Steam feed pump state modeling and steam inlet flow back calculation method, system and equipment based on thermodynamic mechanism and machine learning and medium

Abstract

The invention relates to the technical field of intelligent management of steam feed pumps, in particular to a steam feed pump state modeling and steam inlet flow back calculation method, system and equipment based on a thermodynamic mechanism and machine learning and a medium, and the method comprises the steps that a thermodynamic mechanism model is constructed based on the first law of thermodynamics and an equipment design efficiency curve; real-time operation parameters of the steam feed pump set are obtained, and theoretical parameters are calculated through a thermodynamic mechanism model; the real-time operation parameters and the theoretical parameters are jointly constructed and input into a pre-trained machine learning correction model, and efficiency correction factors are output; calibrating the theoretical isentropic efficiency by using the efficiency correction factor to obtain the actual isentropic efficiency; and based on the actual isentropic efficiency, high-precision values of the steam inlet flow and the steam exhaust enthalpy of the steam turbine are calculated and output through a thermodynamic mechanism model. According to the method, the physical interpretability of the mechanism model and the self-adaptive capacity of the data driving model are fused, and high-precision back calculation of the steam inlet flow and dynamic tracking of the equipment performance state can be achieved.

Description

Technical Field

[0001] This application relates to the field of intelligent management technology for steam-driven feedwater pumps, specifically to a method, system, equipment, and medium for state modeling and inverse calculation of steam flow rate of steam-driven feedwater pumps based on thermodynamic mechanisms and machine learning. Background Technology

[0002] In some large generator sets, steam-driven feedwater pumps are critical auxiliary equipment, and their operating status directly affects the safety and economy of the entire unit. With the advancement of smart power plant construction, utilizing digital twin technology to achieve precise status perception, performance evaluation, and predictive maintenance of critical equipment has become an important development direction in the industry. Establishing a high-fidelity digital twin of the steam-driven feedwater pump, especially enabling reliable back-calculation of the turbine inlet steam flow rate, which is difficult to measure directly and accurately, is of great significance for optimizing operation and providing early warning of faults.

[0003] Currently, modeling methods for steam-driven feedwater pumps are mainly divided into two categories: mechanistic modeling based on physical laws and data-driven modeling based on historical data. Pure mechanistic modeling methods construct a system of equations with clear physical meaning based on fundamental principles such as the first law of thermodynamics and fluid mechanics. They calculate target parameters (such as shaft power, efficiency, and steam flow rate) by inputting measurable parameters (such as pressure, temperature, and flow rate). The structure and parameters of these models are usually derived from equipment design data, exhibiting good interpretability. Pure data-driven modeling methods, such as various neural networks and support vector machines, rely entirely on a large amount of historical operating data. They learn the complex mapping relationship between input and output variables through training, without requiring explicit physical formulas, and demonstrate flexibility in handling nonlinear problems.

[0004] However, both of the aforementioned modeling methods for steam-driven feedwater pumps have inherent limitations: the accuracy of pure mechanistic models heavily depends on the accuracy of their internal efficiency parameters (such as the isentropic efficiency of the turbine), which are typically based on fixed efficiency curves provided during equipment design. In actual operation, due to factors such as equipment aging, scaling, and wear, the actual efficiency will gradually deviate from the design curve, causing the calculation accuracy of mechanistic models based on fixed efficiency curves to decrease after long-term operation, failing to truly reflect the current performance status of the equipment; while pure data-driven models can learn efficiency changes from data, they are entirely dependent on data and lack physical constraints. When operating conditions exceed the range of training data, they may produce prediction results that violate the basic laws of thermodynamics, resulting in poor generalization ability and reliability of the model. Summary of the Invention

[0005] In existing modeling methods for steam-driven feedwater pumps, pure mechanistic models cannot accurately reflect long-term equipment performance degradation due to their reliance on fixed equipment design efficiency curves, while pure data-driven models may produce predictions that violate physical laws when operating conditions exceed the training range due to a lack of physical constraints. This application provides a method, system, equipment, and medium for steam-driven feedwater pump state modeling and inverse calculation of inlet steam flow based on thermodynamic mechanisms and machine learning. By combining thermodynamic mechanism models with machine learning correction models, adaptive tracking of equipment performance degradation is achieved while ensuring the physical interpretability of the model, thereby obtaining high-precision and reliable inverse calculation results of turbine inlet steam flow.

[0006] In a first aspect, this application provides a method for state modeling and inverse calculation of steam inlet flow rate of a steam-driven feedwater pump based on thermodynamic mechanisms and machine learning, including the following steps: S1. A thermodynamic mechanism model of the feedwater pump set is constructed based on the first law of thermodynamics and the equipment design efficiency curve. The thermodynamic mechanism model defines the relationship between the following parameters: The calculation relationship between water supply flow rate, water supply pump inlet pressure, water supply pump outlet pressure, water supply density, and water supply pump shaft power; The mapping relationship between the shaft power of the water pump and the theoretical isentropic efficiency; The calculated relationship between turbine inlet steam enthalpy, condenser pressure, and isentropic enthalpy drop of steam; The calculation relationship between feedwater pump shaft power, isentropic enthalpy drop, isentropic efficiency and turbine inlet steam flow rate and turbine exhaust steam enthalpy; S2. Obtain real-time operating parameters of the steam-driven feedwater pump set, including feedwater flow rate, feedwater pump inlet pressure, feedwater pump outlet pressure, feedwater density, turbine inlet steam enthalpy, and condenser pressure; S3. Based on real-time operating parameters, theoretical parameters are calculated using a thermodynamic mechanism model, including theoretical feedwater pump shaft power, theoretical isentropic enthalpy drop, theoretical isentropic efficiency, theoretical turbine inlet steam flow rate, and theoretical turbine exhaust steam enthalpy. S4. Construct an input feature vector using real-time operating parameters and theoretical parameters, then input it into a pre-trained machine learning correction model. The output is an efficiency correction factor that characterizes the degree of deviation of the device's current performance from its design performance. The machine learning correction model is a regression model built based on the gradient boosting tree algorithm; S5. The theoretical isentropic efficiency is calibrated using an efficiency correction factor to obtain the actual isentropic efficiency. The expression is:

[0007] Indicates theoretical isentropic efficiency; S6. Based on the actual isentropic efficiency, high-precision values ​​of turbine inlet steam flow rate and turbine exhaust steam enthalpy are calculated and output through a thermodynamic mechanism model.

[0008] It should be further noted that the calculation relationship between the feedwater flow rate, feedwater pump inlet pressure, feedwater pump outlet pressure, feedwater density, and feedwater pump shaft power in step S1 is defined by the following formula:

[0009] in, Indicates the power of the water pump shaft; Indicates water supply flow rate; Indicates the outlet pressure of the water pump; Indicates the inlet pressure of the water pump; Indicates the density of the water supply; This indicates the efficiency of the water supply pump, which is a preset constant value or can be obtained by querying the pump's own performance curve based on its operating conditions.

[0010] It should be further explained that in step S1, the mapping relationship between the pump shaft power and the theoretical isentropic efficiency is clarified by querying the equipment design efficiency curve constructed based on the equipment design data. The expression of the equipment design efficiency curve is as follows:

[0011] in, Indicates theoretical isentropic efficiency; Indicates the power of the water pump shaft; The theoretical isentropic efficiency is expressed as a function of the change in the shaft power of the feedwater pump, and is obtained by interpolation of discrete data points or polynomial fitting.

[0012] It should be further noted that the calculation relationship between the turbine inlet enthalpy, condenser pressure, and isentropic enthalpy drop of steam in step S1 is defined by the following formula:

[0013] in, This represents the isentropic enthalpy drop of steam. Indicates the enthalpy of the steam entering the turbine; This represents the isentropic exhaust enthalpy, based on the condenser pressure. Obtained by consulting the table of thermodynamic properties of water vapor.

[0014] It should be further noted that the steam thermodynamic property table is based on the IAPWS-IF97 standard industrial formula.

[0015] It should be further noted that the calculation relationship between the feedwater pump shaft power, isentropic enthalpy drop, isentropic efficiency, and turbine inlet steam flow rate and turbine exhaust steam enthalpy is defined by the following formula:

[0016]

[0017] in, Indicates the steam inlet flow rate of the steam turbine; Indicates the power of the water pump shaft; This indicates isentropic enthalpy decrease; Indicates isentropic efficiency; Indicates the exhaust enthalpy of the steam turbine; This indicates the enthalpy of the steam entering the turbine.

[0018] It should be further explained that in step S4, the machine learning correction model is based on the gradient boosting tree algorithm, which performs regression prediction by integrating multiple decision trees; wherein, each decision tree performs a nonlinear transformation on the input feature vector according to the feature splitting rule, and the efficiency correction factor is finally output by the model. The core formula for the weighted sum of the outputs of all decision trees is expressed as:

[0019] in, This represents the input feature vector; Indicates the first Each decision tree is used for the input feature vector The predicted output; Indicates the first The weight coefficients of each decision tree are obtained by minimizing the loss function during training; This represents the total number of decision trees.

[0020] It should be further noted that when constructing a decision tree, the machine learning correction model calculates the loss function gain resulting from feature splitting. The expression for selecting the optimal split point is:

[0021] in, This represents the loss of the parent node; This indicates the loss of the left subtree; This indicates the loss of the right subtree.

[0022] It should be further noted that the training steps for the machine learning correction model in step S4 are as follows: S401. Collect historical datasets, which contain historical real-time operating parameters of multiple training samples, corresponding measured feedwater pump shaft power and measured steam turbine inlet flow rate; S402. Based on the historical real-time operating parameters of each training sample, the corresponding theoretical parameters are calculated through a thermodynamic mechanism model, including theoretical feedwater pump shaft power, theoretical isentropic enthalpy drop, theoretical isentropic efficiency, theoretical turbine inlet steam flow rate, and theoretical turbine exhaust steam enthalpy. S403. Based on the measured feedwater pump shaft power and measured turbine inlet steam flow rate of each training sample, the actual isentropic efficiency is calculated. The actual isentropic efficiency of each training sample The calculation formula is:

[0023] in, Indicates the first The actual axis power corresponding to each training sample; Indicates the first The actual steam flow rate corresponding to each training sample; Indicates the first The isentropic enthalpy decrease corresponding to each training sample; S404. Calculate the true efficiency correction factor for each training sample, the first... True efficiency correction factor for each training sample The calculation formula is:

[0024] Indicates the first The theoretical isentropic efficiency of each training sample; S405. Using the historical real-time running parameters and corresponding theoretical parameters of each training sample as input features, and the corresponding real efficiency correction factor as the training target label, supervise the training of the machine learning model to obtain a pre-trained machine learning correction model.

[0025] It should be further noted that in step S405, the objective of supervised training is to minimize the objective function, the expression of which is:

[0026] in, Indicates the first The true efficiency correction factor for each training sample; Indicates the machine learning correction model for the first The predicted output value for each training sample; This represents the total number of training samples; This is the loss function, used to measure the difference between the predicted values ​​of the training samples and the training target labels; This is a regularization term used to control the... Sub-model The complexity is reduced to prevent overfitting.

[0027] It should be further explained that the loss function Specifically, it refers to the mean square error function.

[0028] It should be further noted that step S4 also includes an online learning and updating step: periodically using newly generated historical real-time operating parameters, corresponding measured feedwater pump shaft power and measured turbine inlet steam flow to construct new training samples, and using the new training samples to update the model parameters of the machine learning correction model in an incremental learning or periodic full retraining manner.

[0029] It should be further noted that step S7 is also included: continuously recording the efficiency correction factor of the machine learning correction model output. Analyze its 30-day moving average offset. ,when When the preset warning threshold is exceeded, a performance degradation report and maintenance warning are generated and sent to staff.

[0030] It should be further noted that the warning threshold is -5%.

[0031] Secondly, this application provides a system for state modeling and inverse calculation of steam flow rate of a steam-driven feedwater pump based on thermodynamic mechanisms and machine learning, used to implement the above-mentioned method for state modeling and inverse calculation of steam flow rate of a steam-driven feedwater pump, including: The thermodynamic mechanism model building module is used to build a thermodynamic mechanism model of the feedwater pump set based on the first law of thermodynamics and the equipment design efficiency curve. The real-time operating parameter acquisition module is used to acquire the real-time operating parameters of the steam-driven feedwater pump set, including feedwater flow rate, feedwater pump inlet pressure, feedwater pump outlet pressure, feedwater density, turbine inlet steam enthalpy, and condenser pressure. The theoretical parameter calculation module is used to calculate theoretical parameters based on real-time operating parameters using a thermodynamic mechanism model. The efficiency correction factor generation module is used to construct an input feature vector by combining real-time operating parameters and theoretical parameters, input it into a pre-trained machine learning correction model, and output an efficiency correction factor that characterizes the degree of deviation of the current performance of the device from its design performance. The theoretical isentropic efficiency calibration module is used to calibrate the theoretical isentropic efficiency using an efficiency correction factor to obtain the actual isentropic efficiency. The high-precision value calculation module is used to calculate and output high-precision values ​​of turbine inlet steam flow and turbine exhaust steam enthalpy based on actual isentropic efficiency and through a thermodynamic mechanism model.

[0032] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the steps of the above-described method for modeling the state of a steam-driven feedwater pump and calculating the inlet steam flow rate.

[0033] Fourthly, this application provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for modeling the state of a steam-driven feedwater pump and calculating the inlet steam flow rate.

[0034] As can be seen from the above technical solutions, this application has the following advantages: 1. This application combines a mechanistic model based on the first law of thermodynamics with a machine learning correction model based on the gradient boosting tree algorithm. First, the mechanistic model provides a physically consistent theoretical parameter calculation framework based on the design efficiency curve. Then, the machine learning model learns from real-time operating data and theoretical parameters, dynamically outputting an efficiency correction factor to calibrate the theoretical efficiency in the mechanistic model. Ultimately, it achieves high-precision back-calculation of turbine inlet steam flow and exhaust steam enthalpy, combining the physical interpretability of the mechanistic model with the adaptive capability of the data-driven model, significantly improving the back-calculation accuracy under equipment performance degradation conditions.

[0035] 2. This application uses a thermodynamic mechanism model as the core framework for computation, ensuring that the entire computation process follows the basic physical laws. Then, it uses a machine learning model to perform data-driven bias learning only in the efficiency correction stage. This avoids the possibility that a purely data-driven model might produce predictions that violate physical laws such as energy conservation when the operating conditions change drastically or exceed the training range due to a lack of physical constraints. This enhances the robustness and extrapolation reliability of the model.

[0036] 3. This application uses an efficiency correction factor dynamically generated by a machine learning correction model to calibrate the theoretical isentropic efficiency in real time. This effectively overcomes the problem that pure mechanistic models cannot accurately reflect the degradation of actual equipment performance over time due to their reliance on fixed equipment design efficiency curves. This allows the model to adaptively track the current real efficiency level of the equipment, thereby ensuring the accuracy of the flow back calculation results under long-term operation. Attached Figure Description

[0037] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a flowchart of a method for state modeling and inverse calculation of steam flow rate of a steam-driven feedwater pump based on thermodynamic mechanisms and machine learning in one embodiment of this application.

[0039] Figure 2 This is a schematic block diagram of a steam-driven feedwater pump state modeling and steam inlet flow back calculation system based on thermodynamic mechanisms and machine learning in one embodiment of this application.

[0040] Figure 3 This is a schematic diagram of the hardware structure of an electronic device in one embodiment of this application. Detailed Implementation

[0041] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0042] The following describes in detail the steam-driven feedwater pump state modeling and inverse calculation method for steam inlet flow rate involved in this application. Specific details such as particular system structures and technologies are presented for illustrative purposes rather than limiting, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application can also be implemented in other embodiments without these specific details.

[0043] In the steam-driven feedwater pump state modeling and inverse calculation of steam flow rate involved in this application, the term "comprising" indicates the presence of the described feature, integral, step, operation, element, and / or component, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or sets thereof. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0044] To facilitate a clear description of the technical solutions of this application, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" do not necessarily imply that they are different.

[0045] The terms "one embodiment" or "some embodiments" used in this application mean that one or more embodiments of this application include the specific features, structures, or characteristics described in that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this application do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized.

[0046] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0047] The steam-driven feedwater pump state modeling and steam inlet flow calculation method provided in this application embodiment is executed by computer equipment. Correspondingly, the steam-driven feedwater pump state modeling and steam inlet flow calculation system based on thermodynamic mechanism and machine learning runs in the computer equipment.

[0048] Figure 1 This is a flowchart illustrating a method for state modeling and inverse calculation of steam flow rate of a steam-driven feedwater pump based on thermodynamic mechanisms and machine learning, according to an embodiment of this application. Figure 1 The executing entity can be a steam-driven feedwater pump state modeling and inverse steam flow calculation system. Depending on different requirements, the order of steps in this flowchart can be changed, and some steps can be omitted.

[0049] like Figure 1 As shown, the method for state modeling and inverse calculation of steam flow rate of steam-driven feedwater pump based on thermodynamic mechanisms and machine learning includes: Step S1: Based on the first law of thermodynamics and the equipment design efficiency curve, a thermodynamic mechanism model of the feedwater pump set is constructed. The thermodynamic mechanism model defines the relationship between the following parameters: The calculation relationship between water supply flow rate, water supply pump inlet pressure, water supply pump outlet pressure, water supply density, and water supply pump shaft power; The mapping relationship between the shaft power of the water pump and the theoretical isentropic efficiency; The calculated relationship between turbine inlet steam enthalpy, condenser pressure, and isentropic enthalpy drop of steam; The calculation relationship between feedwater pump shaft power, isentropic enthalpy drop, isentropic efficiency and turbine inlet steam flow rate and turbine exhaust steam enthalpy.

[0050] By clearly defining the relationship between feedwater flow rate, pressure, density, and shaft power based on the first law of thermodynamics and the equipment design efficiency curve, establishing a mapping between shaft power and theoretical isentropic efficiency, and constructing a complete calculation chain of inlet steam enthalpy, condenser pressure, isentropic enthalpy drop, and further, inlet steam flow rate and exhaust steam enthalpy, a static mechanism calculation framework for steam-driven feedwater pump sets with a clear structure and explicit physical meaning is established. This framework uses the equipment design efficiency curve as the performance benchmark and links various measurable parameters with target back-calculated parameters through rigorous thermodynamic relationships, providing a reliable theoretical calculation basis and a reference benchmark for deviation analysis for the entire method.

[0051] In some specific embodiments, the calculation relationship between feedwater flow rate, feedwater pump inlet pressure, feedwater pump outlet pressure, feedwater density, and feedwater pump shaft power is defined by the following formula:

[0052] in, Indicates the power of the water pump shaft; Indicates water supply flow rate; Indicates the outlet pressure of the water pump; Indicates the inlet pressure of the water pump; Indicates the density of the water supply; This indicates the efficiency of the water supply pump, which is a preset constant value or can be obtained by querying the pump's own performance curve based on its operating conditions.

[0053] By employing a formula that explicitly includes feedwater flow rate, inlet and outlet pressure, density, and feedwater pump efficiency, the calculation relationship of feedwater pump shaft power is specifically defined. It is also clarified that feedwater pump efficiency can be used as a preset constant or obtained by querying its own performance curve based on operating conditions. This provides a precise and flexible implementation plan for the shaft power calculation link in the entire method. The formula has a clear physical meaning and is based on measurable fluid parameters, which enhances the model's practicality and adaptability to different application scenarios.

[0054] In some specific embodiments, the mapping relationship between the pump shaft power and the theoretical isentropic efficiency is clarified by querying the equipment design efficiency curve constructed based on equipment design data. The expression for the equipment design efficiency curve is as follows:

[0055] in, Indicates theoretical isentropic efficiency; Indicates the power of the water pump shaft; The theoretical isentropic efficiency is expressed as a function of the change in the shaft power of the feedwater pump, and is obtained by interpolation of discrete data points or polynomial fitting.

[0056] By explicitly adopting a digital design efficiency curve based on equipment design data and indexed by the power of the water pump shaft, and obtaining the functional relationship between theoretical isentropic efficiency and shaft power through discrete point interpolation or polynomial fitting, a concrete and operable means is provided for the key step of querying theoretical isentropic efficiency. This transforms the paper-based design curve into a data model that can be automatically processed and queried by a computer, ensuring the automation, accuracy, and efficiency of obtaining the theoretical efficiency benchmark value.

[0057] In some specific embodiments, the calculated relationship between turbine inlet enthalpy, condenser pressure, and isentropic enthalpy drop of steam is defined by the following formula:

[0058] in, This represents the isentropic enthalpy drop of steam. Indicates the enthalpy of the steam entering the turbine; This represents the isentropic exhaust enthalpy, based on the condenser pressure. Obtained by consulting the table of thermodynamic properties of water vapor.

[0059] By using the difference between the turbine inlet steam enthalpy and the isentropic exhaust steam enthalpy obtained by consulting the steam thermodynamic property table based on the condenser pressure, the calculation relationship of the isentropic enthalpy drop of steam is specifically defined. This provides a clear and standardized calculation method for the core thermodynamic parameter of the turbine's theoretical work capacity (i.e., isentropic enthalpy drop). This method relies on accurate physical property lookup to ensure the theoretical correctness and engineering practicality of the enthalpy drop calculation.

[0060] In some specific embodiments, the steam thermodynamic property tables are implemented based on the IAPWS-IF97 standard industrial formula.

[0061] By further clarifying the IAPWS-IF97 standard industrial formula upon which the steam thermodynamic property table is based, the fundamental step of steam property lookup has established an internationally recognized, high-precision calculation standard, fundamentally ensuring the authority, consistency, and reliability of the isentropic exhaust enthalpy value and all subsequent related thermodynamic parameter calculations.

[0062] In some specific embodiments, the calculation relationships between feedwater pump shaft power, isentropic enthalpy drop, isentropic efficiency, and turbine inlet steam flow rate and turbine exhaust steam enthalpy are defined by the following formulas:

[0063]

[0064] in, Indicates the steam inlet flow rate of the steam turbine; Indicates the power of the water pump shaft; This indicates isentropic enthalpy decrease; Indicates isentropic efficiency; Indicates the exhaust enthalpy of the steam turbine; This indicates the enthalpy of the steam entering the turbine.

[0065] By employing specific formulas for calculating turbine inlet steam flow rate using feedwater pump shaft power, isentropic enthalpy drop, and isentropic efficiency, and for calculating turbine exhaust steam enthalpy using inlet steam enthalpy, isentropic enthalpy drop, and isentropic efficiency, clear and closed physical equations are provided for back-calculating key steam parameters from an energy balance perspective. This clarifies the direct mathematical basis for the back-calculation process, enabling high-precision back-calculation to be implemented.

[0066] Step S2: Obtain the real-time operating parameters of the steam-driven feedwater pump set, including feedwater flow rate, feedwater pump inlet pressure, feedwater pump outlet pressure, feedwater density, turbine inlet steam enthalpy, and condenser pressure.

[0067] By acquiring directly or indirectly measurable operating parameters such as feedwater flow rate, feedwater pump inlet and outlet pressure, feedwater density, turbine inlet steam enthalpy, and condenser pressure in real time, a dynamic set of input data reflecting the current actual operating conditions of the equipment is provided for the entire modeling and back-calculation process. This enables mechanism calculations and machine learning corrections to be based on real-time, objective field data, and is an indispensable data input link for realizing the transition from theoretical design analysis to online actual condition assessment and parameter back-calculation.

[0068] Step S3: Based on real-time operating parameters, calculate theoretical parameters using a thermodynamic mechanism model, including theoretical feedwater pump shaft power, theoretical isentropic enthalpy drop, theoretical isentropic efficiency, theoretical turbine inlet steam flow rate, and theoretical turbine exhaust steam enthalpy.

[0069] By inputting real-time operating parameters into the thermodynamic mechanism model, a complete set of theoretical parameters, including theoretical feedwater pump shaft power, theoretical isentropic enthalpy drop, theoretical isentropic efficiency, theoretical turbine inlet steam flow rate, and theoretical turbine exhaust steam enthalpy, are calculated. Based on the current operating conditions and design performance benchmarks, a complete set of "theoretical reference values" is generated. These values ​​characterize the performance that the equipment should have under the current operating conditions when it is in its ideal design state, providing a key benchmark and intermediate quantity for identifying and quantifying the deviation between the actual equipment performance and the design benchmark.

[0070] Step S4: The real-time operating parameters and theoretical parameters are combined to construct an input feature vector, which is then input into a pre-trained machine learning correction model. The output is an efficiency correction factor that characterizes the degree of deviation of the device's current performance from its design performance. The machine learning correction model is a regression model built based on the gradient boosting tree algorithm.

[0071] By constructing a feature vector from real-time operating parameters and theoretical parameters, and inputting it into a pre-trained machine learning correction model based on the gradient boosting tree algorithm, the machine learning correction model outputs an efficiency correction factor that characterizes the degree of deviation of the current performance of the equipment from the design performance. This introduces a data-driven intelligent correction link, which can automatically learn and extract nonlinear efficiency deviation patterns caused by changes in equipment state that cannot be described by fixed mechanism formulas from complex features containing actual operating condition information and theoretical benchmark information, and quantify them with a single correction factor.

[0072] In some specific embodiments, the machine learning correction model is based on the gradient boosting tree algorithm, which performs regression prediction by integrating multiple decision trees; wherein, each decision tree performs a nonlinear transformation on the input feature vector according to a feature splitting rule, and the model finally outputs an efficiency correction factor. The core formula for the weighted sum of the outputs of all decision trees is expressed as:

[0073] in, This represents the input feature vector; Indicates the first Each decision tree is used for the input feature vector The predicted output; Indicates the first The weight coefficients of each decision tree are obtained by minimizing the loss function during training; This represents the total number of decision trees.

[0074] By clarifying that the machine learning correction model is based on the gradient boosting tree algorithm and uses multiple decision trees for prediction, the final output is a weighted sum of the predictions from all decision trees. This reveals the core algorithm architecture and output mechanism of the efficiency correction factor prediction model, and explains how the model improves the accuracy and stability of the final output by fusing the predictions of multiple weak learners through an ensemble learning strategy.

[0075] In some specific embodiments, the machine learning correction model calculates the loss function gain resulting from feature splitting when constructing the decision tree. The expression for selecting the optimal split point is:

[0076] in, This represents the loss of the parent node; This indicates the loss of the left subtree; This indicates the loss of the right subtree.

[0077] By clarifying that the loss function gain value resulting from feature splitting is used as the criterion for selecting the optimal split point when constructing the decision tree of the gradient boosting tree model, this paper reveals the specific optimization principles of feature selection and tree structure construction within the model. It explains how the model automatically constructs efficient decision rules by quantitatively evaluating the contribution of different splitting methods to improving prediction accuracy, thereby improving the overall performance of the model.

[0078] In some specific embodiments, the training steps of the machine learning correction model are as follows: S401. Collect historical datasets, which contain historical real-time operating parameters of multiple training samples, corresponding measured feedwater pump shaft power and measured steam turbine inlet flow rate; S402. Based on the historical real-time operating parameters of each training sample, the corresponding theoretical parameters are calculated through a thermodynamic mechanism model, including theoretical feedwater pump shaft power, theoretical isentropic enthalpy drop, theoretical isentropic efficiency, theoretical turbine inlet steam flow rate, and theoretical turbine exhaust steam enthalpy. S403. Based on the measured feedwater pump shaft power and measured turbine inlet steam flow rate of each training sample, the actual isentropic efficiency is calculated. The actual isentropic efficiency of each training sample The calculation formula is:

[0079] in, Indicates the first The actual axis power corresponding to each training sample; Indicates the first The actual steam flow rate corresponding to each training sample; Indicates the first The isentropic enthalpy decrease corresponding to each training sample; S404. Calculate the true efficiency correction factor for each training sample, the first... True efficiency correction factor for each training sample The calculation formula is:

[0080] Indicates the first The theoretical isentropic efficiency of each training sample; S405. Using the historical real-time running parameters and corresponding theoretical parameters of each training sample as input features, and the corresponding real efficiency correction factor as the training target label, supervise the training of the machine learning model to obtain a pre-trained machine learning correction model.

[0081] By specifying in detail the training steps of the machine learning calibration model, including collecting historical data, calculating theoretical parameters through the mechanistic model, back-calculating the actual efficiency using measured shaft power and steam flow rate and calculating the true efficiency correction factor, and finally conducting supervised training with theoretical parameters and operating parameters as features and the true correction factor as labels, a complete, closed-loop, and physically meaningful model construction methodology is provided. This ensures that the learning objective of the trained machine learning model is a reliable label obtained through independent measured data and thermodynamic principles, thereby guaranteeing the effectiveness of the model calibration capability.

[0082] In some specific embodiments, in step S405, the objective of supervised training is to minimize the objective function, the expression of which is:

[0083] in, Indicates the first The true efficiency correction factor for each training sample; Indicates the machine learning correction model for the first The predicted output value for each training sample; This represents the total number of training samples; This is the loss function, used to measure the difference between the predicted values ​​of the training samples and the training target labels; This is a regularization term used to control the... Sub-model The complexity is reduced to prevent overfitting.

[0084] By explicitly defining the goal of supervised training as minimizing a comprehensive objective function that combines a loss function that measures the difference between the predicted value and the label with a regularization term that controls the model complexity, a mechanism to prevent overfitting is introduced during the model training phase. By balancing fitting accuracy and model complexity in the optimization objective, the trained model not only has a good fit to the training data, but also has a good generalization prediction ability for new data.

[0085] In some specific embodiments, the loss function Specifically, it refers to the mean square error function.

[0086] By further specifying the loss function used in training as the mean squared error function, a standard and mathematically sound error metric in regression problems is adopted. Its convexity and differentiability facilitate the implementation of optimization algorithms, and its characteristic of imposing greater penalties on larger errors helps to drive the model training process to more effectively reduce significant prediction bias.

[0087] In some specific embodiments, step S4 further includes an online learning and updating step: periodically using newly generated historical real-time operating parameters, corresponding measured feedwater pump shaft power, and measured turbine inlet steam flow to construct new training samples, and using the new training samples to update the model parameters of the machine learning correction model in an incremental learning or periodic full retraining manner.

[0088] By adding an online learning and update step for the model, new samples are built regularly using newly generated data, and the model parameters are updated through incremental learning or periodic full retraining. This gives the machine learning correction module the ability to continuously self-optimize and adapt to changes, enabling it to follow the long-term gradual trend of device performance or the evolution of operating modes, thereby maintaining the long-term correction accuracy of the model throughout the entire device lifecycle.

[0089] Step S5: Calibrate the theoretical isentropic efficiency using an efficiency correction factor to obtain the actual isentropic efficiency. The expression is:

[0090] This represents the theoretical isentropic efficiency.

[0091] By utilizing the efficiency correction factor output by the machine learning correction model, the theoretical isentropic efficiency calculated by the mechanistic model is directly calibrated through product, thereby obtaining the actual isentropic efficiency that reflects the current true performance level of the equipment. This achieves dynamic and online correction of the core performance parameter (i.e., isentropic efficiency) of the mechanistic model, and adjusts the design efficiency value representing the ideal state to the actual value that matches the current health state of the equipment in real time and adaptively. This fundamentally solves the problem of long-term calculation accuracy decline caused by the inability of static mechanistic models to reflect the time-varying degradation of equipment performance.

[0092] Step S6: Based on the actual isentropic efficiency, calculate and output high-precision values ​​of turbine inlet steam flow rate and turbine exhaust steam enthalpy through a thermodynamic mechanism model.

[0093] By substituting the actual isentropic efficiency obtained after calibration back into the relevant calculation relationship defined by the thermodynamic mechanism model, the final steam inlet flow rate and steam exhaust enthalpy of the turbine are calculated and output. This achieves the final back calculation using the rigorous physical formula of the mechanism model after obtaining accurate actual efficiency parameters calibrated by data-driven method. This ensures that the output results inherit the adaptive ability of the data-driven model to the equipment state and strictly follow the basic laws of thermodynamics, thereby comprehensively improving the accuracy, reliability and physical consistency of the back calculation results.

[0094] In some specific embodiments, step S7 is also included: continuously recording the efficiency correction factor of the machine learning correction model output. Analyze its 30-day moving average offset. ,when When the preset warning threshold is exceeded, a performance degradation report and maintenance warning are generated and sent to staff.

[0095] By adding steps such as continuously recording the efficiency correction factor and analyzing its 30-day moving average deviation rate, and generating performance degradation reports and maintenance warnings when the deviation rate exceeds a preset warning threshold, the efficiency correction factor calculated and output by the model is elevated to a core indicator for monitoring the health status of equipment performance. Furthermore, by smoothing short-term fluctuations through moving average processing to capture long-term deterioration trends, automatic and quantitative monitoring and early warning of slow equipment performance degradation are achieved.

[0096] In some specific embodiments, the warning threshold is set to -5%.

[0097] By specifying the threshold value for performance degradation warning as -5%, a specific and clear numerical standard is provided for the warning triggering condition, making the generation of performance degradation warnings have a clear, consistent and operable judgment basis, thus enhancing the practicality and manageability of the warning system.

[0098] The following are embodiments of the steam-driven feedwater pump state modeling and inverse calculation system based on thermodynamic mechanisms and machine learning provided in this application. This steam-driven feedwater pump state modeling and inverse calculation system based on thermodynamic mechanisms and machine learning belongs to the same inventive concept as the steam-driven feedwater pump state modeling and inverse calculation methods in the above embodiments. For details not described in detail in the embodiments of the steam-driven feedwater pump state modeling and inverse calculation system, please refer to the embodiments of the steam-driven feedwater pump state modeling and inverse calculation methods based on thermodynamic mechanisms and machine learning.

[0099] like Figure 2 As shown, the steam-driven feedwater pump state modeling and inlet steam flow back calculation system based on thermodynamic mechanisms and machine learning includes: The thermodynamic mechanism model building module is used to build a thermodynamic mechanism model of the feedwater pump set based on the first law of thermodynamics and the equipment design efficiency curve. The real-time operating parameter acquisition module is used to acquire the real-time operating parameters of the steam-driven feedwater pump set, including feedwater flow rate, feedwater pump inlet pressure, feedwater pump outlet pressure, feedwater density, turbine inlet steam enthalpy, and condenser pressure. The theoretical parameter calculation module is used to calculate theoretical parameters based on real-time operating parameters using a thermodynamic mechanism model. The efficiency correction factor generation module is used to construct an input feature vector by combining real-time operating parameters and theoretical parameters, input it into a pre-trained machine learning correction model, and output an efficiency correction factor that characterizes the degree of deviation of the current performance of the device from its design performance. The theoretical isentropic efficiency calibration module is used to calibrate the theoretical isentropic efficiency using an efficiency correction factor to obtain the actual isentropic efficiency. The high-precision value calculation module is used to calculate and output high-precision values ​​of turbine inlet steam flow and turbine exhaust steam enthalpy based on actual isentropic efficiency and through a thermodynamic mechanism model.

[0100] The steam-driven feedwater pump state modeling and inverse steam flow calculation system in this embodiment is used to implement a method for steam-driven feedwater pump state modeling and inverse steam flow calculation based on thermodynamic mechanisms and machine learning.

[0101] This application also provides an electronic device for implementing the various embodiments of this application. Figure 3 To illustrate the hardware structure of an electronic device according to various embodiments of this application, as shown in the following diagram... Figure 3 As shown, the electronic device includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor.

[0102] Those skilled in the art will understand that the electronic device structure involved in the embodiments of this application does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0103] In embodiments of this application, electronic devices include, but are not limited to, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of this application described and / or claimed herein.

[0104] In this application embodiment, the processor can be implemented using at least one of an Application-Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Device (DSPD), a processor, a controller, a microcontroller, a microprocessor, or an electronic unit designed to perform the functions described herein. In some cases, such implementations can be implemented within a controller. For software implementations, implementations such as processes or functions can be implemented with separate software modules that allow the performance of at least one function or operation. The software code can be implemented by a software application (or program) written in any suitable programming language, and the software code can be stored in memory and executed by the controller.

[0105] In addition, the electronic device includes some functional modules not shown, which will not be described in detail here.

[0106] Those skilled in the art will understand that the various aspects of the electronic device provided in this application can be implemented as a system, method, or program product. Therefore, the various aspects of this application can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software aspects, collectively referred to herein as a "circuit," "module," or "system."

[0107] This application also provides a storage medium storing a program product capable of implementing a method for state modeling and inverse calculation of steam flow rate of a steam-driven feedwater pump based on thermodynamic mechanisms and machine learning. In some possible implementations, various aspects of this application can also be implemented as a program product comprising program code that, when run on a terminal device, causes the terminal device to perform the steps described in the "Exemplary Methods" section of this specification according to various exemplary embodiments of this application.

[0108] The storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example,, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0109] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for state modeling and inverse calculation of steam inlet flow rate of a steam-driven feedwater pump based on thermodynamic mechanisms and machine learning, characterized in that, include: S1. A thermodynamic mechanism model of the feedwater pump set is constructed based on the first law of thermodynamics and the equipment design efficiency curve. The thermodynamic mechanism model defines the relationship between the following parameters: The calculation relationship between water supply flow rate, water supply pump inlet pressure, water supply pump outlet pressure, water supply density, and water supply pump shaft power; The mapping relationship between the shaft power of the water pump and the theoretical isentropic efficiency; Calculation relationship between turbine inlet steam enthalpy, condenser pressure and isentropic enthalpy drop of steam; The calculation relationship between feedwater pump shaft power, isentropic enthalpy drop, isentropic efficiency and turbine inlet steam flow rate and turbine exhaust steam enthalpy; S2. Obtain real-time operating parameters of the steam-driven feedwater pump set, including feedwater flow rate, feedwater pump inlet pressure, feedwater pump outlet pressure, feedwater density, turbine inlet steam enthalpy, and condenser pressure; S3. Based on real-time operating parameters, theoretical parameters are calculated using a thermodynamic mechanism model, including theoretical feedwater pump shaft power, theoretical isentropic enthalpy drop, theoretical isentropic efficiency, theoretical turbine inlet steam flow rate, and theoretical turbine exhaust steam enthalpy. S4. Construct an input feature vector using real-time operating parameters and theoretical parameters, then input it into a pre-trained machine learning correction model. The output is an efficiency correction factor that characterizes the degree of deviation of the device's current performance from its design performance. The machine learning correction model is a regression model built based on the gradient boosting tree algorithm; S5. The theoretical isentropic efficiency is calibrated using an efficiency correction factor to obtain the actual isentropic efficiency. The expression is: Indicates theoretical isentropic efficiency; S6. Based on the actual isentropic efficiency, high-precision values ​​of turbine inlet steam flow rate and turbine exhaust steam enthalpy are calculated and output through a thermodynamic mechanism model.

2. The method for state modeling and inverse calculation of steam flow rate of a steam-driven feedwater pump as described in claim 1, characterized in that, In step S1, the calculation relationship between feedwater flow rate, feedwater pump inlet pressure, feedwater pump outlet pressure, feedwater density, and feedwater pump shaft power is defined by the following formula: in, Indicates the power of the water pump shaft; Indicates water supply flow rate; Indicates the outlet pressure of the water pump; Indicates the inlet pressure of the water pump; Indicates the density of the water supply; This indicates the efficiency of the water supply pump, which is a preset constant value or can be obtained by querying the pump's own performance curve based on its operating conditions.

3. The method for state modeling and inverse calculation of steam flow rate of a steam-driven feedwater pump as described in claim 1, characterized in that, In step S1, the calculated relationship between the turbine inlet enthalpy, condenser pressure, and isentropic enthalpy drop of the steam is defined by the following formula: in, This represents the isentropic enthalpy drop of steam. Indicates the enthalpy of the steam entering the turbine; This represents the isentropic exhaust enthalpy, based on the condenser pressure. Obtained by consulting the table of thermodynamic properties of water vapor.

4. The method for state modeling and inverse calculation of steam flow rate of a steam-driven feedwater pump as described in claim 1, characterized in that, In step S1, the calculation relationship between the feedwater pump shaft power, isentropic enthalpy drop, isentropic efficiency, and turbine inlet steam flow rate and turbine exhaust steam enthalpy is defined by the following formula: in, Indicates the steam inlet flow rate of the steam turbine; Indicates the power of the water pump shaft; This indicates isentropic enthalpy drop; Indicates isentropic efficiency; Indicates the exhaust enthalpy of the steam turbine; This indicates the enthalpy of the steam entering the turbine.

5. The method for state modeling and inverse calculation of steam flow rate of a steam-driven feedwater pump as described in claim 1, characterized in that, In step S4, the machine learning correction model is based on the gradient boosting tree algorithm, which performs regression prediction by integrating multiple decision trees. Each decision tree performs a non-linear transformation on the input feature vector according to feature splitting rules, and the model ultimately outputs an efficiency correction factor. The core formula for the weighted sum of the outputs of all decision trees is expressed as: in, This represents the input feature vector; Indicates the first Each decision tree is used for the input feature vector. The predicted output; Indicates the first The weight coefficients of each decision tree are obtained by minimizing the loss function during training; This represents the total number of decision trees.

6. The method for state modeling and inverse calculation of steam flow rate of a steam-driven feedwater pump as described in claim 1, characterized in that, In step S4, the training steps for the machine learning correction model are as follows: S401. Collect historical datasets, which contain historical real-time operating parameters of multiple training samples, corresponding measured feedwater pump shaft power and measured steam turbine inlet flow rate; S402. Based on the historical real-time operating parameters of each training sample, the corresponding theoretical parameters are calculated through a thermodynamic mechanism model, including theoretical feedwater pump shaft power, theoretical isentropic enthalpy drop, theoretical isentropic efficiency, theoretical turbine inlet steam flow rate, and theoretical turbine exhaust steam enthalpy. S403. Based on the measured feedwater pump shaft power and measured turbine inlet steam flow rate of each training sample, the actual isentropic efficiency is calculated. The actual isentropic efficiency of each training sample The calculation formula is: in, Indicates the first The actual axis power corresponding to each training sample; Indicates the first The actual steam flow rate corresponding to each training sample; Indicates the first The isentropic enthalpy decrease corresponding to each training sample; S404. Calculate the true efficiency correction factor for each training sample, the first... True efficiency correction factor for each training sample The calculation formula is: Indicates the first The theoretical isentropic efficiency of each training sample; S405. Using the historical real-time running parameters and corresponding theoretical parameters of each training sample as input features, and the corresponding real efficiency correction factor as the training target label, supervise the training of the machine learning model to obtain a pre-trained machine learning correction model.

7. The method for state modeling and inverse calculation of steam flow rate of a steam-driven feedwater pump as described in claim 6, characterized in that, In step S405, the objective of supervised training is to minimize the objective function, which is expressed as: in, Indicates the first The true efficiency correction factor for each training sample; Indicates the machine learning correction model for the first The predicted output value for each training sample; This represents the total number of training samples; This is the loss function, used to measure the difference between the predicted values ​​of the training samples and the training target labels; This is a regularization term used to control the... Sub-model The complexity is reduced to prevent overfitting.

8. A system for state modeling and inverse calculation of steam flow rate of a steam-driven feedwater pump based on thermodynamic mechanisms and machine learning, characterized in that, The method for implementing the steam-driven feedwater pump state modeling and inverse calculation of steam flow rate as described in any one of claims 1-7 includes: Thermodynamic mechanism model building module is used to build a thermodynamic mechanism model of the feedwater pump set based on the first law of thermodynamics and the equipment design efficiency curve; The real-time operating parameter acquisition module is used to acquire the real-time operating parameters of the steam-driven feedwater pump set, including feedwater flow rate, feedwater pump inlet pressure, feedwater pump outlet pressure, feedwater density, turbine inlet steam enthalpy, and condenser pressure. The theoretical parameter calculation module is used to calculate theoretical parameters based on real-time operating parameters using a thermodynamic mechanism model. The efficiency correction factor generation module is used to construct an input feature vector by combining real-time operating parameters and theoretical parameters, input it into a pre-trained machine learning correction model, and output an efficiency correction factor that characterizes the degree of deviation of the current performance of the device from its design performance. The theoretical isentropic efficiency calibration module is used to calibrate the theoretical isentropic efficiency using an efficiency correction factor to obtain the actual isentropic efficiency. The high-precision value calculation module is used to calculate and output high-precision values ​​of turbine inlet steam flow and turbine exhaust steam enthalpy based on actual isentropic efficiency and through a thermodynamic mechanism model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes a computer program, it implements the steps of the steam-driven feedwater pump state modeling and steam inlet flow rate inverse calculation method as described in any one of claims 1-7.

10. A storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the steam-driven feedwater pump state modeling and steam inlet flow rate inverse calculation method as described in any one of claims 1-7.

Images