Method and system for remotely monitoring and diagnosing organic heat carrier driven by cloud platform

Abstract

The invention discloses a cloud platform-driven organic heat carrier remote monitoring and diagnosis method and system, and relates to the technical field of organic heat carrier monitoring, and the method comprises the steps: positioning key nodes of organic heat carrier circulation, and deploying a multi-dimensional sensor network connected with edge equipment in combination with the operation conditions of the key nodes; the edge equipment continuously collects operation data of the organic heat carrier and uploads the operation data to the cloud platform, and meanwhile inherent characteristics of the organic heat carrier are synchronously transmitted; the cloud platform calls a fault mode set based on inherent feature matching and trains a comprehensive fault diagnosis model in combination with historical records; and carrying out fault identification on real-time data by using the model, and executing an early warning response based on an identification result. Therefore, the technical effects of data acquisition precision, monitoring and diagnosis path diversification, and fault diagnosis accuracy and efficiency improvement are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of organic heat carrier monitoring, and particularly to a remote monitoring and diagnosis method and system for organic heat carriers driven by a cloud platform. Background Art

[0002] In the modern industrial production process, as a common heating equipment, the safe and stable operation of an organic heat carrier boiler is crucial for many industrial processes. Traditionally, the monitoring of the organic heat carrier circulation system mainly relies on manual inspections and some basic local monitoring devices. These methods can often only obtain limited and partial data information, and it is difficult to achieve comprehensive and real-time control of the entire organic heat carrier circulation system.

[0003] Currently, there are certain limitations in the monitoring means for organic heat carrier circulation. On the one hand, it is impossible to accurately locate those key nodes in the target organic heat carrier circulation, making the monitoring work lack pertinence and it is difficult to effectively obtain a large amount of important data. On the other hand, the functions of the monitoring devices and data processing systems used are relatively single, and the collected data cannot be deeply mined and efficiently utilized, making it difficult to build a perfect fault diagnosis system and it is impossible to discover potential fault hazards in a timely and accurate manner. Summary of the Invention

[0004] The present invention provides a remote monitoring and diagnosis method and system for organic heat carriers driven by a cloud platform to solve the technical problems in the prior art such as lack of pertinence in data collection, single monitoring and diagnosis path, and affecting the accuracy and efficiency of fault diagnosis, and to achieve the technical effects of online monitoring, complete monitoring, and good early warning ability.

[0005] In the first aspect, the present invention provides a remote monitoring and diagnosis method for organic heat carriers driven by a cloud platform. Among them, the remote monitoring and diagnosis method for organic heat carriers driven by the cloud platform includes: Locate the key nodes of the target organic heat carrier circulation, and deploy a multi-dimensional sensor network in combination with the current conditions of the target organic heat carrier circulation, wherein the multi-dimensional sensor network is connected to an edge device.

[0006] Activate the edge device to continuously collect the organic heat carrier, upload the obtained operation data of the organic heat carrier to the target cloud platform, and synchronously transmit the inherent characteristics of the organic heat carrier in the target organic heat carrier circulation.

[0007] The target cloud platform performs matching calls of fault modes based on the inherent characteristics, obtains a set of fault modes, and trains a comprehensive fault diagnosis model based on historical records in combination with the set of fault modes.

[0008] The target cloud platform performs fault identification on the operation data of the organic heat carrier collected in real time based on the trained comprehensive fault diagnosis model, and performs early warning response according to the fault identification result.

[0009] In a feasible implementation, locating the key nodes of the target organic heat carrier cycle includes: Interact with the target organic heat carrier cycle to obtain cycle structure information including a cycle topology diagram, design data, and operating parameters.

[0010] Based on the cycle structure information, identify the parts with significant pressure changes, frequent heat exchange, and key control valves.

[0011] Classify and analyze the identified parts, determine the nodes that meet the requirements of monitoring representativeness, status indication, and accessibility, and form a key node set.

[0012] In a feasible implementation, combining the current conditions of the target organic heat carrier cycle, deploy a multi-dimensional sensor network, including: Based on the current conditions of the target organic heat carrier cycle, extract the current sensor form.

[0013] Based on the complement of the preset sensor form and the current sensor form as the target layout form, update and deploy the current sensors of the target organic heat carrier cycle to construct the multi-dimensional sensor network, where the preset sensor form includes at least temperature sensors, pressure sensors, flow sensors, dielectric constant sensors, and viscosity sensors.

[0014] In a feasible implementation, a rule-based expert system is integrated in the target cloud platform, where the expert system includes a plurality of expert subsystems, each expert subsystem corresponds to a type of organic heat carrier, and the operating standard parameter space and fault mode library of the corresponding organic heat carrier are built in.

[0015] In a feasible implementation, the target cloud platform performs matching and calling of fault modes based on the inherent characteristics, obtains a set of fault modes, and combines the set of fault modes to train a comprehensive fault diagnosis model based on historical records, including: According to the inherent characteristics of the organic heat carrier in the target scenario, match the corresponding expert subsystem and call the corresponding set of fault modes.

[0016] Based on the set of fault modes and the historical records stored in the target cloud platform, construct a training data set, establish a mapping relationship between the training data set and the set of fault modes, and obtain M training data subsets corresponding to M fault modes.

[0017] Construct and train M groups of machine learning-based specific fault diagnosis models respectively through the M training data subsets, and fuse the M groups of specific fault diagnosis models to construct the comprehensive fault diagnosis model.

[0018] In a feasible implementation manner, fusing the M groups of specific fault diagnosis models to construct the comprehensive fault diagnosis model includes: Perform grouped integration based on the M groups of specific fault diagnosis models to form M specific diagnosis clusters.

[0019] Establish a fault mode classifier for the fault mode set, and the fault mode classifier is used to identify the fault type corresponding to the real-time data.

[0020] Connect the M output ends of the fault mode classifier to the input ends of the M specific diagnosis clusters, and connect the output ends of the M specific diagnosis clusters to the integrated output layer to obtain the comprehensive fault diagnosis model.

[0021] In a feasible implementation manner, fusing the M groups of specific fault diagnosis models to construct the comprehensive fault diagnosis model further includes: Randomly select N specific fault diagnosis models with a preset proportion from the M groups of specific fault diagnosis models, where N < M.

[0022] Randomly introduce sample data of heterogeneous fault modes, perform transfer learning on the extracted N specific fault diagnosis models respectively, and make the models after transfer learning independent as a verification model group.

[0023] Fuse the verification model group and the M groups of specific fault diagnosis models based on the ensemble learning method to obtain the comprehensive fault diagnosis model.

[0024] In a feasible implementation manner, according to the fault recognition result for early warning response, it further includes: Perform fault classification and level determination on the fault recognition result, and combine preset rules to judge whether to trigger an automatic control response, where the automatic control response at least includes reducing load operation, active shutdown processing, adjusting operation parameters, and enabling redundant backup.

[0025] In a feasible implementation manner, after performing early warning response according to the fault recognition result, it further includes: Construct a visual user interface for displaying the real-time operation status, historical data trend, and fault recognition result of the target organic heat carrier circulation.

[0026] Present the corresponding positions of each fault mode in the target organic heat carrier circulation in a graphical manner on the visual user interface, where the color of the display prompt is driven by the confidence level of the fault recognition result.

[0027] In a second aspect, the present invention further provides a remotely monitored and diagnosed system for organic heat carriers driven by a cloud platform. The remotely monitored and diagnosed system for organic heat carriers driven by the cloud platform includes: A key node positioning and sensor network deployment module, configured to position key nodes in the circulation of the target organic heat carrier, and deploy a multi-dimensional sensor network in combination with the current conditions of the circulation of the target organic heat carrier. The multi-dimensional sensor network is connected to edge devices.

[0028] An organic heat carrier operation data acquisition and upload module, configured to activate the edge devices to continuously acquire the organic heat carrier, upload the obtained operation data of the organic heat carrier to the target cloud platform, and synchronously transmit the inherent characteristics of the organic heat carrier in the circulation of the target organic heat carrier.

[0029] A pattern matching and model training module, configured to enable the target cloud platform to perform matching calls of fault patterns based on the inherent characteristics, obtain a set of fault patterns, and perform comprehensive fault diagnosis model training based on historical records in combination with the set of fault patterns.

[0030] A fault identification and early warning response module, configured to enable the target cloud platform to perform fault identification on the operation data of the organic heat carrier collected in real time based on the trained comprehensive fault diagnosis model, and perform early warning responses according to the fault identification results.

[0031] The present invention discloses a remotely monitored and diagnosed method and system for organic heat carriers driven by a cloud platform, including: identifying key nodes in the target organic heat carrier circulation system, and deploying a multi-dimensional sensor network connected to edge devices in combination with its current operating state to achieve state perception; continuously acquiring the operation process of the organic heat carrier through the edge devices, and synchronously uploading the acquired operation data and the inherent characteristic information of the organic heat carrier to the target cloud platform; at the cloud platform end, calling a matching set of fault patterns based on the uploaded inherent characteristics, and carrying out the training of a comprehensive fault diagnosis model based on this set in combination with historical operation records; using the trained comprehensive fault diagnosis model to perform fault identification on the operation data of the organic heat carrier collected in real time, and triggering a corresponding early warning response mechanism based on the identification results. The remotely monitored and diagnosed method and system for organic heat carriers driven by the cloud platform disclosed by the present invention solve the technical problems of lack of pertinence in data acquisition, single monitoring and diagnosis path, and affecting the accuracy and efficiency of fault diagnosis, and achieve the technical effects of precise data acquisition, diversified monitoring and diagnosis paths, and improved accuracy and efficiency of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a schematic flowchart of the remotely monitored and diagnosed method for organic heat carriers driven by the cloud platform of the present invention.

[0033] Figure 2 This is a schematic structural diagram of the organic heat carrier remote monitoring and diagnosis system driven by the cloud platform of the present invention.

[0034] Explanation of reference numerals: The key node positioning and sensor network deployment module 11, the organic heat carrier operation data acquisition and upload module 12, the pattern matching and model training module 13, and the fault identification and early warning response module 14. Specific implementation manners

[0035] The following will combine the specification drawings and specific implementation manners to elaborate on the above technical solutions in detail to better understand the above technical solutions. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments for explaining the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention. In addition, it should be noted that for the sake of description, only parts related to the present invention rather than all are shown in the drawings.

[0036] Embodiment 1, as Figure 1 This is a flowchart of the organic heat carrier remote monitoring and diagnosis method driven by the cloud platform of the present invention. Among them, the organic heat carrier remote monitoring and diagnosis method driven by the cloud platform includes: S100: Locate the key nodes of the target organic heat carrier circulation, and deploy a multi-dimensional sensor network in combination with the current conditions of the target organic heat carrier circulation, where the multi-dimensional sensor network is connected to the edge device.

[0037] Specifically, the key nodes refer to the parts in the organic heat carrier circulation system that have an important impact on the system operation state, such as the parts with significant pressure change, frequent heat exchange, or key control valves, which are the key points for monitoring. The multi-dimensional sensor network is a network composed of various types of sensors (such as temperature, pressure, flow, etc.), which can collect data from different dimensions. The edge device refers to the device installed at the edge of the sensor network for data acquisition, preliminary processing, and transmission.

[0038] Through the accurate positioning of the key nodes of the organic heat carrier circulation system and the adaptive deployment of the multi-dimensional sensor network, the all-round and real-time perception of the system operation state can be achieved. Among them, the data aggregation and preliminary intelligent processing based on the edge device provide high-quality and low-latency basic data support for the cloud platform remote monitoring and diagnosis, thereby greatly improving the intelligent operation and maintenance ability, operation safety, and energy efficiency level of the organic heat carrier system.

[0039] In some embodiments, locating the key nodes of the target organic heat carrier circulation includes: Interact with the target organic heat carrier circulation to obtain circulation structure information including the circulation topology diagram, design data, and operating parameters; based on the circulation structure information, identify the parts with significant pressure changes, frequent heat exchange, and key control valves; classify and analyze the identified parts to determine the nodes that meet the requirements of monitoring representativeness, status indication, and accessibility, and form a set of key nodes.

[0040] Specifically, the circulation topology diagram refers to a structure diagram that reflects the connection relationships of various devices, pipelines, valves, measuring points, etc. in the organic heat carrier circulation system, usually a P&ID (Process and Instrumentation Diagram). Design data refers to engineering and technical documents such as design parameters, equipment selection, and process flow descriptions of the organic heat carrier circulation system. Operating parameters refer to the data of various physical quantities (such as temperature, pressure, flow rate, valve opening, etc.) collected during actual operation.

[0041] Specifically, the requirement of monitoring representativeness means that the key nodes should be able to reflect the typicality of the overall or local operating state of the system. The requirement of status indication means that the parameter changes of the key nodes should be able to sensitively reflect the state changes such as faults, abnormalities, and energy efficiency. Accessibility requires that the key nodes need to have physical space, interface conditions, or safety conditions, which is helpful for the convenient deployment of sensors or monitoring devices.

[0042] Specifically, first, collect the circulation topology diagram, design data, and current / historical operating parameters of the target organic heat carrier circulation through the system interface or manual input. For example, obtain the P&ID diagram, equipment list, design flow rate, pressure grade, and operating data of the past month of the heat transfer oil circulation system in a chemical plant. Then, use structured data analysis to parse the circulation topology diagram and automatically identify the following parts: nodes with significant pressure changes (such as the outlet of the heater, before and after the pump, and the inlet and outlet of the heat exchanger), nodes with frequent heat exchange (such as between multi-stage heat exchangers and heat user branches), and nodes with key control valves (such as the main circulation valve, branch regulating valve, safety valve, etc.). Optionally, assist in the identification through graph theory algorithms, rule bases, or expert systems.

[0043] Furthermore, further analyze the monitoring representativeness, status indication, and accessibility of the above-identified parts. In other words, based on the monitoring representativeness, preferentially select the nodes that can reflect the main circulation conditions, heat loss, and system safety; based on the status indication, preferentially select the nodes that are sensitive to abnormalities such as leakage, blockage, and overheating; based on the accessibility, exclude the nodes with narrow physical space, severe medium corrosion, or inability to safely install sensors; thus, form the final set of key nodes. Exemplarily, the set of key nodes includes the outlet of the heater, the outlet of the main circulation pump, the inlet of the main heat exchanger, the liquid level port of the expansion tank, the return oil main pipe, etc.

[0044] By interactively acquiring loop structure information and identifying key nodes, it is possible to accurately locate the parts that have the greatest impact on the system's operating status, avoiding data loss or invalid monitoring problems caused by improper selection of monitoring points in traditional monitoring methods, and providing a high-quality data foundation for subsequent data collection and fault diagnosis.

[0045] In some embodiments, a multi-dimensional sensor network is deployed in combination with the current conditions of the target organic heat carrier cycle, including: Based on the current conditions of the target organic heat carrier cycle, a current sensor form is extracted; based on the complement of a preset sensor form and the current sensor form as a target layout form, the current sensors of the target organic heat carrier cycle are updated and deployed to construct the multidimensional sensor network, wherein the preset sensor form includes at least a temperature sensor, a pressure sensor, a flow sensor, a dielectric constant sensor and a viscosity sensor.

[0046] Specifically, the current conditions refer to the actual operating status of the target organic heat carrier circulation system, including the existing sensor layout, equipment status, operating parameters, etc. Among them, the current sensor form refers to the list of installed sensors, including the type, location, quantity and other information of the sensors. The preset sensor form refers to the list of sensor configurations pre-designed according to the ideal monitoring requirements of the organic heat carrier circulation system, exemplarily including temperature sensors, pressure sensors, flow sensors, dielectric constant sensors and viscosity sensors, etc.

[0047] Specifically, first, we combed through on-site inspections, system interfaces, historical archives, and other methods to sort out the various sensors currently installed in the target organic heat carrier circulation system and form a current sensor table. For example, a system currently has 8 temperature sensors (model Pt100, distributed in heaters, heat exchangers, and return oil pipes), 3 pressure sensors, 2 flow meters, and no dielectric constant and viscosity sensors.

[0048] Then, according to the monitoring requirements, call the preset sensor form (for example, each key node should be equipped with 1 temperature, pressure, flow, dielectric constant, and viscosity sensor), and calculate the corresponding complement to obtain the target layout form. For example, the preset requirements are: temperature 10, pressure 10, flow 10, dielectric constant 10, viscosity 10; the current situation is: temperature 8, pressure 3, flow 2, dielectric constant 0, viscosity 0; then the complement is: temperature 2, pressure 7, flow 8, dielectric constant 10, viscosity 10; the target layout form is the above complement result, and the respective installation nodes are clearly defined.

[0049] Furthermore, according to the target layout form, the required types and quantities of sensors are installed at key nodes, such as adding a viscosity sensor at the outlet of the main circulation pump and a dielectric constant sensor at the return oil pipe. Then, the sensors are connected to the edge devices by wire (such as Modbus, CAN) or wireless (such as LoRa, NB-IoT) to complete the construction of a multi-dimensional sensor network. Preferably, for nodes with special working conditions, explosion-proof, high-temperature or corrosion-resistant sensors are selected to ensure long-term stable operation.

[0050] The above process can effectively make up for the shortcomings of the existing sensor layout by deploying a multi-dimensional sensor network based on the current conditions, and avoid the problem of incomplete data collection caused by a single sensor type or unreasonable layout.

[0051] S200: activating the edge device to continuously collect data on the organic heat carrier, uploading the acquired operation data of the organic heat carrier to the target cloud platform, and synchronously transmitting the inherent characteristics of the organic heat carrier in the target organic heat carrier cycle.

[0052] Specifically, the organic heat carrier operation data is data that reflects the operating status of the organic heat carrier circulation system, including real-time data collected by sensors such as temperature, pressure, and flow. Intrinsic characteristics refer to the characteristics of the organic heat carrier itself, such as chemical composition, thermal stability, dielectric constant, etc., which are important references for fault diagnosis. Through this inherent characteristic, the monitoring and diagnosis plan for the target organic heat carrier can be defined in a targeted manner. The target cloud platform is a cloud platform for storing, processing and analyzing data. The target cloud platform can receive data from edge devices and perform further processing and analysis, thereby realizing remote monitoring and diagnosis of organic heat carriers.

[0053] Specifically, in the implementation, the edge devices deployed at the key nodes of the target organic heat carrier circulation system are first started to continuously collect real-time signals from various sensors. For example, temperature, pressure, flow and other data are collected every 5 seconds, and local preliminary verification and formatting are performed. Then, the collected organic heat carrier operation data is uploaded to the designated cloud platform via a wired / wireless network. At the same time, when the organic heat carrier is started for the first time or replaced, the inherent characteristic parameters of the target organic heat carrier are automatically read or manually entered through the edge device and uploaded together with the dynamic operation data.

[0054] The purpose of this process is to ensure that the cloud platform can obtain comprehensive and real-time data, providing a basis for subsequent fault diagnosis and early warning. Through continuous collection and real-time uploading, potential problems can be discovered and early warnings can be issued in a timely manner to avoid the expansion of faults.

[0055] S300: The target cloud platform performs matching calls for fault modes based on the inherent characteristics, obtains a set of fault modes, and combines the set of fault modes to train a comprehensive fault diagnosis model based on historical records.

[0056] Specifically, according to the inherent characteristics of the organic heat carrier, fault modes that match it can be found in the fault mode library of the cloud platform to obtain possible fault types. Among them, the set of fault modes is the set of all possible fault modes related to the current inherent characteristics of the organic heat carrier, ensuring that the training data of the fault diagnosis model is targeted and only includes fault types related to the current system.

[0057] Specifically, the training of the comprehensive fault diagnosis model refers to the fault diagnosis execution structure obtained through supervised training using machine learning algorithms with historical data (including past fault data and normal operation data).

[0058] Through the fault mode matching call based on inherent characteristics and the model training based on historical records, the accuracy and pertinence of the fault diagnosis model can be effectively improved. Compared with traditional methods, this method avoids using general fault modes for diagnosis, making the model more in line with the actual situation of the current system, thus significantly improving the accuracy of fault identification.

[0059] In some embodiments, a rule-based expert system is integrated in the target cloud platform. Among them, the expert system includes a plurality of expert subsystems, each expert subsystem corresponds to a type of organic heat carrier, and the operating standard parameter space and fault mode library corresponding to the organic heat carrier are built-in.

[0060] Specifically, the expert system makes inferences and judgments based on predefined rules and logics. In this solution, the expert system is used to store rules and logic information related to the fault modes of organic heat carriers, and can extract all the fault modes involved in each type of organic heat carrier by matching the inherent characteristics.

[0061] Specifically, a rule-based expert system is integrated in the target cloud platform. The expert system includes multiple expert subsystems, each expert subsystem corresponds to a type of organic heat carrier, and the operating standard parameter space and fault mode library of this type of organic heat carrier are built-in. When the target cloud platform receives the inherent characteristic data of the organic heat carrier, the expert system can select the corresponding expert subsystem according to these characteristics. The expert subsystem has built-in the operating standard parameter space (i.e., the parameter ranges during normal operation) and fault mode library (i.e., known fault types and their characteristics) of the corresponding organic heat carrier, which are used for matching and reference during fault diagnosis.

[0062] In some embodiments, the target cloud platform matches and invokes failure modes based on the inherent characteristics, obtains a set of failure modes, and combines the set of failure modes to train a comprehensive failure diagnosis model based on historical records, including: According to the inherent characteristics of the organic heat carrier in the target scenario, match the corresponding expert subsystem and invoke the corresponding set of failure modes; construct a training data set based on the set of failure modes and the historical records stored in the target cloud platform, and establish a mapping relationship between the training data set and the set of failure modes to obtain M training data subsets corresponding to M failure modes; through the M training data subsets, respectively construct and train M groups of specific failure diagnosis models based on machine learning, and fuse the M groups of specific failure diagnosis models to construct the comprehensive failure diagnosis model.

[0063] Specifically, the set of failure modes refers to all possible failure types and their characteristic descriptions (such as coking, decomposition, oxidation, leakage, etc.) output by the expert system for a specific type of organic heat carrier and its operating conditions. The specific failure diagnosis model is a machine learning model trained for a specific failure mode and can identify and diagnose the failure mode. The comprehensive failure diagnosis model refers to the total diagnosis model obtained by integrating multiple specific failure diagnosis models (such as voting, weighting, ensemble learning, etc.) and has the comprehensive discrimination ability for multiple failure modes.

[0064] Specifically, after the cloud platform receives the inherent characteristics of the organic heat carrier, first, it automatically matches the corresponding expert subsystem. For example, if the inherent characteristics indicate that the organic heat carrier is of the biphenyl type, the biphenyl type expert subsystem is invoked to obtain a set of relevant failure modes such as "coking", "thermal degradation", "oxidative failure", etc. Then, based on the obtained set of failure modes, retrieve and filter the historical operation data and known failure labels in the target cloud platform to construct a training data set. For example, if the historical data records the operation parameters and failure labels related to "biphenyl type organic heat carrier - coking", it is classified into the training data subset of the "coking" mode.

[0065] Furthermore, establish a mapping relationship between each failure mode and its corresponding training data subset to obtain M training data subsets corresponding to M failure modes, and for each training data subset, use machine learning algorithms (such as decision tree, random forest, SVM, neural network, etc.) to train M groups of specific failure diagnosis models respectively. For example, for the "coking" training subset, train model A; for the "thermal degradation" training subset, train model B, and so on. Finally, fuse the obtained M groups of specific failure diagnosis models through ensemble learning and other methods to obtain the comprehensive failure diagnosis model, so as to realize the unified identification and diagnosis of multiple failure modes.

[0066] Through the above process, the following technical effects can be achieved: Automatically match the applicable set of fault modes according to different organic heat carriers and their inherent characteristics, so as to improve the pertinence and diagnostic accuracy of the model; dynamically construct a training data set for multiple fault modes based on historical data and the expert knowledge base to achieve the self - adaptation and continuous optimization of the model; through multi - model integration, improve the comprehensive discrimination ability of the system for complex working conditions and diverse faults, and enhance the robustness and intelligence level of fault diagnosis.

[0067] In some implementation manners, fuse M groups of the specific fault diagnosis models to construct the comprehensive fault diagnosis model, including: Perform grouped integration based on M groups of specific fault diagnosis models to form M specific diagnosis clusters; establish a fault mode classifier for the set of fault modes, and the fault mode classifier is used to identify the fault type corresponding to the real - time data; connect the M output ends of the fault mode classifier to the input ends of the M specific diagnosis clusters, and connect the output ends of the M specific diagnosis clusters to the integrated output layer to obtain the comprehensive fault diagnosis model.

[0068] Specifically, the specific fault diagnosis model is a machine - learning diagnosis model independently trained for a single fault mode. The specific diagnosis cluster is a model set composed of the same type or related specific fault diagnosis models, and is dedicated to deeply discriminating a certain fault mode.

[0069] Specifically, the fault mode classifier is used to preliminarily classify the input real - time operation data and identify the fault mode to which it most likely belongs. The integrated output layer is used to summarize, weight or make decisions on the diagnosis results of each specific diagnosis cluster and output the final comprehensive diagnosis conclusion.

[0070] Specifically, first, group the M groups of specific fault diagnosis models according to the fault modes, and each group forms a specific diagnosis cluster. For example, multiple models under the coking mode form a "coking diagnosis cluster", and multiple models under the thermal degradation mode form a "thermal degradation diagnosis cluster", etc. Then, based on the previously determined set of fault modes, train a multi - classification model (such as softmax classifier, LightGBM, multi - layer perceptron, etc.) for preliminarily identifying the fault mode of the real - time input data and outputting M category probabilities or confidences.

[0071] Furthermore, connect the M output ends (corresponding to M fault modes respectively) of the fault mode classifier to the input ends of the M specific diagnosis clusters one by one. In other words, the real - time data first passes through the fault mode classifier and is distributed to one or more specific diagnosis clusters that are most likely, and the output results are uniformly summarized to the integrated output layer, and the final comprehensive fault diagnosis result is obtained through methods such as weighted voting and confidence fusion. Among them, the output results include fault probability, fault level, etc.

[0072] Through the above process, an efficient and accurate comprehensive fault diagnosis model can be constructed. This model can make full use of the professionalism of each specific fault diagnosis model to achieve a comprehensive diagnosis of multiple fault types. At the same time, the introduction of the fault mode classifier improves the efficiency and accuracy of fault recognition, enabling real-time data to be quickly classified and sent to the corresponding diagnostic clusters for processing, which helps to improve the diagnostic speed and enhance the robustness and adaptability of the model.

[0073] In some implementation manners, when fusing M groups of the specific fault diagnosis models to construct the comprehensive fault diagnosis model, it further includes: Randomly select N specific fault diagnosis models with a preset proportion from M groups of specific fault diagnosis models, where N < M; randomly introduce sample data of heterogeneous fault modes, and perform transfer learning on the extracted N specific fault diagnosis models respectively, and make the models after transfer learning independent as a verification model group; fuse the verification model group and M groups of specific fault diagnosis models based on the ensemble learning method to obtain the comprehensive fault diagnosis model.

[0074] Specifically, first, randomly select N models (N < M) from M groups of specific fault diagnosis models according to a preset proportion (such as 30% - 50%) to ensure model diversity and representativeness; for example, M = 10 and the preset proportion is 40%, then N = 4. Then, for each selected model, randomly introduce additional sample data of heterogeneous fault modes that do not belong to its original training category, and retrain each model using a reinforcement learning strategy (such as rewarding correct recognition of heterogeneous samples and punishing misjudgments) to improve the model's boundary discrimination ability and adaptability to unknown modes. Independently save the N models after transfer learning to form a verification model group.

[0075] Furthermore, adopt the ensemble learning method to fuse the verification model group and the original M groups of specific fault diagnosis models. For example, adopt the Stacking method, with the first layer being M + N basic models and the second layer being the fusion decision layer to output the final diagnosis result.

[0076] Optionally, adopt methods such as weighted voting and confidence weighting for ensemble to improve the overall performance of the comprehensive fault diagnosis model.

[0077] The comprehensive model obtained through the above process can not only utilize the professionalism of the original model, but also improve the overall diagnostic ability through the optimization of the verification model group, effectively enhancing the model's recognition ability and generalization ability for new, complex or unknown fault modes, improving the adaptability and accuracy of the comprehensive fault diagnosis model under actual complex working conditions, and reducing the risk of false alarms and missed detections.

[0078] S400: The target cloud platform performs fault identification on the operation data of the organic heat carrier collected in real time based on the trained comprehensive fault diagnosis model, and conducts early warning responses according to the fault identification results.

[0079] Specifically, the operation data of the organic heat carrier collected in real time is input into the trained comprehensive fault diagnosis model, and the model analyzes the input data to obtain the fault identification results. Optionally, the fault identification results include fault types, fault locations, inferred reasons, etc.

[0080] Specifically, according to the fault identification results output by the model, the cloud platform automatically triggers corresponding early warning response measures, such as: sending fault early warning notifications to operation and maintenance personnel; pushing detailed fault information to the management background; starting emergency response processes, such as automatically switching to standby systems, reducing operating loads, etc.

[0081] Through the above process, the automatic and intelligent identification of faults in the organic heat carrier system is realized. Among them, through centralized management and response by the cloud platform, early warnings can be issued in a timely manner and emergency measures can be linked, significantly reducing the risk of equipment damage and potential safety hazards in operation, supporting remote operation and maintenance management of large-scale and multi-site organic heat carrier systems, thereby helping to improve the overall operation and maintenance efficiency and service quality.

[0082] In some embodiments, conducting early warning responses according to the fault identification results further includes: Classifying and grading the faults in the fault identification results, and judging whether to trigger an automatic control response in combination with preset rules, where the automatic control response at least includes reducing load operation, active shutdown processing, adjusting operation parameters, and enabling redundant backup.

[0083] Specifically, the preset rules are the conditions and criteria set in advance according to the safe operation requirements and management specifications of the organic heat carrier system for judging whether automatic control responses are required. When detecting specific-level or type faults, control measures that can be automatically executed according to the preset rules include, but are not limited to, reducing load operation, active shutdown processing, adjusting operation parameters, enabling redundant backup, etc.

[0084] Specifically, after the comprehensive fault diagnosis model outputs the fault results, first, classify the fault types, and determine the fault levels based on parameters such as the scope of influence and severity of the faults. Then, compare the fault classification and levels with the preset rule library to judge whether the current fault meets the trigger conditions for automatic control responses. If the automatic response conditions are met, the system automatically executes the corresponding control measures. For example: For general-level faults, only send early warning notifications; for important-level faults, automatically adjust operation parameters or reduce load operation; for severe or emergency-level faults, automatically enable redundant backup or perform active shutdown processing.

[0085] Exemplarily, when the comprehensive fault diagnosis model identifies that "the temperature rise of the main circulation pump is abnormal" and determines it as "severe level", the standby pump can be automatically started according to the rules and the main pump can be shut down, while pushing the fault information to the operation and maintenance personnel. If it identifies "temperature sensor drift" and it is "general level", only a warning prompt is issued without automatic control response.

[0086] Through the above process, the intelligence and automation of fault response are realized, and the optimal control measures can be dynamically matched according to the fault type and level, improving the system safety and operation and maintenance efficiency.

[0087] In some embodiments, after making a warning response according to the fault identification result, it further includes: Construct a visual user interface for displaying the real-time operation status, historical data trend and fault identification result of the target organic heat carrier circulation; present the corresponding positions of each fault mode in the target organic heat carrier circulation in a graphical manner on the visual user interface, wherein the color of the display prompt is driven by the confidence level of the fault identification result.

[0088] Specifically, the visual user interface refers to a graphical interaction interface implemented through a computer or a mobile terminal, which is used to display the operation parameters, historical data and fault diagnosis results of the organic heat carrier circulation system in real time. The corresponding position of the fault mode refers to the occurrence location of various faults in specific equipment or links marked in the system structure diagram or flow chart. The confidence level refers to the credibility of the fault identification model for the current determination result. Preferably, it is represented in numerical or hierarchical form and is used to measure the reliability of the fault determination.

[0089] Specifically, continuously collect the operation parameters of the organic heat carrier circulation (such as temperature, pressure, flow rate, etc.) and display them in real time on the visual user interface in various ways such as numbers, curves, and dashboards. When the fault diagnosis model identifies an abnormality, the specific location where the fault occurs can be intuitively displayed on the interface structure diagram in a highlighted, flashing or special icon manner.

[0090] Specifically, according to the confidence level of the fault identification result, the interface can automatically adjust the prompt color. For example: High confidence level (such as ≥80%): displayed in red, indicating a serious warning; Medium confidence level (50% - 80%): displayed in orange or yellow, prompting key attention; Low confidence level (<50%): displayed in blue or gray, only giving a general prompt.

[0091] Through the above process, the visualization of the operation status and fault information of the organic heat carrier circulation is realized, improving the perception efficiency and response speed of the operation and maintenance personnel. Among them, the confidence level-driven color prompt makes the severity of the fault clear at a glance, facilitating hierarchical response and scientific decision-making.

[0092] In summary, the organic heat carrier remote monitoring and diagnosis method driven by the cloud platform provided by the present invention has the following technical effects: By identifying the key nodes in the target organic heat carrier circulation system and combining their current operating states, a multi-dimensional sensor network connected to edge devices is deployed to achieve state perception; through the edge devices, the operation process of the organic heat carrier is continuously collected, and the collected operation data and the inherent characteristic information of the organic heat carrier are synchronously uploaded to the target cloud platform; at the cloud platform end, according to the uploaded inherent characteristics, a matching set of fault modes is called, and based on this set and historical operation records, the training of a comprehensive fault diagnosis model is carried out; the trained comprehensive fault diagnosis model is used to identify faults in the real-time collected operation data of the organic heat carrier, and based on the identification results, the corresponding early warning response mechanism is triggered, so as to achieve the technical effects of accurate data collection, diversified monitoring and diagnosis paths, and improved fault diagnosis accuracy and efficiency.

[0093] Embodiment 2, as Figure 2 is a schematic structural diagram of the organic heat carrier remote monitoring and diagnosis system driven by the cloud platform of the present invention. For example, Figure 1 In the flow schematic diagram of the organic heat carrier remote monitoring and diagnosis method driven by the cloud platform of the present invention, it can be implemented by a structure such as Figure 2 shown.

[0094] Based on the same concept as the organic heat carrier remote monitoring and diagnosis method driven by the cloud platform in the above embodiment, the organic heat carrier remote monitoring and diagnosis system provided by the present invention further includes: A key node positioning and sensor network deployment module 11, configured to locate the key nodes of the target organic heat carrier circulation, and deploy a multi-dimensional sensor network in combination with the current conditions of the target organic heat carrier circulation, wherein the multi-dimensional sensor network is connected to edge devices.

[0095] An organic heat carrier operation data collection and upload module 12, configured to activate the edge devices to continuously collect the organic heat carrier, upload the obtained organic heat carrier operation data to the target cloud platform, and synchronously transmit the inherent characteristics of the organic heat carrier in the target organic heat carrier circulation.

[0096] A mode matching and model training module 13, configured to enable the target cloud platform to perform matching calls of fault modes based on the inherent characteristics, obtain a set of fault modes, and perform comprehensive fault diagnosis model training based on historical records in combination with the set of fault modes.

[0097] A fault identification and early warning response module 14, configured to enable the target cloud platform to identify faults in the real-time collected operation data of the organic heat carrier based on the trained comprehensive fault diagnosis model, and perform early warning responses according to the fault identification results.

[0098] In some embodiments, the critical node positioning and sensor network deployment module 11 includes: A cyclic structure information acquisition unit, configured to interact with the target organic heat carrier cycle to acquire cyclic structure information including a cyclic topology map, design data, and operating parameters.

[0099] A critical part identification unit, configured to identify parts with significant pressure change, frequent heat exchange, and key control valves based on the cyclic structure information.

[0100] A critical node set formation unit, configured to classify and analyze the identified parts, determine nodes that meet monitoring representativeness, status indication, and accessibility, and form a critical node set.

[0101] In some embodiments, the critical node positioning and sensor network deployment module 11 includes: A current situation sensor form extraction unit, configured to extract a current situation sensor form based on the current situation conditions of the target organic heat carrier cycle.

[0102] A sensor update deployment and multi-dimensional sensor network construction unit, configured to use the complement of the preset sensor form and the current situation sensor form as the target layout form to perform update deployment of the current situation sensors of the target organic heat carrier cycle and construct the multi-dimensional sensor network, where the preset sensor form at least includes a temperature sensor, a pressure sensor, a flow sensor, a dielectric constant sensor, and a viscosity sensor.

[0103] In some embodiments, the pattern matching and model training module 13 includes: An expert subsystem matching and fault mode set calling unit, configured to match a corresponding expert subsystem and call a corresponding fault mode set according to the inherent characteristics of the organic heat carrier in the target scenario.

[0104] A training data set construction and mapping relationship establishment unit, configured to construct a training data set based on the fault mode set and historical records stored in the target cloud platform, establish a mapping relationship between the training data set and the fault mode set, and obtain M training data subsets corresponding to M fault modes.

[0105] A specific fault diagnosis model construction and fusion unit, configured to respectively construct and train M groups of specific fault diagnosis models based on machine learning through the M training data subsets, and fuse the M groups of specific fault diagnosis models to construct the comprehensive fault diagnosis model.

[0106] In some implementation manners, the specific fault diagnosis model construction and fusion unit in the pattern matching and model training module 13 includes: A specific diagnosis cluster formation unit is used to perform grouped integration based on M groups of specific fault diagnosis models to form M specific diagnosis clusters.

[0107] A fault mode classifier establishment unit is used to establish a fault mode classifier for the set of fault modes, and the fault mode classifier is used to identify the fault type corresponding to the real-time data.

[0108] A comprehensive fault diagnosis model acquisition unit is used to connect the M output terminals of the fault mode classifier to the input terminals of the M specific diagnosis clusters, and connect the output terminals of the M specific diagnosis clusters to the integrated output layer to obtain the comprehensive fault diagnosis model.

[0109] In some implementation manners, the specific fault diagnosis model construction and fusion unit in the pattern matching and model training module 13 further includes: A specific fault diagnosis model random selection unit is used to randomly select N specific fault diagnosis models with a preset proportion from M groups of specific fault diagnosis models, where N < M.

[0110] A reinforcement learning and verification model group generation unit is used to randomly introduce sample data of heterogeneous fault modes, perform transfer learning on the extracted N specific fault diagnosis models respectively, and make the models after transfer learning independent as a verification model group.

[0111] A comprehensive fault diagnosis model acquisition unit is used to fuse the verification model group and M groups of specific fault diagnosis models based on the ensemble learning method to obtain the comprehensive fault diagnosis model.

[0112] In some embodiments, the execution steps of the fault identification and early warning response module 14 include: performing fault classification and level determination on the fault identification result, and judging whether to trigger an automatic control response in combination with a preset rule, where the automatic control response at least includes reducing load operation, actively shutting down, adjusting operation parameters, and enabling redundant backup.

[0113] In some implementation manners, the execution steps of the fault identification and early warning response module 14 further include: A visualization user interface construction unit is used to construct a visualization user interface for displaying the real-time operation status, historical data trend, and fault identification result of the target organic heat carrier circulation.

[0114] A fault mode position presentation and color driving unit is used to graphically present the corresponding positions of each fault mode in the target organic heat carrier circulation on the visualization user interface, where the color of the display prompt is driven by the confidence level of the fault identification result.

[0115] In some implementations, a rule-based expert system is integrated in the target cloud platform. Among them, the expert system includes a plurality of expert subsystems. Each expert subsystem corresponds to a type of organic heat carrier, and a running standard parameter space and a fault mode library corresponding to the organic heat carrier are built in.

[0116] It should be understood that the key point of the embodiments mentioned in this specification lies in their differences from other embodiments. The specific embodiments in the first-mentioned Embodiment 1 are equally applicable to the organic heat carrier remote monitoring and diagnosis system driven by the cloud platform described in Embodiment 2. For the sake of brevity of the specification, no further elaboration is made here.

[0117] It should be understood that the disclosed embodiments of the present invention and the above descriptions enable those skilled in the art to implement the present invention using the present invention. At the same time, the present invention is not limited to the part of the embodiments mentioned above. It should be understood that those of ordinary skill in the art can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A cloud platform-driven remote monitoring and diagnosis method for organic heat carriers, characterized in that: include: Locating key nodes of the target organic heat carrier cycle, and deploying a multi-dimensional sensor network in combination with the current conditions of the target organic heat carrier cycle, wherein the multi-dimensional sensor network is connected to the edge device; activating the edge device to continuously collect data on the organic heat carrier, uploading the acquired operation data of the organic heat carrier to the target cloud platform, and synchronously transmitting the inherent characteristics of the organic heat carrier in the target organic heat carrier cycle; The target cloud platform performs matching calls on the fault mode based on the inherent features, obtains a set of fault modes, and performs training of a comprehensive fault diagnosis model based on historical records in combination with the set of fault modes; The target cloud platform performs fault identification on the organic heat carrier operation data collected in real time based on the trained comprehensive fault diagnosis model, and performs an early warning response according to the fault identification result.

2. The cloud platform driven organic heat carrier remote monitoring and diagnosis method according to claim 1, characterized in that: Locate the key nodes of the target organic heat carrier cycle, including: Interactive target organic heat carrier cycle to obtain cycle structure information including cycle topology, design data and operating parameters; Based on the cycle structure information, identifying locations where pressure changes significantly, heat exchanges frequently, and key control valves exist; The parts of the identification results are classified and analyzed to determine the nodes that meet the monitoring representativeness, status indication and accessibility requirements to form a set of key nodes.

3. The cloud platform driven organic heat carrier remote monitoring and diagnosis method according to claim 1, characterized in that: Combined with the current conditions of the target organic heat carrier cycle, a multi-dimensional sensor network is deployed, including: Based on the current conditions of the target organic heat carrier cycle, extract the current sensor table; Based on the complement of the preset sensor form and the current sensor form as the target layout form, the current status sensor of the target organic heat carrier cycle is updated and deployed to construct the multidimensional sensor network, wherein the preset sensor form includes at least a temperature sensor, a pressure sensor, a flow sensor, a dielectric constant sensor and a viscosity sensor.

4. The cloud platform driven organic heat carrier remote monitoring and diagnosis method according to claim 1, characterized in that: A rule-based expert system is integrated in the target cloud platform, wherein the expert system includes a plurality of expert subsystems, each expert subsystem corresponds to a type of organic heat carrier, and has a built-in operating standard parameter space and a fault mode library corresponding to the organic heat carrier.

5. The cloud platform driven organic heat carrier remote monitoring and diagnosis method according to claim 4, characterized in that: The target cloud platform performs matching calls for fault modes based on the inherent features, obtains a set of fault modes, and performs training of a comprehensive fault diagnosis model based on historical records in combination with the set of fault modes, including: According to the inherent characteristics of the organic heat carrier in the target scenario, the corresponding expert subsystem is matched and the corresponding fault mode set is called; Constructing a training data set based on the failure mode set and the historical records stored in the target cloud platform, and establishing a mapping relationship between the training data set and the failure mode set, to obtain M training data subsets corresponding to M failure modes; Through the M subsets of the training data, M groups of specific fault diagnosis models based on machine learning are constructed and trained respectively, and the M groups of specific fault diagnosis models are integrated to construct the comprehensive fault diagnosis model.

6. The cloud platform driven organic heat carrier remote monitoring and diagnosis method according to claim 5, characterized in that: The specific fault diagnosis models of the M groups are integrated to construct the comprehensive fault diagnosis model, including: Based on M groups of specific fault diagnosis models, grouping and integration are performed to form M specific diagnosis clusters; Establishing a fault mode classifier for the fault mode set, wherein the fault mode classifier is used to identify the fault type corresponding to the real-time data; The M output ends of the fault mode classifier are connected to the input ends of the M specific diagnosis clusters, and the output ends of the M specific diagnosis clusters are connected to the integrated output layer to obtain the comprehensive fault diagnosis model.

7. The cloud platform driven organic heat carrier remote monitoring and diagnosis method according to claim 6, characterized in that: Fusion of the M groups of specific fault diagnosis models to construct the comprehensive fault diagnosis model also includes: Randomly selecting N specific fault diagnosis models of a preset proportion from M groups of specific fault diagnosis models, wherein N<M; Randomly introduce sample data of heterogeneous fault modes, perform transfer learning on the extracted N specific fault diagnosis models respectively, and separate the models after transfer learning into a verification model group; The verification model group and M groups of specific fault diagnosis models are integrated based on an integrated learning method to obtain the comprehensive fault diagnosis model.

8. The cloud platform driven organic heat carrier remote monitoring and diagnosis method according to claim 1, characterized in that: The early warning response is carried out according to the fault identification results, including: The fault identification result is subjected to fault classification and level determination, and a determination is made in combination with preset rules as to whether an automated control response is triggered, wherein the automated control response at least includes load reduction operation, active shutdown processing, adjustment of operating parameters, and activation of redundant backup.

9. The cloud platform driven organic heat carrier remote monitoring and diagnosis method according to claim 1, characterized in that: According to the fault identification results, early warning response is carried out, and then: Build a visual user interface to display the real-time operating status, historical data trends and fault identification results of the target organic heat carrier cycle; The corresponding position of each fault mode in the target organic heat carrier cycle is graphically presented in the visual user interface, wherein the color of the display prompt is driven by the confidence level of the fault identification result.

10. The cloud platform driven organic heat carrier remote monitoring and diagnosis system is characterized by: A remote monitoring and diagnosis method for organic heat carriers driven by a cloud platform according to any one of claims 1 to 9, comprising: A key node positioning and sensor network deployment module is used to locate the key nodes of the target organic heat carrier cycle and deploy a multi-dimensional sensor network in combination with the current conditions of the target organic heat carrier cycle, wherein the multi-dimensional sensor network is connected to the edge device; An organic heat carrier operation data collection and upload module, used to activate the edge device to continuously collect the organic heat carrier, upload the acquired organic heat carrier operation data to the target cloud platform, and synchronously transmit the inherent characteristics of the organic heat carrier in the target organic heat carrier cycle; A pattern matching and model training module is used for the target cloud platform to match and call the fault pattern based on the inherent characteristics, obtain a set of fault patterns, and perform comprehensive fault diagnosis model training based on historical records in combination with the set of fault patterns; The fault identification and early warning response module is used for the target cloud platform to perform fault identification on the organic heat carrier operation data collected in real time based on the trained comprehensive fault diagnosis model, and to perform early warning response according to the fault identification result.

Images