Power plant equipment intelligent coordination control method and system based on multi-source heterogeneous data

Abstract

The invention discloses an intelligent coordination control method and system for power plant equipment based on multi-source heterogeneous data, and belongs to the technical field of intelligent manufacturing and industrial automation, and the method comprises the steps: deploying a multi-mode sensor network in the power plant equipment, collecting the multi-source heterogeneous data in real time, and carrying out the real-time data preprocessing through an edge calculation node; carrying out collaborative modeling on the preprocessed data by adopting a hybrid analysis framework, predicting an equipment state trend, identifying a fault propagation path, positioning a root cause and optimizing a maintenance decision scheme; the equipment failure probability is evaluated through a fault diagnosis result, a grading early warning mechanism is triggered, and a rule base is updated and optimized in combination with a dynamic knowledge base; a three-dimensional model is constructed by using a digital twinning technology to carry out virtual simulation and remote control, and maintenance guidance is carried out through an augmented reality auxiliary technology. According to the method, efficient real-time monitoring and fault prediction are achieved, the fault diagnosis time and the operation and maintenance cost are remarkably reduced by combining the fault tree model and the digital twinning technology, and the equipment operation safety and reliability are improved.

Description

Technical Field

[0001] This invention relates to the fields of intelligent manufacturing and industrial automation technology, specifically to a method and system for intelligent coordinated control of power plant equipment based on multi-source heterogeneous data. Background Technology

[0002] Traditional power plant operation and maintenance relies on manual inspections, paper-based work orders, and experience-driven decision-making, which has significant limitations: equipment condition monitoring depends on discrete sensors, with low data acquisition frequency and insufficient coverage, making it difficult to capture transient anomalies; various subsystems (such as DCS, SIS, vibration monitoring, etc.) use heterogeneous communication protocols, resulting in severe data silos and a lack of a unified analysis platform; fault diagnosis is mostly based on threshold comparison and rule base matching, which cannot handle multi-source data coupling relationships and complex fault propagation mechanisms, resulting in weak predictive maintenance capabilities; operation and maintenance work order execution relies on manual experience scheduling, leading to low resource allocation efficiency, high risk of unplanned downtime, and over-reliance on periodic maintenance can easily cause equipment redundancy losses, resulting in high overall operation and maintenance costs.

[0003] Existing technologies have several drawbacks, including: the inconsistency of features in the spatiotemporal domain of heterogeneous data such as vibration signals, temperature fields, and visual images based on multi-source data fusion frameworks; the inconsistency between integrated mechanism models (such as ISO standards) and data-driven results (such as the association paths mined by GNNs) in constructing dynamic knowledge graphs of equipment-fault-maintenance rules; and the objective shortcomings of fault tree Bayesian inference based on evidence theory in quantifying the joint probability distribution in scenarios with multiple concurrent faults. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, this invention provides an intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data, which is used to solve the problem of how to achieve intelligent coordinated control and fault prediction of power plant equipment, so as to improve the operating efficiency of the equipment and reduce maintenance costs.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for intelligent coordinated control of power plant equipment based on multi-source heterogeneous data, comprising,

[0007] As a preferred embodiment of the intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data described in this invention, the method includes: deploying a multimodal sensor network in the power plant equipment to collect multi-source heterogeneous data in real time; performing real-time data preprocessing through edge computing nodes; using a hybrid analysis framework to collaboratively model the preprocessed data, predicting equipment status trends, identifying fault propagation paths, locating root causes, and optimizing maintenance decision-making schemes; assessing equipment failure probability through fault diagnosis results, triggering a graded early warning mechanism, and updating and optimizing the rule base in conjunction with a dynamic knowledge base; driving execution with decision output, using digital twin technology to construct a three-dimensional model for virtual simulation and remote control, and providing maintenance guidance through augmented reality-assisted technology.

[0008] As a preferred embodiment of the intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data described in this invention, the multi-modal sensor network includes the acquisition of multi-source heterogeneous data, including multi-dimensional parameters, visual anomaly signals, and acoustic anomaly signals, by installing sensors.

[0009] Preprocessing of heterogeneous data from multiple sources by using edge computing nodes to perform noise reduction, data compression, and protocol conversion to unify the heterogeneous data format;

[0010] Simultaneously, local feature extraction is performed to identify surface defects or abnormal signals on the equipment, and all collected data is stored, with full historical data synchronized to the cloud.

[0011] As a preferred embodiment of the intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data described in this invention, the prediction of equipment status trends includes: using an LSTM-Transformer hybrid model, fusing multi-dimensional parameters, and establishing a multi-step prediction model.

[0012] The model uses a multi-step prediction model to monitor in real time and calculates residual statistics using a sliding window. When the statistics trigger a threshold rule, the point is marked as a potential outlier.

[0013] As a preferred embodiment of the intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data described in this invention, the method for identifying fault propagation paths includes: abstracting equipment as nodes, defining the physical connections and control logic between equipment as edges, constructing an industrial equipment graph, calculating node attention weights, and identifying fault propagation paths.

[0014] As a preferred embodiment of the intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data described in this invention, the root cause localization includes deploying a model based on Pearl causal graphs, combining a rule base with data-driven results, and quantifying the contribution of key factors.

[0015] The expected effects of different maintenance options are evaluated through Monte Carlo simulation, and the optimal cost-benefit ratio option is output.

[0016] As a preferred embodiment of the intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data described in this invention, the method for assessing equipment failure probability includes: constructing a fault tree model, calculating the failure probability of each component, and calculating the posterior probability based on evidence theory.

[0017] A four-level early warning mechanism is set up to provide early warning of equipment failures, and the failure rule base is continuously optimized through expert feedback and online incremental learning.

[0018] As a preferred embodiment of the intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data described in this invention, the maintenance guidance includes: constructing a three-dimensional model of the power plant based on the Unity engine, mapping equipment status data in real time, and performing virtual simulation and remote control.

[0019] The virtual simulation simulates the changes in performance parameters after equipment maintenance, and the remote control is the bidirectional control command transmission between the AR glasses and the DCS system via WebRTC.

[0020] By using AR positioning algorithms and overlaying maintenance guidance animations, the error rate of manual operation can be reduced.

[0021] Another objective of this invention is to provide an intelligent coordinated control system for power plant equipment based on multi-source heterogeneous data.

[0022] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an intelligent coordination control system for power plant equipment based on multi-source heterogeneous data, comprising:

[0023] The data acquisition layer, through the deployment of a multimodal sensor network, collects multi-dimensional parameters, visual anomaly signals, and acoustic anomaly signals of the device in real time; it uses edge computing nodes to perform noise reduction, data compression, and protocol conversion on the raw data to unify heterogeneous data formats; it performs feature extraction locally and stores the processed data and synchronizes it to the cloud;

[0024] The analysis layer employs an LSTM-Transformer hybrid model to fuse multi-dimensional parameters for multi-step trend prediction and detects potential outliers using sliding window residual statistics. It constructs an industrial equipment map, calculates node attention weights to identify fault propagation paths, combines a Pearl causal graph model with a rule base to quantify the contribution of key factors, and locates root cause faults. Finally, it evaluates the cost-effectiveness of different maintenance options through Monte Carlo simulation and outputs the optimal decision.

[0025] The decision-making level calculates the component failure probability based on the fault tree model and triggers early warnings according to four-level thresholds; the fault rule base is dynamically optimized through expert feedback and online incremental learning.

[0026] The execution layer constructs a 3D digital twin model of the power plant based on the Unity engine, mapping equipment status data in real time; it simulates changes in performance parameters after maintenance, and uses the WebRTC protocol to enable bidirectional command transmission between AR glasses and the DCS system; it uses AR positioning algorithms to overlay maintenance animation guidance on physical equipment, reducing human operation errors; and it provides visual maintenance step prompts in conjunction with a dynamic knowledge base.

[0027] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data.

[0028] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data.

[0029] The beneficial effects of this invention are as follows: By deploying a multimodal sensor network in power plant equipment to collect multi-source heterogeneous data in real time and utilizing edge computing for data preprocessing, the efficiency and accuracy of data processing are significantly improved. This step enables high-precision real-time monitoring and abnormal signal capture, quickly identifying potential equipment hazards and providing a reliable data foundation for subsequent collaborative modeling. By adopting a hybrid analysis framework, multi-step prediction of equipment status can be performed, fault propagation paths can be identified, and maintenance decisions can be optimized, thereby improving the operational safety and reliability of the equipment. Furthermore, by assessing the probability of equipment failure through fault tree modeling and implementing an early warning mechanism, timely maintenance measures can be taken to reduce downtime. Finally, combining digital twin technology for virtual simulation and augmented reality-assisted maintenance guidance not only reduces human error but also improves maintenance efficiency and accuracy. Overall, this invention significantly shortens fault diagnosis time and reduces operation and maintenance costs and the incidence of unplanned downtime. Attached Figure Description

[0030] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1This is a flowchart illustrating an intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data, provided as an embodiment of the present invention.

[0032] Figure 2 This is a hierarchical diagram of an intelligent coordinated control system for power plant equipment based on multi-source heterogeneous data, provided as an embodiment of the present invention. Detailed Implementation

[0033] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0034] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides a method for intelligent coordinated control of power plant equipment based on multi-source heterogeneous data, including:

[0035] S1. Deploy a multimodal sensor network in power plant equipment to collect multi-source heterogeneous data in real time and perform real-time data preprocessing through edge computing nodes;

[0036] Specifically, in one embodiment of the present invention, high-precision IoT sensors are installed in key equipment such as steam turbines, generators, and boilers. For example, vibration sensors have a sampling rate of ≥20kHz and temperature sensors have an accuracy of ±0.5℃. These sensors collect multi-source heterogeneous data, including multi-dimensional parameters, visual anomaly signals, and acoustic anomaly signals. The multi-dimensional parameters include historical operating parameters of the equipment, equipment parameters, and external environmental data. Industrial cameras and acoustic sensors are deployed to capture visual and acoustic anomaly signals of the equipment.

[0037] The lightweight EdgeX Foundry framework is used to deploy edge servers for real-time data preprocessing, specifically as follows:

[0038] Noise is eliminated by Kalman filtering and wavelet transform, and the data is compressed to 1 / 5 of the original amount.

[0039] Protocol conversion: Supports unified translation of heterogeneous protocols such as Modbus TCP / IP and OPC UA into MQTT message format;

[0040] Local feature extraction: Real-time identification of visual defects such as cracks and leaks on the surface of equipment based on lightweight CNN models (such as MobileNet-SSD).

[0041] It should be noted that the surface images of the equipment captured by the industrial camera are input into the lightweight CNN model after being denoised by wavelet transform. The model's convolutional layers automatically extract texture gradient features (such as the linear gradient of cracks and the droplet shape of leaks), dynamically adjust the size of the feature receptive field to adapt to defects of different sizes, generate the bounding box of the defect region (such as the coordinates of the crack region in the boiler pipe), and output the defect type and confidence level (such as "pipe crack: 92%)". When the confidence level is >90%, a structured data package (equipment ID, defect coordinates, type label) is pushed to the S2 hybrid analysis framework.

[0042] Data storage: The edge stores the time-series data of the most recent 3 days (sampling frequency 1Hz), and the cloud synchronizes the full historical data.

[0043] In an optional embodiment, the preprocessing can be as follows: the original waveform of the vibration sensor (such as the vibration signal of a steam turbine bearing) is segmented into time windows (such as 50ms); the arithmetic mean of the data in each window is calculated and replaced with the values ​​of all sampling points in that window to suppress high-frequency noise; the filtered data is sampled at fixed intervals (such as keeping 1 out of every 5 points); and the unsampled points are directly discarded to compress the amount of data to 1 / 5 of the original.

[0044] In another optional embodiment, the preprocessing may also include: dividing the vibration signal into segments of fixed length (e.g., 200 sampling points per segment); taking the median value of each segment and replacing all point values ​​within that segment to suppress impulse noise (e.g., occasional interference from the sensor); setting a fixed storage step size (e.g., storing 1 point every 0.1 seconds) and directly discarding the data within the interval; and compressing the data volume to 1 / 5 of the original (20kHz → 10Hz storage).

[0045] S2. Use a hybrid analysis framework to collaboratively model preprocessed data, predict equipment status trends, identify fault propagation paths, locate root causes, and optimize maintenance decision-making schemes.

[0046] Specifically, when receiving structured data packets, an LSTM-Transformer hybrid model is used to fuse multi-dimensional parameters and establish a multi-step prediction model.

[0047] It should be noted that the multi-step forecasting model is divided into initial forecasting, medium-term forecasting, and long-term forecasting.

[0048] The initial prediction is a short-term state prediction (0-30 seconds). The input is the real-time vibration spectrum of the device and the temperature gradient calculated by the edge. The instantaneous fluctuations (such as a sudden increase in bearing temperature of 0.5℃ / second) are captured by the LSTM layer. The output is whether the device will trigger the protection threshold within 30 seconds. If the protection threshold is triggered, the current device will be predicted in the medium term.

[0049] Mid-term trend projection (1-5 minutes): Input the visual defect features extracted by S1 (such as crack propagation rate), associate the coupling effect of multiple devices through the Transformer layer (such as boiler pressure → turbine speed), and output the key parameter change curve (such as the vibration value will exceed the limit by 8% after 2 minutes).

[0050] Long-term risk prediction (5-30 minutes): Input external environmental data and equipment historical degradation records, and predict based on key parameter change curves.

[0051] Compare real-time parameter values ​​with the baseline and calculate the percentage of offset (e.g., current vibration value 70μm → offset rate = (70-50) / 50 = 40%). Introduce a time decay factor; the longer the offset lasts, the higher the degradation. Calculate the coupling degradation index = vibration degradation × temperature influence coefficient + flow degradation × coupling weight. Retrieve historical cases with similar degradation indices (e.g., within ±5%), statistically analyze the predicted failure rate, and output a failure probability heatmap. The heatmap indicates the predicted values ​​for different failures.

[0052] The design is based on an improved CUSUM algorithm for real-time monitoring, upgrading the static statistical model to a working condition adaptive dynamic model. The residual statistics are calculated through a sliding window, and when the statistics trigger a threshold rule, they are marked as potential outliers.

[0053] Furthermore, in one embodiment of the present invention, the equipment is abstracted as nodes, and the physical connection and control logic between the equipment are defined as edges. An industrial equipment graph containing 1,200+ nodes is constructed. The influence strength between nodes is analyzed based on graph neural networks (e.g., the influence coefficient of a steam turbine anomaly on a boiler is 0.85). The attention weight of the nodes is calculated, and the path is traced back along the high-weight edge (high weight is greater than 0.7) (e.g., steam turbine → steam valve → boiler) to identify the fault propagation path.

[0054] In an optional embodiment, identifying fault propagation paths can be achieved by assigning a static influence coefficient to each edge based on an industrial equipment map, making it impossible to dynamically adjust with operating conditions; when any equipment is marked as an abnormal node, path backtracking is initiated; the path is backtracked using breadth-first search (BFS), starting from the abnormal node and traversing adjacent nodes layer by layer outwards, only filtering edges with a weight greater than 0.7, recording forward paths that meet the conditions, checking only directly adjacent nodes, ignoring multiple indirect associations, merging all paths that meet the static weight threshold, and generating a fault propagation chain list.

[0055] In another optional embodiment, identifying the fault propagation path can also be achieved by manually defining several "fault propagation chain" rules based on historical fault reports. Each rule includes a triggering condition, a propagation target (device ID), and a confidence level. When any device malfunctions, all rules in the rule base that start from that device are retrieved. The propagation path is triggered only if the real-time parameters fully meet the threshold conditions in the rule. If the triggered propagation target device also malfunctions, the rules that start from that device are matched to generate multi-level chains, and all activated rule chains are returned.

[0056] Furthermore, in one embodiment of the present invention, a Pearl causal graph model is constructed using the root cause of the fault in the fault propagation path as the node and the directed edge connecting the cause and the result as the directed edge. The contribution of key factors is quantified by combining the rule base and data-driven results, freezing other factors, activating the current root cause alone, and observing the change in the system fault probability. The magnitude of the change is the contribution.

[0057] The expected effects of different maintenance options are evaluated through Monte Carlo simulation. The cost-benefit ratio of each option (the benefit of trouble-free operation time after maintenance / total cost) is calculated, and the option with the best cost-benefit ratio is output.

[0058] In an optional embodiment, locating the root cause of a fault can be achieved by extracting the fault propagation path from the industrial equipment map, retaining the influence coefficient between nodes, calculating the path attenuation weight, setting the initial weight of the root cause node to 1.0, attenuating it layer by layer along the propagation path, multiplying the weight by the influence coefficient of each edge, distributing the degradation value of the fault node in reverse to the root cause node according to the path weight, and outputting the root cause list in descending order of contribution.

[0059] In another optional embodiment, locating the root cause of the fault can also involve encoding the current fault features into a vector, retrieving similar historical cases, retrieving cases with feature similarity > 80% from the cloud historical database, filtering the top 10 cases, counting the frequency of root cause occurrence in similar cases, calculating the contribution, root cause contribution = number of occurrences of the root cause / total number of cases × fault degradation value, and outputting the root cause list in descending order of contribution.

[0060] S3. Assess the probability of equipment failure based on fault diagnosis results, trigger a graded early warning mechanism, and update and optimize the rule base in conjunction with a dynamic knowledge base.

[0061] Specifically, the fault tree model containing 200+ nodes is constructed as follows:

[0062] The top event is equipment shutdown;

[0063] Intermediate events are logical combinations (e.g., "lubrication failure" = "insufficient oil pressure" or "oil deterioration");

[0064] The root cause of the event is the component located by S2 (such as bearing wear);

[0065] Calculate the failure probability of each component, input the real-time state prediction of the equipment from the S2 hybrid model, the potential anomalies marked by the S2 improved CUSUM algorithm, and locate the root cause component ID.

[0066] The basic failure rate of components is retrieved from the historical database, and the basic value is adjusted according to the real-time prediction data of S2. The evidence is weighted in real time. If the amplitude of the vibration sensor exceeds the threshold of ISO 10816 standard, the evidence weight is increased by 30%; if the S1 identifies a bearing housing crack in the visual defect (confidence level 95%), the evidence weight is increased by 40%; if the abnormal noise spectrum in the acoustic signal matches the bearing failure characteristics, the evidence weight is increased by 20%.

[0067] Component failure probability = Base probability × (1 + Σ evidence weight)

[0068] The failure probability is dynamically transmitted from bottom to top through the fault tree, and the probability of the top event is updated in real time.

[0069] Four-level early warning thresholds are set, and the fault rule base is continuously optimized through expert feedback and online incremental learning.

[0070] Level 1 warning (failure probability > 95%): Automatically sends a shutdown command and locks the equipment;

[0071] Level 2 warning (failure probability > 70%): Generate preventative maintenance work orders and dispatch spare parts;

[0072] Level 3 warning (failure probability > 50%): Activate redundant equipment and limit the load on the current equipment.

[0073] Level 4 early warning (failure probability > 30%), increase data collection frequency, and activate real-time monitoring of digital twins.

[0074] S4. Decision-driven execution utilizes digital twin technology to construct a 3D model for virtual simulation and remote control, and provides maintenance guidance through augmented reality-assisted technology.

[0075] Specifically, the maintenance guidelines include building a 3D model of the power plant based on the Unity engine, mapping equipment status data in real time, and performing virtual simulation and remote control;

[0076] The virtual simulation simulates the changes in performance parameters after equipment maintenance, and the remote control is the bidirectional control command transmission between the AR glasses and the DCS system via WebRTC.

[0077] By using AR positioning algorithms, specifically SLAM+UWB fusion positioning, and overlaying maintenance guidance animations, the error rate of manual operation can be reduced.

[0078] Specifically, maintenance personnel wearing AR glasses activate the system, using SLAM (Simultaneous Localization and Mapping) to scan the environment and generate a 3D spatial map. Simultaneously, UWB (Ultra-Wideband) positioning base stations capture the personnel's location in real time (accuracy ±5cm). The faulty component located by the S3 is mapped to the AR view through the digital twin platform, with a red flashing frame superimposed on the surface of the real equipment. Pre-stored maintenance procedures (such as "high-temperature bolt disassembly process") are invoked, and dynamic disassembly and assembly animations (such as torque wrench rotation direction arrows and disassembly sequence numbers) are superimposed on the fault location. After confirming the operation steps through AR gestures, a "close steam valve" command is sent to the DCS system via the WebRTC protocol, and a valve closing progress bar is displayed in real time on the AR interface. The camera monitors human actions, and if the deviation from the standard procedure is >15° (such as incorrect wrench angle), a vibration warning is triggered and the guidance animation is replayed.

[0079] In one optional embodiment, the maintenance instructions may involve affixing dedicated QR code markers to critical equipment (such as turbine casings), deploying low-power BLE 4.0 beacons (10-meter spacing) in the maintenance area, scanning the QR codes with AR glasses to obtain the absolute coordinates of the equipment (error ±20cm), and determining the approximate location of personnel through BLE signal strength (RSSI) triangulation; continuously tracking the QR code position using a camera, switching to gyroscope inertial navigation when the marker moves out of the field of view (cumulative error 1 meter / minute); requiring manual scanning of the equipment's QR code to load the maintenance animation, and being unable to dynamically track moving targets in real time.

[0080] In another optional embodiment, the maintenance guidance may also be as follows: maintenance personnel complete the initial alignment at a known coordinate point (such as the boiler platform entrance) and activate the IMU (Inertial Measurement Unit); the displacement is calculated using accelerometers and gyroscopes (with an error of ±2% per step), and the position needs to be corrected by touching the device's NFC tag every 30 seconds; when the cumulative error of the IMU is >50cm, the area is matched by accessing the plant's WiFi fingerprint database (based on a historical map of signal strength); the maintenance animation is only displayed at a fixed viewpoint and cannot dynamically adjust the projection angle according to the actual position of the device.

[0081] In summary, by implementing the method according to the present invention, fault diagnosis time is reduced from hours to minutes, predictive maintenance coverage of critical equipment is ≥90%, annual operation and maintenance costs are reduced by 15%-20%, and unplanned downtime is reduced by 40%.

[0082] Example 2 is an embodiment of the present invention, which provides an intelligent coordinated control system for power plant equipment based on multi-source heterogeneous data, including:

[0083] The data acquisition layer, through the deployment of a multimodal sensor network, collects multi-dimensional parameters, visual anomaly signals, and acoustic anomaly signals of the device in real time; it uses edge computing nodes to perform noise reduction, data compression, and protocol conversion on the raw data to unify heterogeneous data formats; it performs feature extraction locally and stores the processed data and synchronizes it to the cloud;

[0084] The analysis layer employs an LSTM-Transformer hybrid model to fuse multi-dimensional parameters for multi-step trend prediction and detects potential outliers using sliding window residual statistics. It constructs an industrial equipment map, calculates node attention weights to identify fault propagation paths, combines a Pearl causal graph model with a rule base to quantify the contribution of key factors, and locates root cause faults. Finally, it evaluates the cost-effectiveness of different maintenance options through Monte Carlo simulation and outputs the optimal decision.

[0085] The decision-making level calculates the component failure probability based on the fault tree model and triggers early warnings according to four-level thresholds; the fault rule base is dynamically optimized through expert feedback and online incremental learning.

[0086] The execution layer constructs a 3D digital twin model of the power plant based on the Unity engine, mapping equipment status data in real time; it simulates changes in performance parameters after maintenance, and uses the WebRTC protocol to enable bidirectional command transmission between AR glasses and the DCS system; it uses AR positioning algorithms to overlay maintenance animation guidance on physical equipment, reducing human operation errors; and it provides visual maintenance step prompts in conjunction with a dynamic knowledge base.

[0087] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a power plant equipment intelligent coordinated control method based on multi-source heterogeneous data as proposed in the above embodiments.

[0088] The storage medium proposed in this embodiment belongs to the same inventive concept as the method for intelligent coordinated control of power plant equipment based on multi-source heterogeneous data proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0089] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0090] 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 preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligent coordinated control of power plant equipment based on multi-source heterogeneous data, characterized in that: include, Deploy a multimodal sensor network in power plant equipment to collect multi-source heterogeneous data in real time, and perform real-time data preprocessing through edge computing nodes; A hybrid analysis framework is used to collaboratively model preprocessed data, predict equipment status trends, identify fault propagation paths, locate root causes, and optimize maintenance decision-making schemes. The probability of equipment failure is assessed by the fault diagnosis results, triggering a graded early warning mechanism and updating and optimizing the rule base in conjunction with a dynamic knowledge base. Decision-making output drives execution, utilizing digital twin technology to construct 3D models for virtual simulation and remote control, and employing augmented reality-assisted technology for maintenance guidance.

2. The intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data as described in claim 1, characterized in that: The multimodal sensor network includes the acquisition of multi-source heterogeneous data, including multi-dimensional parameters, visual anomaly signals, and voiceprint anomaly signals, by installing sensors. Preprocessing of heterogeneous data from multiple sources by using edge computing nodes to perform noise reduction, data compression, and protocol conversion to unify the heterogeneous data format; Simultaneously, local feature extraction is performed to identify surface defects or abnormal signals on the equipment, and all collected data is stored, with full historical data synchronized to the cloud.

3. The intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data as described in claim 1, characterized in that: The prediction of device status trends includes using an LSTM-Transformer hybrid model to fuse multi-dimensional parameters and establish a multi-step prediction model. The model uses a multi-step prediction model to monitor in real time and calculates residual statistics using a sliding window. When the statistics trigger a threshold rule, the point is marked as a potential outlier.

4. The intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data as described in claim 1, characterized in that: The method for identifying fault propagation paths includes abstracting devices as nodes, defining the physical connections and control logic between devices as edges, constructing an industrial equipment graph, calculating node attention weights, and identifying fault propagation paths.

5. The intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data as described in claim 4, characterized in that: The root cause identification includes deploying a Pearl causal graph-based model, combining a rule base with data-driven results, and quantifying the contribution of key factors. The expected effects of different maintenance options are evaluated through Monte Carlo simulation, and the optimal cost-benefit ratio option is output.

6. The intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data as described in claim 5, characterized in that: The assessment of equipment failure probability includes, Construct a fault tree model, calculate the failure probability of each component, and calculate the posterior probability based on evidence theory; A four-level early warning mechanism is set up to provide early warning of equipment failures, and the failure rule base is continuously optimized through expert feedback and online incremental learning.

7. The intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data as described in claim 6, characterized in that: The maintenance guidelines include building a 3D model of the power plant based on the Unity engine, mapping equipment status data in real time, and performing virtual simulation and remote control. The virtual simulation simulates the changes in performance parameters after equipment maintenance, and the remote control is the bidirectional control command transmission between the AR glasses and the DCS system via WebRTC. By using AR positioning algorithms and overlaying maintenance guidance animations, the error rate of manual operation can be reduced.

8. A power plant equipment intelligent coordination control system based on multi-source heterogeneous data, employing the power plant equipment intelligent coordination control method based on multi-source heterogeneous data as described in any one of claims 1 to 7, characterized in that, include: The data acquisition layer, through the deployment of a multimodal sensor network, collects multi-dimensional parameters, visual anomaly signals, and acoustic anomaly signals of the device in real time; Edge computing nodes are used to perform noise reduction, data compression, and protocol conversion on the raw data to unify heterogeneous data formats; feature extraction is performed locally, and the processed data is stored and synchronized to the cloud. The analysis layer employs an LSTM-Transformer hybrid model to fuse multi-dimensional parameters for multi-step trend prediction and detects potential outliers using sliding window residual statistics. It constructs an industrial equipment map, calculates node attention weights to identify fault propagation paths, combines a Pearl causal graph model with a rule base to quantify the contribution of key factors, and locates root cause faults. Finally, it evaluates the cost-effectiveness of different maintenance options through Monte Carlo simulation and outputs the optimal decision. The decision-making level calculates the component failure probability based on the fault tree model and triggers early warnings according to the four-level threshold. The fault rule base is dynamically optimized through expert feedback and online incremental learning. The execution layer constructs a 3D digital twin model of the power plant based on the Unity engine, mapping equipment status data in real time; it simulates changes in performance parameters after maintenance, and uses the WebRTC protocol to enable bidirectional command transmission between AR glasses and the DCS system; it uses AR positioning algorithms to overlay maintenance animation guidance on physical equipment, reducing human operation errors; and it provides visual maintenance step prompts in conjunction with a dynamic knowledge base.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the intelligent coordinated control method for power plant equipment based on multi-source heterogeneous data as described in any one of claims 1 to 7.

Images