Semi-physical simulation method and system for online intelligent optimization
Through the online intelligent optimization semi-physical simulation method, combined with the connection between the simulation machine and the DCS system, data preprocessing and model calibration are carried out, and the parameters are automatically adjusted using intelligent algorithms, which solves the insufficient accuracy of simulation models and low optimization efficiency in the simulation technology of nuclear power plants. The problem has achieved efficient and accurate simulation results.
Patent Information
- Application Number
- CN202411023747.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-29
- Publication Date
- 2024-12-06
AI Technical Summary
In the existing simulation technology of nuclear power plants, the simulation model is insufficient and the parameter optimization efficiency is low, resulting in insufficient accuracy and poor real-time simulation results.
The semi-physical simulation method with online intelligent optimization is adopted. Through the connection between the simulation machine and the DCS system, data preprocessing, model calibration and automatic parameter adjustment are carried out, and multiple iterative optimizations are used to improve the accuracy and reliability of the simulation model.
It realizes efficient and accurate simulation model calibration and preprocessing, improves the accuracy and reliability of the simulation model, solves the problems of insufficient simulation model accuracy and low parameter optimization efficiency, and ensures the real-time and stability of the system.
Smart Images

Figure CN119087832A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of nuclear power plant simulation, and in particular to an online intelligent optimized semi-physical simulation method and system. Background Art
[0002] Simulation technology is a simulation model technology that uses simulation hardware and simulation software to reflect system behavior or processes through simulation experiments, with the help of certain numerical calculations and problem solving. Simulation technology can bring huge economic benefits. Simulation is mainly used in engineering and technical fields such as aviation, aerospace, electric power and chemical industry. In the aviation industry, the use of simulation technology has shortened the design and development cycle of large passenger aircraft by 20%.
[0003] In the field of nuclear power, nuclear power plant simulation technology integrates high-tech technologies in many fields such as reactor engineering, thermal power, computer and control technology. The full-range simulator is one of the important achievements of nuclear power plant simulation technology. It can provide the necessary equipment for operator training and qualification examinations, and can provide important reference for commissioning and operation.
[0004] At present, simulation technologies in the nuclear power field all use offline simulation technology, establish simulation models according to the characteristics and functions of the simulated system, and use the real data results of the simulated system to correct the simulation model. The present invention improves the defects of insufficient simulation model accuracy and low parameter optimization efficiency in the prior art. Summary of the invention
[0005] In view of the problems existing in the existing online intelligent optimization semi-physical simulation method and system, the present invention is proposed.
[0006] Therefore, the present invention aims to improve the defects of the prior art, such as insufficient simulation model accuracy and low parameter optimization efficiency.
[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:
[0008] In a first aspect, an embodiment of the present invention provides an online intelligent optimized semi-physical simulation method, which includes building a semi-physical simulation model based on a simulation machine, a DCS system and a simulated object, connecting the simulation machine with the DCS system, and preprocessing the simulated object;
[0009] The preprocessed results are transmitted to the simulation machine to calibrate the semi-physical simulation model, extract non-residual data, and perform secondary preprocessing using the preprocessed results and non-residual data;
[0010] An intelligent algorithm is introduced to analyze the differences in the secondary preprocessing results, automatically adjust the parameters of the semi-physical simulation model, and optimize the accuracy of the semi-physical simulation model through multiple rounds of iterative optimization.
[0011] As a preferred solution of the semi-physical simulation method for online intelligent optimization of the present invention, wherein: the construction of the semi-physical simulation model includes building a simulation machine, a DCS system and a simulated object through hardware connection and software integration;
[0012] The hardware connection includes using a high-bandwidth Ethernet or optical fiber network to connect the simulator to the DCS system through a router and a switch;
[0013] The software integration includes configuring an intelligent algorithm in the DCS system to complete monitoring and control of the simulated object, and using the Modbus communication protocol to complete seamless integration of the simulator and the DCS system.
[0014] As a preferred solution of the semi-physical simulation method for online intelligent optimization of the present invention, wherein: the preprocessing of the simulated object includes collecting an initial data set X0 from the simulated object based on building a semi-physical simulation model, wherein the initial data set includes sensor readings and operating parameters. When the acquisition function is set to f ac When , the initial data set is calculated and collected. The specific calculation formula is:
[0015] X0=f ac (O) = {x 0,i |i=1,2,...,n}
[0016] Among them, x 0,i represents the data of the nth sampling point, n represents the total number of sampling points, O represents the simulated object, X0 represents the initial data set, and f ac represents the acquisition function;
[0017] The wavelet transform is used to remove high-frequency noise to perform denoising on the collected initial data set, and the exponential weighted average method is used to smooth the denoised data. The specific calculation formula for the denoising process is:
[0018] X d =ω(X0)
[0019] Among them, X d represents the smoothed data after denoising, X0 represents the initial data set, and ω represents the wavelet transform function;
[0020] Extract feature vectors from the smoothed data, use principal component analysis to reduce the dimension of the feature vectors, and normalize the reduced feature vectors;
[0021] The calculation formula of the dimension reduction vector of the feature vector is:
[0022] F=f pca (X s )=ω T X s
[0023] Among them, ω T The matrix representing the eigenvectors, X s represents the smoothed data set, F represents the dimension reduction vector, and f pca represents the feature extraction function;
[0024] The specific calculation formula for the normalization process is:
[0025]
[0026] Among them, F represents the dimension reduction vector, F norm represents the normalized reduced dimension vector, f norm represents the normalization function, μ represents the mean, and σ represents the standard deviation.
[0027] As a preferred solution of the semi-physical simulation method for online intelligent optimization of the present invention, wherein: the preprocessed result is transmitted to the simulation machine for judgment, and the preprocessing parameters are dynamically adjusted based on the judgment result. The specific steps are as follows:
[0028] If the denoised data X d The noise level satisfies ||X0-X d When ||<∈1, it means that the denoising is effective and the subsequent processing continues; otherwise, the denoising is performed again, where ∈1 represents the allowable error threshold of the denoising process;
[0029] If the smoothed data set X s The rate of change satisfies When , it means that smoothing is effective and the subsequent processing continues; otherwise, the smoothing window size m is adjusted to perform smoothing again, and ∈2 represents the allowable error threshold of smoothing processing;
[0030] If the principal component contribution rate of the dimension reduction vector F satisfies where λ k If ∈ 3 represents the eigenvalue of the kth principal component and p represents the number of selected principal components, the feature extraction is considered valid and the subsequent processing continues; otherwise, the number of principal components p is adjusted and feature extraction is performed again. ∈ 3 represents the principal component contribution rate threshold of feature extraction.
[0031] As a preferred solution of the online intelligent optimization semi-physical simulation method of the present invention, wherein: the semi-physical simulation model calibration includes the simulation machine using the dimension reduction vector to calibrate the initial simulation model M0, assuming that the calibration function is f c , then the simulation model formula after calibration is:
[0032] M1=f c (M0,F norm )
[0033] Where M1 represents the calibrated simulation model, f c represents the calibration function, M0 represents the initial simulation model, F norm represents the normalized reduced dimensionality vector;
[0034] The secondary preprocessing includes using the calibrated simulation model M1 and the non-residual data X1 to perform calculations, complete the judgment of the semi-physical simulation model, and define the secondary preprocessing function f p2 , the specific calculation formula is:
[0035] Y0=f p2 (M1, X1)
[0036] Among them, Y0 represents the secondary preprocessing result, X1 represents the non-residual data, M1 represents the calibrated simulation model, and f p2 represents the secondary preprocessing function;
[0037] The judgment of the semi-physical simulation model includes fusing the dimensionality reduction vector with the simulation model, and defining the fusion function as f fusion , calculate the fusion result, the specific calculation formula is:
[0038] Y fusion =f fusion (F nonlin , M1)
[0039] Among them, Y fusion represents the fusion result of the dimension reduction vector and the simulation model, f fusion represents the fusion function, M1 represents the simulation model;
[0040] If the fusion result Y fusion The correlation with the simulation model M1 satisfies corr(Y fusion , M1)> the operation parameters of the initial data set, it means that the model fusion is effective and the subsequent processing continues; otherwise, adjust the model fusion parameters and perform fusion again;
[0041] Perform error correction on the fused result and define the error correction function as f error , calculate the secondary preprocessing result after fusion, the specific calculation formula is:
[0042] Y'0=f error (Y fusion , X 1,aligned )
[0043] Among them, Y'0 represents the secondary preprocessing result after fusion, f error Denotes the definition of error correction function, X 1,aligned represents the data after registration, Y fusion Indicates the fusion result of the dimension reduction vector and the simulation model;
[0044] If the root mean square error of the second preprocessing result Y'0 after fusion satisfies RMSE(Y0, X 1,aligned ) < Y0, it indicates that the error correction is effective and the preprocessing is completed; otherwise, adjust the error correction function and perform the correction again.
[0045] As a preferred solution of the hardware-in-the-loop simulation method for online intelligent optimization described in the present invention, wherein: analyzing the difference in the second preprocessing result includes using the second preprocessing result and the actual data error E1 as the input of the intelligent algorithm to automatically adjust the parameters of the simulation model M1;
[0046] When the optimization function of the intelligent algorithm is fa, the result after automatically adjusting the simulation model is M'1 = f a (M2, E1);
[0047] The multi-round iterative optimization includes establishing the optimized simulation model M opt , and outputting the simulation result of M opt through the simulator S to verify the accuracy V of the simulation model. Then the calculation result of the accuracy verification is:
[0048] V = f v (M opt , X)
[0049] Wherein, V represents the accuracy of verifying the simulation model, M opt represents the simulation result, X represents the actual data set, and f v represents the verification function.
[0050] As a preferred solution of the hardware-in-the-loop simulation method for online intelligent optimization described in the present invention, wherein: optimizing the accuracy of the hardware-in-the-loop simulation model includes comparing the result M'1 after automatically adjusting the simulation model with the accuracy V of the verified simulation model. The specific steps are as follows:
[0051] If M'1 < V, then perform normal iteration and output the data of the accuracy of the hardware-in-the-loop simulation model;
[0052] If M'1 ≥ V, then stop the iteration. At this time, perform preprocessing and calibration N times until the normal iteration in the case of M'1 < V is achieved.
[0053] In a second aspect, an embodiment of the present invention provides a hardware-in-the-loop simulation system for online intelligent optimization, which includes: a preprocessing module that builds a hardware-in-the-loop simulation model for the object, connects the simulator to the DCS system, and preprocesses the object to be simulated;
[0054] A calibration module that transmits the result after preprocessing to the simulator for calibrating the hardware-in-the-loop simulation model;
[0055] The optimization module introduces an intelligent algorithm to analyze the differences in the secondary preprocessing results, automatically adjusts the parameters of the semi-physical simulation model, and optimizes the accuracy of the semi-physical simulation model.
[0056] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, any step of the above-mentioned semi-physical simulation method for online intelligent optimization is implemented.
[0057] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the above-mentioned semi-physical simulation method for online intelligent optimization is implemented.
[0058] The beneficial effects of the present invention are as follows: the present invention iteratively optimizes the simulation model parameters by introducing an intelligent algorithm, thereby realizing efficient and accurate simulation model calibration and preprocessing; through multiple rounds of iterative optimization and error correction, the accuracy and reliability of the simulation model are improved, and the problems of insufficient simulation model accuracy and low parameter optimization efficiency in the prior art are solved; the present invention utilizes the seamless integration of a high-bandwidth network and an intelligent algorithm to realize efficient connection and data transmission between a DCS system and a simulation machine, thereby ensuring the real-time performance and stability of the system, providing a more accurate and intelligent solution for semi-physical simulation, and is widely applicable to the simulation and optimization of complex industrial processes. BRIEF DESCRIPTION OF THE DRAWINGS
[0059] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:
[0060] Figure 1 A flowchart of an online intelligent optimized semi-physical simulation method and system is provided for one embodiment of the present invention.
[0061] Figure 2 A simulation schematic diagram of an online intelligent optimization semi-physical simulation method and system provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0062] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.
[0063] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.
[0064] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.
[0065] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.
[0066] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.
[0067] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0068] Example 1
[0069] Reference Figure 1 and Figure 2, which is the first embodiment of the present invention, and provides an online intelligent optimization semi-physical simulation method, comprising:
[0070] S1: Build a semi-physical simulation model based on the simulator, DCS system and the simulated object, connect the simulator to the DCS system, and pre-process the simulated object.
[0071] Among them, building a semi-physical simulation model includes building the simulator, DCS system and the simulated object through hardware connection and software integration;
[0072] Hardware connection includes using high-bandwidth Ethernet or fiber optic network to connect the simulator to the DCS system through routers and switches;
[0073] Software integration includes configuring intelligent algorithms in the DCS system to complete the monitoring and control of the simulated object, and using the Modbus communication protocol to complete the seamless integration of the simulator and the DCS system.
[0074] S1.1: Preprocessing the simulated object includes collecting the initial data set X0 from the simulated object based on building a semi-physical simulation model. The initial data set includes sensor readings and operating parameters. When the acquisition function is set to f ac When , the initial data set is calculated and collected. The specific calculation formula is:
[0075] X0=f ac (O) = {x 0,i |i=1,2,...,n}
[0076] Among them, x 0,i represents the data of the nth sampling point, n represents the total number of sampling points, O represents the simulated object, X0 represents the initial data set, and f ac represents the acquisition function;
[0077] Wavelet transform is used to remove high-frequency noise to denoise the initial data set, and the exponential weighted average method is used to smooth the denoised data. The specific calculation formula for denoising is:
[0078] X d =ω(X0)
[0079] Among them, X d represents the smoothed data after denoising, X0 represents the initial data set, and ω represents the wavelet transform function;
[0080] Extract feature vectors from the smoothed data, use principal component analysis to reduce the dimension of the feature vectors, and normalize the reduced feature vectors;
[0081] The calculation formula of the dimension reduction vector of the feature vector is:
[0082] F=f pca (X s )=ω T X s
[0083] Among them, ω T The matrix representing the eigenvectors, X s represents the smoothed data set, F represents the dimension reduction vector, and f pca represents the feature extraction function;
[0084] The specific calculation formula for normalization is:
[0085]
[0086] Among them, F represents the dimension reduction vector, F norm represents the normalized reduced-dimensional vector, f norm represents the normalization function, μ represents the mean, and σ represents the standard deviation.
[0087] Furthermore, preprocessing the simulated object includes collecting an initial data set from the simulated object based on building a semi-physical simulation model. The initial data set includes sensor readings and operating parameters. For example, if there is a simulated object, its initial data set consists of data from 100 sensor collection points. The data at each sampling point records the temperature, pressure and flow information of the equipment. In actual operation, a sampling point in the initial data set may be recorded as: temperature 75°C, pressure 1.5MPa, flow 30L / min.
[0088] After the initial data set is collected, wavelet transform is used to remove high-frequency noise. For example, assuming that the original data of a certain sampling point has slight fluctuations, after wavelet transform, the noise is filtered out and smoother data is obtained. Then, the denoised data is smoothed using exponential weighted average method to further reduce the fluctuation of the data. For example, the temperature readings of certain points in the original data may be 75℃, 76℃, and 74.5℃. After smoothing, the data readings may be adjusted to 75.2℃, 75.8℃, and 75℃.
[0089] Feature vectors are extracted from the smoothed data, and principal component analysis (PCA) is used to reduce the dimension of the feature vectors. For example, 10 main feature vectors are extracted from the data of 100 sensors. These feature vectors may represent the main change trends of temperature, pressure and flow. Assume that a feature vector (temperature, pressure, flow) in the original data is [75℃, 1.5MPa, 30L / min]. After PCA dimension reduction, it is simplified to [0.8, 1.2, 0.5], which represents the contribution ratio of the main features.
[0090] The eigenvectors after dimensionality reduction are normalized so that the data can be compared on the same scale. For example, if one component in the eigenvector after dimensionality reduction is 0.8, the normalized component may be adjusted to 0.75, so that the values of all eigenvectors are between 0 and 1. The range of standardized data is helpful for subsequent data analysis and processing. The data preprocessing results are shown in Table 1 below:
[0091] Table 1 Data preprocessing results
[0092]
[0093] The table shows the detailed results of the initial data set collected from the simulated object and its eigenvectors and normalized data after wavelet transform denoising and exponentially weighted average smoothing.
[0094] S2: The preprocessed results are transmitted to the simulation machine to calibrate the semi-physical simulation model, extract non-residual data, and perform secondary preprocessing using the preprocessed results and non-residual data.
[0095] The preprocessing results are transmitted to the simulation machine for judgment, and the preprocessing parameters are dynamically adjusted based on the judgment results. The specific steps are as follows:
[0096] If the denoised data X d The noise level satisfies ||X0-X d When ||<∈1, it means that the denoising is effective and the subsequent processing continues; otherwise, the denoising is performed again, where ∈1 represents the allowable error threshold of the denoising process;
[0097] If the smoothed data set X s The rate of change satisfies When , it means that smoothing is effective and the subsequent processing continues; otherwise, the smoothing window size m is adjusted to perform smoothing again, and ∈2 represents the allowable error threshold of smoothing processing;
[0098] If the principal component contribution rate of the dimension reduction vector F satisfies where λ k If ∈ 3 represents the eigenvalue of the kth principal component and p represents the number of selected principal components, the feature extraction is considered valid and the subsequent processing continues; otherwise, the number of principal components p is adjusted and feature extraction is performed again. ∈ 3 represents the principal component contribution rate threshold of feature extraction.
[0099] Furthermore, when calibrating the semi-physical simulation model, specific thresholds are set to ensure the accuracy of data processing. First, the allowable error threshold of denoising is set to 0.05, which means that the noise level of the denoised data must be less than 5%, otherwise denoising is required. Secondly, the allowable error threshold of smoothing is set to 0.02, indicating that the change rate of the smoothed data set must be less than 2%, otherwise the smoothing window size needs to be adjusted and smoothed again. Finally, the principal component contribution rate threshold of feature extraction is set to 85%, that is, the principal component contribution rate of the dimensionality reduction vector must exceed 85%, otherwise the number of principal components needs to be adjusted and feature extraction needs to be performed again. These specific values ensure the effectiveness and accuracy of data preprocessing and provide reliable guarantees for the calibration and optimization of subsequent simulation models.
[0100] S2.1: Calibrate the semi-physical simulation model, including using the dimension reduction vector to calibrate the initial simulation model M0. Let the calibration function be f c , then the simulation model formula after calibration is:
[0101] M1=f c (M0,F norm )
[0102] Where M1 represents the calibrated simulation model, f c represents the calibration function, M0 represents the initial simulation model, F norm represents the normalized reduced dimensionality vector;
[0103] The secondary preprocessing includes using the calibrated simulation model M1 and the non-residual data X1 to perform calculations, complete the judgment of the semi-physical simulation model, and define the secondary preprocessing function f p2 , the specific calculation formula is:
[0104] Y0=f p2 (M1, X1)
[0105] Among them, Y0 represents the secondary preprocessing result, X1 represents the non-residual data, M1 represents the calibrated simulation model, and f p2 represents the secondary preprocessing function;
[0106] The judgment of the semi-physical simulation model includes fusing the dimensionality reduction vector with the simulation model, and defining the fusion function as f fusion , calculate the fusion result, the specific calculation formula is:
[0107] Y fusion =f fusion (F nonlin , M1)
[0108] Among them, Y fusioin represents the fusion result of the dimension reduction vector and the simulation model, ffusion Let the fusion function be denoted as, and the simulation model be denoted as M1;
[0109] If the correlation between the fused result Y fusion and the simulation model M1 satisfies corr(Y fusion , M1) > the operation parameters of the initial data set, it indicates that the model fusion is effective, and subsequent processing can continue; otherwise, adjust the model fusion parameters and perform the fusion again;
[0110] Perform error correction on the fused result. Define the error correction function as f error , and calculate the secondary preprocessing result after fusion. The specific calculation formula is:
[0111] Y'0 = f error (Y fusion , X 1,aligned )
[0112] where Y'0 represents the secondary preprocessing result after fusion, f error represents the defined error correction function, X 1,aligned represents the registered data, and Y fusion represents the fusion result of the dimensionality reduction vector and the simulation model;
[0113] If the root mean square error of the secondary preprocessing result Y'0 after fusion satisfies RMSE(Y0, X 1,aligned ) < Y0, it indicates that the error correction is effective and the preprocessing is completed; otherwise, adjust the error correction function and perform the correction again.
[0114] During the calibration process of the hardware-in-the-loop simulation model, the initial simulation model is adjusted using the dimensionality reduction vector according to specific data to ensure that the model is more in line with the actual data. The calibrated simulation model and non-residual data are used for secondary preprocessing, and the final preprocessing result is obtained through the defined processing function. In the model fusion stage, the dimensionality reduction vector is fused with the simulation model, and the fusion result is corrected according to the preset error criteria to ensure the accuracy and effectiveness of the final processing result.
[0115] S3: Introduce an intelligent algorithm to analyze the differences in the secondary preprocessing results, automatically adjust the parameters of the hardware-in-the-loop simulation model, and optimize the accuracy of the hardware-in-the-loop simulation model through multiple rounds of iteration.
[0116] Among them, analyzing the differences in the secondary preprocessing results includes using the secondary preprocessing results and the actual data error E1 as the input of the intelligent algorithm to automatically adjust the parameters of the simulation model M1;
[0117] When the optimization function of the intelligent algorithm is f a , the result after automatically adjusting the simulation model is M'1 = f a (M2, E1);
[0118] Multi-round iterative optimization includes establishing an optimized simulation model M opt , and outputting M through the simulator S opt . The simulation results of M are verified to obtain the accuracy V of the simulation model. Then, the calculation accuracy verification result is as follows:
[0119] V = f v (M opt , X)
[0120] where V represents the accuracy of verifying the simulation model, M opt represents the simulation results, X represents the actual data set, and f v represents the verification function.
[0121] Furthermore, optimizing the accuracy of the hardware-in-the-loop simulation model includes comparing the result M'1 after automatically adjusting the simulation model with the accuracy V of verifying the simulation model. The specific steps are as follows:
[0122] If M'1 < V, normal iteration is performed, and the data of the hardware-in-the-loop simulation model accuracy is output;
[0123] If M'1 ≥ V, the iteration is stopped. At this time, preprocessing and calibration are performed N times until normal iteration in the case of M'1 < V is achieved.
[0124] In a preferred embodiment, an online intelligent optimization hardware-in-the-loop simulation system includes a preprocessing module that builds a hardware-in-the-loop simulation model for the object, connects the simulator to the DCS system, and preprocesses the object to be simulated;
[0125] a calibration module that transmits the preprocessed result to the simulator for hardware-in-the-loop simulation model calibration;
[0126] an optimization module that introduces an intelligent algorithm to analyze the difference in the secondary preprocessing results, automatically adjusts the parameters of the hardware-in-the-loop simulation model, and optimizes the accuracy of the hardware-in-the-loop simulation model.
[0127] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.
[0128] In summary, the present invention achieves efficient and accurate simulation model calibration and preprocessing by introducing intelligent algorithms to iteratively optimize simulation model parameters. Through multiple rounds of iterative optimization and error correction, the accuracy and reliability of the simulation model are improved, and the problems of insufficient simulation model accuracy and low parameter optimization efficiency in the prior art are solved. The present invention utilizes the seamless integration of high-bandwidth networks and intelligent algorithms to achieve efficient connection and data transmission between the DCS system and the simulation machine, ensures the real-time and stability of the system, provides a more accurate and intelligent solution for semi-physical simulation, and is widely applicable to the simulation and optimization of complex industrial processes.
[0129] Example 2
[0130] Reference Figure 1 and Figure 2 , which is the second embodiment of the present invention, and this embodiment provides a semi-physical simulation method and system for online intelligent optimization. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.
[0131] During the experiment, a semi-physical simulation model was first built, and the simulation machine was connected to the DCS system. Data was transmitted through Ethernet to collect specific data from the simulated object. For example, frequency, temperature, and pressure data were obtained from the sensor, and the number of sampling points per second was recorded as 1000. Wavelet transform is applied to these initial data for denoising, and the allowable error threshold is set to 0.01. After denoising, the exponential weighted average method is used for data smoothing, and the window size is adjusted to 10 data points to ensure that the change rate of the smoothed data set meets the set allowable error threshold of 0.05. The principal component analysis method is used to extract the feature vector of the smoothed data, and the first 5 principal components are selected to ensure that their cumulative contribution rate reaches more than 80%. The extracted feature vectors are normalized to ensure the uniform standard of the data. The initial simulation model is calibrated using the reduced dimension vector. After the calibrated model is tested, it is preprocessed twice together with the non-residual data. In the model fusion stage, the reduced dimension vector is fused with the simulation model according to the set correlation operation parameter of 0.95, and the fusion result is corrected according to the root mean square error standard of 0.02. Finally, the accurate semi-physical simulation model preprocessing results are obtained. The sample data of sensor data acquisition is shown in Table 2 below:
[0132] Table 2 Sensor data collection example data table
[0133] Number of sampling points Frequency (Hz) Temperature (Celsius) Pressure(Pa) 1 50 25 100000 2 55 26 95000 3 52 24 105000 4 53 25.5 98000 5 54 24.8 102000
[0134] The above table shows the data collected from the sensor for five consecutive sampling points. Each sampling point includes the measured values of frequency (in Hertz), temperature (Celsius) and pressure (Pascal). For example, the frequency of the first sampling point is 50Hz, the temperature is 25 degrees Celsius, and the pressure is 100,000 Pascal. These data are used for analysis and processing to support subsequent model calibration and preprocessing work. The present invention is compared with the prior art as shown in Table 3 below:
[0135] Table 3 Comparison between the present invention and the prior art
[0136]
[0137] The technology of the present invention has obvious advantages over the existing technology. It ensures stable and fast data transmission through high-bandwidth network connection, applies advanced data preprocessing methods such as wavelet transform and exponential weighted average method to improve data quality, introduces intelligent optimization algorithm to automatically adjust simulation model parameters and multiple rounds of iterative optimization to ensure model accuracy, and adopts modern hardware and software integration to improve system stability and efficiency, greatly reducing the need for manual operation.
[0138] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.
Claims
1. A semi-physical simulation method for online intelligent optimization, characterized in that: including Build a hardware-in-the-loop simulation model based on a simulator, a DCS system, and the object to be simulated. Connect the simulator to the DCS system and preprocess the object to be simulated; Transmit the preprocessed result to the simulator, perform calibration of the hardware-in-the-loop simulation model, extract non-residual data, and perform secondary preprocessing using the preprocessed result and the non-residual data; Introduce an intelligent algorithm to analyze the differences in the secondary preprocessing results, automatically adjust the parameters of the hardware-in-the-loop simulation model, and optimize the accuracy of the hardware-in-the-loop simulation model through multiple rounds of iteration.
2. The semi-physical simulation method for online intelligent optimization according to claim 1, characterized in that: The building of the hardware-in-the-loop simulation model includes building the simulator, the DCS system, and the object to be simulated through hardware connection and software integration; The hardware connection includes connecting the simulator to the DCS system through a router and a switch using a high-bandwidth Ethernet or fiber optic network; The software integration includes configuring an intelligent algorithm in the DCS system to complete the monitoring and control of the object to be simulated, and using the Modbus communication protocol to complete the seamless integration of the simulator and the DCS system.
3. The semi-physical simulation method for online intelligent optimization according to claim 2, characterized in that: The preprocessing of the simulated object includes collecting an initial data set X0 from the simulated object based on building a semi-physical simulation model, wherein the initial data set includes sensor readings and operating parameters. When the acquisition function is set to f ac When , the initial data set is calculated and collected. The specific calculation formula is: X0=f ac (O)={x 0,i |i=1,2,...,n} Among them, x 0,i represents the data of the nth sampling point, n represents the total number of sampling points, O represents the simulated object, X0 represents the initial data set, and f ac represents the acquisition function; Use wavelet transform to remove high-frequency noise and denoise the initially collected data set, and use the exponential weighted moving average method to smooth the denoised data. The specific calculation formula for the denoising process is: X d =ω(X0) Among them, X d represents the smoothed data after denoising, X0 represents the initial data set, and ω represents the wavelet transform function; Extract feature vectors from the smoothed data, use the principal component analysis method to reduce the dimension of the feature vectors, and normalize the feature vectors after dimension reduction; The calculation formula for the reduced-dimensional vector of the feature vectors is: F=f pca (X s )=ω T X s Among them, ω T The matrix representing the eigenvectors, X s represents the smoothed data set, F represents the dimension reduction vector, and f pca represents the feature extraction function; The specific calculation formula for the normalization process is: Among them, F represents the dimension reduction vector, F norm represents the normalized reduced dimension vector, f norm represents the normalization function, μ represents the mean, and σ represents the standard deviation.
4. The semi-physical simulation method for online intelligent optimization according to claim 3, characterized in that: The transmitting the preprocessed result to the simulator for judgment, and dynamically adjusting the preprocessing parameters based on the judgment result, the specific steps are as follows: If the denoised data X d The noise level satisfies ||X0-X d When ||<∈1, it means that the denoising is effective and the subsequent processing continues; otherwise, the denoising is performed again, where ∈1 represents the allowable error threshold of the denoising process; If the smoothed data set X s The rate of change satisfies When , it means that smoothing is effective and the subsequent processing continues; otherwise, the smoothing window size m is adjusted to perform smoothing again, and ∈2 represents the allowable error threshold of smoothing processing; If the principal component contribution rate of the dimension reduction vector F satisfies where λ k If ∈ 3 represents the eigenvalue of the kth principal component and p represents the number of selected principal components, the feature extraction is considered valid and the subsequent processing continues; otherwise, the number of principal components p is adjusted and feature extraction is performed again. ∈ 3 represents the principal component contribution rate threshold of feature extraction.
5. The semi-physical simulation method for online intelligent optimization according to claim 4, characterized in that: The semi-physical simulation model calibration includes the simulation machine using the dimension reduction vector to calibrate the initial simulation model M0, assuming that the calibration function is f c , then the simulation model formula after calibration is: M1=f c (M0,F norm ) Where M1 represents the calibrated simulation model, f c represents the calibration function, M0 represents the initial simulation model, F norm represents the normalized reduced dimensionality vector; The secondary preprocessing includes using the calibrated simulation model M1 and the non-residual data X1 to perform calculations, complete the judgment of the semi-physical simulation model, and define the secondary preprocessing function f p2 , the specific calculation formula is: Y0=f p2 (M1,X1) Among them, Y0 represents the secondary preprocessing result, X1 represents the non-residual data, M1 represents the calibrated simulation model, and f p2 represents the secondary preprocessing function; The judgment of the semi-physical simulation model includes fusing the dimensionality reduction vector with the simulation model, and defining the fusion function as f fusion , calculate the fusion result, the specific calculation formula is: Y fusion =f fusion (F nonlin ,M1) Among them, Y fusion represents the fusion result of the dimension reduction vector and the simulation model, f fusion represents the fusion function, M1 represents the simulation model; If the fusion result Y fusion The correlation with the simulation model M1 satisfies corr(Y fusion , M1)> the operation parameters of the initial data set, it means that the model fusion is effective and the subsequent processing continues; otherwise, adjust the model fusion parameters and perform fusion again; Perform error correction on the fused result and define the error correction function as f error , calculate the secondary preprocessing result after fusion, the specific calculation formula is: Y′0=f error (Y fusion ,X 1,aligned ) Among them, Y'0 represents the secondary preprocessing result after fusion, f error Denotes the definition of error correction function, X 1,aligned represents the data after registration, Y fusion Indicates the fusion result of the dimension reduction vector and the simulation model; If the root mean square error of the second preprocessing result Y'0 after fusion satisfies RMSE(Y0, X 1,aligned ) < Y0, it indicates that the error correction is effective and the preprocessing is completed; otherwise, adjust the error correction function and perform the correction again.
6. The semi-physical simulation method for online intelligent optimization according to claim 5, characterized in that: The analyzing the differences in the secondary preprocessing results includes taking the error E1 between the secondary preprocessing result and the actual data as the input of the intelligent algorithm, and automatically adjusting the parameters of the simulation model M1; When the optimization function of the intelligent algorithm is fa, the result after automatically adjusting the simulation model is M'1=f a (M2, E1); The multiple rounds of iterative optimization include setting up the optimized simulation model M opt , output M through the simulation machine S opt The simulation results verify the accuracy V of the simulation model, and the calculation accuracy verification result is: V=f v (M opt ,X) Among them, V represents the accuracy of the verification simulation model, M opt represents the simulation result, X represents the actual data set, and f v Represents a validation function.
7. The semi-physical simulation method for online intelligent optimization according to claim 6, characterized in that: The optimizing the accuracy of the hardware-in-the-loop simulation model includes comparing the result M'1 after automatically adjusting the simulation model with the accuracy V of the verification simulation model, the specific steps are as follows: If M'1 < V, then perform normal iteration and output the data of the accuracy of the hardware-in-the-loop simulation model; If M'1 ≥ V, then stop the iteration. At this time, perform preprocessing and calibration N times until normal iteration in the case of M'1 < V is achieved.
8. An online intelligent optimization semi-physical simulation system, based on the online intelligent optimization semi-physical simulation method according to any one of claims 1 to 7, characterized in that: including A preprocessing module that builds a hardware-in-the-loop simulation model for the object, connects the simulator to the DCS system, and preprocesses the object to be simulated; A calibration module that transmits the preprocessed result to the simulator and performs calibration of the hardware-in-the-loop simulation model; An optimization module that introduces an intelligent algorithm to analyze the differences in the secondary preprocessing results, automatically adjusts the parameters of the hardware-in-the-loop simulation model, and optimizes the accuracy of the hardware-in-the-loop simulation model.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, it implements the steps of the hardware-in-the-loop simulation method for online intelligent optimization according to any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the hardware-in-the-loop simulation method for online intelligent optimization according to any one of claims 1 to 7.
