High-pressure heater intelligent alarm system based on digital twinning and expert system and use method of high-pressure heater intelligent alarm system

Abstract

The invention relates to a high-pressure heater intelligent alarm system based on a digital twin and expert system. The high-pressure heater intelligent alarm system comprises a data acquisition layer, a data transmission layer, a data processing layer, a model establishment layer, a model verification layer, a model optimization layer and an application service layer, the invention discloses a use method of a high-pressure heater intelligent alarm system based on a digital twinborn and expert system. The use method comprises the following steps: S1, building a digital twinborn system of a thermodynamic system; s2, solving by adopting a bat algorithm; s3, obtaining deviation of different degrees between predicted output and actual output according to an output image when the fault occurs; s4, by calculating the Euclidean distance between the prediction output array and the actual output array, two parameter physical entity output and twin model output monitored by the service system are integrated into one index; and S5, adopting an alarm analysis form of a fault tree. The method has the advantages that the working condition of the unit is dynamically adjusted, and accurate alarming under different working conditions is supported.

Description

Technical Field

[0001] The present invention relates to the technical field of alarm systems, and in particular to an intelligent alarm system for a high-voltage heater based on a digital twin and an expert system and a method for using the system. Background Art

[0002] After the "dual carbon" goal was proposed, the demand for deep peak regulation of the power grid faced by thermal power units has continued to increase, and at the same time, the requirements for energy saving and consumption reduction, carbon emission reduction, etc. have become increasingly high. Faced with complex situations such as frequent fluctuations in unit operating conditions, increasingly severe environmental pressure, and hidden dangers in the long-term service of equipment, the centralized control operators have problems such as high pressure on monitoring work, late discovery of parameter anomalies, and untimely fault judgment.

[0003] The traditional monitoring mode can no longer adapt to the increasingly complex requirements of safety, environmental protection, economy, etc. Whether from the need to respond to changes in the external situation or from the development of centralized control operation itself, it is imperative to carry out research on smart monitoring systems.

[0004] Common heater failures

[0005] 1. Heater pipe leakage: Leakage is the most common factor leading to failure and shutdown of high-pressure or low-pressure heating systems, among which partition leakage is the most prominent. If the partition leakage is not handled in time, it will lead to a reduction in feed water volume and a drop in temperature, thereby reducing the high-pressure heating efficiency. Moreover, if the leakage is aggravated, most of the feed water will flow directly into the drain system, thus endangering the safety of the steam turbine.

[0006] 2. Part of the tube bundle of the heater is blocked: During the operation of some tube bundles, due to various reasons, the fluid in the tube bundle cannot pass smoothly, resulting in increased flow resistance and reduced flow in the tube bundle, which ultimately affects the normal operation of the equipment.

[0007] 3. Short circuit of the heater inlet and outlet water chambers: The pipe connections of the heater inlet and outlet water chambers are loose, and the baffle screws and bolts are loose, resulting in damage to the partition between the inlet and outlet water chambers, so that part of the incoming water flows directly into the heater water chamber outlet without being heated by the tube bundle, forming a short circuit, which is the so-called inlet and outlet water chamber short circuit. When the water chamber is short-circuited, the outlet water temperature of the high-pressure heater will drop sharply, which will significantly reduce the thermal efficiency of the unit. Summary of the invention

[0008] 1. Technical Problems Solved

[0009] The technical problem to be solved by the present invention is that the operating conditions of the unit in the prior art cannot be dynamically adjusted and cannot support accurate alarms under different operating conditions.

[0010] 2. Technical Solution

[0011] In order to solve the above technical problems, the present invention provides the following technical solutions: an intelligent alarm system for high-voltage heaters based on digital twins and expert systems, comprising a data acquisition layer, a data transmission layer, a data processing layer, a model building layer, a model verification layer, a model optimization layer and an application service layer;

[0012] The data collection layer collects the operation data of the thermal system through sensors and metering instruments;

[0013] The data transmission layer transmits the collected data to the background of the digital twin system;

[0014] The data processing layer pre-processes, cleans and stores the collected data;

[0015] The model building layer builds a virtual digital twin model of the actual physical entity by analyzing and modeling the operating data of the actual thermal system;

[0016] The model verification layer compares and verifies the model of the digital twin system with the actual thermal system, and verifies the accuracy of the model through error testing between the twin model and the actual output of the physical entity;

[0017] The model optimization layer optimizes the hyperparameters of the deep forest through an improved bat algorithm to optimize the digital twin system model;

[0018] The application service layer performs real-time monitoring and fault detection.

[0019] Furthermore, the data collection layer collects temperature, pressure and flow information through sensors and measuring instruments.

[0020] The method for using the intelligent alarm system for high-voltage heater based on digital twin and expert system includes the following steps:

[0021] S1. Build a digital twin system for the thermal system;

[0022] S2, using bat algorithm to solve;

[0023] S3. From the output image when the fault occurs, it can be seen that there are different degrees of deviation between the predicted output and the actual output;

[0024] S4. By calculating the Euclidean distance between the predicted output array and the actual output array, the two parameters of the service system monitoring, the physical entity output and the twin model output, are integrated into one indicator;

[0025] S5. The fault tree alarm analysis method is adopted, and its main function is to monitor the abnormal changes of some important parameters during the operation of the unit.

[0026] Furthermore, the steps in S1 specifically include:

[0027] S11. Build data collection layer: collect the actual operation data of thermal system through sensors and metering instruments, including temperature, pressure, flow and other information;

[0028] S12. Build the data transmission layer: transmit the collected data to the background of the digital twin system. The data transmission layer needs to ensure the reliability and security of data transmission. In the same local area network, use the socket module programming in Python to realize the data transmission between the twin model and the physical entity through wired mode;

[0029] S13. Build the data processing layer: pre-process, clean and store the collected data for subsequent analysis and modeling;

[0030] S14, build model building layer: by analyzing and modeling the actual thermal system operation data, build a virtual digital twin model of the actual physical entity, build a deep forest model of the steam turbine system through Python, and use the operation history data of the physical entity for training to ensure the applicability of the model;

[0031] S15. Build the model verification layer: Compare and verify the model of the digital twin system with the actual thermal system, and verify the accuracy of the model through error testing between the twin model and the actual output of the physical entity;

[0032] S16. Build the model optimization layer: To further improve the accuracy of the model, the hyperparameters of the deep forest are optimized through the improved bat algorithm to optimize the digital twin system model;

[0033] S17, application service layer: by comparing the output of the twin model with the output of the simulation machine, the similarity function method is introduced to integrate a parameter that can characterize the operating status as the analysis result of the digital twin system, and it is applied to the operation of the actual thermal system to achieve real-time monitoring and fault detection.

[0034] Furthermore, the S2 specifically includes the following steps:

[0035] S21, initialize the position and speed of the bat group, and set relevant parameters, including the bat group size, maximum number of iterations, jump step length, jump speed, minimum frequency, and maximum frequency.

[0036] S22. Calculate the fitness value of each bat.

[0037] S23. For each bat, calculate the new position and new speed, and update the bat position and speed.

[0038] S24. For each bat, the position is updated or not updated randomly according to a certain probability, that is, if ri>Ai, the position is not updated, Ai is the pulse emission frequency of bat i, ri is a random number between 0 and 1, otherwise the position is updated.

[0039] S25. Update the global optimal solution x best.

[0040] S26. Repeat steps S22-S25 until the stop condition is reached.

[0041] 3. Beneficial Effects

[0042] The advantages of the present invention compared with the prior art are:

[0043] 1. The operating parameter threshold is dynamically adjusted according to the unit operating conditions, supporting accurate alarms under different operating conditions;

[0044] 2. Discover abnormal conditions of the unit operation faster than the operators, leaving them enough time to handle the abnormal conditions;

[0045] 3. The model logic rules integrate multiple alarm algorithms such as over-limit warning, rate alarm, shaking alarm, residual alarm, etc., and support logical calculation with cache, change rate, and difference comparison, which is different from the fixed value alarm of DCS;

[0046] 4. Utilize advanced technologies such as big data and artificial intelligence to achieve the integration of mechanism rules + artificial intelligence (AI) algorithms + expert knowledge, and support comprehensive fault alarms based on multi-parameter joint judgment and mechanism model + AI model fusion judgment, which is different from the single parameter alarm of DCS;

[0047] 5. Automatically conduct preliminary analysis after a fault occurs, automatically eliminate meter faults and parameter reactions under special working conditions, push corresponding operation instructions, and assist operators in handling abnormalities;

[0048] 6. The fault diagnosis module based on the expert system expresses and solidifies the professional knowledge and rich experience, operating procedures, equipment design data, etc. of the operating experts into an expert fault knowledge base. The expert fault knowledge base has fault identification rules to ensure the accuracy of judgment of various fault types. It identifies and diagnoses equipment faults through reasoning and analysis methods, and gives fault severity and maintenance suggestions. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a structural diagram of the present invention;

[0050] Figure 2 This is a water outlet temperature curve diagram of an embodiment of the present invention;

[0051] Figure 3 A valve opening curve diagram of an embodiment of the present invention;

[0052] Figure 4 is a flow chart of an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] It should be noted that all directional indications such as up, down, left, right, front, back, etc. in the embodiments of the present invention are only used to explain the relative position relationship, movement status, etc. between the components in a certain specific posture as shown in the accompanying drawings. If the specific posture changes, the directional indication will also change accordingly.

[0055] In addition, in the present invention, descriptions such as "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0056] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the invention.

[0057] The present invention will now be further described with reference to the accompanying drawings.

[0058] Example

[0059] According to the five-dimensional structure of the digital twin system, a complete digital twin system includes five dimensions, namely physical entity, digital model, data, service system, and interactive connection. It is implemented in the digital twin system of the steam turbine system in order to provide effective support for the reliability and availability of the steam turbine system.

[0060] Through an in-depth analysis of the five-dimensional architecture of the digital twin system of the thermal system, it can be pointed out that the steps for building the digital twin system of the thermal system are as follows.

[0061] Build the data collection layer: collect the actual operation data of the thermal system through sensors and metering instruments, including temperature, pressure, flow and other information. This article reads the relevant parameters from the physical entity of the thermal power unit to ensure the accuracy and real-time nature of the data.

[0062] Build a data transmission layer: The collected data is transmitted to the background of the digital twin system. The data transmission layer needs to ensure the reliability and security of data transmission. In the same local area network, the socket module programming in Python is used to realize the data transmission between the twin model and the physical entity through wired means.

[0063] Build a data processing layer: pre-process, clean and store the collected data for subsequent analysis and modeling.

[0064] Build the model building layer: By analyzing and modeling the operating data of the actual thermal system, a virtual digital twin model of the actual physical entity is established. This paper uses Python to build a deep forest model of the steam turbine system and uses the operating history data of the physical entity for training to ensure the applicability of the model.

[0065] Build a model verification layer: Compare and verify the model of the digital twin system with the actual thermal system, and verify the accuracy of the model through error testing between the twin model and the actual output of the physical entity.

[0066] Build a model optimization layer: To further improve the accuracy of the model, the hyperparameters of the deep forest are optimized through the improved bat algorithm to optimize the digital twin system model.

[0067] Application service layer: By comparing the output of the twin model with the output of the simulation machine, the similarity function method is introduced to integrate a parameter that can characterize the operating status as the analysis result of the digital twin system, and it is applied to the operation of the actual thermal system to achieve real-time monitoring and fault detection.

[0068] Each decision tree in the deep forest is built based on a hierarchical structure and uses different feature extraction methods, which enables the deep forest algorithm to better capture the essential characteristics of the data; each decision tree in the random forest is balanced, that is, the number of samples in each leaf node is almost the same, while the decision trees in the deep forest are unbalanced, which can further improve the generalization ability of the model; a simple voting method is used in the random forest for classification, while the deep forest uses the average value among multiple decision trees for prediction, which can improve the stability and accuracy of the prediction.

[0069] The bat algorithm is used to solve the problem. The basic idea of ​​the bat algorithm is to use a group of bats as search agents in the search space. Each bat represents a search solution. By simulating the flight and hunting process of bats, the optimal solution is gradually searched. During the search process, the position and speed of each bat are constantly updated, and certain mechanisms are also used to guide the search process.

[0070] 1. Initialize the position and speed of the bat group, and set related parameters, including the bat group size, maximum number of iterations, jump step length, jump speed, minimum frequency, and maximum frequency.

[0071] 2. Calculate the fitness value of each bat.

[0072] 3. For each bat, calculate the new position and new speed, and update the bat position and speed,

[0073] 4. For each bat, the position is updated or not updated randomly according to a certain probability, that is, if ri>Ai, the position is not updated, Ai is the pulse emission frequency of bat i, ri is a random number between 0 and 1, otherwise the position is updated.

[0074] 5. Update the global optimal solution x best.

[0075] 6. Repeat steps 2-5 until the stopping condition is reached.

[0076] The core concept of the bat algorithm is the speed and position change of the bat. The fitness function of the bat algorithm can be determined according to the specific problem, usually the objective function that needs to be minimized. In the iterative process of the algorithm, after each position update, the best solution among all bats' current positions is selected as the global optimal solution xbest. The stopping condition of the algorithm can be reaching the maximum number of iterations or finding a good enough solution.

[0077] From the output image when the fault occurs, it can be seen that there are different degrees of deviation between the predicted output and the actual output. In order to measure the degree of deviation between the predicted output and the actual output in order to monitor system faults, the similarity function, the Euclidean distance, is selected to measure the degree of deviation. The Euclidean distance function is used to calculate the distance between vectors, and its calculation method is as follows.

[0078] Assume that there are two n-dimensional vectors x and y, and the elements of each vector can be expressed as x = [x1, x2, ..., xn] and y = [y1, y2, ..., yn], then the Euclidean distance function can be expressed as:

[0079]

[0080] By calculating the Euclidean distance between the predicted output array and the actual output array, the two parameters monitored by the service system, physical entity output and twin model output, are integrated into one indicator.

[0081] The experimental condition of this paper selects the steady-state process fault of the high-pressure booster. Under the steady-state load, the fault is set to 6% leakage of the #1 high-pressure booster piping system. The actual output value of the physical entity and the output value of the twin model change. By analyzing the model output, it can be seen that when the #1 high-pressure booster has a 6% leakage fault in the piping system, the drain temperature of the #1 high-pressure booster has a significant decrease in the physical entity. In comparison, the outlet water temperature of the three high-pressure boosters has a smaller decrease. The most obvious change is the drain valve opening. In the physical entity, the drain valve opening of the #1 high-pressure booster rises rapidly from the beginning of the fault until it reaches 100% opening.

[0082] The fault diagnosis module based on the expert system adopts the alarm analysis form of the fault tree. Its main function is to monitor the abnormal changes of some important parameters during the operation of the unit. When the parameters deviate from the normal value or some abnormal signs are found, these fault information will be prompted to attract the attention of relevant personnel. The operation fault diagnosis process requires the knowledge of the operation procedures and the experience of the operation experts. The module can analyze the alarm information online, find out the cause of the fault, and propose the best solution for the fault. According to the needs of the user, the system provides the reasoning process and expert explanation.

[0083] The fault diagnosis module based on the expert system will build a root fault knowledge base according to the rules defined by the system. When a key alarm item appears, it can start the automatic reasoning process according to the fault knowledge base structure, combine historical data, quickly locate the root cause of the alarm, determine the initial cause of the fault alarm event, and prompt the operator.

[0084] First, the alarm information is entered and modified through the software; then the knowledge base file (rule base) is defined, and alarm reasoning is initiated after an alarm occurs, and expert reasoning is performed through process data and intermediate variables or manual input; finally, the cause of the fault and operating instructions are obtained.

[0085] The fault diagnosis is explained by taking the leakage of No. 1 high-pressure heater (hereinafter referred to as high-pressure heater) as an example. The fault knowledge base of the leakage of No. 1 high-pressure heater is obtained based on the analysis of operation experts.

[0086] There are 5 fault diagnosis conditions for leakage of No. 1 high pressure heater. If all 5 conditions are met at the same time, it indicates that No. 1 high pressure heater is leaking.

[0087] Condition 1 is that the feedwater pump inlet flow is greater than the sum of the feedwater flow and the cooling water flow. The value in this condition is calculated by the SIS measuring point. When the difference between the calculated feedwater pump inlet flow and the feedwater flow and the cooling water flow is greater than the set value, the condition is triggered. The set value is determined based on historical query and operation expert analysis.

[0088] Condition 2 is that the temperature rise of the No. 1 high-pressure heater becomes smaller. The temperature rise of the high-pressure heater is generally related to the unit load, and the temperature rise of the high-pressure heater will change under different loads. This condition is triggered when the difference between the standard value and the actual value of the temperature rise of the No. 1 high-pressure heater is greater than the set value, and the set value is determined based on historical data or operation expert analysis. The calculation process of the expected value of the temperature rise of the No. 1 high-pressure heater: The SIS historical data of the unit load and the temperature rise of the No. 1 high-pressure heater during normal operation (data at the time of shutdown and failure are cleared) is trained using the LSTM network to obtain a model of the temperature rise of the No. 1 high-pressure heater and the unit load during normal operation. The real-time expected value of the feedwater pump current can be obtained by calling the model with real-time load.

[0089] Condition 3 is that the feedwater pump current is greater than the expected value. When the high-pressure heater leaks, the feedwater pump inlet flow rate will increase in order to maintain the same load of the unit. This condition is triggered when the difference between the expected value and the actual value of the total current of the feedwater pump is greater than the set value, which is determined based on historical data or operation expert analysis. Calculation process of the expected value of the feedwater pump current: The SIS historical data of the unit load and the total current of the feedwater pump (data at the time of shutdown and failure are cleared) is trained using the LSTM network to obtain a model of the feedwater pump current and the unit load. The real-time expected value of the feedwater pump current can be obtained by calling the model with real-time load.

[0090] Condition 4 is that the No. 1 high-pressure heater drain temperature is greater than the expected value. The high-pressure heater drain temperature is generally related to the unit load, and the high-pressure heater temperature rise will vary under different loads. This condition is triggered when the difference between the standard value and the actual value of the No. 1 high-pressure heater drain temperature is greater than the set value, which is determined based on historical data or operation expert analysis. The calculation process of the expected value of the No. 1 high-pressure heater drain temperature: The SIS historical data of the unit load and the total current of the feedwater pump (data at the time of shutdown and failure are cleared) is trained using the LSTM network to obtain a model of the feedwater pump current and the unit load. The real-time expected value of the feedwater pump current can be obtained by calling the model with real-time load.

[0091] Condition 5 is the increase in the liquid level of high pressure heater No. 1. The increase in the liquid level of high pressure heater No. 1 is identified through logic configuration.

[0092] With the continuous development of intelligent improvement of power generation process, more and more intelligent modules are integrated into power generation production. Intelligent diagnosis to identify the operating efficiency and equipment status of thermal power units has become an indispensable part of intelligent power generation. Intelligent diagnosis of thermal power units can realize online monitoring and fault diagnosis of process systems, equipment operating status and control systems, and can further provide fault handling operation guidance or automatically handle faults, realize self-recovery of equipment and functional faults, and improve monitoring efficiency and production safety.

[0093] The invention and its implementation methods are described, and such description is not restrictive. The drawings show only one implementation method of the invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and implementation method similar to the technical solution without creativity, which should fall within the protection scope of the invention.

Claims

1. High-voltage heater intelligent alarm system based on digital twin and expert system, characterized by: It includes data collection layer, data transmission layer, data processing layer, model building layer, model verification layer, model optimization layer and application service layer; The data collection layer collects the operation data of the thermal system through sensors and metering instruments; The data transmission layer transmits the collected data to the background of the digital twin system; The data processing layer pre-processes, cleans and stores the collected data; The model building layer builds a virtual digital twin model of the actual physical entity by analyzing and modeling the operating data of the actual thermal system; The model verification layer compares and verifies the model of the digital twin system with the actual thermal system, and verifies the accuracy of the model through error testing between the twin model and the actual output of the physical entity; The model optimization layer optimizes the hyperparameters of the deep forest through an improved bat algorithm to optimize the digital twin system model; The application service layer performs real-time monitoring and fault detection.

2. The high-voltage heater intelligent alarm system based on digital twin and expert system according to claim 1 is characterized in that: The data collection layer collects temperature, pressure and flow information through sensors and measuring instruments.

3. The method for using the intelligent alarm system for high-voltage heater based on digital twin and expert system is characterized in that: The following steps are involved: S1. Build a digital twin system for the thermal system; S2, using bat algorithm to solve; S3. From the output image when the fault occurs, it can be seen that there are different degrees of deviation between the predicted output and the actual output; S4, by calculating the Euclidean distance between the predicted output array and the actual output array, the two parameters of the service system monitoring, the physical entity output and the twin model output, are integrated into one indicator; S5. The fault tree alarm analysis method is adopted, and its main function is to monitor the abnormal changes of some important parameters during the operation of the unit.

4. The method for using the high-voltage heater intelligent alarm system based on digital twin and expert system according to claim 3 is characterized in that: The steps in S1 specifically include: S11. Build data collection layer: collect the actual operation data of thermal system through sensors and metering instruments, including temperature, pressure, flow and other information; S12. Build the data transmission layer: transmit the collected data to the background of the digital twin system. The data transmission layer needs to ensure the reliability and security of data transmission. In the same local area network, use the socket module programming in Python to realize the data transmission between the twin model and the physical entity through wired mode; S13. Build the data processing layer: pre-process, clean and store the collected data for subsequent analysis and modeling; S14, build model building layer: by analyzing and modeling the actual thermal system operation data, build a virtual digital twin model of the actual physical entity, build a deep forest model of the steam turbine system through Python, and use the operation history data of the physical entity for training to ensure the applicability of the model; S15. Build the model verification layer: Compare and verify the model of the digital twin system with the actual thermal system, and verify the accuracy of the model through error testing between the twin model and the actual output of the physical entity; S16. Build the model optimization layer: To further improve the accuracy of the model, the hyperparameters of the deep forest are optimized through the improved bat algorithm to optimize the digital twin system model; S17, application service layer: by comparing the output of the twin model with the output of the simulation machine, the similarity function method is introduced to integrate a parameter that can characterize the operating status as the analysis result of the digital twin system, and it is applied to the operation of the actual thermal system to achieve real-time monitoring and fault detection.

5. The method for using the high-voltage heater intelligent alarm system based on digital twin and expert system according to claim 3 is characterized in that: The S2 specifically includes the following steps: S21, initialize the position and speed of the bat group, and set relevant parameters, including the bat group size, maximum number of iterations, jump step length, jump speed, minimum frequency, and maximum frequency. S22. Calculate the fitness value of each bat. S23. For each bat, calculate the new position and new speed, and update the bat position and speed. S24. For each bat, the position is updated or not updated randomly according to a certain probability, that is, if ri>Ai, the position is not updated, Ai is the pulse emission frequency of bat i, ri is a random number between 0 and 1, otherwise the position is updated. S25. Update the global optimal solution x best. S26. Repeat steps S22-S25 until the stop condition is reached.

Images