A method, apparatus, device, and storage medium for correcting measurement data of a thermal system.

Abstract

This invention discloses a method, apparatus, device, and storage medium for correcting measurement data of a thermal system, relating to the field of thermal system measurement and processing technology. The method includes the following steps: obtaining a simulation calculation model of the thermal system; constructing a known parameter correction model and an unknown parameter estimation model through a physical neural network; connecting the outputs of the known parameter correction model and the unknown parameter estimation model with the input of the simulation calculation model to obtain a fast correction model; inputting real-time measurement data and real-time operating data into the trained fast correction model to obtain the corresponding correction values ​​and unknown parameter estimates. Compared to classical data coordination methods, the method proposed in this invention effectively avoids the problems of slow calculation speed and poor iterative convergence, making it more suitable for the measurement parameter correction calculation of complex thermal systems in engineering application scenarios.

Description

Technical Field

[0001] This invention relates to the field of thermal system measurement and processing technology, and in particular to a method, apparatus, equipment and storage medium for correcting thermal system measurement data. Background Technology

[0002] For thermal systems in the energy and industrial sectors, operational economy and safety are crucial factors for maintenance personnel. Performance monitoring and operational control of thermal systems rely on the accurate measurement of various parameters within the system, such as temperature, pressure, and flow rate. However, due to limitations in the accuracy of measuring equipment, environmental interference, and instrument malfunctions, actual measurement data always contains unknown deviations, thus affecting the reliability of the thermal system analysis and calculation results based on them.

[0003] Currently, data reconciliation methods are commonly used to reduce errors in the original measurement data. This method applies the smallest possible correction to each measurement to satisfy all known physical constraints of the system, thus making the measurement data more consistent with the actual operating state of the system. However, many engineering applications involve highly complex thermal systems with numerous devices, intricate physical constraints, and high nonlinearity, making the application of data reconciliation methods difficult. These systems often suffer from slow algorithm convergence and poor iterative convergence, negatively impacting the general applicability of this approach. Summary of the Invention

[0004] Based on the shortcomings of the existing technology, the present invention provides a method, device, equipment and storage medium for correcting measurement data of thermal systems. It solves the problem that many thermal systems in engineering application scenarios are quite complex, with a large number of devices, complex physical constraints and high nonlinearity, which makes the application of data coordination methods difficult and often results in slow algorithm convergence.

[0005] The present invention adopts the following technical solution:

[0006] In a first aspect, the present invention provides a method for correcting measurement data of a thermal system, comprising the following steps:

[0007] Each device in the thermal system is modeled to obtain a corresponding simulation model; the heat balance equation corresponding to each simulation model is obtained, and the residuals of multiple heat balance equations are arranged into a vector. Based on this vector, the simulation calculation model of the thermal system is obtained; wherein, the parameters in the heat balance equation include known parameters whose specific data can be determined and unknown parameters whose specific data cannot be determined. The known parameters include the measurement parameters of the thermal system and the operating parameters of the thermal system.

[0008] A known parameter correction model and an unknown parameter estimation model are constructed using a physical neural network. The known parameter correction model takes known parameter values ​​as input and outputs the corresponding correction values. The unknown parameter estimation model takes known parameter values ​​as input and outputs the corresponding unknown parameter values.

[0009] By connecting the outputs of the known parameter correction model and the unknown parameter estimation model with the input of the simulation calculation model, a fast correction model is obtained.

[0010] Obtain multiple sets of historical known parameter values ​​and corresponding historical correction values ​​of the thermal system under different operating conditions, construct a dataset, and train the fast correction model using the dataset;

[0011] Real-time measurement data of thermal system parameters and real-time operational data of operating parameters are collected. The real-time measurement data and real-time operational data are then input into the trained fast calibration model to obtain the corresponding calibration values ​​and estimated values ​​of unknown parameters.

[0012] Preferably, the simulation calculation model of the thermal system obtained based on this vector is specifically as follows:

[0013] f(x,u) = 0;

[0014] In the formula, f is the residual of multiple heat balance equations, x is a known parameter, and u is an unknown parameter.

[0015] Preferably, obtaining multiple sets of historical known parameter values ​​and corresponding historical correction values ​​under different operating conditions in the thermal system includes the following steps:

[0016] Multiple sets of historically known parameter values ​​are input into the simulation calculation model, and an optimization problem is constructed through a data coordination algorithm.

[0017] The optimization problem is solved to obtain the corresponding historical correction values.

[0018] Preferably, the optimization problem is constructed using a data coordination algorithm, and the specific optimization problem is as follows:

[0019]

[0020] In the formula, x * The correction value for the known parameter, ξ(x) * To optimize the objective function, S x Let be the covariance matrix between known parameters.

[0021] Preferably, training the fast correction model using the dataset includes the following steps:

[0022] Define the loss function for the fast correction model;

[0023] The fast correction model is trained based on the loss function of the Adam optimizer and the fast correction model;

[0024] The loss function of the fast calibration model includes a parameter correction value loss function and a system equilibrium equation loss function, the parameter correction value loss function being as follows:

[0025]

[0026] In the formula, L x Let R(x) = x be the parameter correction loss function. * The correction value for the known parameters. Let ||·|| represent the corresponding historical correction value in the dataset, where ||·|| denotes the norm of the corresponding vector;

[0027] The loss function of the system equilibrium equation is as follows:

[0028] L f =||f(R(x),E(x))|| =||f(x) * ,u)||;

[0029] In the formula, L f Let E(x) = u be the loss function of the system equilibrium equation, and let E(x) = u be the value of the unknown parameter obtained by using the unknown parameter estimation model E on the input parameters.

[0030] The loss function for the fast correction model is shown below:

[0031] L = L x +C·L f ;

[0032] In the formula, L is the loss function of the fast calibration model, and C is the weight of the loss function of the system equilibrium equation.

[0033] Preferably, the known parameter correction model and the unknown parameter estimation model are fully connected neural networks, multilayer perceptrons, residual neural networks, sparse connection networks, or autoencoders.

[0034] Preferably, the acquisition of multiple sets of historical known parameter values ​​and corresponding historical correction values ​​under different operating conditions in the thermal system further includes:

[0035] Obtain the standard values ​​of multiple known parameters in a thermal system under any reference operating condition;

[0036] The standard value is used to simulate errors, resulting in multiple error measurement values, which are shown below:

[0037]

[0038] In the formula, Let x0 be the standard value, Δ be the error measurement value, and S be the error vector with the same length as x0. x Let N(0, S) be the covariance matrix between known parameters. x Let S be a vector with mean 0 and covariance S. x It follows a multivariate normal distribution.

[0039] Let multiple standard values ​​be historical correction values, and multiple error measurement values ​​be historical known parameter values.

[0040] In a second aspect, the present invention provides a thermal system measurement data correction device, comprising:

[0041] The acquisition module is used to model each device in the thermal system to obtain the corresponding simulation model; to acquire the heat balance equation corresponding to each simulation model, to arrange the residuals of multiple heat balance equations into a vector, and to obtain the simulation calculation model of the thermal system based on the vector; wherein, the parameters in the heat balance equation include known parameters whose specific data can be determined and unknown parameters whose specific data cannot be determined, and the known parameters include the measurement parameters of the thermal system and the operating parameters of the thermal system;

[0042] The module is used to construct a known parameter correction model and an unknown parameter estimation model through a physical neural network. The known parameter correction model takes known parameter values ​​as input and outputs the corresponding correction values; the unknown parameter estimation model takes known parameter values ​​as input and outputs the corresponding unknown parameter values.

[0043] The connection module is used to connect the outputs of the known parameter correction model and the unknown parameter estimation model to the input of the simulation calculation model to obtain a fast correction model;

[0044] The training module is used to obtain multiple sets of historical known parameter values ​​and corresponding historical correction values ​​of the thermal system under different operating conditions, construct a dataset, and train the fast correction model using the dataset;

[0045] The calibration module is used to collect real-time measurement data of the thermal system's measurement parameters and real-time operating data of the operating parameters. It inputs the real-time measurement data and real-time operating data into the trained fast calibration model to obtain the corresponding calibration values ​​and estimated values ​​of unknown parameters.

[0046] Thirdly, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described method for correcting thermal system measurement data.

[0047] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for correcting thermal system measurement data.

[0048] Compared with the prior art, the above-mentioned at least one technical solution adopted by the present invention can achieve the following beneficial effects:

[0049] This invention first identifies the known and unknown parameters of a thermodynamic system. Then, it constructs a known parameter correction model and an unknown parameter estimation model using a physical neural network, effectively addressing the limitations of traditional numerical methods in handling complex nonlinear physical problems. The known parameter correction model, the unknown parameter estimation model, and the simulation calculation model of the thermodynamic system are connected to obtain a fast correction model. This fast correction model takes the corrected parameter values ​​and the estimated unknown parameter values ​​as input and calculates the residual vectors of each heat balance equation in the system. By training the fast correction model with a dataset, it can quickly calculate the corrected values ​​of the known parameters and the estimated values ​​of the unknown parameters, ensuring that they satisfy the heat balance equations defined by the fundamental physical laws of the thermodynamic system. Compared to classical data reconciliation methods, the method proposed in this invention effectively avoids the problems of slow calculation speed and poor iterative convergence, making it more suitable for the measurement parameter correction calculation of complex thermodynamic systems in engineering applications. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart of a method for correcting measurement data of a thermal system according to the present invention;

[0052] Figure 2 This is a structural diagram of the rapid correction model for thermal system measurement data of the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] In order to solve the problems of the existing technology, referring to Figure 1 This invention provides a method for correcting measurement data of a thermal system, specifically including the following steps:

[0055] S1: Construct a simulation calculation model of the thermal system.

[0056] First, a model of the thermal system deployed in the actual engineering scenario needs to be created to obtain a simulation model of the thermal system. Various devices in the system are abstracted into standard equipment simulation models, including heat exchangers, turbine stages, and pumps. Each equipment simulation model has its corresponding heat balance equations, including flow balance equations, energy balance equations, and pressure balance equations.

[0057] For the thermodynamic system under study, a simulation model is constructed based on the principle of heat balance. Specifically, the residuals of all heat balance equations in the thermodynamic system are arranged into a vector f, thus forming the simulation model of the thermodynamic system. These equations involve various parameters in the system, which are divided into two categories: measured parameters obtained by setting up measuring points, and operating parameters defined according to design values. Operating parameters, such as heat transfer coefficients and turbine stage efficiency, are called known parameters x. Other parameters whose values ​​cannot be determined are called unknown parameters u. The function of the simulation model is to solve for the unknown parameters u by substituting the known parameters x into the balance equation f(x,u).

[0058] Therefore, the simulation model of the thermal system can be characterized as the following multivariate equations:

[0059] f(x,u)=0 (1);

[0060] The aforementioned multivariate equations hold true when both known and unknown parameters are precise values, satisfying all equilibrium equations. If the residuals are not zero, it indicates errors in the measured values ​​of the known parameters or the estimated values ​​of the unknown parameters, requiring correction. Furthermore, for the thermodynamic system simulation model constructed according to the above steps, it should be ensured that, given a fixed known parameter x, all unknown parameters u of the system can be solved using the residual f of the equilibrium equations; that is, the constructed thermodynamic system simulation model can successfully solve all the system's heat balance equations.

[0061] S2: Construct a benchmark dataset for the calibration of thermal system measurement data.

[0062] Based on the measurement data obtained from the thermal system, two different methods can be used to acquire and construct a calibration benchmark dataset for the thermal system measurement data.

[0063] Method 1: First, acquire historical measurement data from multiple devices in the thermal system, including measured values ​​from various types of instruments such as pressure, temperature, flow rate, and power, as well as operational parameter data, denoted as vector x. Then, substitute this data into the aforementioned thermal system simulation model and employ a standard data coordination algorithm to construct a data coordination optimization problem for the simulation model. Specifically, the following mathematical optimization problem needs to be constructed and solved:

[0064]

[0065] Where, x * Here are the correction values ​​for each known parameter, ξ(x). * To optimize the objective function, S x Given the covariance matrix between known parameters, the element s in the i-th row and j-th column of this matrix can be determined as follows: ij :

[0066]

[0067] In the formula, σ i ρ is the standard deviation of the i-th known parameter, which can be obtained by transforming the parameter's instrument accuracy or its measurement uncertainty; ij The correlation coefficient between the i-th and j-th known parameters can be obtained through statistical analysis of the historical value distribution of the measurement data. When relevant information is lacking, ρ can be directly set. ij =0. Using the optimization algorithm to solve the optimization problem (2), the optimal solution x can be obtained. * This value is the correction value of the known parameters at each measuring point in the thermal system.

[0068] For each set of historical measurement data, the known parameter x is processed using the method described above to calculate its correction value x. * With x as input, x * For output, construct a calibration benchmark dataset for thermal system measurement data.

[0069] Method 2: When historical data is limited, generate a simulation dataset based on the baseline operating condition of the thermal system for data augmentation. Specifically, suppose the standard parameter values ​​of a certain baseline operating condition of the thermal system can be arranged into a vector x0, and generate the simulation error parameter value x according to the following formula:

[0070]

[0071] Where Δ is an error vector with the same length as x0, and this vector follows a zero-mean vector with a covariance matrix of S. x The value follows a multivariate normal random distribution. For each baseline operating condition parameter value x0, a series of error measurements can be generated using the method described above. Collect all reference parameter values ​​and error measurement pairs, using the error measurement values ​​as the basis. Using x0 as the input and the reference parameter value as the output, construct a calibration reference dataset for thermal system measurement data.

[0072] The calibration benchmark datasets for thermal system measurement data constructed using the two methods described above can be merged into a single dataset. Let the input parameters in this dataset be denoted as vector x, and the output parameters (target values) as vector x. * Let X be the set of all input parameters and Y be the set of all output parameters.

[0073] S3: Divide the thermal system measurement data into training and validation sets.

[0074] For a fast calibration model of thermal system measurement data, the input is a vector set X of measurement data with unknown errors. This vector set is randomly shuffled and divided into a model training set X' at a ratio of 4:1. train With validation set X eval And accordingly, the corresponding training target set Y is obtained. train With the verification target set Y eval In particular, if the dataset used for model training is constructed using the two methods described above, the aforementioned 4:1 ratio should be applied separately to the dataset portions using the two methods.

[0075] S4: Build a rapid calibration model for rapid calibration of thermal system measurement data.

[0076] Physics-Informed Neural Networks (PINNs) are a novel machine learning method that combines deep learning techniques with classical physics theories. It embeds physical laws as constraints into the training process of the neural network, thus enabling the prediction or correction of the behavior of physical systems using limited experimental data and known physical rules. This approach effectively addresses the limitations of traditional numerical methods in handling complex nonlinear physical problems, while also overcoming the dependence of data-driven methods like deep learning on large amounts of high-quality data. Therefore, applying PINNs to the calibration of measurement data in thermodynamic systems can improve the accuracy and reliability of these measurements, thereby enhancing the system's operational efficiency and safety. Compared to classical data coordination methods, PINNs effectively address practical issues such as algorithm convergence difficulties and slow computation speeds, making them more suitable for real-time measurement data calibration in complex thermodynamic systems and providing an effective solution for improving the reliability of thermodynamic system operation and control.

[0077] like Figure 2As shown, the rapid correction model for rapid correction of measurement data in a thermal system consists of three parts: a known parameter correction model R, an unknown parameter estimation model E, and a thermal system simulation model. First, a neural network is used to construct a model from the input parameter x to the target value x. * First, a known parameter correction model is constructed, which is used to calculate the correction values ​​of the known parameters of the system. Second, a neural network is also used to construct an unknown parameter estimation model from the input parameter x to the vector u, which is used to calculate the predicted values ​​of the unknown parameters in the system. Finally, the thermal system simulation calculation model constructed in the previous steps is connected to the output of the two prediction models mentioned above, using the parameter correction values ​​x predicted by the aforementioned network. * The estimated value of the unknown parameter u is used as input to calculate the residual vector f of each heat balance equation in the system.

[0078] The network model used for known parameter correction and unknown parameter estimation can employ any effective neural network architecture, including standard fully connected neural networks (FCNs) or multilayer perceptrons (MLPs), residual neural networks (ResNets), sparsely connected networks, autoencoders, etc. Furthermore, the arrangement of the two models can also take different forms, in addition to... Figure 2 The examples shown predict x with x as input. * Besides the pattern of u, the network structure can also be rearranged to first predict the unknown parameter u using the known parameter x, and then simultaneously predict x using both x and u as inputs. * This is to simulate the conventional method of thermal balance calculation.

[0079] The above three model components are interconnected, forming a unified rapid correction model for thermal system measurement data. The loss function of this model consists of two weighted components: the parameter correction value loss function L... x With the system equilibrium equation loss function L f Among them, the parameter correction value loss function L x The calculation formula is as follows:

[0080]

[0081] Where R(x)=x * This represents the corrected value obtained by applying a known parameter correction model R to the input parameter x. Let Y be the corresponding target value of the output parameter set Y in the benchmark dataset; ||·|| represents the norm of the corresponding vector. In the constructed model, any effective norm calculation method can be used, such as the p-norm defined by the following formula:

[0082]

[0083] In the formula, p > 0 is a fixed parameter, and a is an arbitrary n-dimensional vector. Usually, when applied to the rapid correction model of thermal system measurement data, p = 2 is taken in the above formula, that is, the 2-norm is used to calculate the parameter correction value loss function in formula (5).

[0084] Another part of the loss function in the fast correction model of thermal system measurement data is the system equilibrium equation loss function L. f The calculation formula is as follows:

[0085] L f =||f(R(x),E(x))|| =||f(x) * ,u)|| (7);

[0086] Where E(x)=u represents the value of the unknown parameter obtained by using the unknown parameter estimation model E on the input parameters, and the definition of the norm operation ||·|| must be consistent with the norm definition used in equation (5). The parameter correction value loss function and the system equilibrium equation loss function are combined into the loss function of the rapid correction model of thermodynamic system measurement data, which is defined as follows:

[0087] L = L x +C·L f (8);

[0088] In the formula, C is the weight of the loss function of the system equilibrium equation, which can be determined according to the overall situation of instrument error in the thermal system.

[0089] S5: A fast calibration model for training thermal system measurement data.

[0090] The fast calibration model for thermal system measurement data is trained using the optimization function defined in equation (8) to provide calibration results for the thermal system measurement data. During training, the optimizer is set to Adam, the initial learning rate is set to 0.01, and 100 training steps are performed. Subsequently, every 100 training steps, the learning rate is reduced to 1 / 10 of its original value. The appropriate total number of training steps is controlled according to the complexity of the thermal system and the size of the dataset. When the system's loss function L is within the validation set (X... eval ,Y eval Training ends when the value on the learning rate no longer decreases significantly due to the decrease in the learning rate.

[0091] S6: Rapid correction calculation of thermal system measurement data.

[0092] According to the network structure process described in step S4, real-time measurement data and real-time operation data of the thermal system are acquired and arranged sequentially into a known parameter vector x. This vector is then input into the fast correction physical neural network model for thermal system measurement data trained in the previous step to predict the corrected parameter value vector x. *And the estimated value of the unknown parameter u, thereby completing the rapid correction calculation process of the thermal system measurement data.

[0093] Based on the same concept, the present invention also provides a thermal system measurement data correction device, including an acquisition module, a construction module, a connection module, a training module and a correction module.

[0094] The acquisition module is used to model each device in the thermal system and obtain the corresponding simulation model; it obtains the heat balance equation corresponding to each simulation model, arranges the residuals of multiple heat balance equations into a vector, and obtains the simulation calculation model of the thermal system based on the vector; the parameters in the heat balance equation include known parameters whose specific data can be determined and unknown parameters whose specific data cannot be determined. The known parameters include the measurement parameters of the thermal system and the operating parameters defined according to the design values ​​of the thermal system.

[0095] The building module is used to construct known parameter correction models and unknown parameter estimation models through physical neural networks. The known parameter correction model takes known parameter values ​​as input and the corresponding correction values ​​as output; the unknown parameter estimation model takes unknown parameter values ​​as input and the corresponding unknown parameter values ​​as output.

[0096] The connection module is used to connect the outputs of the known parameter calibration model and the unknown parameter estimation model to the input of the simulation calculation model to obtain a fast calibration model.

[0097] The training module is used to obtain multiple sets of historical known parameter values ​​and corresponding historical correction values ​​of the thermal system under different operating conditions, construct a dataset, and train the fast correction model using the dataset.

[0098] The calibration module is used to collect real-time measurement data of the thermal system's measurement parameters and real-time operating data of the operating parameters. The real-time measurement data and real-time operating data are input into the trained fast calibration model to obtain the corresponding calibration values ​​and estimated values ​​of unknown parameters.

[0099] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described method for correcting measurement data of a thermal system.

[0100] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described method for correcting measurement data of a thermal system.

[0101] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0102] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for correcting measurement data of a thermal system, characterized in that, Includes the following steps: Each device in the thermal system is modeled to obtain a corresponding simulation model; the heat balance equation corresponding to each simulation model is obtained, and the residuals of multiple heat balance equations are arranged into a vector. Based on this vector, the simulation calculation model of the thermal system is obtained; wherein, the parameters in the heat balance equation include known parameters whose specific data can be determined and unknown parameters whose specific data cannot be determined. The known parameters include the measurement parameters of the thermal system and the operating parameters of the thermal system. A known parameter correction model and an unknown parameter estimation model are constructed using a physical neural network. The known parameter correction model takes known parameter values ​​as input and outputs the corresponding correction values. The unknown parameter estimation model takes known parameter values ​​as input and outputs the corresponding unknown parameter values. By connecting the outputs of the known parameter correction model and the unknown parameter estimation model with the input of the simulation calculation model, a fast correction model is obtained. Obtain multiple sets of historical known parameter values ​​and corresponding historical correction values ​​of the thermal system under different operating conditions, construct a dataset, and train the fast correction model using the dataset; Real-time measurement data of thermal system parameters and real-time operational data of operating parameters are collected. The real-time measurement data and real-time operational data are then input into the trained fast calibration model to obtain the corresponding calibration values ​​and estimated values ​​of unknown parameters.

2. The method for correcting measurement data of a thermal system as described in claim 1, characterized in that, The simulation calculation model of the thermal system is obtained based on this vector, and the specific simulation calculation model is as follows: f(x,u) = 0; In the formula, f is the residual of multiple heat balance equations, x is a known parameter, and u is an unknown parameter.

3. The method for correcting measurement data of a thermal system as described in claim 2, characterized in that, The acquisition of multiple sets of historical known parameter values ​​and corresponding historical correction values ​​under different operating conditions in the thermal system includes the following steps: Multiple sets of historically known parameter values ​​are input into the simulation calculation model, and an optimization problem is constructed through a data coordination algorithm. The optimization problem is solved to obtain the corresponding historical correction values.

4. The method for correcting measurement data of a thermal system as described in claim 3, characterized in that, The optimization problem is constructed using a data coordination algorithm, and the specific optimization problem is as follows: In the formula, x * The correction value for the known parameter, ξ(x) * To optimize the objective function, S x Let be the covariance matrix between known parameters.

5. The method for correcting measurement data of a thermal system as described in claim 4, characterized in that, The training of the fast correction model using the dataset includes the following steps: Define the loss function for the fast correction model; The fast correction model is trained based on the loss function of the Adam optimizer and the fast correction model; The loss function of the fast calibration model includes a parameter correction value loss function and a system equilibrium equation loss function, the parameter correction value loss function being as follows: In the formula, L x Let R(x) = x be the parameter correction loss function. * The correction value for the known parameters. Let ||·|| represent the corresponding historical correction value in the dataset, where ||·|| denotes the norm of the corresponding vector; The loss function of the system equilibrium equation is as follows: L f =||f(R(x),E(x))||=||f(x * ,u)||; In the formula, L f Let E(x) = u be the loss function of the system equilibrium equation, and let E(x) = u be the value of the unknown parameter obtained by using the unknown parameter estimation model E on the input parameters. The loss function for the fast correction model is shown below: L=L x +C·L f ; In the formula, L is the loss function of the fast calibration model, and C is the weight of the loss function of the system equilibrium equation.

6. The method for correcting measurement data of a thermal system as described in claim 1, characterized in that, The known parameter correction model and the unknown parameter estimation model are fully connected neural networks, multilayer perceptrons, residual neural networks, sparse connection networks, or autoencoders.

7. The method for correcting measurement data of a thermal system as described in claim 1, characterized in that, The acquisition of multiple sets of historical known parameter values ​​and corresponding historical correction values ​​under different operating conditions in the thermal system also includes: Obtain the standard values ​​of multiple known parameters in a thermal system under any reference operating condition; The standard value is used to simulate errors, resulting in multiple error measurement values, which are shown below: In the formula, Let x0 be the standard value, Δ be the error measurement value, and S be the error vector with the same length as x0. x Let N(0, S) be the covariance matrix between known parameters. x Let S be a vector with mean 0 and covariance S. x The multivariate normal distribution; Let multiple standard values ​​be historical correction values, and multiple error measurement values ​​be historical known parameter values.

8. A device for correcting measurement data of a thermal system, characterized in that, include: The acquisition module is used to model each device in the thermal system and obtain the corresponding simulation model; The heat balance equation corresponding to each simulation model is obtained, and the residuals of multiple heat balance equations are arranged into a vector. Based on this vector, the simulation calculation model of the thermal system is obtained. The parameters in the heat balance equation include known parameters whose specific data can be determined and unknown parameters whose specific data cannot be determined. The known parameters include the measurement parameters of the thermal system and the operating parameters of the thermal system. The module is used to construct a known parameter correction model and an unknown parameter estimation model through a physical neural network. The known parameter correction model takes known parameter values ​​as input and outputs the corresponding correction values; the unknown parameter estimation model takes known parameter values ​​as input and outputs the corresponding unknown parameter values. The connection module is used to connect the outputs of the known parameter correction model and the unknown parameter estimation model to the input of the simulation calculation model to obtain a fast correction model; The training module is used to obtain multiple sets of historical known parameter values ​​and corresponding historical correction values ​​of the thermal system under different operating conditions, construct a dataset, and train the fast correction model using the dataset; The calibration module is used to collect real-time measurement data of the thermal system's measurement parameters and real-time operating data of the operating parameters. It inputs the real-time measurement data and real-time operating data into the trained fast calibration model to obtain the corresponding calibration values ​​and estimated values ​​of unknown parameters.

9. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the thermal system measurement data correction method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the thermal system measurement data correction method according to any one of claims 1-7.

Images