Steam turbine high-pressure regulating valve nonlinear flow modeling method and device based on multiple complex factors and storage medium

Abstract

The invention discloses a turbine high-pressure control valve nonlinear flow modeling method and device based on various complex factors and a storage medium, and relates to the field of thermal equipment modeling and emulation.The method achieves precise modeling through four steps that firstly, seven kinds of core factors such as the valve position opening degree, the steam temperature and pressure, the viscosity, the flow speed and the accumulative operation duration are selected; collecting data through a distributed sensor and preprocessing the data through weighting standardization; secondly, three correction factors such as an in-valve flow field coupling factor are introduced, and a nonlinear core flow model is constructed; real-time error compensation is realized in combination with an Elman neural network and an improved Sigmoid function, and a final model is formed; and finally, verifying the precision through MAE and RMSE, and optimizing key parameters by adopting a particle swarm algorithm. According to the method, the multi-factor coupling effect is comprehensively described, the data reliability and the model prediction precision are improved, core support is provided for fine control and energy efficiency optimization of the high-pressure regulating valve of the steam turbine, and the method has important engineering value.

Description

Technical Field

[0001] This invention relates to the field of thermal equipment modeling and simulation, specifically to a method, equipment, and storage medium for modeling the nonlinear flow of a steam turbine high-pressure regulating valve based on multiple complex factors. Background Technology

[0002] As the core power equipment in energy systems such as thermal power and nuclear power, the operating efficiency and stability of steam turbines directly determine the level of energy utilization and system safety. The high-pressure regulating valve, as a key component for regulating steam flow and pressure in steam turbines, requires accurate modeling of its flow characteristics as a crucial prerequisite for achieving refined unit control, energy efficiency optimization, and condition monitoring. During the variable operating conditions of steam turbines, the flow rate of the high-pressure regulating valve is affected by multiple complex factors, including valve opening, steam thermodynamic parameters, fluid physical properties, and equipment aging status, exhibiting strong nonlinearity, time-varying characteristics, and coupling. This places stringent demands on the comprehensiveness and accuracy of flow modeling.

[0003] Existing methods for modeling the flow rate of high-pressure regulating valves in steam turbines are mostly based on traditional physical mechanisms, with a focus on constructing simplified models around valve opening and inlet / outlet pressure. These methods generally suffer from the following limitations: First, they consider influencing factors in a one-sided manner. Most models only focus on basic parameters such as valve opening and inlet / outlet pressure, neglecting the coupling effect of steam viscosity and flow velocity, the interaction of temperature and pressure, and the aging and degradation of valves after long-term operation. This results in models that cannot adapt to flow rate changes under complex operating conditions, leading to insufficient fitting accuracy. Second, data preprocessing methods are simplistic. Traditional methods often employ single linear normalization or mean-standard deviation normalization, making it difficult to balance the need for dimension elimination and preservation of the original data distribution. This easily leads to data distortion, affecting the reliability of subsequent modeling. Third, error compensation mechanisms are lacking. Pure physical mechanism models cannot cover random errors and system residual errors under complex operating conditions, resulting in significant flow rate prediction deviations. Fourth, parameter optimization methods are inefficient, relying heavily on empirical assignment or simple iterative algorithms. This fails to achieve globally optimal matching of key parameters, resulting in poor robustness and significant accuracy fluctuations under various operating conditions.

[0004] As the energy industry continues to demand higher efficiency and control precision from steam turbines, traditional modeling methods are increasingly inadequate to meet the practical needs of deep peak shaving and energy conservation. For example, under complex operating conditions such as variable load operation and start-up / shutdown processes, the prediction errors of existing models often exceed the engineering allowable range, leading to delayed response and insufficient control precision of high-pressure control valves, which in turn affects the operational stability and energy utilization efficiency of the steam turbine. Therefore, developing a high-pressure control valve nonlinear flow modeling method that can comprehensively characterize the coupling effects of multiple factors, optimize data preprocessing processes, possess dynamic error compensation capabilities, and have an efficient parameter optimization mechanism has become an urgent need to solve current technical bottlenecks and promote the upgrading of steam turbine control technology. This method has significant engineering value and practical significance for improving the overall operational level of energy systems. Summary of the Invention

[0005] The purpose of this invention is to provide a method, device, and storage medium for nonlinear flow modeling of turbine high-pressure control valves based on multiple complex factors. This method involves collecting and preprocessing data on core influencing factors such as valve opening degree and steam temperature and pressure using a distributed sensor array. Three correction factors, including the valve internal flow field coupling factor, are introduced to construct a nonlinear core flow model. Combined with real-time error compensation using an Elman neural network, and iterative optimization of the nonlinear correction index and standardized weight coefficients using a particle swarm optimization algorithm, the nonlinear mapping relationship between the turbine high-pressure control valve flow and complex factors is ultimately accurately characterized.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for modeling the nonlinear flow of a turbine high-pressure regulating valve based on multiple complex factors, characterized by the following steps:

[0008] S1: Valve position opening, inlet steam pressure and temperature, outlet steam pressure, steam viscosity, inlet steam velocity and cumulative running time are selected as core influencing factors. Real-time data are collected synchronously through a distributed sensor array and preprocessed.

[0009] S2: Using preprocessed data as input, introduce the valve in-valve flow field coupling factor that characterizes the coupling effect of viscosity and flow velocity, the temperature and pressure synergistic correction factor that corrects the temperature and pressure interaction, and the aging decay factor that quantifies valve aging decay. Combine the inherent flow coefficient and the nonlinear correction index of the opening to construct a nonlinear core flow model.

[0010] S3: Based on the construction of a nonlinear core flow model, a real-time error compensation term calculated by an Elman neural network is introduced. The preprocessed data and core flow value are taken as inputs, and the difference between the measured flow value and the core value is taken as the output. An improved Sigmoid function is used as the activation function to construct the final flow model.

[0011] S4: Select measured data under multiple working conditions, and verify the model accuracy through mean absolute error and root mean square error. If the error value exceeds the set value, the model parameters are iteratively updated using the particle swarm algorithm with the goal of minimizing the root mean square error, until the error reaches the standard and the final model is obtained.

[0012] Step S1 specifically involves selecting valve opening degree, inlet steam pressure, inlet steam temperature, outlet steam pressure, steam viscosity, inlet steam velocity, and cumulative operating time as core influencing factors, and using a distributed sensor array to synchronously collect real-time data at a fixed frequency. The collected data is first processed by removing outliers using a 3σ adaptive threshold, and then mapped to the [0,1] interval using a standardized formula, as follows:

[0013] ;

[0014] In the formula, For the first The original data collected from the core influencing factors; For the first Data of the core influencing factors after standardization; For the first The historical minimum values ​​of the core influencing factors; For the first The historical maximum values ​​of the core influencing factors; For the first Standardized weight coefficients of the core influencing factors; For the first Historical average values ​​of the core influencing factors; For the first The historical standard deviation of the core influencing factors.

[0015] The distributed sensor array configuration in step S1 is as follows: a pressure sensor and a temperature sensor are installed on the pipeline before the valve at 1 / 3 of the distance from the valve interface; a pressure sensor is installed on the pipeline after the valve at 1 / 3 of the distance from the valve interface; a viscosity sensor and a flow velocity sensor are installed in the stable flow field region on the inner wall of the valve body; and the valve controller integrates an opening sensor and a timing module.

[0016] Step S2 specifically involves using the preprocessed data from S1 as input and introducing an in-valve flow field coupling factor. Temperature and pressure synergistic correction factor and aging degradation factor Constructing a nonlinear core model as follows:

[0017] ;

[0018] In the formula, This represents the non-linear core flow value. This is the inherent flow coefficient; Valve position opening; This is a nonlinear correction index for the opening degree; The pressure is the steam pressure before the valve. This refers to the steam pressure after the valve. The density of the vapor; The coupling factor of the flow field inside the valve; This is a temperature and pressure co-correction factor; It is an aging degradation factor.

[0019] The The formula used to characterize the coupling effect between viscosity and flow velocity is as follows:

[0020] ;

[0021] In the formula, The coupling factor of the flow field inside the valve; Vapor viscosity; The steam flow rate inside the valve; The pressure is the steam pressure before the valve. This refers to the standardized valve opening. This is the coupling effect adjustment coefficient; The reference steam flow rate; Pi; It is a natural constant;

[0022] The The formula used to correct for the interaction between temperature and pressure is as follows:

[0023] ;

[0024] In the formula, This is a temperature and pressure co-correction factor; The temperature of the steam before the valve; The reference steam temperature; Temperature correction index; The pressure is the steam pressure before the valve. The reference steam pressure; This is a pressure correction index; This refers to the steam pressure after the valve. Pi;

[0025] The aging degradation factor The formula used to quantify the aging degradation effect of valves is as follows:

[0026] ;

[0027] In the formula, It is an aging degradation factor; It is a natural constant; The main coefficient for aging degradation; Cumulative runtime; This represents the average valve opening. This is the aging degradation correction factor; The baseline runtime.

[0028] Based on the construction of a nonlinear core flow model, a real-time error compensation term calculated by an Elman neural network is introduced. Taking preprocessed data and core flow values ​​as input, and the difference between the measured flow value and the core value as output, an improved Sigmoid function is used as the activation function. The final model is as follows:

[0029] ;

[0030] In the formula, For the final model; This represents the non-linear core flow value. This is for real-time error compensation.

[0031] in The Elman neural network calculates the data using S1 preprocessed data and core traffic values. The input is the difference between the measured flow rate and the core value, and the output is the difference between the measured flow rate and the core value. There are 15-25 hidden layer nodes, and the activation function is a modified Sigmoid function, the formula of which is as follows:

[0032] ;

[0033] In the formula, To improve the output value of the Sigmoid function; The input value for the activation function; It is a natural constant; Adjust the parameters for the shape.

[0034] Step S4 specifically involves selecting multi-condition measured data and calculating the model output Q and the MAE and RMSE of the measured values; if or Then the particle swarm optimization algorithm is used for iterative updates. , The parameters are iterated 100-200 times with the goal of minimizing RMSE until the error reaches the target, and the final model is obtained.

[0035] The particle swarm optimization (PSO) algorithm parameters are: population size 30-50 particles, and inertia weight linearly decreasing from 0.9 to 0.4. Its particle velocity The upper limit is 20% of the parameter's value range, and its calculation formula is as follows:

[0036] ;

[0037] In the formula, For the first During the nth iteration, the 1st The first particle Speed ​​in each dimension; For the first During the nth iteration, the 1st The first particle Speed ​​in each dimension; The inertial weight decreases linearly from 0.9 to 0.4. The acceleration factor is fixed at 2.0; A random number within the interval [0,1]; For the first During the nth iteration, the 1st The first particle The optimal position of an individual in each dimension; For the first During the nth iteration, the 1st The first particle Current position in each dimension; For the first In the nth iteration, the entire particle swarm... The global optimal position in each dimension.

[0038] A computer device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, cause the at least one processor to perform the method according to any one of claims 1 to 7.

[0039] A non-transitory computer-readable storage medium storing computer instructions that, when executed by at least one processor, cause the at least one processor to perform the method as described in any one of claims 1 to 7.

[0040] This nonlinear flow modeling method for the high-pressure regulating valve of a steam turbine accurately depicts the nonlinear mapping relationship between the high-pressure regulating valve flow and complex influencing factors, as detailed below:

[0041] First, a distributed sensor array is used to achieve comprehensive and synchronous perception of key influencing factors: sensors are configured in a specific spatial layout for key parameters such as valve opening, inlet / outlet steam pressure, inlet steam temperature, steam viscosity, inlet steam flow rate, and cumulative running time to ensure the relevance and timeliness of data acquisition; after outliers are removed by 3σ adaptive thresholding, the collected data is mapped to the [0,1] interval by a weighted normalization formula that integrates linear normalization and mean standard deviation normalization, which not only eliminates the difference in dimensions, but also balances the influence of different normalization methods through weight coefficients, providing high-quality input data for subsequent modeling.

[0042] Secondly, a nonlinear core flow model with coupled multiple correction factors is constructed: based on preprocessed data, it breaks through the limitations of traditional models that only consider basic parameters, and introduces three key factors, including the valve in-valve flow field coupling factor ξ, which quantifies the synergistic effect of viscosity and flow velocity by integrating the nonlinear relationship between steam viscosity, flow velocity, inlet valve pressure and valve opening; the temperature and pressure synergistic correction factor η, which corrects the nonlinear influence of temperature and pressure interaction on flow rate by the ratio exponent of inlet valve temperature and pressure to the baseline value and the sine function term of inlet and outlet valve pressure; and the aging attenuation factor ζ, which accurately characterizes the performance degradation effect of valves after long-term operation by combining the exponential and logarithmic relationship between cumulative running time and average valve opening. The core flow model is constructed to achieve a quantitative characterization of the main nonlinear factors.

[0043] Furthermore, an Elman neural network is introduced to achieve real-time error compensation: considering that the core model is difficult to fully cover the residual errors under complex working conditions, the preprocessed data and core flow values ​​are used as inputs, and the difference between the measured flow value and the core value is used as the output. An improved Sigmoid function is used as the activation function to construct an error compensation network ΔQ, which dynamically offsets the systematic and random errors of the core model, forming a final flow model that takes into account both physical mechanisms and data-driven approaches.

[0044] Finally, measured data under different operating conditions were selected, and the mean absolute error and root mean square error were used as evaluation indicators. If the error exceeded the set threshold, the key parameters of the nonlinear correction of the opening were updated iteratively by the particle swarm optimization algorithm with the goal of minimizing the RMSE. After 100 to 200 iterations, the error was brought to the standard, and finally a flow model with high accuracy under multiple operating conditions was obtained, which provides core support for the precise control and condition monitoring of the turbine high-pressure regulating valve.

[0045] Compared with the prior art, the beneficial effects of the present invention are:

[0046] 1. This invention selects seven core influencing factors and introduces valve internal flow field coupling factor, temperature and pressure synergistic correction factor and aging attenuation factor to accurately quantify the nonlinear coupling relationship between multiple factors, thereby achieving a panoramic characterization of the influence mechanism of high-speed valve flow and greatly improving the model's adaptability to complex operating conditions.

[0047] 2. This invention proposes a weighted standardization method that integrates linear normalization and mean-standard deviation normalization. By balancing the advantages of the two methods through the standardization weight coefficients, and combining it with 3σ adaptive threshold to remove outliers, it not only eliminates the dimensional differences of different influencing factors, but also preserves the original distribution characteristics of the data. This provides highly reliable and consistent input data for subsequent modeling, laying the foundation for high-precision models.

[0048] 3. This invention innovatively introduces an Elman neural network to construct a real-time error compensation term, using an improved Sigmoid function as the activation function. With preprocessed data and core traffic values ​​as input, it dynamically offsets the errors of the core model, significantly reducing the residual error of traffic prediction and improving the prediction accuracy and stability of the model.

[0049] 4. This invention aims to minimize the root mean square error. It uses a particle swarm optimization algorithm to iteratively update the nonlinear correction exponent of the opening and the standardized weight coefficient ᵢ. By linearly decreasing the inertia weight and reasonably setting the acceleration coefficient and population size, it achieves efficient global search of parameters, ensuring that the model can reach the error standard under different working conditions, and significantly improving the robustness and engineering applicability of the model. Attached Figure Description

[0050] Figure 1 This is a flowchart of a nonlinear flow modeling method for a turbine high-pressure regulating valve based on multiple complex factors, according to the present invention. Detailed Implementation

[0051] The technical solutions of the present invention will now be described in detail with reference to the accompanying drawings.

[0052] like Figure 1 As shown, S1: Valve position opening, inlet steam pressure and temperature, outlet steam pressure, steam viscosity, inlet steam velocity and cumulative running time are selected as core influencing factors. Real-time data are collected synchronously through a distributed sensor array and preprocessed.

[0053] Step S1 specifically involves selecting valve opening degree, inlet steam pressure, inlet steam temperature, outlet steam pressure, steam viscosity, inlet steam velocity, and cumulative operating time as core influencing factors, and using a distributed sensor array to synchronously collect real-time data at a fixed frequency. The collected data is first processed by removing outliers using a 3σ adaptive threshold, and then mapped to the [0,1] interval using a standardized formula, as follows:

[0054] ;

[0055] In the formula, For the first The original data collected from the core influencing factors; For the first Data of the core influencing factors after standardization; For the first The historical minimum values ​​of the core influencing factors; For the first The historical maximum values ​​of the core influencing factors; For the first Standardized weight coefficients of the core influencing factors; For the first Historical average values ​​of the core influencing factors; For the first The historical standard deviation of the core influencing factors.

[0056] The distributed sensor array configuration in step S1 is as follows: a pressure sensor and a temperature sensor are installed on the pipeline before the valve at 1 / 3 of the distance from the valve interface; a pressure sensor is installed on the pipeline after the valve at 1 / 3 of the distance from the valve interface; a viscosity sensor and a flow velocity sensor are installed in the stable flow field region on the inner wall of the valve body; and the valve controller integrates an opening sensor and a timing module.

[0057] S2: Using preprocessed data as input, introduce the valve in-valve flow field coupling factor that characterizes the coupling effect of viscosity and flow velocity, the temperature and pressure synergistic correction factor that corrects the temperature and pressure interaction, and the aging decay factor that quantifies valve aging decay. Combine the inherent flow coefficient and the nonlinear correction index of the opening to construct a nonlinear core flow model.

[0058] Constructing a nonlinear core model as follows:

[0059] ;

[0060] In the formula, This represents the non-linear core flow value. This is the inherent flow coefficient; Valve position opening; This is a nonlinear correction index for the opening degree; The pressure is the steam pressure before the valve. This refers to the steam pressure after the valve. The density of the vapor; The coupling factor of the flow field inside the valve; This is a temperature and pressure co-correction factor; It is an aging degradation factor.

[0061] The The formula used to characterize the coupling effect between viscosity and flow velocity is as follows:

[0062] ;

[0063] In the formula, The coupling factor of the flow field inside the valve; Vapor viscosity; The steam flow rate inside the valve; The pressure is the steam pressure before the valve. This refers to the standardized valve opening. This is the coupling effect adjustment coefficient; The reference steam flow rate; Pi; It is a natural constant;

[0064] The The formula used to correct for the interaction between temperature and pressure is as follows:

[0065] ;

[0066] In the formula, This is a temperature and pressure co-correction factor; The temperature of the steam before the valve; The reference steam temperature; Temperature correction index; The pressure is the steam pressure before the valve. The reference steam pressure; This is a pressure correction index; This refers to the steam pressure after the valve. Pi;

[0067] The aging degradation factor The formula used to quantify the aging degradation effect of valves is as follows:

[0068] ;

[0069] In the formula, It is an aging degradation factor; It is a natural constant; The main coefficient for aging degradation; Cumulative runtime; This represents the average valve opening. This is the aging degradation correction factor; The baseline runtime.

[0070] S3: Based on the construction of a nonlinear core flow model, a real-time error compensation term calculated by an Elman neural network is introduced. The preprocessed data and core flow value are taken as inputs, and the difference between the measured flow value and the core value is taken as the output. An improved Sigmoid function is used as the activation function to construct the final flow model.

[0071] The final model is constructed as follows:

[0072] ;

[0073] In the formula, For the final model; This represents the non-linear core flow value. This is for real-time error compensation.

[0074] in The Elman neural network calculates the data using S1 preprocessed data and core traffic values. The input is the difference between the measured flow rate and the core value, and the output is the difference between the measured flow rate and the core value. There are 15-25 hidden layer nodes, and the activation function is a modified Sigmoid function, the formula of which is as follows:

[0075] ;

[0076] In the formula, To improve the output value of the Sigmoid function; The input value for the activation function; It is a natural constant; Adjust the parameters for the shape.

[0077] S4: Select measured data under multiple working conditions, and verify the model accuracy through mean absolute error and root mean square error. If the error value exceeds the set value, the model parameters are iteratively updated using the particle swarm algorithm with the goal of minimizing the root mean square error, until the error reaches the standard and the final model is obtained.

[0078] Step S4 specifically involves selecting multi-condition measured data and calculating the model output Q and the MAE and RMSE of the measured values; if or Then the particle swarm optimization algorithm is used for iterative updates. , The parameters are iterated 100-200 times with the goal of minimizing RMSE until the error reaches the target, and the final model is obtained.

[0079] The particle swarm optimization (PSO) algorithm parameters are: population size 30-50 particles, and inertia weight linearly decreasing from 0.9 to 0.4. Its particle velocity The upper limit is 20% of the parameter's value range, and its calculation formula is as follows:

[0080] ;

[0081] In the formula, For the first During the nth iteration, the 1st The first particle Speed ​​in each dimension; For the first During the nth iteration, the 1st The first particle Speed ​​in each dimension; The inertial weight decreases linearly from 0.9 to 0.4. The acceleration factor is fixed at 2.0; A random number within the interval [0,1]; For the first During the nth iteration, the 1st The first particle The optimal position of an individual in each dimension; For the first During the nth iteration, the 1st The first particle Current position in each dimension; For the first In the nth iteration, the entire particle swarm... The global optimal position in each dimension.

Claims

1. A method for modeling the nonlinear flow of a turbine high-pressure regulating valve based on multiple complex factors, characterized in that, Includes the following steps: S1: Valve position opening, inlet steam pressure and temperature, outlet steam pressure, steam viscosity, inlet steam velocity and cumulative running time are selected as core influencing factors. Real-time data are collected synchronously through a distributed sensor array and preprocessed. S2: Using preprocessed data as input, introduce the valve in-valve flow field coupling factor that characterizes the coupling effect of viscosity and flow velocity, the temperature and pressure synergistic correction factor that corrects the temperature and pressure interaction, and the aging decay factor that quantifies valve aging decay. Combine the inherent flow coefficient and the nonlinear correction index of the opening to construct a nonlinear core flow model. S3: Based on the construction of a nonlinear core flow model, a real-time error compensation term calculated by an Elman neural network is introduced. The preprocessed data and core flow value are taken as inputs, and the difference between the measured flow value and the core value is taken as the output. An improved Sigmoid function is used as the activation function to construct the final flow model. S4: Select measured data under multiple working conditions, and verify the model accuracy through mean absolute error and root mean square error. If the error value exceeds the set value, the model parameters are iteratively updated using the particle swarm algorithm with the goal of minimizing the root mean square error, until the error reaches the standard and the final model is obtained.

2. The method for modeling nonlinear flow of a turbine high-pressure regulating valve based on multiple complex factors as described in claim 1, characterized in that, Step S1 specifically involves selecting valve opening degree, inlet steam pressure, inlet steam temperature, outlet steam pressure, steam viscosity, inlet steam velocity, and cumulative operating time as core influencing factors, and using a distributed sensor array to synchronously collect real-time data at a fixed frequency. The collected data is first processed by removing outliers using a 3σ adaptive threshold, and then mapped to the [0,1] interval using a standardized formula, as follows: ； In the formula, For the first The original data collected from the core influencing factors; For the first Data of the core influencing factors after standardization; For the first The historical minimum values ​​of the core influencing factors; For the first The historical maximum values ​​of the core influencing factors; For the first Standardized weight coefficients of the core influencing factors; For the first Historical average values ​​of the core influencing factors; For the first The historical standard deviation of the core influencing factors.

3. The method for modeling nonlinear flow of a turbine high-pressure regulating valve based on multiple complex factors as described in claim 2, characterized in that, The distributed sensor array configuration in step S1 is as follows: a pressure sensor and a temperature sensor are installed on the pipeline before the valve at 1 / 3 of the distance from the valve interface; a pressure sensor is installed on the pipeline after the valve at 1 / 3 of the distance from the valve interface; a viscosity sensor and a flow velocity sensor are installed in the stable flow field region on the inner wall of the valve body; and the valve controller integrates an opening sensor and a timing module.

4. The method for modeling nonlinear flow of a turbine high-pressure regulating valve based on multiple complex factors as described in claim 1, characterized in that, Step S2 specifically involves using the preprocessed data from S1 as input and introducing an in-valve flow field coupling factor. Temperature and pressure synergistic correction factor and aging degradation factor Constructing a nonlinear core model as follows: ； In the formula, This represents the non-linear core flow value. This is the inherent flow coefficient; Valve position opening; This is a nonlinear correction index for the opening degree; The pressure is the steam pressure before the valve. This refers to the steam pressure after the valve. The density of the vapor; The coupling factor of the flow field inside the valve; This is a temperature and pressure co-correction factor; It is an aging degradation factor.

5. The method for modeling nonlinear flow of a turbine high-pressure regulating valve based on multiple complex factors according to claim 4, characterized in that, The The formula used to characterize the coupling effect between viscosity and flow velocity is as follows: ； In the formula, The coupling factor of the flow field inside the valve; Vapor viscosity; The steam flow rate inside the valve; The pressure is the steam pressure before the valve. This refers to the standardized valve opening. This is the coupling effect adjustment coefficient; The reference steam flow rate; Pi; It is a natural constant; The The formula used to correct for the interaction between temperature and pressure is as follows: ； In the formula, This is a temperature and pressure co-correction factor; The temperature of the steam before the valve; The reference steam temperature; Temperature correction index; The pressure is the steam pressure before the valve. The reference steam pressure; This is a pressure correction index; This refers to the steam pressure after the valve. Pi; The aging degradation factor The formula used to quantify the aging degradation effect of valves is as follows: ； In the formula, It is an aging degradation factor; It is a natural constant; The main coefficient for aging degradation; Cumulative runtime; This represents the average valve opening. This is the aging degradation correction factor; The baseline runtime.

6. The method for modeling nonlinear flow of a turbine high-pressure regulating valve based on multiple complex factors as described in claim 1, characterized in that, Based on the construction of a nonlinear core flow model, a real-time error compensation term calculated by an Elman neural network is introduced. Taking preprocessed data and core flow values ​​as input, and the difference between the measured flow value and the core value as output, an improved Sigmoid function is used as the activation function. The final model is as follows: ； In the formula, For the final model; This represents the non-linear core flow value. This is for real-time error compensation. in The Elman neural network calculates the data using S1 preprocessed data and core traffic values. The input is the difference between the measured flow rate and the core value, and the output is the difference between the measured flow rate and the core value. There are 15-25 hidden layer nodes, and the activation function is a modified Sigmoid function, the formula of which is as follows: ； In the formula, To improve the output value of the Sigmoid function; The input value for the activation function; It is a natural constant; Adjust the parameters for the shape.

7. The method for modeling nonlinear flow of a turbine high-pressure regulating valve based on multiple complex factors as described in claim 1, characterized in that, Step S4 specifically involves selecting multi-condition measured data and calculating the model output Q and the MAE and RMSE of the measured values; if or The particle swarm optimization algorithm is then used to iteratively update the nonlinear correction exponent for the opening degree. Standardized weighting coefficients With the goal of minimizing RMSE, the model is iterated 100-200 times until the error reaches the target, and the final model is obtained.

8. The method for modeling nonlinear flow of a turbine high-pressure regulating valve based on multiple complex factors as described in claim 7, characterized in that, The particle swarm optimization (PSO) algorithm parameters are: population size 30-50 particles, and inertia weight linearly decreasing from 0.9 to 0.

4. Its particle velocity The upper limit is 20% of the parameter's value range, and its calculation formula is as follows: ； In the formula, For the first During the nth iteration, the 1st The first particle Speed ​​in each dimension; For the first During the nth iteration, the 1st The first particle Speed ​​in each dimension; The inertial weight decreases linearly from 0.9 to 0.

4. The acceleration factor is fixed at 2.0; A random number within the interval [0,1]; For the first During the nth iteration, the 1st The first particle The optimal position of an individual in each dimension; For the first During the nth iteration, the 1st The first particle Current position in each dimension; For the first In the nth iteration, the entire particle swarm... The global optimal position in each dimension.

9. A computer device, the device comprising: At least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, cause the at least one processor to perform the method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing computer instructions that, when executed by at least one processor, cause the at least one processor to perform the method as described in any one of claims 1 to 7.

Images