is crucial to the reliability and continuity of power supply. Under existing technical conditions, thermal power plant equipment monitoring faces many challenges. Traditional monitoring methods often rely on manual inspections and limited sensor data collection, and there are problems such as incomplete monitoring coverage and low data collection frequency, which make it impossible to capture subtle changes in the operating status of equipment in a timely manner and to discover potential faults in a timely manner. At the same time, in terms of equipment management, there is a lack of effective data analysis and management strategies. It is difficult to conduct in-depth mining and comprehensive analysis of a large amount of monitoring data, and it is impossible to accurately assess the health status and operating trends of equipment. The correlation and mutual influence between different equipment have not been fully understood and utilized. Once a certain equipment fails, it is difficult to quickly determine the scope of its impact on the entire system, resulting in untimely and inaccurate maintenance decisions, and the maintenance process takes a long time, which reduces the efficiency of equipment management and affects the power generation efficiency and economic benefits of thermal power plants. In addition, the existing early warning mechanism is not perfect, the early warning information is not intuitive and timely, and the operation and maintenance personnel find it difficult to obtain accurate equipment abnormality information in a timely manner and cannot quickly make effective response measures, which further aggravates the negative impact of equipment failure.

[0003] The existing technology has technical problems such as insufficient comprehensive and accurate monitoring of the operating status of thermal power plant equipment, low equipment management efficiency and difficulty in ensuring reliability. Summary of the invention

[0004] The present application provides an AI-based immersive equipment status monitoring method and system for thermal power plants, which is used to solve the technical problems in the prior art that the operating status monitoring of thermal power plant equipment is not comprehensive and accurate enough, the equipment management efficiency is low, and the reliability is difficult to ensure.

[0005] In view of the above problems, the present application provides an AI-based immersive equipment status monitoring method and system for thermal power plants.

[0006] The first aspect of the present application provides an AI-based immersive equipment status monitoring method for a thermal power plant, the method comprising:

[0007] After the monitoring sensors are deployed on the equipment of the thermal power plant, a digital twin model of the thermal power plant is established using the collected equipment data and monitoring sensors, wherein the digital twin model establishes signal mapping with the equipment and the monitoring sensors respectively; each device is taken as the smallest clustering center, and attraction clustering is performed according to the equipment topological relationship in the digital twin model to establish a multi-level clustering result; the monitoring data set of the monitoring sensor is read, and the monitoring data set is synchronized to the digital twin model; the independent anomaly recognition network is activated, and the independent anomaly recognition network is used to perform anomaly recognition based on the monitoring data set to establish an independent anomaly recognition result; the equipment criticality is established according to the independent anomaly recognition result, and the equipment criticality includes the main equipment criticality, the auxiliary equipment criticality and the associated equipment criticality; the weighted propagation chain is reconstructed according to the equipment criticality and the multi-level clustering result, and the weighted propagation chain and the monitoring data set are used to perform joint analysis of equipment anomalies to establish a joint warning of equipment anomalies; the joint warning of equipment anomalies is immersively visualized through the digital twin model.

[0008] The second aspect of the present application provides an AI-based immersive equipment status monitoring system for a thermal power plant, the system comprising:

[0009] The digital twin model establishment module is used to establish a digital twin model of the thermal power plant by using the collected equipment data and monitoring sensors after the monitoring sensors are deployed on the equipment of the thermal power plant, wherein the digital twin model establishes signal mapping with the equipment and the monitoring sensors respectively; the multi-level clustering result establishment module is used to take each device as the smallest cluster center, perform attraction clustering according to the equipment topological relationship in the digital twin model, and establish a multi-level clustering result; the monitoring data set reading module is used to read the monitoring data set of the monitoring sensor and synchronize the monitoring data set to the digital twin model; the independent anomaly recognition result establishment module is used to activate the independent anomaly recognition network, and utilize The independent anomaly identification network is used to perform anomaly identification based on the monitoring data set to establish an independent anomaly identification result; an equipment criticality establishment module is used to establish equipment criticality according to the independent anomaly identification result, and the equipment criticality includes main equipment criticality, auxiliary equipment criticality and associated equipment criticality; an equipment anomaly joint warning establishment module is used to reconstruct a weighted propagation chain according to the equipment criticality and the multi-level clustering results, and use the weighted propagation chain and the monitoring data set to perform joint analysis of equipment anomalies and establish a joint warning of equipment anomalies; a visualization presentation module is used to perform immersive visualization of the equipment anomaly joint warning through the digital twin model.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] After the monitoring sensors are deployed on the equipment of the thermal power plant, the digital twin model of the thermal power plant is established using the collected equipment data and monitoring sensors; the monitoring data set of the monitoring sensors is read and synchronized to the digital twin model; the independent anomaly recognition network is activated to establish an independent anomaly recognition result; the equipment criticality is established according to the independent anomaly recognition result; the weighted propagation chain is reconstructed according to the equipment criticality and the multi-level clustering results, and the equipment anomaly joint analysis is performed to establish the equipment anomaly joint warning; the equipment anomaly joint warning is presented in an immersive and visual manner through the digital twin model. The technical effect of realizing all-round and accurate monitoring of the operating status of the thermal power plant equipment and effectively improving the efficiency and reliability of equipment management is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0013] Figure 1 A schematic flow chart of an AI-based immersive equipment status monitoring method for a thermal power plant provided in an embodiment of the present application;

[0014] Figure 2 A schematic diagram of the structure of an AI-based immersive equipment status monitoring system for a thermal power plant provided in an embodiment of the present application.

[0015] Explanation of the accompanying drawings: digital twin model establishing module 10, multi-level clustering result establishing module 20, monitoring data set reading module 30, independent anomaly identification result establishing module 40, equipment criticality establishing module 50, equipment anomaly joint warning establishing module 60, visualization presentation module 70. DETAILED DESCRIPTION

[0016] This application provides an AI-based immersive equipment status monitoring method and system for thermal power plants, which is used to solve the technical problems in the prior art that the operating status monitoring of thermal power plant equipment is not comprehensive and accurate enough, the equipment management efficiency is low, and the reliability is difficult to ensure.

[0017] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0018] Embodiment 1, as Figure 1 As shown, the present application provides an AI-based immersive equipment status monitoring method for a thermal power plant, the method comprising:

[0019] Step S100: After the monitoring sensors are deployed on the equipment of the thermal power plant, a digital twin model of the thermal power plant is established using the collected equipment data and the monitoring sensors, wherein the digital twin model establishes signal mapping with the equipment and the monitoring sensors respectively.

[0020] Specifically, the collected equipment data (such as time series data such as temperature, pressure, vibration, etc.) is preprocessed using the wavelet transform algorithm. Wavelet transform can decompose the signal into sub-bands of different frequencies, effectively remove noise interference, extract feature information from the data, and enhance the availability of data. Then, the convolutional neural network (CNN) in deep learning is used to process the equipment data after wavelet transform. CNN automatically extracts local features in the data through the convolution layer, reduces the data dimension through the pooling layer, and integrates feature information through the fully connected layer, thereby learning the complex mapping relationship between the equipment operation status and data features. For example, through the training of a large amount of historical data, CNN can identify the normal or abnormal operation mode of the equipment corresponding to different combinations of parameters such as temperature and pressure. In terms of establishing signal mapping between the digital twin model and the equipment and monitoring sensors, an algorithm based on model predictive control (MPC) is adopted. The MPC algorithm calculates the optimal control signal based on the prediction of the future operation status of the equipment by the digital twin model and the data collected by the current monitoring sensor. On the one hand, the signal is fed back to the actual equipment to adjust the operating parameters of the equipment and realize the optimized control of the equipment; on the other hand, it acts on the digital twin model, so that it continuously updates and corrects the model parameters according to the feedback of the actual equipment, ensuring the high consistency between the model and the equipment. At the same time, in order to ensure the real-time and accuracy of signal mapping, the Kalman filter algorithm is used to filter and estimate the state of the data collected by the sensor. The Kalman filter can obtain the optimal state estimate at the current moment through recursive calculation based on the state estimate value at the previous moment and the measured value at the current moment, effectively reducing the error in the data transmission process, so that the digital twin model can obtain the real operating status information of the equipment in a timely and accurate manner, thereby realizing stable and reliable signal mapping between the equipment and the monitoring sensor. Through the collaborative work of the above multiple algorithms, the digital twin model of the thermal power plant is successfully constructed, and an effective signal mapping relationship between it and the equipment and monitoring sensors is established, providing strong support for the efficient monitoring and management of thermal power plant equipment. Step S200: Take each device as the smallest clustering center, perform attraction clustering according to the equipment topological relationship in the digital twin model, and establish a multi-level clustering result.

[0021] Step S200: Take each device as the smallest cluster center, perform attraction clustering according to the device topological relationship in the digital twin model, and establish a multi-level clustering result.

[0022] Specifically, after the construction of the digital twin model of the thermal power plant is completed, the equipment clustering analysis stage is immediately entered. In this step, each device is set as the smallest cluster center, and the equipment topology relationship information contained in the digital twin model is fully utilized to perform the attraction clustering operation. The equipment topology relationship describes in detail the complex association patterns such as physical connection, energy transfer, and signal interaction between devices. Through the hierarchical clustering algorithm, the similarity or distance measurement between each device is calculated, and the attraction degree between the devices is determined based on this measurement. In the attraction clustering process, devices with strong topological associations with the cluster center devices (such as direct connection, frequent energy interaction, etc.) will be preferentially attracted and clustered together to form initial clusters. As the clustering process continues to advance, these initial clusters will be further merged or adjusted according to the equipment topology relationship between each other, and a multi-level clustering structure will be gradually constructed, and finally a multi-level clustering result covering different levels of equipment combinations will be obtained. This result can clearly show the grouping of equipment in the thermal power plant system, provide a more detailed and systematic basis for subsequent equipment status monitoring, abnormal analysis, and management decisions, and help to fully grasp the equipment operation status from the overall to the local.

[0023] Step S300: Read the monitoring data set of the monitoring sensor, and synchronize the monitoring data set to the digital twin model.

[0024] Specifically, in the equipment status monitoring process of thermal power plants, monitoring sensors continuously collect various parameter information during the operation of the equipment to form a monitoring data set. These monitoring data sets are read regularly or in real time, and their data covers key operating indicators such as temperature, pressure, vibration amplitude, flow, etc. of the equipment. After reading the data, the monitoring data set is quickly synchronized to the digital twin model through a high-speed data transmission channel and a carefully designed data interface. To ensure the accuracy and integrity of data transmission, data verification algorithms such as cyclic redundancy check (CRC) are used to verify the transmitted data. At the same time, during the data synchronization process, data caching technology is used to cope with short interruptions or fluctuations in the data transmission process to ensure that the data can be stably transmitted to the digital twin model. Once the data is successfully synchronized, the digital twin model will immediately update its own status based on the new data, so as to always maintain a high degree of consistency with the actual operating status of the equipment, and provide real-time and accurate data support for subsequent model-based equipment status evaluation, abnormality identification and predictive analysis operations, ensuring that the entire monitoring system can timely and accurately reflect the equipment operation status and effectively prevent the occurrence of potential failures.

[0025] Step S400: activating an independent anomaly recognition network, using the independent anomaly recognition network to perform anomaly recognition based on the monitoring data set, and establishing an independent anomaly recognition result.

[0026] Specifically, the independent anomaly recognition network is first activated. This network has powerful data processing capabilities and is specifically designed to conduct in-depth analysis of the monitoring data set synchronized from the monitoring sensors to the digital twin model. The independent anomaly recognition network includes a feature extraction layer, which uses algorithms such as principal component analysis (PCA) and wavelet transform to extract representative features from the monitoring data set, including time domain features (such as data change trends, periodicity, etc.), frequency domain features (such as spectrum distribution, harmonic components, etc.) and time series embedding features (such as data correlation and self-similarity in time series). These feature extraction results are synchronized to the distributed anomaly detection subnetwork, in which the extracted features are compared and calculated with the pre-trained normal mode based on the principle of the autoencoder. The autoencoder can reconstruct the input data by learning from normal data. When the deviation between the features in the monitoring data and the reconstruction results exceeds the preset threshold, it is judged as an anomaly. In this way, an independent anomaly recognition result is established, and the possible anomalies in the monitoring data set are accurately located, providing a key basis for subsequent equipment criticality assessment and anomaly joint analysis, effectively helping operation and maintenance personnel to promptly discover potential risks in equipment operation.

[0027] Step S500: Establishing equipment criticality according to the independent abnormality identification result, wherein the equipment criticality includes main equipment criticality, auxiliary equipment criticality and associated equipment criticality.

[0028] Specifically, the construction of equipment criticality is carried out based on the independent anomaly identification results obtained above. First, the proportion of abnormal values ​​in the entire monitoring data is deeply analyzed for the independent anomaly identification results to determine the abnormal proportion. Then, the monitoring data set is segmented by creating a time window, and the maximum, minimum and average values ​​of the operating parameters (such as temperature, pressure, speed, etc.) in each window are calculated. The window fluctuation factor is calculated using these statistical values, and the fluctuation factor is obtained by weighted calculation combined with the timing coefficient. At the same time, the load data of the equipment is obtained, and the load factor is obtained by analyzing its ratio to the rated load, and the maintenance cycle factor is determined according to the proximity between the equipment calibration maintenance cycle and the current time node. Combining these operating stability characteristics and the role characteristics of the equipment in the entire system (such as whether it is a core power generation equipment, auxiliary support equipment, or the degree of association with other key equipment, etc.), finally a comprehensive evaluation system covering the criticality of main equipment, auxiliary equipment and related equipment is established, so as to accurately quantify the importance of different equipment in the operation of thermal power plants, provide key basis for subsequent weighted propagation chain reconstruction and targeted equipment management, and help to reasonably allocate operation and maintenance resources to ensure the stable and efficient operation of thermal power plants.

[0029] Step S600: reconstruct a weighted propagation chain according to the equipment criticality and the multi-level clustering result, use the weighted propagation chain and the monitoring data set to perform joint analysis of equipment anomalies, and establish a joint warning of equipment anomalies.

[0030] Specifically, a graph structure is constructed based on the criticality of the equipment and the multi-level clustering results. The equipment is regarded as a node, and the connection relationship between the equipment (based on the topological relationship and clustering results) is regarded as an edge. For the main equipment, a higher node weight is given according to its criticality, and for the equipment closely associated with the main equipment (determined according to the scope of association), its node weight is also adjusted accordingly. The initial setting of the edge weight is based on factors such as the physical connection strength between the equipment and the frequency of data interaction. Then, the constructed graph structure is trained using the graph neural network algorithm to learn the feature representation of the node. Through the message passing mechanism, the node can obtain the information of the neighboring node and update its own status. During the training process, the importance of the node feature is adjusted according to the criticality of the equipment, so that the model pays more attention to the information of the key equipment and its related nodes. At the same time, based on the monitoring data set, the operation data of the equipment is input into the graph neural network algorithm as the node feature, so that the model can learn the graph structure feature pattern under the normal operation and abnormal state of the equipment. In the weighted propagation chain reconstruction stage, the propagation chain is optimized according to the node weight and edge weight learned by the graph neural network algorithm. Increase the influence of key equipment in the propagation chain, adjust the weight distribution of information transmission between equipment, so that abnormal information can be propagated along a more reasonable path. For the joint analysis of equipment anomalies, the trained graph neural network algorithm is used to predict the real-time monitoring data set. When the model predicts that a certain node (equipment) has the possibility of anomaly, the back propagation algorithm is used to calculate the impact of the anomaly on other equipment based on the weighted propagation chain, and the anomaly risks of multiple devices are comprehensively considered to determine the joint anomaly score. Finally, a threshold is set according to the joint anomaly score. When the score exceeds the threshold, a joint warning of equipment anomalies is established to prompt operation and maintenance personnel to deal with potential faults in a timely manner, thereby ensuring the stable operation of thermal power plant equipment. Through this graph neural network-based method, equipment criticality and clustering results can be used more effectively to achieve accurate joint warning of anomalies.

[0031] Step S700: The device abnormality joint warning is immersively visualized through the digital twin model.

[0032] Specifically, the joint warning of equipment anomalies is realized by means of digital twin models in an immersive visual presentation. With its powerful rendering engine and interactive functions, the digital twin model displays the joint warning information of equipment anomalies in an intuitive, vivid and immersive way. The virtual scene of thermal power plant equipment is constructed by using three-dimensional modeling technology. In this scene, the abnormal equipment and its surrounding related equipment are marked with eye-catching colors (such as red for severe anomalies, yellow for moderate anomalies, and orange for mild anomalies) and dynamic effects (such as flashing, pulsation, etc.), clearly presenting the location and range of the anomaly. At the same time, the key parameters, abnormal values, and duration of abnormal equipment are displayed in detail through charts (such as line charts showing the trend of abnormal parameters over time, bar charts comparing the degree of abnormality of different equipment, etc.) and data labels, so that operation and maintenance personnel can quickly obtain comprehensive and critical abnormal data. With the help of virtual reality (VR) or augmented reality (AR) technology, operation and maintenance personnel can observe abnormal equipment from different angles, interactively view detailed information of equipment, and even simulate troubleshooting steps, which greatly improves the understanding and handling ability of abnormal situations, so that they can make response decisions more quickly and accurately to ensure the safe and stable operation of thermal power plant equipment.

[0033] In a possible implementation, step S500 further includes:

[0034] Step S510: Perform an outlier value ratio analysis based on the independent anomaly identification results to establish an anomaly ratio.

[0035] Step S520: Acquire the operational stability characteristics of the equipment, where the operational stability characteristics are constructed by a fluctuation factor, a load factor, and a maintenance cycle factor.

[0036] Step S530: Acquire the device role characteristics of the device, and establish the device criticality according to the operation stability characteristics, the device role characteristics and the abnormality ratio.

[0037] Specifically, an analysis of the proportion of outliers is carried out based on the results generated by the independent anomaly identification network to determine the anomaly ratio. After the independent anomaly identification network conducts a comprehensive inspection of the monitoring data set, it marks all data points that are judged to be abnormal. By accurately counting the number of these abnormal data points and comparing them with the total data volume of the monitoring data set, the proportion of outliers in the entire data set is obtained. This anomaly ratio intuitively quantifies the degree to which the equipment operation data deviates from the normal state, provides a key basis for the subsequent evaluation of the equipment's operating status and stability, and is an important reference indicator for judging whether there are potential problems with the equipment and the severity of the problem. A higher anomaly ratio means that the frequency of abnormal situations during the operation of the equipment is high, which may indicate that the equipment is facing a greater risk and requires further in-depth analysis and attention; conversely, a lower anomaly ratio indicates that the equipment is relatively stable, but continuous monitoring is still required to prevent the occurrence of abnormal situations.

[0038] Obtaining the operational stability characteristics of the equipment is a key link in assessing the criticality of the equipment. This characteristic is jointly constructed by the fluctuation factor, load factor and maintenance cycle factor. The calculation of the fluctuation factor requires the creation of a time window, and the monitoring data set is divided into several window segments in chronological order. For the operating parameters (such as temperature, pressure, vibration, etc.) in each window segment, the maximum, minimum and average values ​​are calculated respectively. Using these statistical values, the window fluctuation factor is calculated, and then it is weighted and calculated in combination with the timing coefficient to obtain the fluctuation factor that can accurately reflect the degree of fluctuation of the equipment operating parameters. The larger the fluctuation factor, the more unstable the equipment operation state is, and it may face a higher risk of failure. The acquisition of the load factor depends on the load data of the equipment. The load factor is determined by analyzing the proportional relationship between the current load of the equipment and the rated load. Equipment running at high load usually bears greater pressure, and its failure probability may be relatively high. Therefore, the load factor is of great significance in assessing the criticality of the equipment. The maintenance cycle factor is determined according to the proximity between the equipment's calibrated maintenance cycle and the current time node. For equipment that is close to its calibrated maintenance cycle, its operating stability may decrease due to factors such as component wear and aging, and its maintenance cycle factor will increase accordingly, indicating that the equipment requires more attention and maintenance.

[0039] After obtaining the equipment role characteristics of the equipment, the equipment criticality is established by integrating the operation stability characteristics and abnormality ratio. The equipment role characteristics cover the functional positioning of the equipment in the overall operation system of the thermal power plant, such as whether it is a core power generation equipment, a key auxiliary equipment or a general associated equipment. For core power generation equipment, even if its operation stability characteristics and abnormality ratio are relatively good, it will still be given a higher weight when calculating the equipment criticality due to its key position in the power generation process. In the comprehensive calculation process, the abnormality ratio is the dominant factor because it directly reflects the abnormality level of the current operation of the equipment. The fluctuation factor, load factor and maintenance cycle factor in the operation stability characteristics are used as auxiliary factors to correct and supplement the abnormality ratio. For example, if the abnormality ratio of a certain equipment is high, the fluctuation factor is large, the load is heavy and it is close to the maintenance cycle, its equipment criticality will be significantly improved and it will be judged as a high-criticality equipment, such as a main equipment; on the contrary, the equipment with a low abnormality ratio and good operation stability characteristics has a relatively low criticality and may be classified as an auxiliary equipment or an associated equipment. In this way, a comprehensive and accurate equipment criticality system is established, providing a key basis for the subsequent weighted propagation chain reconstruction and equipment management decision-making.

[0040] In a possible implementation, step S520 further includes:

[0041] Step S521: Create a time window, and use the time window to split the monitoring data set window.

[0042] Step S522: Calculate the maximum value, minimum value and average value of the operating parameters of each window segmentation result.

[0043] Step S523: Calculate the window fluctuation factor using the maximum value, minimum value and average value, and establish the fluctuation factor after weighting the window fluctuation factor using the timing coefficient.

[0044] Step S524: Obtain load data of the equipment, perform a proportional analysis of the rated load based on the load data, and establish a load factor.

[0045] Step S525: Obtain the calibration maintenance cycle of the equipment, and establish a maintenance cycle factor according to the proximity between the current time node and the calibration maintenance cycle.

[0046] Step S526: constructing an operation stability feature using the fluctuation factor, the load factor and the maintenance cycle factor.

[0047] Specifically, creating a time window is an important starting step in analyzing the stable operation characteristics of equipment. The creation of a time window needs to comprehensively consider factors such as the characteristics of equipment operation, the frequency of data collection, and the time scale accuracy required for analysis. For example, for large thermal power plant equipment with relatively slow operating status changes, a longer time window can be set, such as in hours; while for equipment with more frequent operating parameter fluctuations, a shorter time window can be used, such as a few minutes. After determining the length of the time window, slide on the monitoring data set in turn at fixed time intervals to divide the continuous monitoring data into a series of non-overlapping (or partially overlapping, depending on specific needs) sub-datasets, each of which corresponds to a snapshot of the equipment operation data within a specific time window. This window segmentation method can convert long-term series monitoring data into multiple data set fragments with time locality, making the subsequent analysis of the data in each window more targeted, helping to deeply explore the operating laws of the equipment at different time stages, and laying the foundation for accurately calculating the statistical values ​​of operating parameters and further evaluating the stability of equipment operation.

[0048] For each result obtained by time window segmentation, the maximum, minimum and average values ​​of its operating parameters are calculated. The operating parameters include various key indicators that can characterize the operating status of the equipment, such as temperature, pressure, vibration amplitude, rotation speed, flow rate, etc. For these operating parameter data in each time window, the maximum and minimum values ​​are determined by traversing all data points. The maximum value represents the highest level reached by the operating parameter in this specific time period, and the minimum value reflects the lowest level. They together define the fluctuation range of the operating parameter in this time window. At the same time, the operating parameter values ​​of all data points in the window are added and divided by the number of data points to obtain the average value, which can reflect the central trend or typical level of the operating parameter in this time window. For example, in a time window of temperature data, the maximum value may represent the highest temperature experienced by the equipment during this period, the minimum value is the lowest temperature, and the average value roughly reflects the average temperature condition of the equipment during this period. The calculation of these maximum, minimum and average values ​​provides basic data for the subsequent evaluation of the equipment's operating stability and performance, which helps to gain a deep understanding of the characteristics of the equipment's operating parameter changes in different time windows, so as to more accurately judge the equipment's working status.

[0049] The window fluctuation factor is calculated using the maximum, minimum and average values ​​of the operating parameters in each window segmentation result. The calculation method is to divide the difference between the maximum and minimum values ​​by the average value to obtain a value that can reflect the relative fluctuation degree of the operating parameters in the time window, which is the window fluctuation factor. Then, the timing coefficient is introduced to perform weighted calculation on the window fluctuation factor to establish the fluctuation factor. The timing coefficient is determined according to the position of the time window in the entire monitoring sequence. The closer the window is to the current time, the larger its coefficient is. The exponential decay function is used to set the timing coefficient. The coefficient of the recent window is close to 1 and gradually decreases over time. The final fluctuation factor is obtained by multiplying the fluctuation factor of each window by the corresponding timing coefficient and then summing them up, which further highlights the impact of the recent operating status on the current stability of the device, because the recent operating conditions can better reflect the current true status of the device.

[0050] Obtain the real-time load data of the equipment during operation. These load data accurately reflect the current workload of the equipment. Subsequently, the obtained load data is proportionally analyzed with the pre-set rated load of the equipment. The ratio obtained by dividing the real-time load data by the rated load is the load factor. For example, if the rated load of a certain equipment is 1000 kilowatts, and the current real-time monitored load is 800 kilowatts, then the load factor of the equipment at this time is 800÷1000=0.8. The load factor intuitively shows the degree of closeness between the current operating load of the equipment and the designed load capacity. When the load factor is close to 1, it means that the equipment is running at high load and is under great pressure. In this case, the wear rate of the equipment components may accelerate and the risk of failure will increase accordingly; while a small load factor indicates that the equipment is running at a light load, but if it is in a state of too low load for a long time, the operating efficiency of the equipment may be affected. Therefore, the load factor is one of the key indicators for measuring the stable operation characteristics of the equipment. It provides an important basis for comprehensively evaluating the operating status and stability of the equipment, and helps the operation and maintenance personnel to understand the working intensity of the equipment in a timely manner so as to make reasonable operation adjustments or maintenance plans.

[0051] Get the calibration maintenance cycle of the equipment. This cycle is determined by the equipment manufacturer based on multiple factors such as the equipment's design structure, material properties, expected service life, and normal operation wear law. It is an important time reference for preventive maintenance of the equipment. Next, calculate the closeness of the current time node to the calibration maintenance cycle, which can be measured by calculating the time difference between the two. For example, if the calibration maintenance cycle of a certain equipment is 1000 hours and it has been running for 950 hours, then there are 50 hours left before the next maintenance. The ratio of this remaining time to the total maintenance cycle (50÷1000=0.05) can be used as an indicator to measure the proximity. According to this proximity, establish a maintenance cycle factor. The higher the proximity (i.e., the closer to the next maintenance time), the larger the maintenance cycle factor. This is because as the equipment's operating time approaches the calibration maintenance cycle, the performance of the internal parts of the equipment may gradually decline due to long-term wear, fatigue, aging, etc., and the operating stability is reduced, and the possibility of failure increases. Therefore, the maintenance cycle factor can intuitively reflect the urgency of the equipment's current need for maintenance and overhaul, and provide a key reference for operation and maintenance personnel to reasonably arrange equipment maintenance plans, helping to ensure that the equipment is always in good operating condition and reduce losses caused by sudden failures.

[0052] Assign a corresponding weight to each factor. The determination of the weight can be based on expert experience, analysis of historical data, or an in-depth understanding of the equipment operation mechanism. For example, after analyzing a large amount of historical data, it was found that the fluctuation factor accounted for 40% of the weight of the equipment failure, the load factor accounted for 30%, and the maintenance cycle factor accounted for 30%. Then, multiply the fluctuation factor by its corresponding weight, the load factor by its weight, and the maintenance cycle factor by its weight. Finally, add the three weighted factors to obtain the operation stability characteristic value. In this way, the fluctuation factor, load factor, and maintenance cycle factor are organically combined to form a characteristic indicator that can comprehensively and comprehensively reflect the degree of equipment operation stability, providing a scientific and accurate basis for subsequent equipment criticality assessment. In practical applications, the weights can be dynamically adjusted according to factors such as equipment type and operating environment to improve the accuracy and adaptability of operation stability characteristic assessment.

[0053] In a possible implementation, step S600 further includes:

[0054] Step S610: Acquire the main equipment criticality among the equipment criticalities, locate the main equipment using the main equipment criticality, and configure the associated range of the main equipment.

[0055] Step S620: Perform propagation chain screening of multi-level clustering results based on the main equipment and the associated scope, and establish a selected propagation chain.

[0056] Step S630: Obtain the basic connection weight of the selected propagation chain, and establish additional connection weights based on the device criticality.

[0057] Step S640: Perform weighted fitting of the selected propagation chain according to the basic connection weight and the additional connection weight, and reconstruct the weighted propagation chain.

[0058] Specifically, the criticality information of equipment is accurately captured from the constructed equipment criticality database, and the criticality of the main equipment is screened out. After obtaining the criticality, the criticality of the main equipment is compared with the preset precise threshold range using an intelligent comparison algorithm. Once the criticality value of a certain equipment successfully falls into the set high criticality range, it is immediately locked as the main equipment. When the main equipment is determined, the associated range configuration task is immediately started. With the help of digital equipment connection maps and the use of depth-first search algorithms in graph theory, we dig deep from the physical connection level to include those devices that are closely connected to the main equipment through high-strength mechanical connections, large-capacity power transmission lines or high-speed data links, and are directly involved in energy conversion and material transportation, such as boilers that continuously supply energy to generator sets and substations responsible for power distribution; from the perspective of functional coordination, based on predefined functional coordination logic rules, we find out the equipment that provides key auxiliary functions such as cooling, lubrication, monitoring and feedback to ensure the efficient operation of the main equipment, such as the cooling circulation system that ensures the smooth operation of the unit and the sensor array that monitors the operating parameters in real time; from the dimension of information interaction, by analyzing the communication protocol between devices, we track the automation control units that frequently interact with the main equipment for commands and status synchronization, and comprehensively integrate this information to tailor a scientific and reasonable range of association for the main equipment, laying a solid foundation for the subsequent in-depth analysis of the potential impacts between equipment.

[0059] Given the main equipment and its associated scope, first call out the multi-level clustering results, which are obtained through clustering and stratification based on comprehensive factors such as physical connection, functional collaboration, and information interaction between equipment. Define the search area with the main equipment as the center and the associated scope as the radius, and use the intelligent traversal algorithm to go deep into it. From the physical connection, track the path of energy transmission through pipelines and lines with the main equipment, such as the connection of auxiliary devices that power the main equipment; from the functional collaboration level, explore the equipment links that ensure the stable operation of the main equipment and assist the core functions, such as the connection of auxiliary equipment that provides cooling and lubrication; from the information interaction dimension, lock the equipment connection that frequently sends and receives instructions and feedback status with the main equipment. Logically sort out the connections that meet the conditions under these dimensions and connect them in series to screen out the key paths, and thus establish the selected propagation chain to lay a solid foundation for subsequent analysis.

[0060] Obtain the basic connection weight of the selected propagation chain. For each connection link in the selected propagation chain, deeply explore its physical connection characteristics, such as the connection method between devices (whether it is directly connected or indirectly connected through intermediate devices), the stability of the connection (durability of the connection components, reliability of the connection interface, etc.) and the complexity of the connection (the number of connection points involved, the complexity of the connection line, etc.). At the same time, fully consider factors related to data interaction, including the rate of data transmission, the size of data traffic, the real-time requirements of data transmission, and the bit error rate during data transmission. Combined with the fault propagation instances recorded in the historical operation data, the frequency of each connection link being involved in past fault events is counted. Combining these physical connection characteristics, data interaction factors, and historical fault propagation probabilities, etc., a weight calculation model is used to accurately calculate and obtain the basic connection weight for each connection link in the selected propagation chain. After determining the basic connection weight, proceed to build additional connection weights based on the previously established device criticality. A detailed analysis is conducted on the criticality of the main equipment, auxiliary equipment and associated equipment in the equipment criticality. For the main equipment, since it is in a core position in the entire thermal power plant equipment system and plays a key supporting role in the power generation production process, once a failure occurs, it may cause the paralysis of the entire system, so a higher weight value is given in the additional connection weight. Auxiliary equipment provides necessary auxiliary support for the normal operation of the main equipment, and its failure will also have a certain impact on production. The corresponding weight is assigned according to the importance of its auxiliary function. The weight of the associated equipment is determined according to the closeness of its association with the main equipment and auxiliary equipment and the potential impact in the process of fault propagation. Based on the equipment criticality, an additional connection weight is constructed for each connection link in the selected propagation chain, thereby providing a comprehensive and accurate weight basis for subsequent weighted fitting.

[0061] The obtained basic connection weights and additional connection weights are integrated. The basic connection weights are calculated based on the actual parameters of the physical connection between each device in the transmission chain. Among them, the pipeline connection is based on the parameters such as pipe diameter, pipe wall thickness, material pressure resistance, etc., and its transmission capacity is quantified by fluid mechanics related formulas; the cable connection determines the conductive efficiency weight according to the material conductivity, cross-sectional area, insulation performance and carrying current intensity according to Ohm's law; the mechanical connection is based on the tightening torque, friction coefficient, and rigidity strength of the connection part, and the stability weight is calculated using the material mechanics model. These parameters comprehensively reflect the inherent characteristics of the physical connection between devices. The additional connection weight is set based on the criticality of the device. The higher the criticality of the device, the greater the additional connection weight of the link connected to it, which means that the link has a stronger influence on fault propagation, key information transmission, etc. After completing the weight integration, the weighted algorithm is used to implement the fitting operation, and a variant of the shortest path algorithm based on graph theory is adopted. This algorithm fully considers the basic connection weight and the additional connection weight in the calculation process and replans the connection path between the nodes of the transmission chain. For links with high basic connection weights, the algorithm ensures that their transmission advantages are retained during the propagation process, so that they remain efficient in conventional operation links such as energy transmission and signal transmission; for links with prominent additional connection weights, the algorithm focuses on strengthening their key node role in fault propagation and key information guidance. Through multiple iterative operations, the weight distribution of each link is continuously adjusted until the structure of the propagation chain accurately matches the energy flow direction, fault propagation path, and signal interaction mechanism in the actual equipment operation, corrects the unreasonable connections or weak links that may exist in the original propagation chain, and finally successfully reconstructs the weighted propagation chain to provide support for subsequent accurate joint anomaly positioning and efficient risk warning.

[0062] In a possible implementation, step S600 further includes:

[0063] Step S650: extracting features of the monitoring data set using the feature extraction layer of the independent anomaly recognition network, and establishing feature extraction results, wherein the feature extraction results include time domain features, frequency domain features, and time series embedding features.

[0064] Step S660: Synchronize the feature extraction result to the distributed anomaly detection subnetwork of the independent anomaly recognition network, perform anomaly calculation of the autoencoder, and establish the independent anomaly recognition result.

[0065] Specifically, the built-in feature extraction layer of the independent anomaly recognition network is activated to perform deep feature extraction operations on the monitoring data set to construct accurate and comprehensive feature extraction results, which cover time domain features, frequency domain features, and time series embedded features. Faced with complex and massive monitoring data, the feature extraction layer uses a series of advanced algorithms and technologies to work in an orderly manner. In the time domain analysis stage, based on the sliding window mechanism, it applies algorithms such as mean calculation, variance analysis, and autocorrelation function solution to the time series data formed by the equipment operating parameters (such as vibration amplitude, temperature value, pressure size, etc.), accurately capturing the fluctuation trend, discreteness, and periodic change law of the data over time, so as to extract time domain features, which can intuitively present the operation stability of the equipment in the short term. Then, the feature extraction layer switches to the frequency domain analysis module, and uses the fast Fourier transform (FFT) algorithm to efficiently convert the time domain signal into a frequency domain expression. In this process, through the detailed analysis of the spectrum, the spectrum peak detection and harmonic analysis technology are used to identify abnormal phenomena such as abnormal fluctuations in specific frequencies during the operation of the motor, high-frequency spikes caused by mechanical structure resonance, and uneven distribution of spectrum energy, and then extract frequency domain features that reflect potential equipment failure hazards. These features are key indicators for diagnosing fault problems closely related to speed and vibration frequency. At the same time, the feature extraction layer introduces time series embedding, builds a deep neural network model, and maps high-dimensional and complex time series data to low-dimensional space. Through model training, the long-term dependencies and hidden dynamic patterns between data are learned and captured, and the low-dimensional vector output by the embedding layer is used to represent the time series embedding features, which can deeply reveal the potential laws in the operation of the equipment. By combining the above three analysis methods, the feature extraction layer finally successfully established a feature extraction result that comprehensively covers the time domain, frequency domain, and time series embedding features, laying a solid foundation for the subsequent accurate identification of abnormal equipment status.

[0066] The extracted results, including time domain, frequency domain and time series embedded features, are quickly transmitted to the distributed anomaly detection sub-network in the independent anomaly identification network. The autoencoder in the sub-network starts to work immediately, first encoding the input data packet, using the dimensionality reduction algorithm to compress the high-dimensional complex data into a low-dimensional latent space to extract key features, and then decoding and reconstructing the original data. Under normal circumstances, the reconstructed data is similar to the original feature results, and the difference is within the fine-tuned threshold; if the equipment has abnormalities such as increased vibration and sudden temperature changes, the reconstruction error exceeds the threshold, and the sub-network activates the algorithm based on the error, comprehensively considers multiple factors, and accurately locates the anomaly by data point and level, and finally obtains an independent anomaly identification result to assist in subsequent repairs and maintenance.

[0067] In a possible implementation, step S600 further includes:

[0068] Step S670: Utilize the weighted propagation chain to screen the monitoring data set and establish an available data set.

[0069] Step S680: using the available data sets to identify abnormal features of the equipment respectively, and performing joint abnormality calculation based on the weighted propagation chain on the abnormal feature identification results to establish a joint abnormality calculation result.

[0070] Step S690: Establishing a joint warning of equipment abnormality according to the joint abnormality calculation result.

[0071] Specifically, the weighted propagation chain, a key tool constructed in the early stage, is used to accurately screen the massive monitoring data sets. The weighted propagation chain, with its precise structure after integrating the basic connection weights and additional connection weights, focuses on those monitoring data links that are closely connected to key equipment and have a key influence on the fault propagation path. According to the weight distribution of each node in the propagation chain, data that is highly indicative of the overall operating status of the equipment is preferentially screened out, and irrelevant or interfering data is eliminated, thereby establishing a more accurate usable data set, providing a high-quality data foundation for subsequent in-depth analysis.

[0072] Use available data sets to identify abnormal features of each device, and use a variety of data analysis techniques and algorithms to conduct in-depth mining for various operating parameters of each device in the data set, such as temperature, pressure, and vibration frequency. For example, use time domain analysis methods to observe the trend of parameter changes over time and determine whether there are abnormal fluctuations or mutations; use frequency domain analysis technology to convert the signal spectrum to analyze whether there are components beyond the normal frequency range to identify resonance or other abnormal frequency phenomena that may exist in the device; at the same time, combine statistical analysis methods to calculate statistical indicators such as the mean, variance, and standard deviation of the parameters, and compare them with the statistical benchmark values ​​under normal operating conditions to determine whether there are significant deviations. After completing the identification of abnormal features of the equipment, these scattered abnormal feature identification results are used for joint abnormal calculation based on the weighted propagation chain. According to the weight information of the connection between each device in the weighted propagation chain, the mutual correlation between the devices in physical structure and operation logic is considered. For devices in the core position with higher connection weights, their abnormal features are given a larger weight in the joint calculation, because the abnormalities of these devices are more likely to have a significant impact; while for devices with relatively low connection weights, the weights of their abnormal features are adjusted accordingly. Through the designed weighted fusion algorithm, the abnormal characteristics of different devices are comprehensively considered according to their importance in the propagation chain, the propagation and diffusion process of faults between devices along the weighted propagation chain is simulated, and the abnormality degree of the entire equipment system is comprehensively quantified. Finally, the joint abnormality calculation results are successfully established, providing accurate data support for subsequent joint early warning of equipment abnormalities.

[0073] Retrieve multiple sets of preset warning thresholds, which are based on the in-depth analysis of a large number of equipment failure data in the past and the accurate summary of the rich practical experience of senior experts. They are carefully classified for different degrees of joint abnormal conditions. Then, the joint abnormal calculation results are strictly compared with these thresholds one by one. Once the calculation results are found to exceed the corresponding threshold range, it means that there are potential risks in the equipment system that cannot be ignored. At this time, the all-round early warning mechanism is quickly started. On the one hand, a striking warning pop-up window pops up on the operation interface of the monitoring center. The pop-up window displays the abnormal equipment information, the degree of abnormality and the possible scope of impact in concise and clear charts and texts; on the other hand, a customized notification is sent through the SMS group function to inform them of the emergency and the initial response measures to be taken. At the same time, the sound and light alarm device on the scene is also activated, sending a strong sound and light signal to remind the surrounding personnel to pay attention to safety and report in time. Through this multi-channel linkage early warning method, it is ensured that the operation and maintenance team can know the equipment abnormality at the first time, respond in time, minimize the losses caused by equipment failure, and ensure the smooth operation of the entire equipment system.

[0074] In a possible implementation, step S690 further includes:

[0075] Step S691: performing equipment abnormality trend fitting based on the available data set, and establishing equipment abnormality trend fitting results.

[0076] Step S692: establishing an abnormality fitting chain according to the equipment abnormality trend fitting result and the equipment abnormality joint warning.

[0077] Step S693: The abnormal fitting chain is immersively visualized through the digital twin model, and a fitting warning is issued.

[0078] Specifically, we conduct an in-depth analysis of the available data sets and extract the representative operating parameters of each device, such as the temperature change curve, pressure fluctuation data, and dynamic information of vibration amplitude during the operation of the equipment. Then, we use professional data processing and analysis techniques, use the moving average method to smooth the data to reduce short-term fluctuation interference, and use the difference method to process non-stationary data to meet the fitting requirements. Then, we select the appropriate fitting model based on the data characteristics. If the data shows a linear trend, the linear regression model is used. If there is obvious periodicity, the Fourier series model is adapted. The processed data is input into the selected model for calculation and optimization, and the model parameters are accurately determined, so as to establish a fitting result that can accurately reflect the abnormal trend of the equipment, providing a key basis for the subsequent in-depth analysis of the health status and potential risks of the equipment.

[0079] The abnormal trend fitting results of the integrated equipment are combined with the previously established equipment abnormal joint warning information to build an abnormal fitting chain. The equipment abnormal trend fitting results provide the potential direction of the development of equipment abnormalities, while the equipment abnormal joint warning reflects the actual abnormal status of the current equipment. Based on the inherent connection between the two, the devices that may be affected are connected in the logical order of abnormal propagation based on the physical connection relationship and data interaction link between the devices to form an abnormal fitting chain. For example, if a key device shows an abnormality in the joint warning, and the trend fitting results show that its abnormality will affect downstream devices, then in the abnormal fitting chain, the path of this abnormal propagation will be clearly displayed, helping operation and maintenance personnel to understand in advance the chain reaction that may be caused by equipment failures, so as to take timely countermeasures.

[0080] Using digital twin model technology, the abnormal fitting chain is presented in an immersive and visual manner. In the digital twin model, the equipment will be displayed in the form of a highly realistic three-dimensional model, and the equipment in the abnormal fitting chain will be highlighted in different colors, flashing effects or other visualization methods according to the degree of abnormality. For example, equipment with serious abnormalities will be displayed as red flashing, while equipment with less impact will be prompted in yellow. At the same time, fitting warnings will be reported based on the trend information and warning levels in the abnormal fitting chain. The warning information will be presented to the operation and maintenance personnel in a variety of forms such as eye-catching text prompts and voice broadcasts to ensure that they can obtain equipment abnormality information in a timely and accurate manner, make preparations in advance, and ensure the safe and stable operation of thermal power plant equipment.

[0081] Embodiment 2 is based on the same inventive concept as the AI-based immersive equipment status monitoring method for thermal power plants in the aforementioned embodiment. Figure 2 As shown, the present application provides an AI-based immersive equipment status monitoring system for thermal power plants. The system and method embodiments in the present application are based on the same inventive concept. The system includes:

[0082] The digital twin model establishment module 10 is used to establish a digital twin model of the thermal power plant using the collected equipment data and monitoring sensors after the monitoring sensors are deployed on the equipment of the thermal power plant, wherein the digital twin model establishes signal mapping with the equipment and the monitoring sensors respectively.

[0083] The multi-level clustering result establishing module 20 is used to take each device as the smallest cluster center, perform attraction clustering according to the device topological relationship in the digital twin model, and establish a multi-level clustering result.

[0084] The monitoring data set reading module 30 is used to read the monitoring data set of the monitoring sensor and synchronize the monitoring data set to the digital twin model.

[0085] The independent anomaly identification result establishing module 40 is used to activate the independent anomaly identification network, use the independent anomaly identification network to perform anomaly identification based on the monitoring data set, and establish an independent anomaly identification result.

[0086] The equipment criticality establishment module 50 is used to establish equipment criticality according to the independent abnormality identification result, and the equipment criticality includes main equipment criticality, auxiliary equipment criticality and associated equipment criticality.

[0087] The equipment abnormality joint warning establishment module 60 is used to reconstruct the weighted propagation chain according to the equipment criticality and the multi-level clustering result, use the weighted propagation chain and the monitoring data set to perform equipment abnormality joint analysis, and establish equipment abnormality joint warning.

[0088] The visualization presentation module 70 is used to perform immersive visualization presentation of the equipment abnormality joint warning through the digital twin model.

[0089] Furthermore, the device criticality establishing module 50 includes:

[0090] The abnormality ratio establishing unit is used to perform abnormal value ratio analysis according to the independent abnormality identification result and establish the abnormality ratio.

[0091] The operation stability feature acquisition unit is used to acquire the operation stability feature of the equipment, wherein the operation stability feature is constructed by a fluctuation factor, a load factor and a maintenance cycle factor.

[0092] The device role characteristic acquisition unit is used to acquire the device role characteristic of the device, and establish the device criticality according to the operation stability characteristic, the device role characteristic and the abnormality ratio.

[0093] Furthermore, the operation stability feature acquisition unit includes:

[0094] The window segmentation unit is used to create a time window and use the time window to segment the monitoring data set window.

[0095] The operating parameter calculation unit is used to calculate the maximum value, minimum value and average value of the operating parameters of each window segmentation result.

[0096] The fluctuation factor establishing unit is used to calculate the window fluctuation factor by using the maximum value, the minimum value and the average value, and to establish the fluctuation factor after weighted calculation of the window fluctuation factor by the time series coefficient.

[0097] The load factor establishing unit is used to obtain load data of the equipment, perform a proportional analysis of the rated load according to the load data, and establish a load factor.

[0098] The maintenance cycle factor establishing unit is used to obtain the calibrated maintenance cycle of the equipment and establish the maintenance cycle factor according to the proximity between the current time node and the calibrated maintenance cycle.

[0099] The operation stability feature construction unit is used to construct the operation stability feature by using the fluctuation factor, the load factor and the maintenance cycle factor.

[0100] Furthermore, the equipment abnormality joint warning establishment module 60 includes:

[0101] The associated range configuration unit is used to obtain the main equipment criticality in the equipment criticality, locate the main equipment by using the main equipment criticality, and configure the associated range of the main equipment.

[0102] A selected propagation chain establishment unit is used to screen the propagation chains of the multi-level clustering results based on the main equipment and the associated scope, and to establish a selected propagation chain.

[0103] The additional connection weight establishing unit is used to obtain the basic connection weight of the selected propagation chain and establish the additional connection weight based on the device criticality.

[0104] The weighted propagation chain reconstruction unit is used to perform weighted fitting of the selected propagation chain according to the basic connection weight and the additional connection weight, and reconstruct the weighted propagation chain.

[0105] Furthermore, the independent abnormality identification result establishing module 40 includes:

[0106] The feature extraction result establishing unit is used to use the feature extraction layer of the independent anomaly recognition network to extract features of the monitoring data set and establish feature extraction results, wherein the feature extraction results include time domain features, frequency domain features and time series embedding features.

[0107] An anomaly calculation unit is used to synchronize the feature extraction result to the distributed anomaly detection subnetwork of the independent anomaly recognition network, perform anomaly calculation of the autoencoder, and establish the independent anomaly recognition result.

[0108] Furthermore, the equipment abnormality joint warning establishment module 60 includes:

[0109] The available data set establishing unit is used to screen the monitoring data set by using the weighted propagation chain to establish an available data set.

[0110] The joint abnormality calculation result establishing unit is used to respectively identify abnormal features of the equipment through the available data sets, and perform joint abnormality calculation based on the weighted propagation chain on the abnormality feature identification results to establish the joint abnormality calculation results.

[0111] The equipment abnormality joint warning establishing unit is used to establish the equipment abnormality joint warning according to the joint abnormality calculation result.

[0112] Furthermore, the equipment abnormality joint early warning establishing unit also includes:

[0113] The device abnormal trend fitting result establishing unit performs device abnormal trend fitting based on the available data set to establish the device abnormal trend fitting result.

[0114] The abnormal fitting chain establishing unit is used to establish the abnormal fitting chain according to the abnormal trend fitting result of the equipment and the abnormal joint warning of the equipment.

[0115] A fitting early warning output unit is used to present the abnormal fitting chain in an immersive and visual manner through the digital twin model and to issue a fitting early warning.

[0116] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0117] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0118] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. An AI-based immersive equipment status monitoring method for thermal power plants, characterized in that: The method comprises: After the monitoring sensors are deployed on the equipment of the thermal power plant, a digital twin model of the thermal power plant is established using the collected equipment data and the monitoring sensors, wherein the digital twin model establishes signal mapping with the equipment and the monitoring sensors respectively; Taking each device as the smallest cluster center, performing attraction clustering according to the device topological relationship in the digital twin model, and establishing a multi-level clustering result; Reading a monitoring data set of the monitoring sensor, and synchronizing the monitoring data set to the digital twin model; activating an independent anomaly recognition network, using the independent anomaly recognition network to perform anomaly recognition based on the monitoring data set, and establishing an independent anomaly recognition result; Establishing equipment criticality according to the independent abnormality identification result, wherein the equipment criticality includes main equipment criticality, auxiliary equipment criticality and associated equipment criticality; Reconstructing a weighted propagation chain according to the equipment criticality and the multi-level clustering result, using the weighted propagation chain and the monitoring data set to perform joint analysis of equipment anomalies, and establishing a joint warning of equipment anomalies; The equipment abnormality joint warning is presented in an immersive and visual manner through the digital twin model.

2. The AI-based immersive equipment status monitoring method for thermal power plants according to claim 1, characterized in that: The establishing of equipment criticality according to the independent abnormality identification result comprises: Perform an outlier value ratio analysis based on the independent anomaly identification results to establish an anomaly ratio; Obtaining an operational stability feature of the equipment, wherein the operational stability feature is constructed by a fluctuation factor, a load factor, and a maintenance cycle factor; The device role characteristics of the device are obtained, and the device criticality is established according to the operation stability characteristics, the device role characteristics and the abnormality ratio.

3. The AI-based immersive equipment status monitoring method for thermal power plants according to claim 2, characterized in that: The acquisition of the operational stability characteristics of the device includes: Creating a time window, and using the time window to split the monitoring data set window; Calculate the maximum, minimum and average values ​​of the operating parameters of each window segmentation result; Calculate the window fluctuation factor using the maximum value, the minimum value and the average value, and establish the fluctuation factor after weighting the window fluctuation factor using the time series coefficient; Obtaining load data of the equipment, performing a proportional analysis of the rated load according to the load data, and establishing a load factor; Obtain the calibration maintenance cycle of the equipment, and establish a maintenance cycle factor according to the proximity between the current time node and the calibration maintenance cycle; An operation stability feature is constructed using the fluctuation factor, the load factor and the maintenance cycle factor.

4. The AI-based immersive equipment status monitoring method for thermal power plants according to claim 1, characterized in that: The reconstructing the weighted propagation chain according to the device criticality and the multi-level clustering result includes: Obtaining the criticality of a main device among the device criticalities, locating the main device using the criticality of the main device, and configuring the associated range of the main device; Performing propagation chain screening of multi-level clustering results based on the main equipment and the associated scope, and establishing a selected propagation chain; Obtaining a basic connection weight for the selected propagation chain and establishing an additional connection weight based on the device criticality; The weighted fitting of the selected propagation chain is performed according to the basic connection weight and the additional connection weight, and the weighted propagation chain is reconstructed.

5. The AI-based immersive equipment status monitoring method for thermal power plants according to claim 1, characterized in that: The using the independent anomaly recognition network to perform anomaly recognition based on the monitoring data set to establish an independent anomaly recognition result includes: Using the feature extraction layer of the independent anomaly recognition network to extract features of the monitoring data set, and establishing feature extraction results, wherein the feature extraction results include time domain features, frequency domain features, and time series embedding features; The feature extraction result is synchronized to the distributed anomaly detection subnetwork of the independent anomaly recognition network, anomaly calculation of the autoencoder is performed, and the independent anomaly recognition result is established.

6. The AI-based immersive equipment status monitoring method for thermal power plants according to claim 1, characterized in that: The method of using the weighted propagation chain and the monitoring data set to perform joint analysis of equipment anomalies and establish joint warning of equipment anomalies includes: Using the weighted propagation chain to screen the monitoring data set to establish an available data set; Using the available data sets, the abnormal features of the equipment are respectively identified, and the abnormal feature identification results are subjected to joint abnormality calculation based on a weighted propagation chain to establish a joint abnormality calculation result; A joint warning of equipment abnormality is established according to the joint abnormality calculation result.

7. The AI-based immersive equipment status monitoring method for thermal power plants according to claim 6, characterized in that: The method further comprises: Performing equipment abnormality trend fitting based on the available data set, and establishing equipment abnormality trend fitting results; Establishing an abnormality fitting chain according to the abnormality trend fitting result of the equipment and the abnormality joint warning of the equipment; The abnormal fitting chain is immersively visualized through the digital twin model, and a fitting warning is issued.

8. AI-based immersive equipment status monitoring system for thermal power plants, characterized by: The system is used to implement the AI-based immersive equipment status monitoring method for a thermal power plant according to any one of claims 1 to 7, and the system comprises: A digital twin model establishment module is used to establish a digital twin model of the thermal power plant by using the collected equipment data and monitoring sensors after the monitoring sensors are deployed on the equipment of the thermal power plant, wherein the digital twin model establishes signal mapping with the equipment and the monitoring sensors respectively; A multi-level clustering result establishment module, used to take each device as the smallest cluster center, perform attraction clustering according to the device topological relationship in the digital twin model, and establish a multi-level clustering result; A monitoring data set reading module, used to read the monitoring data set of the monitoring sensor and synchronize the monitoring data set to the digital twin model; An independent anomaly identification result establishing module, used to activate the independent anomaly identification network, use the independent anomaly identification network to perform anomaly identification based on the monitoring data set, and establish an independent anomaly identification result; An equipment criticality establishment module, used to establish equipment criticality according to the independent abnormality identification result, wherein the equipment criticality includes main equipment criticality, auxiliary equipment criticality and associated equipment criticality; An equipment abnormality joint warning establishment module is used to reconstruct a weighted propagation chain according to the equipment criticality and the multi-level clustering result, use the weighted propagation chain and the monitoring data set to perform equipment abnormality joint analysis, and establish equipment abnormality joint warning; A visualization presentation module is used to perform immersive visualization presentation of the equipment abnormality joint warning through the digital twin model.

9. The AI-based immersive equipment status monitoring system for thermal power plants according to claim 8, characterized in that: The equipment criticality establishment module includes: An abnormality ratio establishing unit, used to perform abnormal value ratio analysis according to the independent abnormality identification result and establish an abnormality ratio; An operation stability feature acquisition unit, used to acquire an operation stability feature of the equipment, wherein the operation stability feature is constructed by a fluctuation factor, a load factor, and a maintenance cycle factor; The device role characteristic acquisition unit is used to acquire the device role characteristic of the device, and establish the device criticality according to the operation stability characteristic, the device role characteristic and the abnormality ratio.

10. The AI-based immersive equipment status monitoring system for thermal power plants according to claim 9, characterized in that: The operation stability feature acquisition unit comprises: A window segmentation unit, used to create a time window, and use the time window to segment the monitoring data set window; An operating parameter calculation unit, used to calculate the maximum value, minimum value and average value of the operating parameters of each window segmentation result; A fluctuation factor establishing unit, used to calculate the window fluctuation factor by using the maximum value, the minimum value and the average value, and to establish the fluctuation factor after weighted calculation of the window fluctuation factor by a time series coefficient; A load factor establishing unit, used to obtain load data of the equipment, perform a proportional analysis of the rated load according to the load data, and establish a load factor; A maintenance cycle factor establishment unit, used to obtain a calibrated maintenance cycle of the equipment and establish a maintenance cycle factor according to the proximity between the current time node and the calibrated maintenance cycle; The operation stability feature construction unit is used to construct the operation stability feature by using the fluctuation factor, the load factor and the maintenance cycle factor.

11. The AI-based immersive equipment status monitoring system for thermal power plants according to claim 8, characterized in that: The equipment abnormality joint warning establishment module includes: an associated range configuration unit, used to obtain the criticality of a main device among the device criticalities, locate the main device using the criticality of the main device, and configure an associated range of the main device; A selected propagation chain establishment unit is configured to screen the propagation chains of the multi-level clustering results based on the main equipment and the associated scope, and establish a selected propagation chain; an additional connection weight establishing unit, configured to obtain a basic connection weight of the selected propagation chain and establish an additional connection weight based on the device criticality; The weighted propagation chain reconstruction unit is used to perform weighted fitting of the selected propagation chain according to the basic connection weight and the additional connection weight, and reconstruct the weighted propagation chain.

12. The AI-based immersive equipment status monitoring system for thermal power plants according to claim 8, characterized in that: The independent abnormality identification result establishment module includes: A feature extraction result establishing unit, used to extract features of the monitoring data set using the feature extraction layer of the independent anomaly recognition network, and establish feature extraction results, wherein the feature extraction results include time domain features, frequency domain features, and time series embedding features; An anomaly calculation unit is used to synchronize the feature extraction result to the distributed anomaly detection subnetwork of the independent anomaly recognition network, perform anomaly calculation of the autoencoder, and establish the independent anomaly recognition result.

13. The AI-based immersive equipment status monitoring system for thermal power plants according to claim 8, characterized in that: The equipment abnormality joint warning establishment module includes: An available data set establishing unit, used for screening the monitoring data set by using the weighted propagation chain to establish an available data set; A joint abnormality calculation result establishing unit, used to respectively identify abnormal features of the equipment through the available data sets, and perform joint abnormality calculation based on the weighted propagation chain on the abnormality feature identification results to establish a joint abnormality calculation result; The equipment abnormality joint warning establishing unit is used to establish the equipment abnormality joint warning according to the joint abnormality calculation result.

14. The AI-based immersive equipment status monitoring system for thermal power plants according to claim 13, characterized in that: The equipment abnormality joint early warning establishing unit also includes: An equipment abnormality trend fitting result establishing unit is configured to perform equipment abnormality trend fitting based on the available data set and establish an equipment abnormality trend fitting result; An abnormal fitting chain establishing unit, used for establishing an abnormal fitting chain according to the abnormal trend fitting result of the equipment and the abnormal joint warning of the equipment; A fitting early warning output unit is used to present the abnormal fitting chain in an immersive and visual manner through the digital twin model and to issue a fitting early warning.

Images