Method and apparatus for constructing multi-physics digital twin simulation model of power equipment

Abstract

To provide a method and device for constructing a multi-physics digital twin simulation model of a power facility.SOLUTION: Determining a target simulation model according to the first parameter set of the electric power equipment in the same operating state, and constructing and training a second sample data set by combining the target neural network model and the original first parameter set to generate an improved target neural network model, so as to improve the accuracy and speed of model construction, and determining a multi-physics digital twin simulation model of the electric power equipment in the target neural network model and the improved target neural network model according to the second parameter set of the electric power equipment in different operating states, the multi-physics digital twin simulation model being used for multi-physics digital twin simulation of the electric power equipment.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to the field of model construction technology, and more particularly to a method and apparatus for constructing a multi-physics digital twin simulation model of power equipment. [Background technology]

[0002] The more complex the operating state of a unit equipment, the more complex the equipment structure configuration, the higher the requirements for the accuracy of the calculation results, and the more precise the requirements for the construction of the corresponding mathematical simulation model, and the more the calculation time required for the analysis and solution process of the corresponding simulation model doubles.In response to the need for online real-time monitoring function, a simulation calculation time that is too long not only poses great difficulties in achieving real-time effects, but also limits the improvement of functions such as safety evaluation of each device included in the unit based on the simulation model, and failure analysis and prediction.

[0003] In the process of building a digital twin simulation model for power equipment, the construction of early data sets in the prior art mostly relies on conventional multiphysics simulation, which requires a long time for training networks with large demands. At the same time, although the prior art has high segmentation accuracy when building real-time simulation models, it still has some limitations in terms of processing detailed information. Summary of the Invention [Problem to be solved by the invention]

[0004] In view of this, the present invention provides a method and device for building a multiphysics digital twin simulation model of power equipment, which solves the problem that in the process of building a digital twin simulation model of power equipment, the construction of early data sets mostly relies on conventional multiphysics simulation, and there are still some limitations in terms of processing detailed information when building a real-time simulation model. [Means for solving the problem]

[0005] According to a first aspect, the present invention provides a method for constructing a multi-physics digital twin simulation model of electric power equipment, the method including: acquiring a first parameter set in a same operating state of the electric power equipment, and determining a target simulation model based on the first parameter set; acquiring an actual operating dataset of the electric power equipment, and constructing a first sample dataset based on the actual operating dataset of the electric power equipment and the target simulation model; using the first sample dataset to perform training and generate a target neural network model; constructing a second sample dataset based on the target neural network model and the first sample dataset; establishing an improved target neural network model based on the target neural network model and the second sample dataset; acquiring a second parameter set in a different operating state of the electric power equipment, and determining a multi-physics digital twin of the electric power equipment from the target neural network model and the improved target neural network model based on the second parameter set.

[0006] The method for constructing a multiphysics digital twin simulation model of a power equipment according to the present invention first determines a target simulation model to be used in subsequent simulations using a first parameter set for the same operating state of the power equipment, and then combines the actual operating data set of the power equipment to construct a first sample data set and train the target neural network model. Next, combines the target neural network model with the original first parameter set to construct a second sample data set and train the target neural network model to obtain an improved target neural network model, thereby improving the accuracy and speed of constructing the improved target neural network model. Finally, using a second parameter set for a different operating state of the power equipment, the resulting target neural network model and the improved target neural network model are used to determine the final multiphysics digital twin simulation model for the power equipment. Therefore, by implementing the present invention, the problems of time-consuming sample data set construction and over-reliance on conventional multiphysics simulations are solved, and the improved target neural network model also solves the problem of limitations in detailed information processing when constructing a real-time simulation model.

[0007] In one alternative embodiment, obtaining a first parameter set in the same operating state of the power equipment and determining a target simulation model based on the first parameter set includes obtaining a two-dimensional data set and a three-dimensional data set for the first parameter set, constructing an initial two-dimensional physics simulation model based on the two-dimensional data set and constructing an initial three-dimensional physics simulation model based on the three-dimensional data set, calculating an average error value based on the two-dimensional data set and the three-dimensional data set, and determining a target simulation model from the initial two-dimensional physics simulation model and the initial three-dimensional physics simulation model based on the average error value.

[0008] The method for constructing a multi-physics digital twin simulation model for power equipment according to the present invention first constructs an initial 2D physics simulation model and an initial 3D physics simulation model using 2D data sets and 3D data sets included in a first parameter set for the same operating state of the power equipment, respectively, and then determines one of the constructed initial 2D physics simulation model and initial 3D physics simulation model as a final target simulation model by calculating average error values ​​of the 2D data sets and the 3D data sets, which provides support for the construction of subsequent sample data sets.

[0009] In one optional embodiment, performing training using the first sample dataset and generating a target neural network model includes: performing a normalization process on the first sample dataset to obtain a third sample dataset; performing training based on the third sample dataset to generate an initial neural network model; obtaining a third parameter set in a new operating state of the power equipment, and inputting the third sample dataset into the initial neural network model for calculation to obtain a feature dataset; performing a de-normalization process and validity verification on the feature dataset; and obtaining the target neural network model if the result of the validity verification satisfies requirements.

[0010] The method for constructing a multi-physics digital twin simulation model for power equipment according to the present invention performs training using the constructed first sample data set, and generates a corresponding target neural network model if the result of validity verification of the feature data set output from the model meets the requirements, thereby providing support for solving the problems of taking time to construct the sample data set and relying too much on conventional multi-physics simulation.

[0011] In one alternative embodiment, establishing an improved subject neural network model based on the subject neural network model and the second sample dataset includes obtaining a mesh hyperparameter set for the subject neural network model, and training and generating the improved subject neural network model based on the mesh hyperparameter set and the second sample dataset.

[0012] The method for constructing a multi-physics digital twin simulation model of power equipment according to the present invention involves transitioning a mesh hyperparameter set of a target neural network model to an improved target neural network model, and training the target neural network model using the mesh hyperparameter set to obtain a trained target neural network model, thereby improving the accuracy and speed of constructing the improved target neural network model from both the sample generation route and neural network type.

[0013] In one alternative embodiment, obtaining a second parameter set in a different operating state of the power equipment and determining a multi-physics digital twin simulation model of the power equipment from the target neural network model and the improved target neural network model based on the second parameter set includes obtaining a second parameter set in a different operating state of the power equipment and performing a correlation analysis on the second parameter set with the target neural network model and the improved target neural network model, respectively, to obtain a first correlation coefficient and a second correlation coefficient; comparing the first correlation coefficient and the second correlation coefficient; and determining the multi-physics digital twin simulation model of the power equipment as the target neural network model if the first correlation coefficient is greater than the second correlation coefficient; and determining the multi-physics digital twin simulation model of the power equipment as the improved target neural network model if the first correlation coefficient is less than the second correlation coefficient.

[0014] The method for constructing a multi-physics digital twin simulation model of power equipment according to the present invention performs correlation analysis on the second parameter sets in different operating states of the power equipment with the constructed target neural network model and the improved target neural network model, and uses the model with a high correlation coefficient as the final multi-physics digital twin simulation model of the power equipment, thereby providing support for realizing the construction of a multi-physics digital twin simulation model of the power equipment.

[0015] In one alternative embodiment, the method further includes obtaining state of health information of the electric power equipment based on a multi-physics digital twin simulation model of the electric power equipment.

[0016] The method for constructing a multiphysics digital twin simulation model of electric power equipment according to the present invention makes it possible to obtain more accurate health status information of electric power equipment in real time using the final determined multiphysics digital twin simulation model of electric power equipment.

[0017] According to a second aspect, the present invention provides an apparatus for constructing a multi-physics digital twin simulation model of electric power equipment, the apparatus including: a first acquisition determination module for acquiring a first parameter set in a same operating state of the electric power equipment and determining a target simulation model based on the first parameter set; an acquisition and construction module for acquiring an actual operating dataset of the electric power equipment and constructing a first sample dataset based on the actual operating dataset of the electric power equipment and the target simulation model; a training module for performing training using the first sample dataset and generating a target neural network model; a construction module for constructing a second sample dataset based on the target neural network model and the first sample dataset; an establishment module for establishing an improved target neural network model based on the target neural network model and the second sample dataset; and a second acquisition determination module for acquiring a second parameter set in a different operating state of the electric power equipment and determining a multi-physics digital twin simulation model of the electric power equipment from the target neural network model and the improved target neural network model based on the second parameter set.

[0018] According to a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other; computer commands stored in the memory; and the processor executing the computer commands to perform the method for constructing a multi-physics digital twin simulation model of an electric power facility according to the first aspect or any one of the corresponding embodiments.

[0019] According to a fourth aspect, the present invention provides a computer-readable recording medium storing computer commands for causing a computer to execute the method for constructing a multi-physics digital twin simulation model of an electric power facility according to the first aspect or any one of the corresponding embodiments.

[0020] According to a fifth aspect, the present invention provides a computer program product including computer commands for causing a computer to execute the method for building a multi-physics digital twin simulation model of an electric power facility according to the first aspect or any one of the corresponding embodiments. [Brief explanation of the drawings]

[0021] In order to more clearly explain the technical solutions in the embodiments or prior art, the following briefly introduces the drawings that need to be used in the description of the embodiments or prior art. Obviously, the drawings described below are some embodiments of the present invention, and those skilled in the art can also obtain other drawings based on these drawings without any creative efforts.

[0022] [Figure 1] 1 is a flowchart of a method for building a multi-physics digital twin simulation model of a power facility according to an embodiment of the present invention. [Figure 2] 1 is a flowchart of another method for building a multi-physics digital twin simulation model of power equipment according to an embodiment of the present invention. [Figure 3] 10 is a flowchart of yet another method for building a multi-physics digital twin simulation model of power equipment according to an embodiment of the present invention. [Figure 4] FIG. 1 is a schematic diagram illustrating the generation process of a U-Net model and a U-net++ model according to an embodiment of the present invention. [Figure 5] 1 is a flowchart for determining a model for constructing a simulation data set according to an embodiment of the present invention. [Figure 6] FIG. 1 is a structural block diagram of a construction device for a multiphysics digital twin simulation model of a power facility according to an embodiment of the present invention. [Figure 7] FIG. 1 is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention; DETAILED DESCRIPTION OF THE INVENTION

[0023] In order to clarify the objectives, technical solutions and advantages of the embodiments of the present invention, the technical solutions of the embodiments of the present invention will be described clearly and completely below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only some of the embodiments of the present invention, and not all of the embodiments. Any other embodiments obtained by those skilled in the art based on the embodiments of the present invention without any creative effort are all within the scope of protection of the present invention.

[0024] As the core of real-time computation for digital twin power plants, agent models can be combined with data collected by on-site sensors to form an intelligent operation and maintenance system. As a preliminary step in the construction of digital twin systems, agent models pose the greatest challenges in computational accuracy and the time required to construct large sample sets. The U-net neural network is an improved version of a fully convolutional neural network in deep learning. It can effectively train and learn even with limited training samples. Its excellent data augmentation and segmentation capabilities have already achieved successful application results in image semantic domain segmentation, 3D video data semantic domain segmentation, and super-resolution image generation. U-Net++ can handle detailed segmentation tasks well, and supports multi-scale segmentation and class imbalance problems with excellent performance. U-Net++ models have high interpretability and clear visualization results, which are useful for further analysis and interpretation of detection results. While multiphysics neural network computations based on U-Net exist, no examples of computations based on U-Net++ exist. The present invention can be used to generate a real-time simulation module of the temperature field of power equipment in a digital twin power plant, using data collected by on-site sensors as input and the results of the simulation model as real-time output, thereby realizing real-time dynamic display of the temperature field and fluid field.

[0025] According to an embodiment of the present invention, there is provided an embodiment of a method for building a multi-physics digital twin simulation model of an electric power utility, wherein the steps illustrated in the flowcharts of the figures may be performed, for example, in a computer system as a series of computer-executable commands, and although the flowcharts show a logical order, in some cases the steps shown or described may be performed in a different order.

[0026] In this embodiment, a method for constructing a multi-physics digital twin simulation model of power equipment is provided, which can be used in electronic devices such as computers, mobile phones, tablets, etc. FIG. 1 is a flowchart of the method for constructing a multi-physics digital twin simulation model of power equipment according to an embodiment of the present invention. As shown in FIG. 1, the flow includes the following steps:

[0027] Step S101: Obtain a first parameter set in the same operating state of the power equipment, and determine a target simulation model based on the first parameter set.

[0028] Specifically, the target simulation model may be one of a two-dimensional simulation model or a three-dimensional simulation model.

[0029] Step S102: Obtain an actual operation data set of the power equipment, and construct a first sample data set based on the actual operation data set of the power equipment and the target simulation model.

[0030] Specifically, a target simulation model can generate a predetermined number of corresponding simulation data sets.

[0031] Furthermore, the simulation data set and the actual operation data set of the power facility are linked together to construct a first sample data set.

[0032] Step S103: Use the first sample data set to train and generate a target neural network model.

[0033] Here, the target neural network model is the U-Net model.

[0034] Specifically, the constructed first sample dataset is used to train the U-Net model until a U-Net model that satisfies the conditions is obtained.

[0035] Step S104: Construct a second sample data set based on the target neural network model and the first sample data set.

[0036] Specifically, a sufficient amount of sample data sets is generated using a target neural network model, and the first sample data sets are combined to form a multi-sample second sample data set.

[0037] Step S105: Establish an improved object neural network model based on the object neural network model and the second sample data set.

[0038] Here, the improved target neural network model is the U-net++ model.

[0039] Specifically, the obtained target neural network models are linked together and trained using the constructed second sample dataset until a U-net++ model that satisfies the conditions is obtained.

[0040] Step S106: Obtain a second set of parameters in different operating states of the power equipment, and determine a multi-physics digital twin simulation model of the power equipment among the target neural network model and the improved target neural network model based on the second set of parameters.

[0041] Specifically, the second parameter sets obtained in different operating states of the power equipment are input to the target neural network model and the improved target neural network model, respectively, and calculations are performed, and errors are compared for the calculated results. Finally, the model with the smallest error is selected from the target neural network model and the improved target neural network model as the final multi-physics digital twin simulation model of the power equipment.

[0042] In this embodiment, the method for building a multiphysics digital twin simulation model for a power equipment involves first determining a target simulation model to be used in subsequent simulations using a first parameter set for the same operating state of the power equipment, then combining the actual operating data set of the power equipment to build a first sample data set and training the target neural network model. Next, combining the target neural network model with the original first parameter set to build a second sample data set and training the target neural network model to obtain an improved target neural network model, thereby improving the accuracy and speed of building the improved target neural network model. Finally, using a second parameter set for a different operating state of the power equipment, the resulting target neural network model and the improved target neural network model are used to determine the final multiphysics digital twin simulation model for the power equipment. Therefore, implementing this invention solves the problems of time-consuming sample data set construction and over-reliance on conventional multiphysics simulation, and further solves the problem of limitations in detailed information processing when building a real-time simulation model using the improved target neural network model.

[0043] In this embodiment, a method for constructing a multi-physics digital twin simulation model of power equipment is provided, which can be used in electronic devices such as computers, mobile phones, tablets, etc. FIG. 2 is a flowchart of the method for constructing a multi-physics digital twin simulation model of power equipment according to an embodiment of the present invention. As shown in FIG. 2, the flow includes the following steps:

[0044] Step S201: Obtain a first parameter set in the same operating state of the power equipment, and determine a target simulation model based on the first parameter set.

[0045] Specifically, the above step S201 includes the following steps.

[0046] Step S2011: A two-dimensional data set and a three-dimensional data set for the first parameter set are obtained.

[0047] Specifically, a first parameter set for the same operating state of the power equipment includes a two-dimensional data set A and a three-dimensional data set B for the same operating state.

[0048] Step S2012: An initial two-dimensional physics simulation model is constructed based on the two-dimensional data set, and an initial three-dimensional physics simulation model is constructed based on the three-dimensional data set.

[0049] Specifically, an initial 2D physics simulation model corresponding to the power equipment is constructed using 2D data set A. At the same time, an initial 3D physics simulation model corresponding to the power equipment is constructed using 3D data set B.

[0050] Step S2013: Calculate an average error value based on the two-dimensional data set and the three-dimensional data set.

[0051] TIFF2026009847000002.tif41164

[0052] TIFF2026009847000003.tif29164

[0053] Step S2014: Based on the average error value, a target simulation model is determined from the initial two-dimensional physics simulation model and the initial three-dimensional physics simulation model.

[0054] TIFF2026009847000004.tif12164

[0055] TIFF2026009847000005.tif12164

[0056] TIFF2026009847000006.tif12164

[0057] Step S202: obtain an actual operation data set of the power equipment, and construct a first sample data set according to the actual operation data set of the power equipment and the target simulation model. For details, refer to step S102 in the embodiment shown in Figure 1, and no further description will be given here.

[0058] Step S203: Use the first sample data set to train and generate a target neural network model.

[0059] Specifically, the above step S203 includes the following steps. Step S2031: A normalization process is performed on the first sample data set to obtain a third sample data set.

[0060] Specifically, the normalization processing method is determined based on different first sample data sets of different power equipment, such as a normalization method based on mean and variance, but the embodiment of the present invention is not specifically limited thereto as long as it meets the needs.

[0061] TIFF2026009847000007.tif41164

[0062] Step S2032: Perform training based on the third sample data set and generate an initial neural network model.

[0063] Specifically, the third sample dataset is divided into a training set and a test set.

[0064] Furthermore, the training set is input to the U-Net neural network for training until the neural network reaches the best convergence state and the corresponding initial neural network model is obtained, and mesh hyperparameters such as learning rate, batch size, and optimization algorithm are adjusted and selected during the training process.

[0065] Step S2033: A third parameter set in a new operating state of the power equipment is obtained, and the third sample data set is input to the initial neural network model for calculation to obtain a feature data set.

[0066] Specifically, the parameters in the new operating state of the power equipment are input to the initial neural network model with the best convergence, calculations are performed, and the corresponding feature dataset is output.

[0067] Step S2034: The feature dataset is subjected to denormalization and validity verification, and if the result of the validity verification satisfies the requirements, a target neural network model is obtained.

[0068] TIFF2026009847000008.tif35164

[0069] Furthermore, several sets of data are randomly selected from the test set, and the error between them and the feature dataset after denormalization output from the initial neural network model is compared. If the error is 10% or less, it is proven that the target neural network model obtained by training is valid.

[0070] Furthermore, if the error is large, select data with large errors from the multiple sets of data and add them from the test set to the training set until a target neural network model that meets the error requirement is generated, and return to step S2031 to continue training.

[0071] Step S204: Construct a second sample data set based on the target neural network model and the first sample data set, the details of which can be seen in step S104 of the embodiment shown in Fig. 1, and will not be further described here.

[0072] Step S205: Establish an improved target neural network model based on the target neural network model and the second sample data set, the details of which can be seen in step S105 in the embodiment shown in Fig. 1, and will not be further described here.

[0073] Step S206: Obtain a second set of parameters under different operating conditions of the power equipment, and determine a multi-physics digital twin simulation model of the power equipment among the target neural network model and the improved target neural network model based on the second set of parameters. For details, please refer to step S106 in the embodiment shown in Figure 1, and no further description will be given here.

[0074] In this embodiment, the method for constructing a multiphysics digital twin simulation model for power equipment includes first constructing an initial 2D physics simulation model and an initial 3D physics simulation model using 2D and 3D datasets included in a first parameter set for the same operating state of the power equipment. Next, one of the constructed initial 2D physics simulation model and the initial 3D physics simulation model is determined as the final target simulation model by calculating the average error value of the 2D and 3D datasets. Then, a first sample dataset is constructed and trained by linking it with actual operating datasets of the power equipment until a target neural network model corresponding to the result of validity verification of the feature dataset output from the model satisfies requirements is generated. Furthermore, a second sample dataset is constructed and trained by linking the target neural network model with the original first parameter set to obtain an improved target neural network model, thereby improving the accuracy and speed of constructing the improved target neural network model. Finally, the final multiphysics digital twin simulation model of the power equipment is determined from the obtained target neural network model and the improved target neural network model according to the second parameter set in different operating states of the power equipment, which is used for the multiphysics digital twin simulation of the power equipment. Therefore, by implementing the present invention, the problems of time-consuming construction of sample data sets and over-reliance on conventional multiphysics simulations are solved, and the problem that the improved target neural network model still has some limitations in terms of processing detailed information when constructing a real-time simulation model is also solved.

[0075] In this embodiment, a method for constructing a multi-physics digital twin simulation model of power equipment is provided, which can be used in electronic devices such as computers, mobile phones, tablets, etc. FIG. 3 is a flowchart of the method for constructing a multi-physics digital twin simulation model of power equipment according to an embodiment of the present invention. As shown in FIG. 3, the flow includes the following steps:

[0076] Step S301: Obtain a first parameter set for the same operating state of the power equipment, and determine a target simulation model based on the first parameter set. For details, refer to step S201 in the embodiment shown in Figure 2, and no further description will be given here.

[0077] Step S302: obtain an actual operation data set of the power equipment, and construct a first sample data set based on the actual operation data set of the power equipment and the target simulation model. For details, refer to step S102 in the embodiment shown in Figure 1, and no further description will be given here.

[0078] Step S303: Use the first sample data set to train and generate a target neural network model, the details of which can be seen in step S203 in the embodiment shown in Figure 2, and will not be further described here.

[0079] Step S304: Construct a second sample data set based on the target neural network model and the first sample data set, the details of which can be seen in step S104 in the embodiment shown in Fig. 1, and will not be further described here.

[0080] Step S305: Establish an improved object neural network model based on the object neural network model and the second sample data set.

[0081] Specifically, the above step S305 includes the following steps.

[0082] Step S3051: A mesh hyperparameter set of the target neural network model is obtained.

[0083] Specifically, the mesh hyperparameter set may include mesh hyperparameters such as the learning rate, batch size, and optimization algorithm of the target neural network model.

[0084] Step S3052: Train and generate an improved target neural network model based on the mesh hyperparameter set and the second sample dataset.

[0085] Specifically, a normalization process is performed on the second sample data set, and the normalized data is divided into a training set and a validation set.

[0086] Then, until an improved target neural network model that satisfies the error condition is obtained, the mesh hyperparameter set is transferred to the U-net++ model and compared and adjusted as the initial parameters, and the training set is input to the U-net++ neural network for training. The specific process can be seen from the above steps S2032 to S2034, and will not be further described here.

[0087] Step S306: Obtain a second set of parameters in different operating states of the power equipment, and determine a multi-physics digital twin simulation model of the power equipment among the target neural network model and the improved target neural network model based on the second set of parameters.

[0088] Specifically, the above step S306 includes the following steps.

[0089] Step S3061: Obtain a second parameter set under different operating conditions of the power equipment, and perform correlation analysis on the second parameter set with the target neural network model and the improved target neural network model, respectively, to obtain a first correlation coefficient and a second correlation coefficient.

[0090] Specifically, the second parameter sets in different operating states of the power equipment are subjected to correlation analysis with the constructed target neural network model and the improved target neural network model, respectively.

[0091] Furthermore, Pearson's correlation coefficient is selected as the index to perform correlation analysis, and the corresponding first correlation coefficient and second correlation coefficient are obtained.

[0092] Step S3062: The first correlation coefficient and the second correlation coefficient are compared.

[0093] Step S3063: If the first correlation coefficient is greater than the second correlation coefficient, determine the multi-physics digital twin simulation model of the power equipment as the target neural network model.

[0094] Specifically, if the first correlation coefficient is greater than the second correlation coefficient, it indicates that there is a strong correlation between the second parameter set and the target neural network model, and in this case, the target neural network model is the final multi-physics digital twin simulation model of the power equipment.

[0095] Step S3064: If the first correlation coefficient is smaller than the second correlation coefficient, determine the multi-physics digital twin simulation model of the power equipment as the improved target neural network model.

[0096] Specifically, if the first correlation coefficient is smaller than the second correlation coefficient, it indicates that there is a strong correlation between the second parameter set and the improved target neural network model, and in this case, the improved target neural network model is the final multi-physics digital twin simulation model of the power equipment.

[0097] Step S307: Obtain health status information of the power equipment based on the multi-physics digital twin simulation model of the power equipment.

[0098] Specifically, the finalized multi-physics digital twin simulation model of the power plant equipment is placed in the power plant's online monitoring system, and the digital model is evaluated in real time through feedback of online monitoring data from the actual equipment, enabling accurate acquisition of information on the health status of the power plant equipment.

[0099] In this embodiment, the method for constructing a multi-physics digital twin simulation model for power equipment involves first determining a target simulation model to be used in subsequent simulations using a first parameter set for the same operating state of the power equipment. Then, a first sample data set is constructed and trained using an actual operating data set for the power equipment to obtain a target neural network model. Next, the mesh hyperparameter set of the target neural network model is transformed into an improved target neural network model, and the target neural network model is trained using the mesh hyperparameter set to obtain a trained target neural network model. This improves the accuracy and speed of constructing the improved target neural network model based on both the sample generation route and neural network type. Furthermore, a correlation analysis is performed between the constructed target neural network model and the improved target neural network model for a second parameter set for a different operating state of the power equipment, and the model with the highest correlation coefficient is used as the final multi-physics digital twin simulation model for the power equipment. Finally, the determined multi-physics digital twin simulation model for the power equipment in the team can provide more accurate health status information for the power equipment in real time. Therefore, by implementing the present invention, the problems of time-consuming construction of sample data sets and excessive reliance on conventional multiphysics simulations are solved, and further, the problem that there are still some limitations in terms of processing detailed information when constructing a real-time simulation model through the improved target neural network model is solved.

[0100] One example provides a method for building a rapid simulation of a multiphysics digital twin of a power facility, including the following steps:

[0101] Calculates 1, 2 and 3D models. Based on the research subject, the same input and output variables are selected and calculations are performed. The differences in the simulation data are compared to determine whether the 2D model calculation can replace the 3D model.

[0102] 2. Generate a U-Net model as shown in Figure 4. 2.1 Based on on-site operation data, find the values ​​of the independent variables and dependent variables that satisfy the selected model. 2.2 Based on the simulation model, a predetermined number of simulation data are generated and constructed as a data set together with historical operation data. 2.3 Normalize the sample dataset and split the dataset into a training set and a validation set. 2.4Train the U-Net model and select mesh hyperparameters such as learning rate, batch size, and optimization algorithm. 2.5 Perform quick calculations on the validation set and training set and output feature data. 2.6 Perform denormalization and validity verification on the data. 2.7 Generate the neural network model.

[0103] 3. Generate the U-net++ model as shown in Figure 4. 3.1 Based on the U-Net neural network model, a sufficient amount of dataset is generated, and the original datasets are combined to form a multi-sample dataset. 3.2 Sample set data is normalized and the data is divided into a training set and a validation set. 3.3 The mesh hyperparameters such as learning rate, batch size, and optimization algorithm of the U-net model are transferred to the U-net++ model and compared and adjusted as the initial parameters. 3.4 Perform quick calculations on the validation set and training set, and output feature data. 3.5 Perform denormalization and validity verification on the data. 3.6 Generate the neural network model.

[0104] 4. Compare models. Multiple sets of data are calculated, errors are compared, and the model with the smallest error is placed as the final model in the on-site online digital twin monitoring system.

[0105] This method for rapidly building a multi-physics digital twin simulation of power equipment is based on the concept and current state of building a real-time simulation model of temperature fields using traditional neural networks. Before building the models, 2D and 3D simulation models corresponding to the power equipment are first built, and they are calculated based on the same operating state parameters to determine whether the error is small. Next, mesh hyperparameters are transformed based on the first basic neural network to generate abundant samples and provide them to the second neural network, improving the accuracy and speed of the agent model from both the sample generation root and neural network type. This method can be used to create a real-time simulation module of the temperature field of power equipment in a digital twin power plant. It uses data collected by on-site sensors as input and the results of the simulation model as real-time output, realizing real-time dynamic display of temperature and fluid fields.

[0106] In one alternative embodiment, a specific example is provided based on the method for rapidly building a simulation of a multi-physics digital twin of a power facility according to the above example.

[0107] Specifically, the research targets two-dimensional and three-dimensional oil-immersed transformers, and the model has two oil inlets, one oil outlet, and a heat source with multiple windings. The input features are the inlet 1 flow velocity, the inlet 2 flow velocity, the heat source density, the inlet 1 temperature, and the inlet 2 temperature, and the maximum heat source temperature and the outlet temperature are selected as output features.

[0108] TIFF2026009847000009.tif95164

[0109] TIFF2026009847000010.tif203164

[0110] TIFF2026009847000011.tif119164

[0111] 4. Compare models. For each of the two models, a sample data set covering multiple sets of driving conditions is selected, and a correlation analysis is performed on each of the two models. The Pearson correlation coefficient is selected as the index, and the model with the largest correlation coefficient is selected as the agent model with the highest accuracy for the research subject.

[0112] 5. Arrange the site. The final generated agent model is placed in the power plant's online monitoring system, and the digital model can be evaluated in real time by feedback of online monitoring data from the actual equipment, thereby obtaining accurate information on the equipment's health status.

[0113] This embodiment further provides an apparatus for constructing a multi-physics digital twin simulation model of a power facility, which is used to realize the above-described embodiment and preferred embodiment, and the contents already described will not be further explained. The term "module" used below may refer to a combination of software and / or hardware that realizes a preset function. Although the apparatus described in the following embodiment is preferably realized by software, it is also possible and conceivable to realize it by hardware or a combination of software and hardware.

[0114] This embodiment provides an apparatus for constructing a multi-physics digital twin simulation model of power equipment. As shown in FIG. 6 , the apparatus includes: a first acquisition determination module 601 for acquiring a first set of parameters in a same operating state of the power equipment and determining a target simulation model based on the first set of parameters; an acquisition and construction module 602 for acquiring an actual operating dataset of the power equipment and constructing a first sample dataset based on the actual operating dataset of the power equipment and the target simulation model; a training module 603 for performing training using the first sample dataset and generating a target neural network model; a construction module 604 for constructing a second sample dataset based on the target neural network model and the first sample dataset; an establishment module 605 for establishing an improved target neural network model based on the target neural network model and the second sample dataset; and a second acquisition determination module 606 for acquiring a second set of parameters in a different operating state of the power equipment and determining a multi-physics digital twin simulation model of the power equipment from the target neural network model and the improved target neural network model based on the second parameter set.

[0115] In some alternative embodiments, the first acquisition determination module 601 includes a first acquisition unit for acquiring a two-dimensional data set and a three-dimensional data set in a first parameter set, a construction unit for constructing an initial two-dimensional physics simulation model based on the two-dimensional data set and constructing an initial three-dimensional physics simulation model based on the three-dimensional data set, a calculation unit for calculating an average error value based on the two-dimensional data set and the three-dimensional data set, and a first determination unit for determining a target simulation model from the initial two-dimensional physics simulation model and the initial three-dimensional physics simulation model based on the average error value.

[0116] In some alternative embodiments, the training module 603 includes: a processing unit for performing a normalization process on the first sample data set to obtain a third sample data set; a first training unit for performing training based on the third sample data set to generate an initial neural network model; an acquisition calculation unit for obtaining a third parameter set in a new operating state of the power equipment and inputting the third parameter set into the initial neural network model for calculation to obtain a feature data set; and a processing verification unit for performing a de-normalization process and validity verification on the feature data set, and obtaining a target neural network model if the result of the validity verification meets requirements.

[0117] In some alternative embodiments, the establishment module 605 includes a second obtaining unit for obtaining a mesh hyperparameter set of the target neural network model, and a second training unit for training based on the mesh hyperparameter set and the second sample dataset and generating an improved target neural network model.

[0118] In some alternative embodiments, the second acquisition determination module 606 includes an acquisition analysis unit for acquiring a second set of parameters under different operating conditions of the power equipment and performing a correlation analysis of the second set of parameters with the target neural network model and the improved target neural network model, respectively, to obtain a first correlation coefficient and a second correlation coefficient; a comparison unit for comparing the first correlation coefficient and the second correlation coefficient; a second determination unit for determining the multi-physics digital twin simulation model of the power equipment as the target neural network model if the first correlation coefficient is greater than the second correlation coefficient; and a third determination unit for determining the multi-physics digital twin simulation model of the power equipment as the improved target neural network model if the first correlation coefficient is less than the second correlation coefficient.

[0119] In some alternative embodiments, the apparatus further includes an acquisition module for acquiring power equipment health information based on a multi-physics digital twin simulation model of the power equipment.

[0120] The further functional description of each of the above modules and units is the same as the corresponding embodiment described above, so it will not be further described here.

[0121] In this embodiment, the construction device for the multi-physics digital twin simulation model of the power facility is shown in the form of a functional unit, where a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or a device capable of providing the above functions.

[0122] An embodiment of the present invention further provides a computer device having the device for constructing a multi-physics digital twin simulation model of a power equipment shown in FIG. 6 above.

[0123] Referring to FIG. 7, FIG. 7 is a structural diagram of a computer device according to an alternative embodiment of the present invention. As shown in FIG. 7, the computer device includes one or more processors 10, memory 20, and interfaces (including high-speed and low-speed interfaces) for connecting each component. Each component is communicatively connected to each other using different buses and can be mounted on a common mainboard or in other ways as needed. The processor can process commands executed within the computer device (including external commands stored in or on memory to display GUI graphics information on an external input / output device (e.g., a display device connected to the interface)). In some alternative embodiments, multiple processors and / or multiple buses can be used along with multiple memories, if necessary. Similarly, multiple computer devices can be connected to provide some of the operations required by each device (e.g., as a server array, a group of blade servers, or a multiprocessor system). FIG. 7 shows one processor 10 as an example.

[0124] The processor 10 may be a central processing unit, a network processor, or a combination thereof. Here, the processor 10 may further include a hardware chip. The hardware chip may be an application specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0125] Here, the memory 20 stores commands executable by the at least one processor 10 to cause the at least one processor 10 to execute and realize the methods shown in the above embodiments.

[0126] The memory 20 may include a program storage area and a data storage area. The program storage area may store an operating system and / or application programs required for at least one function, and the data storage area may store data configured in response to use of the computer device. The memory 20 may include high-speed random access memory and may further include non-transitory memory, such as at least one magnetic disk memory, flash memory, or other non-transitory solid-state memory. In some alternative embodiments, the memory 20 may include memory located remotely from the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local network, a mobile communication network, and combinations thereof. The memory 20 may include volatile memory, such as random access memory; the memory may include non-volatile memory, such as flash memory, a hard disk, or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0127] The computer device further includes a communications interface 30 for use in communications between the computer device and other facilities or communications networks.

[0128] The embodiments of the present invention further provide a computer-readable recording medium, and the methods according to the embodiments of the present invention described above may be implemented in hardware, firmware, or as computer code that can be recorded on a storage medium, or that can be downloaded over a network and stored on a remote storage medium or a non-transitory device-readable storage medium and stored on a local storage medium, whereby the methods described herein may be processed by such software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Here, the storage medium may be a magnetic disk, an optical disk, a read-only memory, a random memory, a flash memory, a hard disk, a solid-state drive, etc., and may further include a combination of the above types of memory. As can be understood, a computer, a processor, a microprocessor controller, or programmable hardware may include a storage component that can store or receive software or computer code, and when the software or computer code is accessed and executed by the computer, processor, or hardware, it performs the methods shown in the embodiments.

[0129] Some aspects of the present invention may be implemented as a computer program product, such as computer program commands, which, when executed by a computer, can invoke or provide the method and / or technical solution according to the present invention through the operation of the computer. As will be understood by those skilled in the art, computer program commands may exist in a computer-readable medium in a variety of forms, including, but not limited to, a source file, an executable file, an installation package file, etc. Accordingly, computer program commands may be executed by a computer in a variety of ways, including, but not limited to, the computer directly executing the commands, compiling the commands and then executing a corresponding compiled program, reading and executing the commands, or reading and installing the commands and then executing a corresponding installed program. The computer-readable medium may be any available computer-readable storage medium or communication medium accessible by a computer.

[0130] Although the embodiments of the present invention have been described with reference to the drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and all such modifications and variations are intended to be included within the scope of the claims.

Claims

1. A method for building a multi-physics digital twin simulation model of an electric power facility, comprising: Obtaining a first parameter set in the same operating state of the power equipment, and determining a target simulation model based on the first parameter set; acquiring an actual operation data set of a power equipment, and constructing a first sample data set based on the actual operation data set of the power equipment and the target simulation model; training and generating a target neural network model using the first sample data set; constructing a second sample data set based on the subject neural network model and the first sample data set; establishing an improved subject neural network model based on the subject neural network model and the second sample data set; obtaining a second parameter set in a different operating state of the electric power equipment, and determining a multi-physics digital twin simulation model of the electric power equipment from the target neural network model and the improved target neural network model based on the second parameter set; Obtaining a first parameter set in the same operating state of the power equipment and determining a target simulation model based on the first parameter set includes: acquiring two-dimensional and three-dimensional data sets for the first parameter set; constructing an initial two-dimensional physics simulation model based on the two-dimensional data set, and constructing an initial three-dimensional physics simulation model based on the three-dimensional data set; calculating a mean error value based on the two-dimensional data set and the three-dimensional data set; determining the target simulation model from the initial two-dimensional physics simulation model and the initial three-dimensional physics simulation model based on the average error value; obtaining a second parameter set in a different operating state of the electric power equipment, and determining a multi-physics digital twin simulation model of the electric power equipment from the target neural network model and the improved target neural network model based on the second parameter set; Obtaining the second parameter set in different operating states of the power equipment, and performing a correlation analysis of the second parameter set with the target neural network model and the improved target neural network model, respectively, to obtain a first correlation coefficient and a second correlation coefficient; comparing the first correlation coefficient with the second correlation coefficient; If the first correlation coefficient is greater than the second correlation coefficient, determining the multi-physics digital twin simulation model of the power equipment as the target neural network model; and determining the multi-physics digital twin simulation model of the electric power equipment as the improved target neural network model if the first correlation coefficient is less than the second correlation coefficient.

2. Training and generating a target neural network model using the first sample data set includes: performing a normalization process on the first sample data set to obtain a third sample data set; training and generating an initial neural network model based on the third sample data set; acquiring a third parameter set in a new operating state of the power equipment, and inputting the third parameter set into the initial neural network model to perform a calculation to obtain a feature dataset; 2. The method according to claim 1, further comprising: performing a denormalization process and validity verification on the feature dataset; and obtaining the target neural network model if a result of the validity verification satisfies a requirement.

3. Establishing an improved subject neural network model based on the subject neural network model and the second sample data set includes: obtaining a mesh hyperparameter set for the target neural network model; and training and generating the improved target neural network model based on the mesh hyperparameter set and the second sample data set.

4. The method comprises:

10. The method of claim 1, further comprising: obtaining power equipment state of health information based on a multi-physics digital twin simulation model of the power equipment.

5. A construction device for a multi-physics digital twin simulation model of a power facility, a first acquisition determination module for acquiring a first parameter set in the same operating state of the power equipment and determining a target simulation model based on the first parameter set; an acquisition and construction module for acquiring an actual operating data set of a power facility and constructing a first sample data set based on the actual operating data set of the power facility and the target simulation model; a training module for training using the first sample data set and generating a target neural network model; a construction module for constructing a second sample data set based on the target neural network model and the first sample data set; an establishment module for establishing an improved subject neural network model based on the subject neural network model and the second sample data set; a second acquisition determination module for acquiring a second set of parameters in different operating states of the electric power equipment, and determining a multi-physics digital twin simulation model of the electric power equipment from the target neural network model and the improved target neural network model based on the second set of parameters; The first acquisition decision module: a first acquisition unit for acquiring two-dimensional and three-dimensional data sets for the first parameter set; a construction unit for constructing an initial two-dimensional physics simulation model based on the two-dimensional data set and for constructing an initial three-dimensional physics simulation model based on the three-dimensional data set; a computing unit for computing an average error value based on the two-dimensional data set and the three-dimensional data set; a first determination unit for determining the target simulation model from the initial two-dimensional physics simulation model and the initial three-dimensional physics simulation model based on the average error value; The second acquisition decision module: an acquisition and analysis unit for acquiring the second parameter set in different operating states of the power equipment, and performing correlation analysis of the second parameter set with the target neural network model and the improved target neural network model, respectively, to obtain a first correlation coefficient and a second correlation coefficient; a comparison unit for comparing the first correlation coefficient with the second correlation coefficient; a second determination unit for determining the multi-physics digital twin simulation model of the power equipment as the target neural network model when the first correlation coefficient is greater than the second correlation coefficient; and a third determination unit for determining the multiphysics digital twin simulation model of the electric power equipment as the improved target neural network model when the first correlation coefficient is smaller than the second correlation coefficient.

6. A computer device comprising: a memory; and a processor; the memory and the processor are communicatively connected to each other; computer commands are stored in the memory; and the processor executes the computer commands to perform the method for constructing a multi-physics digital twin simulation model of electric power equipment according to any one of claims 1 to 4.

7. A computer-readable recording medium storing computer commands for causing a computer to execute the method for building a multiphysics digital twin simulation model of electric power equipment according to any one of claims 1 to 4.

8. A computer program product comprising computer commands for causing a computer to execute the method for constructing a multiphysics digital twin simulation model of electric power equipment according to any one of claims 1 to 4.

Images