Coal mill diagnosis method and system based on digital twinning and BP neural network

Abstract

The application belongs to the technical field of coal mill fault detection, and discloses a coal mill diagnosis method and system based on digital twinning and BP neural network. The method comprises the following steps: a twinning model mapped with the coal mill is constructed; the output of the twinning model is obtained through data interaction between the twinning model and the coal mill, and the target twinning model is obtained by comparing the values of each first parameter of the twinning model with the values of each first parameter collected from the coal mill; a plurality of training samples are obtained based on the simulation of the target twinning model, and the target prediction model is obtained by training the BP neural network prediction model; the target twinning model obtains the real-time parameters of the coal mill to obtain each output, and the values of each first parameter of the target twinning model are compared with the values of each first parameter output by the coal mill to confirm that the two are consistent in operation; and when the two are consistent in operation, all the outputs are input into the target prediction model to predict the fault of the coal mill. The application realizes the fault prediction of the coal mill, thereby reducing the frequency of faults and reducing the power generation cost of the power plant.

Description

Technical Field

[0001] This invention relates to the field of coal mill fault detection technology, specifically to a coal mill diagnostic method and system based on digital twins and BP neural networks. Background Technology

[0002] A thermal power plant is a power plant consisting of boiler units, steam turbines, and generators, which convert the chemical energy of fuel combustion into thermal energy, mechanical energy, and ultimately electrical energy. In actual power generation, the fuel is mostly coal, and the corresponding boiler units are coal-fired boiler units. To improve the coal combustion efficiency of the power plant, a coal mill is added before the coal-fired boiler unit to grind the raw coal into pulverized coal and transport it to the coal-fired boiler.

[0003] To reduce the power generation costs of thermal power plants, it is necessary to minimize equipment failures and avoid equipment downtime due to malfunctions. Statistical analysis of coal-fired boiler unit failures shows that insufficient coal supply, coal blockage, or coal shortages in the coal mill are the main causes of abnormalities in coal-fired boiler units. However, due to the complexity of the coal mill's operating environment, its operating status often has significant uncertainties. Therefore, in actual operation, it is impossible to effectively monitor its operating status based on changes in primary parameters. In this case, the corresponding coal mill failure can only be confirmed when a coal-fired boiler failure occurs and repairs are carried out; by then, the corresponding cost losses have already been incurred. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for diagnosing coal mills based on digital twins and BP neural networks, so as to improve the technical problem that the existing technology cannot effectively monitor the operating status of coal mills and thus cannot effectively avoid corresponding cost losses.

[0005] To achieve the above objectives, the present invention proposes the following technical solution: A coal mill diagnostic method based on digital twins and BP neural networks includes: A twin model mapped to the coal mill is constructed. The governing equations include: dynamic differential equations for primary air temperature, primary air flow rate, outlet air-coal mixture temperature, raw coal, pulverized coal, and the coal mill current equation. The inputs to the twin model include the physical parameters as independent variables in each governing equation, and the outputs include a first parameter corresponding to the coal mill's operation and a second parameter simulated by the twin model. The first parameter includes: current, primary air flow rate, primary air temperature, coal feed rate, and outlet temperature. The second parameter includes: average internal air temperature, standard deviation of internal air temperature, maximum internal air temperature, and minimum internal air temperature. The output quantities of the twin model are obtained based on the data interaction between the twin model and the coal mill. The values ​​of the first parameters output by the twin model are compared with the values ​​of the first parameters collected from the coal mill to iteratively optimize the twin model, thereby obtaining the target twin model. Based on the target twin model simulation, several training samples are obtained, and the BP neural network prediction model is iteratively trained with the prediction accuracy being greater than a preset threshold, thereby obtaining the target prediction model; The target twin model acquires the real-time parameters of the coal mill as input to output various output quantities. Based on the comparison between the values ​​of the first parameters output by the target twin model and the values ​​of the first parameters output by the coal mill to confirm that the target twin model and the coal mill are operating in accordance with each other, all the output quantities are input to the target prediction model to predict the failure of the coal mill.

[0006] Furthermore, the construction of the twin model mapped to the coal mill also includes: A geometric model is constructed based on the structural, material, and dimensional parameters of the coal mill. A physical interface and each of the control equations are added to the geometric model to construct the physical model; wherein, the physical interface is used to input input quantities, which are obtained by data interaction with the coal mill or by simulation; Add constraint rules to the physical model to construct the twin model; wherein the constraint rules include: primary air flow rate constraint, primary air temperature constraint, coal mill outlet temperature range constraint, coal storage constraint, and outlet coal powder fineness constraint.

[0007] Furthermore, the comparison of the values ​​of each of the first parameters output by the target twin model with the values ​​of each of the first parameters output by the coal mill to confirm that the target twin model is consistent with the operation of the coal mill includes: Obtain the set S of values ​​of each of the first parameters based on the output of the target twin model. VE S is the set of values ​​S of each of the first parameters output by the coal mill. PE ; judge At that time, it was confirmed that the target twin model was consistent with the operation of the coal mill; wherein, T S The preset threshold; Conversely, if it is confirmed that the coal mill has not experienced a fault exceeding any constraint rule within the target twin model, the target twin model is re-optimized based on the coal mill.

[0008] Furthermore, including: The dynamic differential equation for the primary air temperature is: ; in, The inertial constant of the temperature from the hot / cold air damper to the primary air is... The temperature of the primary air. The rate of change of primary air temperature. The specific heat capacity of the primary air is [not specified]. This is the maximum airflow of the cold air damper. This refers to the opening degree of the cold air valve. The temperature of the primary air cooling system. For the specific heat capacity of the primary air, This is the maximum airflow of the hot air damper. This refers to the opening degree of the hot air valve. The temperature of the primary hot air. The specific heat capacity of the primary air; The dynamic differential equation for the primary airflow is: ; in, The inertial constant of the primary airflow from the hot / cold air damper. For one airflow, This represents the rate of change of primary airflow. The dynamic differential equation for the temperature of the outlet air-powder mixture is: ; in, The total specific heat capacity of the coal mill participating in heat exchange; For the quality of raw coal, For the quality of pulverized coal, The quality of the metal used for heat exchange inside the coal mill. The rate of change of temperature of the air-coal mixture at the coal mill outlet. The total heat flowing into the coal mill per unit time. The total heat flowing out of the coal mill per unit time; The dynamic differential equation for the raw coal is: ; in, This represents the rate of change in the quality of raw coal. For coal feed rate, The conversion coefficient between raw coal and pulverized coal per unit time; The dynamic differential equation for pulverized coal is: ; in, This refers to the pulverized coal flow rate at the coal mill outlet. The rate of change in the mass of pulverized coal inside the coal mill; The current equation for the coal mill is: ; in, For the coal mill current, , , All are coefficients.

[0009] Furthermore, the step of obtaining several training samples based on the target twin model simulation, and iteratively training the BP neural network prediction model with a prediction accuracy greater than a preset threshold to obtain the target prediction model includes: The output of the BP neural network prediction model is defined as the fault type; wherein, the fault type includes: insufficient coal fault, coal blockage fault, and high outlet temperature fault; The BP neural network prediction model is set to have 9 input nodes, 3 output nodes, and 10 hidden layer nodes.

[0010] Furthermore, including: Select the appropriate maintenance strategy based on the fault prediction results; The twin model is modified and optimized according to the maintenance strategy until the twin model recovers from the fault; The coal mill is maintained based on the optimized maintenance strategy.

[0011] A coal mill diagnostic system based on digital twins and BP neural networks includes: A twin model construction module is used to construct a twin model mapped to the coal mill. The governing equations include: a dynamic differential equation for primary air temperature, a dynamic differential equation for primary air flow rate, a dynamic differential equation for the outlet air-coal mixture temperature, a dynamic differential equation for raw coal, a dynamic differential equation for pulverized coal, and a coal mill current equation. The inputs to the twin model include the physical parameters that serve as independent variables in each of the governing equations. The outputs include a first parameter corresponding to the operation of the coal mill and a second parameter simulated by the twin model. The first parameter includes: current, primary air flow rate, primary air temperature, coal feed rate, and outlet temperature. The second parameter includes: average internal air temperature, standard deviation of internal air temperature, maximum internal air temperature, and minimum internal air temperature. The twin model optimization module is used to obtain each of the output quantities of the twin model based on the data interaction between the twin model and the coal mill, and to compare the values ​​of each first parameter output by the twin model with the values ​​of each first parameter collected from the coal mill to iteratively optimize the twin model, thereby obtaining the target twin model; The prediction model optimization module is used to obtain several training samples based on the target twin model simulation, and to iteratively train the BP neural network prediction model with the goal of achieving a prediction accuracy greater than a preset threshold, thereby obtaining the target prediction model. The entity fault prediction module is used to enable the target twin model to obtain the real-time parameters of the coal mill as input quantities to output various output quantities, and to input all the output quantities into the target prediction model to predict the fault of the coal mill by comparing the values ​​of the first parameters output by the target twin model with the values ​​of the first parameters output by the coal mill to confirm that the target twin model and the coal mill are operating in accordance with each other.

[0012] Furthermore, it also includes: The geometric model building module is used to build a geometric model based on the structural, material, and dimensional parameters of the coal mill. A physical model construction module is used to add physical interfaces and control equations to the geometric model to construct the physical model; wherein, the physical interface is used to input input quantities, which are obtained by data interaction with the coal mill or by simulation; The constraint rule adding module is used to add constraint rules to the physical model to construct the twin model; wherein, the constraint rules include: primary air flow rate constraint, primary air temperature constraint, coal mill outlet temperature range constraint, coal storage constraint, and outlet coal powder fineness constraint.

[0013] Furthermore, including: The data acquisition module is used to acquire a set S of values ​​of each of the first parameters output based on the target twin model. VE S is the set of values ​​S of each of the first parameters output by the coal mill. PE ; The consistency judgment module is used to determine... At that time, it was confirmed that the target twin model was consistent with the operation of the coal mill; wherein, T S The preset threshold; The target twin model optimization module is used to re-optimize the target twin model based on the coal mill when the two are inconsistent in operation and it is confirmed that the coal mill has not experienced a fault that exceeds any constraint rule in the target twin model.

[0014] Furthermore, including: The maintenance strategy acquisition module is used to select the corresponding maintenance strategy based on the fault prediction results. The maintenance strategy optimization module is used to optimize the twin model based on the maintenance strategy until the twin model recovers from the fault. The coal mill maintenance module is used to maintain the coal mill based on the optimized maintenance strategy.

[0015] Beneficial effects: As can be seen from the above technical solutions, the technical solution of the present invention provides a coal mill diagnosis method based on digital twin and BP neural network to improve the defects of the existing technology in coal mill fault monitoring.

[0016] Considering the complexity of the coal mill's operating environment and the uncertainty of its operating status, making it impossible to provide fault prediction from the perspective of the physical coal mill, the inventors considered a virtual approach, leveraging relevant learning algorithms based on their adaptive and self-learning advantages to predict the coal mill's operating status in complex scenarios. Based on this consideration, this technical solution introduces a BP neural network prediction model, which is specifically optimized for the coal mill scenario to obtain a target prediction model. However, during the training of the BP neural network prediction model, considering the need for supervision based on the fault status of the physical coal mill, the existing method of obtaining training samples using random algorithms is not applicable. Furthermore, considering the complexity of coal mill operation, the training samples obtained from historical databases, commonly used in existing supervised learning algorithms, cannot train a more optimized prediction model. Simultaneously, some important parameters strongly correlated with the coal mill's operating status and affecting the accuracy of fault prediction cannot be obtained through sensors deployed on the coal mill, leading to technical limitations in the data types within the historical database corresponding to the physical coal mill during BP neural network prediction model optimization. Therefore, this technical solution introduces digital twin technology to construct a twin model mapped to the physical coal mill. Specifically, in the construction of the twin model, considering that abnormal coal quantity and abnormal outlet temperature are common fault types, abnormal coal quantity manifests as coal blockage, coal shortage, or insufficient coal, mostly caused by insufficient airflow, which will be reflected in the mill current; abnormal outlet temperature is mostly caused by excessively large opening of abnormal hot air inlet and excessively low initial coal humidity, which will cause coal powder blockage, etc. Therefore, the dynamic differential equations of primary air temperature, primary air flow, outlet air-coal mixture temperature, raw coal, coal powder, and mill current are used as the basic control equations to control the operation of the entire twin model; and the current, primary air flow, primary air temperature, coal feed rate, and outlet temperature that can be obtained from the actual operation of the mill (these are the first parameters) are used, and the average internal air temperature, standard deviation of internal air temperature, maximum internal air temperature, and minimum internal air temperature that cannot be obtained from the actual operation of the mill but can be output by simulation (these are the second parameters) are also introduced as the output of the twin model (which are also the training samples when training the BP neural network prediction model and the input of the target prediction model when predicting faults). Furthermore, during the optimization of the twin model, the values ​​of the first parameters of the actual coal mill are directly used as the benchmark for iterative training. This avoids the complexity of the training process when introducing new benchmark values ​​and helps improve training accuracy. On the other hand, when performing fault prediction based on the target prediction model, considering the operating fluctuations or mechanical wear of the coal mill, the twin model is also monitored in real time based on the values ​​of the first parameters of the coal mill during the prediction process to ensure the consistency of their operation. This ensures the reliability of the twin model and the prediction accuracy of the target prediction model.

[0017] Therefore, this technical solution takes BP neural network prediction as its core, combines digital twin technology, and considers anomaly analysis of coal mills. Through real-time interaction between the coal mill, the target twin model, and the target prediction model, it realizes the fault prediction of the coal mill, thereby achieving cost reduction and efficiency improvement in thermal power plants.

[0018] It should be understood that all combinations of the foregoing concepts and the additional concepts described in more detail below can be considered part of the inventive subject matter of this disclosure, provided that such concepts do not contradict each other.

[0019] The foregoing and other aspects, embodiments, and features of the teachings of the present invention will be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the invention, such as features and / or beneficial effects of exemplary embodiments, will become apparent from the following description or may be learned through practice of specific embodiments according to the teachings of the present invention. Attached Figure Description

[0020] The accompanying drawings are not intended to be drawn to scale. In the drawings, each identical or nearly identical component shown in the various figures may be denoted by the same reference numeral. For clarity, not every component is labeled in each figure. Embodiments of various aspects of the invention will now be described by way of example and with reference to the accompanying drawings, wherein: Figure 1 This is a flowchart of the coal mill diagnostic method described in this embodiment; Figure 2 A flowchart for constructing a twin model; Figure 3 A flowchart for constructing a BP neural network prediction model; Figure 4 A flowchart for determining the consistency between the target twin model and the coal mill operation; Figure 5 This is a flowchart for coal mill maintenance. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art.

[0022] The terms "first," "second," and similar words used in the specification and claims of this patent application do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, unless the context clearly indicates otherwise, the singular forms of "an," "a," or "the," etc., do not indicate a quantity limitation, but rather indicate the presence of at least one. Terms such as "comprising" or "including" indicate that the element or object preceding "comprising" encompasses the features, wholes, steps, operations, elements, and / or components listed following "comprising" or "including," and do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or sets thereof. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0023] Coal mill malfunctions are one of the main causes of failures in coal-fired boiler units. However, due to the complexity of the coal mill's operating environment and process, it is impossible to predict potential malfunctions based on its operating status during actual production. This makes it difficult to effectively avoid potential coal mill malfunctions, thus creating a technical bottleneck for reducing production costs in thermal power plants. Therefore, this embodiment aims to provide a coal mill diagnostic method based on digital twins and BP neural networks to improve the aforementioned deficiencies in existing coal mill operation monitoring technologies.

[0024] The following description, in conjunction with the accompanying drawings, details the coal mill diagnostic method based on digital twins and BP neural networks disclosed in this embodiment.

[0025] like Figure 1 As shown, the method includes: Step S102: Construct a twin model that maps to the coal mill.

[0026] As a specific implementation method, such as Figure 2 As shown, the construction of the twin model includes: Step S102.2: Construct a geometric model based on the structural parameters, material parameters, and dimensional parameters of the coal mill.

[0027] In this step, the structural and dimensional parameters are obtained based on specifications and actual measurements, and the material parameters are also obtained based on specifications. Specifically, the structural parameters involve the following structures: variable frequency drive motor, rotary separator, rotary blades, sealed air duct, middle frame, loading frame, primary air inlet, grinding roller, rotary nozzle ring, grinding disc, lower frame, lower frame sealing ring, planetary gear reducer, motor, and hydraulic tie rod.

[0028] Step S102.4: Add physical interfaces and control equations to the geometric model to construct the physical model.

[0029] The physical interface is used to input input quantities, which are obtained through data interaction with the coal mill or through simulation.

[0030] Considering the types of pulverizer malfunctions, they are categorized into two types: abnormal coal quantity and excessively high outlet temperature. Abnormal coal quantity can occur in two ways: coal blockage, coal shortage, or insufficient coal supply. The malfunction is mainly caused by insufficient airflow or a faulty conveyor system. When there is a coal shortage or insufficient coal supply, the pulverizer current decreases accordingly; when there is slight coal blockage inside the machine, the pulverizer current increases accordingly. Excessively high outlet temperature can be caused by factors such as excessively large opening of the primary hot air inlet and excessively low initial moisture content of the coal. Excessively high outlet temperature can cause coal dust blockage, which can lead to spontaneous combustion or explosion of the coal dust.

[0031] Based on this consideration, the control equations are set as follows: dynamic differential equations for primary air temperature, primary air flow rate, outlet air-coal mixture temperature, raw coal, pulverized coal, and mill current. The inputs to the twin model are set as the physical parameters that serve as independent variables in each of the control equations. The outputs are set as a first parameter corresponding to the mill's operation and a second parameter simulated by the twin model. The first parameter includes: current, primary air flow rate, primary air temperature, coal feed rate, and outlet temperature. The second parameter includes: average internal air temperature, standard deviation of internal air temperature, maximum internal air temperature, and minimum internal air temperature.

[0032] Specifically, the expressions for each of the control equations are as follows, which also include the input quantities of the twin model.

[0033] The dynamic differential equation for the primary air temperature is: ; in, The inertial constant of the temperature from the hot / cold air damper to the primary air is... The temperature of the primary air. The rate of change of primary air temperature. The specific heat capacity of the primary air is [not specified]. This is the maximum airflow of the cold air damper. This refers to the opening degree of the cold air valve. The temperature of the primary air cooling system. For the specific heat capacity of the primary air, This is the maximum airflow of the hot air damper. This refers to the opening degree of the hot air valve. The temperature of the primary hot air. The specific heat capacity of the primary air; The dynamic differential equation for the primary airflow is: ; in, The inertial constant of the primary airflow from the hot / cold air damper. For one airflow, This represents the rate of change of primary airflow. The dynamic differential equation for the temperature of the outlet air-powder mixture is: ; in, The total specific heat capacity of the coal mill participating in heat exchange; For the quality of raw coal, For the quality of pulverized coal, The quality of the metal used for heat exchange inside the coal mill. The rate of change of temperature of the air-coal mixture at the coal mill outlet. The total heat flowing into the coal mill per unit time. The total heat flowing out of the coal mill per unit time; The dynamic differential equation for the raw coal is: ; in, This represents the rate of change in the quality of raw coal. For coal feed rate, The conversion coefficient between raw coal and pulverized coal per unit time; The dynamic differential equation for pulverized coal is: ; in, This refers to the pulverized coal flow rate at the coal mill outlet. The rate of change in the mass of pulverized coal inside the coal mill; The current equation for the coal mill is: ; in, For the coal mill current, , , All are coefficients.

[0034] Step S102.6: Add constraint rules to the physical model to construct the twin model.

[0035] The constraint rules set based on the control equation in step S102.4 include: primary air flow rate constraint, primary air temperature constraint, coal mill outlet temperature range constraint, coal storage constraint, and outlet coal powder fineness constraint.

[0036] The twin model is built through steps S102.2 to S102.6. In particular, since the output of the twin model includes a second parameter that cannot be directly obtained from the coal mill, and this second parameter is related to the operation fault of the coal mill, it is beneficial to improve the training of the subsequent BP neural network prediction model and the accuracy of the target prediction model in fault prediction.

[0037] Step S104: Based on the data interaction between the twin model and the coal mill, obtain each of the output quantities of the twin model, and compare the values ​​of each first parameter output by the twin model with the values ​​of each first parameter collected from the coal mill to iteratively optimize the twin model, thereby obtaining the target twin model.

[0038] In this step, since the value of the first parameter in the coal mill is directly used as the optimization target, the complexity of introducing a new optimization benchmark is avoided, and the accuracy of the twin model optimization is guaranteed.

[0039] In practice, one or more of the following modules—WiFi, 5G, and GPRS—are used for data interaction between the two.

[0040] In the interaction between the coal mill and the twin model, the set of attributes of the coal mill can be represented as A = {a} m |1≤i≤M}, where a m This refers to the m-th attribute of the coal mill, such as its specifications, material properties, function, and capacity, where M represents the number of attributes; the set of working conditions can be represented as C = {c n |1≤n≤N}, where c n This is the nth conditional parameter, such as the target value set for external environmental parameters, where N represents the number of parameters. The set of coal mill operating status parameters can be represented as X = {x} i |1≤i≤W}, where x i This represents the i-th real state of the equipment, such as normal state, coal feed pipe blockage, grinding roller stoppage, and hot and cold air valve opening misalignment. W represents the number of parameters. During the twin model simulation, the physical entity attribute parameters A of the coal mill and the current working conditions C are loaded into the virtual model, and the operating status data X of the coal mill is input into the twin model. The digital twin virtual model drives the coal mill to work virtually in the computer, obtaining the simulation operating status data of the coal mill evolving over time. While starting the virtual model to obtain simulation data, the signal connection between the physical entity of the coal mill and the virtual model is checked to ensure real-time data transmission and interaction.

[0041] As an alternative implementation method, the data collected from the coal mill is preprocessed to filter out transmission anomalies such as occasional abnormal data or blank data, thereby improving the effectiveness of data interaction.

[0042] Step S106: Based on the target twin model simulation, obtain several training samples, and iteratively train the BP neural network prediction model with the goal of achieving a prediction accuracy greater than a preset threshold, thereby obtaining the target prediction model.

[0043] As a specific implementation method, combined with Figure 3 As shown, the optimization of the BP neural network prediction model includes: Step S106.2: Set the output of the BP neural network prediction model as the fault type.

[0044] In this step, the fault types include: insufficient coal supply, coal blockage, and high outlet temperature.

[0045] Step S106.4: Set the number of input nodes of the BP neural network prediction model to 9, the number of output nodes to 3, and the number of hidden layer nodes to 10.

[0046] In this step, it is considered that a neural network with an infinite number of hidden layer nodes can approximate any continuous function within the mapping interval. However, increasing the number of hidden layer nodes makes the neural network structure more complex, increases the computational workload, and prolongs the training time. Therefore, referring to the empirical formula for calculating the number of hidden layer nodes, a BP neural network of 9-10-3 is constructed.

[0047] Step S108: The target twin model acquires the real-time parameters of the coal mill as input to output each of the output quantities. Based on the comparison between the values ​​of each of the first parameters output by the target twin model and the values ​​of each of the first parameters output by the coal mill to confirm that the target twin model and the coal mill are operating in agreement, all the output quantities are input to the target prediction model to predict the fault of the coal mill.

[0048] As a specific implementation method, when determining the consistency between the target twin model and the coal mill operation, such as... Figure 4 As shown, it includes: Step S108.2: Obtain the set S of values ​​of each of the first parameters output by the target twin model. VE S is the set of values ​​S of each of the first parameters output by the coal mill. PE .

[0049] Step S108.4, Judgment At that time, it was confirmed that the target twin model was consistent with the operation of the coal mill; wherein, T S This is a preset threshold.

[0050] Step S108.6: Conversely, if it is confirmed that the coal mill has not experienced a fault exceeding any constraint rule in the target twin model, the target twin model is re-optimized based on the coal mill.

[0051] In this step, the reliability of the target twin model and the accuracy of the prediction results of the target prediction model are improved by simultaneously using a coal mill to monitor and calibrate the target twin model in real time during the fault prediction process.

[0052] In specific implementation, when the fault prediction is based on the target prediction model, the corresponding operational changes in the target twin model are as follows: (1) When simulating the coal shortage fault of the coal mill, the fault cause can be set as the coal feed pipe blockage. When the coal feed pipe is blocked, the coal feed of the coal mill is reduced sharply, the amount of raw coal entering the coal mill is reduced, resulting in a reduction in the grinding power of the coal mill, a reduction in the amount of raw coal and pulverized coal in the mill, a reduction in the pulverized coal flow rate at the mill outlet to keep in line with the coal feed rate, and a reduction in the mill outlet temperature. This leads to the opening of the cold air valve and the closing of the hot air valve, which in turn reduces the primary air temperature and causes the mill outlet temperature to drop. (2) When simulating the coal blockage fault of the coal mill, the fault cause can be set as the grinding roller stops running, the conversion coefficient between raw coal and pulverized coal drops to zero, thus causing coal blockage. When the coal blockage fault of the coal mill occurs, the amount of raw coal in the coal mill increases, the pulverized coal flow rate at the outlet decreases, resulting in a reversal in the coal feed rate. The blockage of the primary air passage in the coal mill leads to a decrease in the coal mill outlet temperature. The feedback of the cold air valve decreases, while the feedback of the hot air valve increases. Because the cold air valve operates faster than the hot air valve, the primary air flow first decreases and then increases. (3) When the simulated coal mill outlet temperature is too high, the fault can be set as the hot air valve opening being too large or the cold air valve opening being too small, resulting in a mismatch between hot and cold air, which in turn leads to an excessively high primary air temperature at the mill inlet, resulting in an excessively high coal mill outlet temperature. The cold air valve opens rapidly, while the hot air valve decreases its opening to reduce the primary air temperature and thus reduce the mill outlet temperature. However, the cold air valve operates faster than the hot air valve, and the primary air flow first increases and then decreases. The coal powder flow rate at the mill outlet is affected by the primary air flow rate, first increasing and then decreasing. The coal mill current is affected by the amount of coal stored in the mill and decreases. The coal feed rate is affected by the coal powder flow rate at the outlet, first decreasing and then increasing.

[0053] As a preferred implementation method, such as Figure 5 As shown, after confirming the fault, the following steps are also performed: Step S202: Select the appropriate maintenance strategy based on the fault prediction results.

[0054] Step S204: Optimize the twin model according to the maintenance strategy until the twin model recovers from the fault.

[0055] Step S206: Perform maintenance on the coal mill based on the optimized maintenance strategy.

[0056] Furthermore, based on steps S202 to S206, the selection and optimization of maintenance strategies were realized, thereby improving the efficiency of coal mill maintenance and calibration.

[0057] As described above, this embodiment provides a method for predicting coal mill faults. It combines BP neural network prediction and digital twins to achieve effective fault prediction. Specifically, a target twin model is used to generate training samples for the BP neural network prediction model, improving the simplicity of sample acquisition and the training accuracy of the BP neural network prediction model. Furthermore, during fault diagnosis, the target twin model is adjusted in real-time based on the coal mill, further enhancing prediction accuracy. After fault identification, maintenance strategies are selected and optimized based on the twin model, improving the effectiveness of coal mill maintenance. This ultimately reduces the overall cost losses caused by coal mill faults, facilitating cost reduction and efficiency improvement in thermal power plants.

[0058] The aforementioned program can run in a processor or be stored in memory (or a computer-readable storage medium). Computer-readable media includes both permanent and non-permanent, removable and non-removable media, and information storage can be achieved by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include temporary computer-readable media, such as modulated data signals and carrier waves.

[0059] These computer programs may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes can be implemented using different modules, and different steps can be implemented using different modules.

[0060] This embodiment provides such a system, which can be called a coal mill diagnostic system based on digital twins and BP neural networks. The system includes: A twin model construction module is used to construct a twin model mapped to the coal mill. The governing equations include: a dynamic differential equation for primary air temperature, a dynamic differential equation for primary air flow rate, a dynamic differential equation for the outlet air-coal mixture temperature, a dynamic differential equation for raw coal, a dynamic differential equation for pulverized coal, and a coal mill current equation. The inputs to the twin model include the physical parameters that serve as independent variables in each of the governing equations. The outputs include a first parameter corresponding to the operation of the coal mill and a second parameter simulated by the twin model. The first parameter includes: current, primary air flow rate, primary air temperature, coal feed rate, and outlet temperature. The second parameter includes: average internal air temperature, standard deviation of internal air temperature, maximum internal air temperature, and minimum internal air temperature.

[0061] The twin model optimization module is used to obtain the output quantities of the twin model based on the data interaction between the twin model and the coal mill, and compare the values ​​of the first parameters output by the twin model with the values ​​of the first parameters collected from the coal mill to iteratively optimize the twin model, thereby obtaining the target twin model.

[0062] The prediction model optimization module is used to obtain several training samples based on the target twin model simulation, and to iteratively train the BP neural network prediction model with the goal of achieving a prediction accuracy greater than a preset threshold, thereby obtaining the target prediction model.

[0063] The entity fault prediction module is used to enable the target twin model to obtain the real-time parameters of the coal mill as input quantities to output various output quantities, and to input all the output quantities into the target prediction model to predict the fault of the coal mill by comparing the values ​​of the first parameters output by the target twin model with the values ​​of the first parameters output by the coal mill to confirm that the target twin model and the coal mill are operating in accordance with each other.

[0064] The steps of the system used to implement the methods in the above embodiments have already been described and will not be repeated here.

[0065] For example, the system also includes: The geometric model building module is used to build a geometric model based on the structural, material, and dimensional parameters of the coal mill.

[0066] The physical model construction module is used to add physical interfaces and control equations to the geometric model to construct the physical model; wherein, the physical parameters at the physical interface are obtained by data interaction with the coal mill or by simulation.

[0067] The constraint rule adding module is used to add constraint rules to the physical model to construct the twin model; wherein, the constraint rules include: primary air flow rate constraint, primary air temperature constraint, coal mill outlet temperature range constraint, coal storage constraint, and outlet coal powder fineness constraint.

[0068] For example, the system also includes: The data acquisition module is used to acquire a set S of values ​​of each of the first parameters output based on the target twin model. VE S is the set of values ​​S of each of the first parameters output by the coal mill. PE .

[0069] The consistency judgment module is used to determine... At that time, it was confirmed that the target twin model was consistent with the operation of the coal mill; wherein, T S This is a preset threshold.

[0070] The target twin model optimization module is used to re-optimize the target twin model based on the coal mill when the two are inconsistent in operation and it is confirmed that the coal mill has not experienced a fault that exceeds any constraint rule in the target twin model.

[0071] For example, the system also includes: The maintenance strategy acquisition module is used to select the appropriate maintenance strategy based on the fault prediction results.

[0072] The maintenance strategy optimization module is used to optimize the twin model based on the maintenance strategy until the twin model recovers from the fault.

[0073] The coal mill maintenance module is used to maintain the coal mill based on the optimized maintenance strategy.

[0074] Since the system is built based on the above method, when predicting coal mill failures, the system will also combine the coal mill, the target twin model, and the target prediction model, which will effectively reduce the frequency of coal mill failures and help reduce costs in thermal power plants.

[0075] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Those skilled in the art can make various modifications and refinements without departing from the spirit and scope of the invention. Therefore, the scope of protection of the present invention shall be determined by the claims.

Claims

1. A coal mill diagnostic method based on digital twin and BP neural network, characterized in that, include: A twin model mapped to the coal mill is constructed. The governing equations include: dynamic differential equations for primary air temperature, primary air flow rate, outlet air-powder mixture temperature, raw coal, pulverized coal, and the coal mill current equation. Inputs include physical parameters used as independent variables in each of the governing equations. Outputs include a first parameter corresponding to the coal mill's operation and a second parameter simulated by the twin model. The first parameter includes: current, primary air flow rate, primary air temperature, coal feed rate, and outlet temperature. The second parameter includes: average internal air temperature, standard deviation of internal air temperature, maximum internal air temperature, and minimum internal air temperature. The construction of the twin model mapped to the coal mill further includes: constructing a geometric model based on the structural parameters, material parameters, and dimensional parameters of the coal mill; adding a physical interface and each of the control equations to the geometric model to construct a physical model; wherein the physical interface is used to input input quantities, which are obtained through data interaction with the coal mill or through simulation; adding constraint rules to the physical model to construct the twin model; wherein the constraint rules include: primary air flow rate constraint, primary air temperature constraint, coal mill outlet temperature range constraint, coal storage constraint, and outlet coal powder fineness constraint; The output quantities of the twin model are obtained based on the data interaction between the twin model and the coal mill. The values ​​of the first parameters output by the twin model are compared with the values ​​of the first parameters collected from the coal mill to iteratively optimize the twin model, thereby obtaining the target twin model. Obtain the set S of values ​​of each of the first parameters based on the output of the target twin model. VE S is the set of values ​​S of each of the first parameters output by the coal mill. PE ;judge At that time, it was confirmed that the target twin model was consistent with the operation of the coal mill; wherein, T S The threshold is set as a preset threshold; otherwise, if it is confirmed that the coal mill has not experienced a fault exceeding any constraint rule in the target twin model, the target twin model is re-optimized based on the coal mill. Based on the target twin model simulation, several training samples are obtained, and the BP neural network prediction model is iteratively trained with the prediction accuracy being greater than a preset threshold, thereby obtaining the target prediction model; The target twin model acquires the real-time parameters of the coal mill as input to output various output quantities. Based on the comparison between the values ​​of the first parameters output by the target twin model and the values ​​of the first parameters output by the coal mill to confirm that the target twin model and the coal mill are operating in accordance with each other, all the output quantities are input to the target prediction model to predict the failure of the coal mill.

2. The coal mill diagnostic method based on digital twin and BP neural network according to claim 1, characterized in that, include: The dynamic differential equation for the primary air temperature is: ； in, The inertial constant of the temperature from the hot / cold air damper to the primary air is... The temperature of the primary air. The rate of change of primary air temperature. The specific heat capacity of the primary air is [not specified]. This is the maximum airflow of the cold air damper. This refers to the opening degree of the cold air valve. The temperature of the primary air cooling system. The specific heat capacity of the primary air is... This is the maximum airflow of the hot air damper. This refers to the opening degree of the hot air valve. The temperature of the primary hot air. The specific heat capacity of the primary air; The dynamic differential equation for the primary airflow is: ； in, The inertial constant of the primary airflow from the hot / cold air damper. For one airflow, This represents the rate of change of primary airflow. The dynamic differential equation for the temperature of the outlet air-powder mixture is: ； in, The total specific heat capacity of the coal mill participating in heat exchange; For the quality of raw coal, For the quality of pulverized coal, The quality of the metal used for heat exchange inside the coal mill. The rate of change of temperature of the air-coal mixture at the coal mill outlet. The total heat flowing into the coal mill per unit time. The total heat flowing out of the coal mill per unit time; The dynamic differential equation for the raw coal is: ； in, This represents the rate of change in the quality of raw coal. For coal feed rate, The conversion coefficient between raw coal and pulverized coal per unit time; The dynamic differential equation for pulverized coal is: ； in, This refers to the pulverized coal flow rate at the coal mill outlet. The rate of change in the mass of pulverized coal inside the coal mill; The current equation for the coal mill is: ； in, For the coal mill current, , , All are coefficients.

3. The coal mill diagnostic method based on digital twin and BP neural network according to claim 1, characterized in that, The process involves simulating and obtaining several training samples based on the target twin model, and iteratively training the BP neural network prediction model with a prediction accuracy greater than a preset threshold to obtain the target prediction model, including: The output of the BP neural network prediction model is defined as the fault type; wherein, the fault type includes: insufficient coal fault, coal blockage fault, and high outlet temperature fault; The BP neural network prediction model is set to have 9 input nodes, 3 output nodes, and 10 hidden layer nodes.

4. The coal mill diagnostic method based on digital twin and BP neural network according to claim 1, characterized in that, include: Select the appropriate maintenance strategy based on the fault prediction results; Based on the correction of the twin model according to the maintenance strategy, the maintenance strategy is optimized until the twin model recovers from the fault; The coal mill is maintained based on the optimized maintenance strategy.

5. A coal mill diagnostic system based on digital twins and BP neural networks, used in the coal mill diagnostic method based on digital twins and BP neural networks as described in claim 1, characterized in that, include: A twin model construction module is used to construct a twin model mapped to the coal mill. The governing equations include: a dynamic differential equation for primary air temperature, a dynamic differential equation for primary air flow rate, a dynamic differential equation for the outlet air-coal mixture temperature, a dynamic differential equation for raw coal, a dynamic differential equation for pulverized coal, and a coal mill current equation. The inputs to the twin model include the physical parameters that serve as independent variables in each of the governing equations. The outputs include a first parameter corresponding to the operation of the coal mill and a second parameter simulated by the twin model. The first parameter includes: current, primary air flow rate, primary air temperature, coal feed rate, and outlet temperature. The second parameter includes: average internal air temperature, standard deviation of internal air temperature, maximum internal air temperature, and minimum internal air temperature. The twin model optimization module is used to obtain each of the output quantities of the twin model based on the data interaction between the twin model and the coal mill, and to compare the values ​​of each first parameter output by the twin model with the values ​​of each first parameter collected from the coal mill to iteratively optimize the twin model, thereby obtaining the target twin model; The prediction model optimization module is used to obtain several training samples based on the target twin model simulation, and to iteratively train the BP neural network prediction model with the goal of achieving a prediction accuracy greater than a preset threshold, thereby obtaining the target prediction model. The entity fault prediction module is used to enable the target twin model to obtain the real-time parameters of the coal mill as input quantities to output various output quantities, and to input all the output quantities into the target prediction model to predict the fault of the coal mill by comparing the values ​​of the first parameters output by the target twin model with the values ​​of the first parameters output by the coal mill to confirm that the target twin model and the coal mill are operating in accordance with each other.

6. The coal mill diagnostic system based on digital twin and BP neural network according to claim 5, characterized in that, Also includes: The geometric model building module is used to build a geometric model based on the structural, material, and dimensional parameters of the coal mill. A physical model construction module is used to add physical interfaces and control equations to the geometric model to construct the physical model; wherein, the physical interface is used to input input quantities, which are obtained by data interaction with the coal mill or by simulation; The constraint rule adding module is used to add constraint rules to the physical model to construct the twin model; wherein, the constraint rules include: primary air flow rate constraint, primary air temperature constraint, coal mill outlet temperature range constraint, coal storage constraint, and outlet coal powder fineness constraint.

7. The coal mill diagnostic system based on digital twin and BP neural network according to claim 5, characterized in that, include: The data acquisition module is used to acquire a set S of values ​​of each of the first parameters output based on the target twin model. VE S is the set of values ​​S of each of the first parameters output by the coal mill. PE ; The consistency judgment module is used to determine... At that time, it was confirmed that the target twin model was consistent with the operation of the coal mill; wherein, T S The preset threshold; The target twin model optimization module is used to re-optimize the target twin model based on the coal mill when the two are inconsistent in operation and it is confirmed that the coal mill has not experienced a fault that exceeds any constraint rule in the target twin model.

8. The coal mill diagnostic system based on digital twin and BP neural network according to claim 5, characterized in that, include: The maintenance strategy acquisition module is used to select the corresponding maintenance strategy based on the fault prediction results. The maintenance strategy optimization module is used to optimize the maintenance strategy based on the correction of the twin model, until the twin model recovers from the fault. The coal mill maintenance module is used to maintain the coal mill based on the optimized maintenance strategy.

Images