Self-healing control method and system for chemical device based on self-adaptive fault prediction

Abstract

The invention discloses a chemical device self-healing control method and system based on self-adaptive fault prediction, and relates to the technical field of chemical intelligent control, and the method comprises the steps: connecting a sensor network, and collecting the operation parameters of a chemical device in real time; inputting a long-short-term memory generative adversarial network model to carry out fault identification and state prediction; when fault information is detected, a fault identification result and state prediction data are transmitted to a decision support module to carry out device operation process parameter compensation analysis, self-healing control parameters are obtained, and real-time interactive self-healing control is carried out. The technical problems of high unplanned shutdown rate and insufficient safety risk control caused by failure prediction lag of the chemical device in the prior art are solved, advanced diagnosis and process parameter dynamic compensation of the abnormal state of the chemical device are realized, the production safety and the equipment reliability are effectively improved, and the production cost is reduced. And the non-planned shutdown risk is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of chemical intelligent control, and particularly relates to a self-healing control method and system for chemical plants based on adaptive fault prediction. Background Art

[0002] In the field of chemical production, the stable operation of chemical plants is of vital importance. However, there are many defects in the existing fault monitoring and handling technologies for chemical plants. On the one hand, traditional fault prediction means rely on manual experience or simple threshold judgment, and cannot accurately capture the complex change trends of the operating parameters of the plant, resulting in lagging fault prediction. It is often discovered only when a fault has occurred or is about to occur, which makes the unplanned shutdown rate of chemical plants remain high. This not only causes production interruption and brings huge economic losses, but may also trigger a series of chain reactions, threatening production safety. On the other hand, in the face of a fault, there is a lack of an effective dynamic compensation mechanism for process parameters, and it is difficult to reasonably adjust the plant in the first time when a fault occurs, resulting in insufficient safety risk control.

[0003] The prior art has the technical problems of lagging fault prediction for chemical plants, resulting in a high unplanned shutdown rate and insufficient safety risk control. Summary of the Invention

[0004] The present application provides a self-healing control method and system for chemical plants based on adaptive fault prediction, which is used to solve the technical problems of lagging fault prediction for chemical plants in the prior art, resulting in a high unplanned shutdown rate and insufficient safety risk control.

[0005] In view of the above problems, the present application provides a self-healing control method and system for chemical plants based on adaptive fault prediction.

[0006] In the first aspect of the present application, a self-healing control method for chemical plants based on adaptive fault prediction is provided. The method includes:

[0007] Connect a sensor network to collect the operating parameters of the chemical plant in real time, including temperature, pressure, flow rate, liquid level, and vibration; after cleaning and normalizing the collected operating parameters in real time, input them into a long short-term memory generative adversarial network model for fault identification and state prediction; when fault information is detected, transmit the fault identification result and state prediction data to a decision support module for compensating analysis of the operating process parameters of the plant to obtain self-healing control parameters, and perform real-time interactive self-healing control on the chemical plant according to the self-healing control parameters.

[0008] In the second aspect of the present application, a self-healing control system for chemical plants based on adaptive fault prediction is provided. The system includes:

[0009] The operating parameter acquisition module is used to connect to the sensor network and collect the operating parameters of the chemical plant in real time, including temperature, pressure, flow rate, liquid level, and vibration. The fault identification module is used to clean and normalize the real-time collected operating parameters, and then input them into the long short-term memory generative adversarial network model for fault identification and status prediction. The real-time interactive self-healing control module is used to, when detecting fault information, transmit the fault identification results and status prediction data to the decision support module for compensating analysis of the operating process parameters of the plant, obtain the self-healing control parameters, and perform real-time interactive self-healing control on the chemical plant according to the self-healing control parameters.

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

[0011] Connect to the sensor network and collect the operating parameters of the chemical plant in real time; clean and normalize the real-time collected operating parameters, and then input them into the long short-term memory generative adversarial network model for fault identification and status prediction; when detecting fault information, transmit the fault identification results and status prediction data to the decision support module for compensating analysis of the operating process parameters of the plant, obtain the self-healing control parameters, and perform real-time interactive self-healing control on the chemical plant according to the self-healing control parameters. It achieves the technical effects of realizing the advanced diagnosis of abnormal states of chemical plants and the dynamic compensation of process parameters, effectively improving production safety and equipment reliability, and reducing the risk of unplanned shutdowns. Description of the Drawings

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0013] Figure 1 Schematic flow chart of the self-healing control method for chemical plants based on adaptive fault prediction provided in the embodiments of this application;

[0014] Figure 2 Schematic structural diagram of the self-healing control system for chemical plants based on adaptive fault prediction provided in the embodiments of this application.

[0015] Description of the reference numerals: Operating parameter acquisition module 10, Fault identification module 20, Real-time interactive self-healing control module 30. Detailed Embodiments

[0016] This application provides a self-healing control method and system for chemical plants based on adaptive fault prediction, which is used to solve the technical problems in the prior art of delayed fault prediction of chemical plants, resulting in a high unplanned downtime rate and insufficient safety risk control.

[0017] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative work are within the scope of protection of this application.

[0018] Example 1, as Figure 1 As shown, the present application provides a self-healing control method for a chemical plant based on adaptive fault prediction, the method comprising:

[0019] Step S100: Connecting to a sensor network to collect operating parameters of a chemical plant in real time, including temperature, pressure, flow, liquid level, and vibration.

[0020] Specifically, by connecting a sensor network to chemical plants, a comprehensive, real-time data collection system is established. Temperature sensors precisely monitor heat changes within the plant, pressure sensors provide feedback on pressure within the system, flow sensors record material flow velocity, liquid level sensors monitor liquid height in real time, and vibration sensors capture equipment vibrations. These sensors work together to continuously collect key parameters during chemical plant operation, providing raw data support for subsequent analysis.

[0021] Step S200: After cleaning and normalizing the real-time collected operating parameters, they are input into the long short-term memory generative adversarial network model for fault identification and status prediction.

[0022] Specifically, the original operating parameters collected from the sensors may contain interference information such as noise and outliers, so a cleaning operation must be performed first to remove these invalid data to ensure the accuracy and reliability of the data. Next, the data of different magnitudes and dimensions are unified into a specific interval through normalization processing, making the data easier to process by the model. After that, the processed data is input into the long short-term memory generative adversarial network model. This model relies on the powerful ability of the long short-term memory network (LSTM) to capture complex patterns in time series data, combined with the adversarial training mechanism of the generative adversarial network (GAN) discriminator and generator, to accurately mine the nonlinear features in the data, thereby realizing the identification of chemical plant failures and the prediction of future operating status, and discovering potential risks in advance.

[0023] Step S300: When a fault message is detected, the fault identification result and status prediction data are transmitted to the decision support module for compensating and analyzing the process parameters of the device operation, obtaining self-healing control parameters, and performing real-time interactive self-healing control on the chemical device according to the self-healing control parameters.

[0024] Specifically, after a fault message is detected, the fault identification result and status prediction data are transmitted to the decision support module. This module first accurately locates information such as the location and type of the fault based on these data. Then, through process parameter tracing, it finds the process parameters related to the fault and obtains preliminary compensation process parameters. To more precisely address the fault, the module also identifies abnormal features based on the fault message, searches in the predefined compensation strategy library using the abnormal features and compensation process parameters as indexes, compensates the abnormal amount of the compensation process parameters, and finally obtains self-healing control parameters that can enable the chemical device to resume normal operation. These parameters cover key information such as process parameters, compensation operation parameters, and compensation control quantities. Finally, using digital twin technology, a connection is established with the physical equipment of the chemical device, the self-healing control parameters are fed back to the digital twin module, and through the real-time interaction of this module with the physical equipment, control instructions are sent to the physical equipment to achieve real-time adjustment of the process parameters of the chemical device, complete real-time interactive self-healing control, and ensure the continuous and stable operation of the chemical device.

[0025] In a possible implementation manner, step S200 further includes:

[0026] Step S210: Construct a long short-term memory generative adversarial network.

[0027] Step S220: Collect the operation record data of the chemical device and establish a training data set and a test data set.

[0028] Step S230: Use the training data set and the test data set to train and converge the long short-term memory generative adversarial network to obtain a long short-term memory generative adversarial network model.

[0029] Step S240: Based on online learning technology, automatically update the model parameters of the long short-term memory generative adversarial network model according to the latest collected operation data.

[0030] Specifically, a Long Short-Term Memory Generative Adversarial Network (LSTM-GAN) is constructed. The Long Short-Term Memory Network (LSTM) is good at processing time series data and can capture the complex patterns of the operating parameters of chemical plants changing over time. It effectively solves the problem of vanishing gradients or exploding gradients in traditional recurrent neural networks through its unique gating mechanism, ensuring the memory and learning of long-term dependence information. The Generative Adversarial Network (GAN) consists of a generator and a discriminator, and the two are trained against each other. In this network, LSTM and GAN are combined. The generator generates samples similar to the real operating data by virtue of the capabilities of LSTM, while the discriminator tries to distinguish between the generated samples and the real samples, and through continuous gaming, improves the learning and expression capabilities of the entire network for data features.

[0031] To provide sufficient and effective data for network training, the operating record data of chemical plants is collected. These data cover parameters such as temperature, pressure, flow rate, liquid level, and vibration of the plant under different operating conditions. After the data is collected, it is divided into a training data set and a test data set according to a certain ratio. The training data set is used to let the Long Short-Term Memory Generative Adversarial Network learn the patterns and rules in the data, while the test data set is used to evaluate the performance of the trained network and test its generalization ability for unknown data.

[0032] The training data set is input into the constructed Long Short-Term Memory Generative Adversarial Network, and the generator and the discriminator start adversarial training. The generator continuously adjusts its own parameters and tries to generate more realistic samples to deceive the discriminator; the discriminator is also constantly learning to improve its ability to distinguish between real samples and generated samples. Through repeated iterative training, the loss function of the network is continuously reduced until it reaches a convergence state. At this time, the performance of the network tends to be stable, and the Long Short-Term Memory Generative Adversarial Network model is obtained. Then, the test data set is used to evaluate the model to ensure that it has good prediction accuracy and generalization ability.

[0033] To enable the model to adapt to the continuously changing operating state of chemical plants, an online learning technology is introduced. As the chemical plant continues to operate, the latest operating data is continuously collected. These new data are automatically incorporated into the model, and the parameters of the Long Short-Term Memory Generative Adversarial Network model are updated according to the online learning algorithm. In this way, the model can timely learn the new features and new rules in the process of plant operation and always maintain the accuracy and effectiveness of chemical plant fault identification and state prediction.

[0034] In a possible implementation manner, step S210 further includes:

[0035] Step S211: An input layer that receives time series data.

[0036] Step S212: LSTM layer: As part of the generator, it is used to capture complex patterns in time series.

[0037] Step S213: GAN layer: It includes two parts, a discriminator and a generator, and conducts adversarial training to capture non-linear features.

[0038] Step S214: Output layer: Outputs the fault identification result and the prediction result.

[0039] Specifically, the input layer is connected to numerous sensors distributed throughout the chemical plant. It continuously receives time series data collected from sensors such as temperature, pressure, flow rate, liquid level, and vibration. These data carry real-time status information during the operation of the chemical plant. The input layer preliminarily organizes and distributes these data to prepare for subsequent processing.

[0040] As an important part of the generator, the LSTM layer has a unique gating mechanism that can deeply explore complex patterns in time series data. For example, the operation of a chemical plant often has a certain periodicity and coherence. The LSTM layer can remember the change trends of operation parameters over a past period of time and distinguish which information is long-term dependent and which is short-term fluctuating. In this way, it can accurately capture the time-dependent relationships between different parameters, laying a foundation for more accurate subsequent analysis.

[0041] The discriminator and the generator in the GAN layer work together to further optimize the model performance through adversarial training. The generator uses the information captured by the LSTM layer and tries to generate simulated data similar to the real data. The discriminator, on the other hand, endeavors to distinguish the differences between these simulated data and the real data transmitted from the input layer. During this process, the generator and the discriminator continuously play a game. In order to deceive the discriminator, the generator will continuously adjust the way of generating data to make it closer to the real data; the discriminator will also continuously improve its discrimination ability to accurately distinguish. Through this adversarial training, the model can better capture the non-linear features in the data, thereby gaining a deeper understanding of the operating state of the chemical plant.

[0042] The output layer gives corresponding feedback based on the prediction result after the model is processed by the previous layers. It will output the fault identification result, clearly indicating whether there is a fault in the chemical plant currently and the type of the fault; at the same time, it will also output the prediction result of the operating state, such as the change trend of parameters in the future period of time, etc. These output results provide an important basis for the maintenance and management of the chemical plant, helping the staff to take measures in a timely manner to ensure the stable operation of the chemical plant.

[0043] In a possible implementation manner, step S240 further includes:

[0044] Step S241: Align the real-time collected operation data with the historical data in time series.

[0045] Step S242: Use a sliding window mechanism to screen the data in the time-aligned sequence to obtain newly added changed data.

[0046] Step S243: Use the newly added changed data to update and train the long short-term memory generative adversarial network model and adjust the model parameters.

[0047] Specifically, since the operation of a chemical plant is a continuous dynamic process, there are differences between the newly collected real-time data and the historical data in the time dimension. To effectively utilize the information contained in the historical data, it is necessary to align the real-time collected operation data with the historical data in time series, accurately match the data points at different times according to the timestamps of the data, and ensure that the data corresponds one by one in chronological order. A unified time coordinate is established for the data, enabling data from different periods to be compared and analyzed within the same framework, laying a foundation for subsequent exploration of data change trends and characteristics.

[0048] Use a sliding window mechanism to screen the data in the time-aligned sequence. It has a fixed size. As time goes by, the window slides point by point on the data sequence, and each slide will include a certain number of data points. By setting reasonable window sizes and sliding steps, the newly added changed data within the window is screened out. These newly added changed data represent the latest state change information during the operation of the chemical plant, which is different from the data used in previous model training and can reflect the real-time dynamic changes of the plant operation.

[0049] Use the obtained newly added changed data for the update training of the long short-term memory generative adversarial network model. Based on the previous training of the model, with the newly added changed data as the input, the outputs and errors of each layer of the model are recalculated. Through the backpropagation algorithm, the error is propagated backward from the output layer to all levels of the network, thereby adjusting the model parameters. For example, adjust the weights and gating parameters of the LSTM layer to enable the model to better capture the time-dependent relationships in the newly added data; optimize the parameters of the generator and discriminator in the GAN layer to enhance the model's learning ability for non-linear features. By continuously updating and training with the newly added changed data, the model can timely adapt to the changes in the operation state of the chemical plant, maintain the accurate identification of faults and the precise prediction ability of the operation state, and provide a strong guarantee for the stable operation of the chemical plant.

[0050] In a possible implementation manner, step S300 further includes:

[0051] Step S310: Locate the fault information according to the fault identification result and the state prediction data.

[0052] Step S320: Trace the process parameters based on the fault information to obtain compensated process parameters.

[0053] Step S330: Identify abnormal characteristics according to the fault information, and compensate for the abnormal amount of the compensated process parameters to obtain the self-healing control parameters.

[0054] Specifically, the fault information is accurately located by using the fault identification results and state prediction data. The fault identification results will clarify the type of the fault, such as whether it is a device hardware fault, an abnormal process parameter, or a fault caused by a material problem. The state prediction data contains information such as the development trend of the fault and the possible affected range. First, the fault identification results are analyzed to judge the general direction of the fault, such as determining whether it is a problem in terms of temperature, pressure, flow rate, etc. If the fault identification results show that the temperature is abnormal, then combined with the temperature change trend and the involved area in the state prediction data, the fault range is further narrowed. For example, if the state prediction data indicates that the temperature in a certain area continues to rise and is about to exceed the safety threshold, then the fault can be locked on the relevant equipment or process flow in this area. The structural layout and operation logic of the chemical plant will also be referred to, the correlation between each part will be analyzed, and comprehensive information from multiple aspects will be used to accurately locate the specific location where the fault occurs, such as a specific reactor, a certain section of pipeline, a specific sensor, etc., and at the same time evaluate the severity of the fault to judge whether it is a minor abnormality or a serious fault that may cause the plant to shut down, providing a key basis for subsequent processing decisions.

[0055] Based on the located fault information, the process parameters are traced. The operation of the chemical plant depends on process parameters, and these parameters are interrelated and interact with each other. Along the path where the fault occurs, the process parameters related to this fault are searched. For example, if the located fault is that the temperature in the reactor is too high, then the parameters related to the temperature control of the reactor will be traced, such as the feed flow rate, the cooling medium flow rate, the reaction heat, etc. By comparing and analyzing the historical data of these parameters with the data during normal operation, the parameters that need to be adjusted under the current fault condition are determined, so as to obtain the compensated process parameters, initially providing a parameter adjustment direction for solving the fault.

[0056] Identify abnormal characteristics based on fault information. Different faults will present different abnormal characteristics. For example, a temperature fault may be manifested as a sharp rise or fall in temperature, and a pressure fault may be a sudden change or continuous abnormal fluctuation in pressure. Utilize these abnormal characteristics and combine them with compensation process parameters to perform abnormal quantity compensation. Pre-define a compensation strategy library that stores compensation strategies corresponding to various fault types. Use the abnormal characteristics and compensation process parameters as indexes to search in the compensation strategy library, select an appropriate compensation method according to the specific situation of the fault, determine the compensation amplitude and specific operations, and finally obtain the self-healing control parameters that can enable the chemical plant to resume normal operation. These parameters cover precise process parameters, compensation operation parameters, and compensation control quantities, providing key support for the self-healing control of the chemical plant.

[0057] In a possible implementation manner, step S330 further includes:

[0058] Step S331: Pre-define a compensation strategy library that stores compensation strategies corresponding to various fault types.

[0059] Step S332: Use the abnormal characteristics and compensation process parameters as compensation indexes to search in the compensation strategy library to obtain the self-healing control parameters, including process parameters, compensation operation parameters, and compensation control quantities.

[0060] Specifically, before the chemical plant is put into operation, based on an in-depth understanding of the chemical plant and the analysis of a large amount of historical data, carry out the pre-definition work of the compensation strategy library. Comprehensively sort out various types of faults that may occur in the chemical plant, such as typical fault types like abnormal temperature in the reactor, pipeline pressure imbalance, and material flow blockage. For each fault, formulate detailed compensation strategies based on chemical process principles, equipment operation characteristics, and past actual operation experience. For the fault of too high temperature in the reactor, the compensation strategy may be to reduce the proportion of reactive substances in the feed and at the same time increase the flow rate of the cooling medium; for the fault of too high pipeline pressure, it may be to open a specific pressure relief valve and adjust the output power of the relevant pump to change the material delivery volume. After these strategies are systematically sorted out, they are stored in the compensation strategy library, and each fault type corresponds to a corresponding compensation strategy, forming a complete and orderly set of strategies, providing a reliable reserve of response solutions for the chemical plant when faults occur during operation.

[0061] After the fault information is located and the abnormal features and compensation process parameters are obtained, the abnormal features and compensation process parameters are combined into a unique compensation index, and a detailed search is carried out in the predefined compensation strategy library with the help of this index. The abnormal features precisely describe the specific manifestations of the fault, such as a sharp rise in temperature, continuous fluctuation of pressure, etc.; the compensation process parameters clarify the process direction that needs to be adjusted for preliminary judgment. Based on this information, the fault type and its corresponding compensation strategy that match it are screened out from the strategy library. Among the numerous possible strategies, further analysis is carried out to determine the strategy that best suits the current fault situation. Once the appropriate strategy is found, the self-healing control parameters are extracted from it. These parameters include precise process parameters, such as specific temperature setting values, pressure stability values, etc.; compensation operation parameters, clearly indicating the direction of operation, whether to increase the feed rate, decrease the valve opening, or increase the motor speed, etc.; and compensation control amounts, specifying in detail the amplitude of parameter adjustment to ensure that the adjustment can effectively solve the fault without causing excessive impact on the normal operation of the chemical plant, thereby realizing the self-healing control of the chemical plant.

[0062] In a possible implementation manner, step S300 further includes:

[0063] Step S340: Establish a physical device connection with the chemical plant through digital twin, and the digital twin includes the mapping relationship between the virtual model and the physical device.

[0064] Step S350: Feed back the self-healing control parameters to the digital twin module, and issue control instructions through real-time interaction between the digital twin module and the physical device. The control instructions are used to adjust the process parameters of the physical device according to the self-healing control parameters.

[0065] Specifically, by establishing a connection with physical devices through digital twins, digital twin technology constructs a virtual model corresponding to the physical devices of a chemical plant. This virtual model details and simulates various aspects of the physical devices, such as their structures, operating principles, and technological processes, achieving an accurate mapping relationship between the virtual model and the physical devices. With the help of a sensor network, various sensors distributed on the physical devices, such as temperature sensors, pressure sensors, and flow sensors, will collect the operating data of the devices in real time, including parameters such as temperature, pressure, flow rate, and liquid level, and transmit this data to the virtual model through a wired or wireless communication network. The virtual model dynamically adjusts its own state based on this real-time data to ensure consistency with the operating state of the physical devices. At the same time, the virtual model can also perform simulation analysis and prediction, deduce possible future situations of the physical devices, and discover potential problems in advance. Moreover, this mapping relationship is two-way. Not only can the data of the physical devices be transmitted to the virtual model, but the analysis results and control instructions of the virtual model can also be fed back to the physical devices, enabling real-time monitoring and precise control of the physical devices, providing strong support for the stable operation and self-healing of chemical plants.

[0066] With the help of digital twin technology, precise regulation of physical devices by self-healing control parameters is achieved. After the self-healing control parameters are generated, these parameters are promptly fed back to the digital twin module. The digital twin module connects the virtual model and the physical devices and, based on the pre-established mapping relationship between the virtual model and the physical devices, parses and converts the self-healing control parameters into control instructions that can be recognized and executed by the physical devices. For example, if the self-healing control parameter requires increasing the temperature set value of a certain reactor, the digital twin module will simulate this parameter adjustment process in the virtual model, confirm the feasibility of the adjustment and the possible impacts, and then send the instruction accurately to the temperature control system of the reactor through the real-time interaction channel with the physical device. After receiving the instruction, this control system will automatically adjust the power of the heating equipment, change the flow rate or temperature of the feed, etc., thereby adjusting the process parameters of the physical device according to the self-healing control parameters. Through the efficient information processing and real-time interaction capabilities of the digital twin module, the entire process ensures the accurate transmission of control instructions and the precise response of physical devices, enabling the chemical plant to quickly return to normal operation and effectively improving the safety, stability, and efficiency of chemical production.

[0067] Embodiment 2, based on the same inventive concept as the self-healing control method for chemical plants based on adaptive fault prediction in the foregoing embodiment, as Figure 2 shown, this application provides a self-healing control system for chemical plants based on adaptive fault prediction. The system in the embodiments of this application and the method embodiments are based on the same inventive concept. Among them, the system includes:

[0068] The operating parameter acquisition module 10 is used to connect to the sensor network and collect the operating parameters of the chemical plant in real time, including temperature, pressure, flow rate, liquid level, and vibration.

[0069] The fault identification module 20 is used to clean and normalize the real-time collected operating parameters, and then input them into the long short-term memory generative adversarial network model for fault identification and state prediction.

[0070] The real-time interactive self-healing control module 30 is used to, when detecting fault information, transmit the fault identification result and state prediction data to the decision support module for compensating analysis of the operating process parameters of the device, obtain the self-healing control parameters, and perform real-time interactive self-healing control on the chemical plant according to the self-healing control parameters.

[0071] Furthermore, the system is also used to implement the following functions:

[0072] Construct a long short-term memory generative adversarial network; collect the operating record data of the chemical plant, establish a training data set and a test data set; use the training data set and the test data set to train and converge the long short-term memory generative adversarial network to obtain a long short-term memory generative adversarial network model; based on online learning technology, automatically update the model parameters of the long short-term memory generative adversarial network model according to the latest collected operating data.

[0073] Furthermore, the system is also used to implement the following functions:

[0074] The input layer receives time series data; the LSTM layer: as part of the generator, is used to capture complex patterns in the time series; the GAN layer: includes two parts, the discriminator and the generator, and performs adversarial training to capture non-linear features; the output layer: outputs the fault identification result and the prediction result.

[0075] Furthermore, the system is also used to implement the following functions:

[0076] Align the real-time collected operating data with the historical data in time series; use the sliding window mechanism to screen the time-aligned sequence to obtain the newly added changed data; use the newly added changed data to update and train the long short-term memory generative adversarial network model and adjust the model parameters.

[0077] Furthermore, the system is also used to implement the following functions:

[0078] Locate the fault information according to the fault identification result and state prediction data; trace the process parameters according to the fault information to obtain the compensated process parameters; identify the abnormal features according to the fault information and perform abnormal amount compensation on the compensated process parameters to obtain the self-healing control parameters.

[0079] Further, the system is also used to implement the following functions:

[0080] Predefine a compensation strategy library that stores compensation strategies corresponding to multiple fault types; use the abnormal features and compensation process parameters as compensation indexes to search in the compensation strategy library to obtain the self-healing control parameters, including process parameters, compensation operation parameters, and compensation control quantities.

[0081] Further, the system is also used to implement the following functions:

[0082] Establish a physical device connection with the chemical plant through digital twin, where digital twin includes the mapping relationship between the virtual model and the physical device; feedback the self-healing control parameters to the digital twin module, and issue control instructions through real-time interaction between the digital twin module and the physical device. The control instructions are used to adjust the process parameters of the physical device according to the self-healing control parameters.

[0083] It should be noted that the above order of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above describes specific embodiments of this specification. Additionally, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0084] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included within the protection scope of the present application.

[0085] This specification and the drawings are only exemplary descriptions of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A self-healing control method for chemical plants based on adaptive fault prediction, characterized in that, Including: Connect to the sensor network to collect the operating parameters of the chemical plant in real time, including temperature, pressure, flow rate, liquid level, and vibration; After cleaning and normalizing the real-time collected operating parameters, input them into the long short-term memory generative adversarial network model for fault identification and status prediction; When fault information is detected, transfer the fault identification results and status prediction data to the decision support module for compensating analysis of the device operating process parameters to obtain self-healing control parameters, and perform real-time interactive self-healing control on the chemical plant according to the self-healing control parameters.

2. The self-healing control method for chemical plants based on adaptive fault prediction according to claim 1, characterized in that Before inputting into the long short-term memory generative adversarial network model for fault identification and status prediction, it includes: Construct a long short-term memory generative adversarial network; Collect the operating record data of the chemical plant and establish a training data set and a test data set; Use the training data set and the test data set to train and converge the long short-term memory generative adversarial network to obtain a long short-term memory generative adversarial network model; Based on the online learning technology, automatically update the model parameters of the long short-term memory generative adversarial network model according to the latest collected operating data.

3. The self-healing control method for chemical plants based on adaptive fault prediction according to claim 2, wherein, The long short-term memory generative adversarial network includes: An input layer that receives time series data; An LSTM layer: As part of the generator, it is used to capture complex patterns in the time series; A GAN layer: It includes two parts, a discriminator and a generator, and performs adversarial training to capture non-linear features; An output layer: Outputs the fault identification results and prediction results.

4. The self-healing control method for chemical plants based on adaptive fault prediction according to claim 2, characterized in that Based on the online learning technology, automatically update the model parameters of the long short-term memory generative adversarial network model according to the latest collected operating data, including: Align the real-time collected operating data with the historical data in time series; Adopt a sliding window mechanism to screen the time-aligned sequence data to obtain newly added changed data; Use the newly added changed data to update and train the long short-term memory generative adversarial network model and adjust the model parameters.

5. The self-healing control method for chemical plants based on adaptive fault prediction according to claim 1, characterized in that Transfer the fault identification results and status prediction data to the decision support module for compensating analysis of the device operating process parameters to obtain self-healing control parameters, including: Locate the fault information according to the fault identification results and status prediction data; Trace the process parameters according to the fault information to obtain the compensation process parameters; Identify the abnormal features according to the fault information and perform abnormal amount compensation on the compensation process parameters to obtain the self-healing control parameters.

6. The self-healing control method for chemical plants based on adaptive fault prediction according to claim 5, wherein, Identify the abnormal features according to the fault information and perform abnormal amount compensation on the compensation process parameters, including: Pre-define a compensation strategy library, and the compensation strategy library stores compensation strategies corresponding to various fault types; Use the abnormal features and compensation process parameters as compensation indexes to search in the compensation strategy library to obtain the self-healing control parameters, including process parameters, compensation operation parameters, and compensation control amounts.

7. The self-healing control method for chemical plants based on adaptive fault prediction according to claim 6, characterized in that After obtaining the self-healing control parameters, it includes: Establish a physical device connection with the chemical plant through digital twin, and the digital twin includes the mapping relationship between the virtual model and the physical device; Feed the self-healing control parameters back into the digital twin module, and issue control instructions through real-time interaction between the digital twin module and the physical device. The control instructions are used to adjust the process parameters of the physical device according to the self-healing control parameters.

8. A self-healing control system for chemical plants based on adaptive fault prediction, characterized in that, The system is used to implement the self-healing control method for chemical plants based on adaptive fault prediction according to any one of claims 1-7. The system includes: An operating parameter acquisition module, which is used to connect to the sensor network and collect the operating parameters of the chemical plant in real time, including temperature, pressure, flow rate, liquid level, and vibration; A fault identification module, which is used to clean and normalize the real-time collected operating parameters, and then input them into a long short-term memory generative adversarial network model for fault identification and state prediction; A real-time interactive self-healing control module, which is used to transfer the fault identification results and state prediction data to the decision support module for analysis of the compensation of the device operation process parameters to obtain self-healing control parameters when fault information is detected, and perform real-time interactive self-healing control on the chemical plant according to the self-healing control parameters.

Images