GIL multi-physics field reconstruction method under shell sparse sensing data constraint

Abstract

The invention relates to a GIL multi-physics field reconstruction method under the constraint of shell sparse sensing data, and the method comprises the steps: constructing a high-fidelity multi-physics field numerical simulation model of an electromagnetic field-temperature field-flow field based on the typical working condition of a GIL; constructing a temperature field-flow velocity field reduced-order model, and generating a base and a time evolution coefficient of a low-rank space according to the reduced-order model; calculating the relationship among the sampling matrix, the sensor data and the screen snapshot according to the actual sensor position, and generating final sensing data according to the relationship; and constructing a neural network agent model for GIL multi-physics field reconstruction based on the final sensing data and the base and time evolution coefficient. According to the method, the double-limitation dilemma of high-voltage equipment such as GIL in field-level state detection is solved through a machine learning algorithm, that is, the problem that the stability and accuracy of a traditional reconstruction algorithm are degraded under the conditions that the sensor type is single and the spatial arrangement is limited is solved.

Description

Technical Field

[0001] This application relates to the technical fields of electrical equipment and machine learning, and particularly relates to a GIL multi-physical field reconstruction method under the constraint of sparse shell sensing data. Background Art

[0002] With the continuous improvement of the digitalization requirements of China's power grid and the continuous evolution of Digital Twin (DT) technology, reliably and real-timely reflecting the full-field information of equipment in the virtual digital space has become a hot topic in the intelligent operation and maintenance of power transmission and transformation equipment. After decades of development, the digitalization of power transmission and transformation equipment has formed two major pillars mainly based on numerical simulation and online monitoring. The former discretely calculates the multi-physical field control equations through methods such as finite element / finite volume, and the latter real-timely monitors the equipment status through sensing data. However, traditional numerical simulation has bottlenecks in computing efficiency and defects in convergence stability, making it difficult to meet the millisecond-level response requirements of digital twins; at the same time, field-level status monitoring faces problems such as a sharp increase in the cost of secondary sensor layout and deterioration of electromagnetic compatibility. The field reconstruction technology is an important way to transition from traditional digital technology to digital twin applications. In the offline stage of this method, a set of spatially low-rank bases are obtained through a model reduction technique based on high-fidelity numerical simulation, and then combined with real-time sensing data to perform online inversion of the time evolution coefficients (which can be reduced to a matrix inversion problem), thereby inferring the full-field status of the equipment. However, for high-voltage enclosed equipment such as Gas-Insulated Line (GIL), the layout positions of sensors are limited, and invasive sensors may affect the normal and stable operation of the equipment, which results in a poor condition number of the inversion matrix and the inability to monitor the internal SF6 gas flow rate. Therefore, this patent proposes a highly robust method for reconstructing the internal temperature field-flow field of GIL based on sparse temperature sensing data on the shell. Summary of the Invention

[0003] This application provides a GIL multi-physical field reconstruction method under the constraint of sparse shell sensing data to solve problems in related technologies, such as the limited layout positions of sensors for high-voltage enclosed equipment like gas-insulated metal-enclosed transmission lines, and the fact that invasive sensors may affect the normal and stable operation of the equipment, which leads to a poor condition number of the inversion matrix and the inability to monitor the internal SF6 gas flow rate.

[0004] The first aspect of the present application provides a GIL multi - physical - field reconstruction method under the constraint of sparse sensing data of the shell, which is applied to the model - building stage and includes the following steps: Based on the typical operating conditions of the gas - insulated metal - enclosed transmission line (GIL), construct a high - fidelity multi - physical - field numerical simulation model of the electromagnetic field - temperature field - flow field, and generate the simulation data of the GIL according to the high - fidelity multi - physical - field numerical simulation model; Based on the simulation data, construct a reduced - order model of the temperature field - flow velocity field, and generate the basis of the low - rank space and the time - evolution coefficients according to the reduced - order model; Calculate the relationship between the sampling matrix, the sensor data, and the screen snapshot according to the actual sensor positions, so as to generate the final sensing data according to the relationship; Based on the final sensing data, the basis, and the time - evolution coefficients, construct a neural network proxy model for the GIL multi - physical - field reconstruction.

[0005] Optionally, in an embodiment of the present application, the internal SF6 flow pattern formula under the typical operating conditions of the gas - insulated metal - enclosed transmission line (GIL) is:

[0006]

[0007] where Q 11 is the convective heat transfer between the shell and the outside air, Q7 is the radiative heat transfer between the shell and the outside air, Q 21 is the convective heat transfer between the conductor through the interlayer SF6 and the shell, Q 22 is the radiative heat transfer between the conductor and the shell.

[0008] Optionally, in an embodiment of the present application, the step of constructing a reduced - order model of the temperature field - flow velocity field based on the simulation data and generating the basis of the low - rank space and the time - evolution coefficients according to the reduced - order model includes: Based on the simulation data, collect the screen snapshots of the temperature field and the flow velocity field at a preset period, so as to generate a snapshot matrix according to the screen snapshots; Pre - process the snapshot matrix to perform standardization processing on the physical fields of each spatial point to generate a processed snapshot matrix; Perform singular - value decomposition on the processed snapshot matrix to generate decomposition data, and linearly scale the time - evolution coefficient matrix according to the decomposition data to generate the basis of the low - rank space and the time - evolution coefficients.

[0009] Optionally, in an embodiment of the present application, the scaling formula of the time - evolution coefficient matrix is:

[0010]

[0011] where is the time - evolution coefficient corresponding to the POD mode at time j t, and For The continuous form of the evolution coefficient corresponding to the POD mode, where T represents the temperature field, u represents the x-component of the velocity field, and v represents the y-component of the velocity field.

[0012] Optionally, in an embodiment of the present application, the relationship formula among the sampling matrix, the sensor data, and the screen snapshot is:

[0013]

[0014] where C is the sampling matrix, ∈ represents the error between the measured value and the full-order model data, Ψ = [ψ1,..., ψ r is the POD principal mode corresponding to the reduced-order model, and a i is the time evolution coefficient vector at time t i moment.

[0015] Optionally, in an embodiment of the present application, the construction formula of the neural network surrogate model for field reconstruction is:

[0016]

[0017] where N u is the batch_size, is the vector of evolution coefficients corresponding to the Ψ φ POD mode fitted by the neural network, and is the corresponding Ground_Truth.

[0018] In the second aspect of the present application, an embodiment provides a GIL multi-physical field reconstruction method under the constraint of shell sparse sensing data, which is applied to the model application stage and includes the following steps: obtaining sensing data corresponding to actual sensor positions and bases and time evolution coefficients; inputting the sensing data and the bases and time evolution coefficients into a pre-constructed neural network surrogate model for field reconstruction to reconstruct GIL according to the neural network surrogate model for field reconstruction, where the neural network surrogate model for field reconstruction is constructed from the sensing data and the bases and time evolution coefficients.

[0019] The third - aspect embodiment of this application provides a GIL multi - physical - field reconstruction device under the constraint of sparse sensing data of the housing, which is applied to the model - building stage and includes: a first building module, configured to construct a high - fidelity multi - physical - field numerical simulation model of the electromagnetic field - temperature field - flow field based on the typical operating conditions of the gas - insulated metal - enclosed transmission line (GIL), and generate simulation data of the GIL according to the high - fidelity multi - physical - field numerical simulation model; a second building module, configured to construct a reduced - order model of the temperature field - flow velocity field based on the simulation data, and generate a basis of the low - rank space and time - evolution coefficients according to the reduced - order model; a calculation module, configured to calculate the relationship between the sampling matrix, sensor data, and screen snapshots according to the actual sensor positions, so as to generate final sensing data according to the relationship; a third building module, configured to construct a neural - network proxy model for the GIL multi - physical - field reconstruction based on the final sensing data, the basis, and the time - evolution coefficients.

[0020] Optionally, in an embodiment of this application, the internal SF6 flow - pattern formula under the typical operating conditions of the gas - insulated metal - enclosed transmission line (GIL) is:

[0021]

[0022] where Q 11 is the convective heat transfer between the housing and the outside air, Q7 is the radiative heat transfer between the housing and the outside air, Q 21 is the convective heat transfer between the conductor through the interlayer SF6 and the housing, Q 22 is the radiative heat transfer between the conductor and the housing.

[0023] Optionally, in an embodiment of this application, the second building module includes: an acquisition unit, configured to collect screen snapshots of the temperature field and the flow velocity field at a preset period based on the simulation data, so as to generate a snapshot matrix according to the screen snapshots; a pre - processing unit, configured to pre - process the snapshot matrix to perform normalization processing on the physical fields at each spatial point and generate a processed snapshot matrix; a generation unit, configured to perform singular - value decomposition on the processed snapshot matrix, generate decomposition data, and perform linear scaling on the time - evolution coefficient matrix according to the decomposition data to generate the basis of the low - rank space and the time - evolution coefficients.

[0024] Optionally, in an embodiment of this application, the scaling formula of the time - evolution coefficient matrix is:

[0025]

[0026] where is the time - evolution coefficient corresponding to the POD mode at time j t and For The continuous form of the evolution coefficient corresponding to the POD mode, where T represents the temperature field, u represents the x-component of the velocity field, and v represents the y-component of the velocity field.

[0027] Optionally, in an embodiment of the present application, the relationship formula among the sampling matrix, the sensor data, and the screen snapshot is:

[0028]

[0029] Where C is the sampling matrix, ∈ represents the error between the measured value and the full-order model data, Ψ = [ψ1,..., ψ r is the POD principal mode corresponding to the reduced-order model, and a i is the time evolution coefficient vector at time t i .

[0030] Optionally, in an embodiment of the present application, the construction formula of the neural network surrogate model for field reconstruction is:

[0031]

[0032] Where N u is the batch_size, is the vector of evolution coefficients corresponding to the Ψ φ POD mode fitted by the neural network, and is the corresponding Ground_Truth.

[0033] An embodiment of the fourth aspect of the present application provides a GIL multi-physical field reconstruction device under the constraint of shell sparse sensing data, which is applied to the model application stage, including: an acquisition module for acquiring sensing data corresponding to the actual sensor positions and the basis and time evolution coefficients; a reconstruction module for inputting the sensing data and the basis and time evolution coefficients into a pre-constructed neural network surrogate model for field reconstruction to reconstruct GIL according to the neural network surrogate model for field reconstruction, where the neural network surrogate model for field reconstruction is constructed from the sensing data and the basis and time evolution coefficients.

[0034] An embodiment of the fifth aspect of the present application provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the program to implement the GIL multi-physical field reconstruction method under the constraint of shell sparse sensing data as described in the above embodiments.

[0035] In a sixth aspect embodiment of the present application, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, which when executed by a processor, implements the GIL multi-physical field reconstruction method under the constraint of sparse shell sensing data as described above.

[0036] Embodiments of the present application can solve the "double constraint" dilemma faced by high-voltage equipment such as GIL in field-level state detection through machine learning algorithms, that is, under the conditions of single type of sensors and limited spatial layout (taking the shell temperature monitoring in the present application as an example), the problem of degradation of the stability and accuracy of traditional reconstruction algorithms, and propose a highly robust reconstruction method for the GIL temperature field - flow velocity field jointly driven by online monitored shell temperature data and offline reduced-order models, meeting the basic requirements for constructing a GIL digital twin. Thus, it solves the problems in the related art that the layout positions of sensors for high-voltage enclosed equipment such as gas-insulated metal-enclosed transmission lines are limited, and intrusive sensors may affect the normal and stable operation of the equipment, resulting in a poor condition number of the inversion matrix and the inability to monitor the internal SF6 gas flow velocity, etc.

[0037] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be understood through the practice of the present application. Description of the Drawings

[0038] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description of the embodiments in conjunction with the drawings, where:

[0039] Figure 1 FIG. is a flowchart of a GIL multi-physical field reconstruction method under the constraint of sparse shell sensing data according to an embodiment of the present application applied to the model construction stage;

[0040] Figure 2 FIG. is a basic structural diagram of the thermal circuit of a GIL multi-physical field reconstruction method under the constraint of sparse shell sensing data according to an embodiment of the present application;

[0041] Figure 3 FIG. is a basic structural diagram of the field reconstruction surrogate model of a GIL multi-physical field reconstruction method under the constraint of sparse shell sensing data according to an embodiment of the present application;

[0042] Figure 4 FIG. is a flowchart of a GIL multi-physical field reconstruction method under the constraint of sparse shell sensing data according to an embodiment of the present application applied to the model application stage;

[0043] Figure 5 FIG. is a structural schematic diagram of a GIL multi-physical field reconstruction device under the constraint of sparse shell sensing data according to an embodiment of the present application applied to the model construction stage;

[0044] Figure 6 Schematic structural diagram of an application of a GIL multi - physical - field reconstruction device under the constraint of sparse shell sensing data to the model application stage according to an embodiment of the present application;

[0045] Figure 7 Schematic structural diagram of an electronic device according to an embodiment of the present application. Detailed implementation manners

[0046] The embodiments of the present application will be described in detail below. Examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are intended to explain the present application, and should not be construed as a limitation to the present application.

[0047] The GIL multi - physical - field reconstruction method under the constraint of sparse shell sensing data according to an embodiment of the present application will be described below with reference to the drawings. In the related technology mentioned in the above - mentioned background art, for high - voltage enclosed devices such as gas - insulated metal - enclosed transmission lines (GILs), the layout positions of sensors are limited, and invasive sensors may affect the normal and stable operation of the devices, which results in problems such as a poor condition number of the inversion matrix and the inability to monitor the internal SF6 gas flow rate. The present application provides a GIL multi - physical - field reconstruction method under the constraint of sparse shell sensing data. In this method, a machine - learning algorithm can be used to solve the "double - constraint" dilemma faced by high - voltage devices such as GILs in field - level state detection, that is, the problem of the degradation of the stability and accuracy of traditional reconstruction algorithms under the conditions of single - type sensors and limited spatial layout, and a highly robust reconstruction method for the GIL temperature field - flow field driven by the collaboration of online monitored shell temperature data and an offline reduced - order model is proposed to meet the basic requirements for constructing a GIL digital twin. Thus, the problems in the related technology, such as the limited layout positions of sensors for high - voltage enclosed devices such as gas - insulated metal - enclosed transmission lines, the possible impact of invasive sensors on the normal and stable operation of the devices, the poor condition number of the inversion matrix, and the inability to monitor the internal SF6 gas flow rate, are solved.

[0048] Specifically, Figure 1 Schematic flow diagram of a GIL multi - physical - field reconstruction method under the constraint of sparse shell sensing data provided by an embodiment of the present application.

[0049] As Figure 1 shown, the GIL multi - physical - field reconstruction method under the constraint of sparse shell sensing data includes the following steps:

[0050] In step S101, based on the typical working conditions of the gas - insulated metal - enclosed transmission line GIL, a high - fidelity multi - physical - field numerical simulation model of the electromagnetic field - temperature field - flow field is constructed, and simulation data of the GIL is generated according to the high - fidelity multi - physical - field numerical simulation model.

[0051] In the actual implementation process, the embodiment of the present application can construct a high-fidelity multi-physics field numerical simulation model of the electromagnetic field - temperature field - flow velocity field for typical operating conditions of GIL. Among them, for the heat transfer part, heat conduction, heat convection, and heat radiation need to be considered simultaneously. The flow pattern of the internal SF6 gas and the numerical simulation model are pre-judged through a thermal circuit model and the Grashof number. The basic structure of the thermal circuit is as Figure 2 shown.

[0052] The Q 21 of convective heat transfer between the conductor and the shell through the interlayer SF6 satisfies:

[0053]

[0054] The radiative heat transfer Q 22 between the conductor and the shell satisfies:

[0055]

[0056] The convective heat transfer Q 11 between the shell and the outside air satisfies:

[0057] Q 11 = πD1h(T1 - T amb )

[0058] The radiative heat transfer Q 12 between the shell and the outside satisfies:

[0059] Q 12 = πD1ε1σ(T1 4 - T amb 4 )

[0060] Among them, in an embodiment of the present application, according to the law of conservation of energy, the internal SF6 flow pattern formula under typical operating conditions of the gas-insulated metal-enclosed transmission line GIL is:

[0061]

[0062] Among them, Q 11 is the convective heat transfer between the shell and the outside air, Q7 is the radiative heat transfer between the shell and the outside air, Q 21 is the convective heat transfer between the conductor and the shell through the interlayer SF6, and Q 22 is the radiative heat transfer between the conductor and the shell.

[0063] This is a system of non-linear equations, and T1 and T2 can be obtained through the iterative method.

[0064] According to the temperature difference between the outer shell and the conductor calculated by the thermal circuit method, the Grashof number of the internal SF6 fluid is calculated. The Grashof number is independent of the heat transfer parameter compared with the Rayleigh number. Therefore, it is more reasonable to judge the natural convection flow pattern in the closed space. When its value is greater than 4.56×10 9 it indicates that the gas flow pattern is fully developed into turbulence.

[0065]

[0066] A multi-physics simulation model of GIL is established in COMSOL. In order to better capture the near-wall characteristics of the fluid, the kω-SST turbulence model is selected. The natural convection in the closed cavity is completely driven by thermal buoyancy. Therefore, the simulation of the near-wall characteristics directly affects the accuracy of the simulation model.

[0067] In step S102, based on the simulation data, a reduced-order model of the temperature field-velocity field is constructed, and the basis of the low-rank space and the time evolution coefficient are generated according to the reduced-order model.

[0068] In the actual execution process, the embodiment of the present application can construct a reduced-order model of the temperature field-velocity field based on the simulation data, and generate the basis of the low-rank space and the time evolution coefficient according to the reduced-order model, so as to prepare for the construction of the subsequent data set.

[0069] Optionally, in an embodiment of the present application, based on the simulation data, a reduced-order model of the temperature field-velocity field is constructed, and the basis of the low-rank space and the time evolution coefficient are generated according to the reduced-order model, including: based on the simulation data, collecting screen snapshots of the temperature field and the velocity field at a preset period to generate a snapshot matrix according to the screen snapshots; preprocessing the snapshot matrix to standardize the physical field of each spatial point to generate a processed snapshot matrix; performing singular value decomposition on the processed snapshot matrix to generate decomposition data, and linearly scaling the time evolution coefficient matrix according to the decomposition data to generate the basis of the low-rank space and the time evolution coefficient.

[0070] It can be understood that the preset period in the embodiment of the present application can be 15s. The embodiment of the present application can generate a screen snapshot matrix of the temperature field and the velocity field according to the transient simulation results obtained by the above steps, that is, the simulation data. φ∈{T, u, v}, where is the screen snapshot of a certain physical field at the t j th moment, N s represents the dimension of the spatial grid, and N t represents the number of snapshots. The first r-order low-rank representation of the snapshot matrix is obtained by using singular value decomposition (SVD) / proper orthogonal decomposition (POD). where It is the projection basis of the low-dimensional space, and each column represents a different spatial mode. The magnitude of represents the importance of different modes. represents the time evolution coefficient of the low-dimensional space. The latter two usually form the time evolution coefficient matrix A.

[0071] Specifically, the embodiments of the present application can collect screen snapshots of the temperature field and the flow velocity field every 15 s from the initial state to the steady state based on simulation data, so as to generate a snapshot matrix according to the screen snapshots:

[0072]

[0073] Among them, is the screen snapshot of a certain physical field at the t j moment, N s represents the dimension of the spatial grid, and N t represents the number of snapshots, and the former is much larger than the latter.

[0074] Before using the explanation algorithm, preprocess the snapshot matrix to standardize the physical field of each spatial point and generate a processed snapshot matrix. Since the scales of the temperature field and the velocity field are very different, the order of magnitude of the former is 10 2 , and the latter is 10 -1 , and the scaling factor here is denoted as SC_before_SVD, and the scaling ratio formula is:

[0075]

[0076] Perform singular value decomposition (SVD) on the above processed snapshot matrix to generate decomposition data, which is usually called proper orthogonal decomposition (POD) in the field of reduced-order modeling:

[0077]

[0078] Among them, is the projection basis of the low-dimensional space, and each column represents a different spatial mode. is a diagonal matrix, and its magnitude represents the importance of different modes. represents the time evolution coefficient of the low-dimensional space. The latter two usually form the time evolution coefficient matrix A. The diagonal elements of decay rapidly, and the model can be reasonably truncated at the r-th order according to the low-rank property of the data. The proportion of the remaining energy after truncation at m is calculated by the following formula:

[0079]

[0080] Further, the embodiments of the present application can linearly scale the time evolution coefficient matrix according to the decomposed data to generate the basis of the low-rank space and the time evolution coefficient.

[0081] Among them, in an embodiment of the present application, since the activation function in the neural network adopted by the present application is ReLU(), that is, the network output is greater than 0. At this time, the scaling is given by the formula, denoted as SC_after_SVD. The scaling formula of the time evolution coefficient matrix is:

[0082]

[0083] Among them, is the time evolution coefficient corresponding to the POD mode at time j t , and is the continuous form of the evolution coefficient corresponding to the POD mode.

[0084] In step S103, calculate the relationship between the sampling matrix, the sensor data and the screen snapshot according to the actual sensor position, so as to generate the final sensing data according to the relationship.

[0085] In the actual execution process, the embodiments of the present application can calculate the sampling matrix C, the relationship between the sensor data and the screen snapshot according to the actual sensor position, so as to generate the final sensing data according to the relationship.

[0086] Among them, in an embodiment of the present application, the relationship formula between the sampling matrix, the sensor data and the screen snapshot is:

[0087]

[0088] Among them, C is the sampling matrix, ∈ represents the error between the measured value and the full-order model (FOM) data, Ψ = [ψ1,..., ψ r is the POD main mode corresponding to the reduced-order model, and a i is the time evolution coefficient vector at time i t.

[0089] For standard linear sampling, the C matrix is an orthonormal matrix: At this time, when the sensing data is known, according to the above formula, the time evolution coefficient can be solved by finding the pseudoinverse of Θ, and further the information of the entire field can be inversely calculated to reconstruct the entire physical field. This process is summarized as the following formula:

[0090]

[0091] At this time, the upper limit of the error of the reconstructed field is:

[0092]

[0093] The condition number of the sampling matrix will amplify the error between the sensor measurement values and the simulation data. For the object targeted by this application, the sensor positions are usually arranged on the outer shell of the closed device, which makes the condition number of Θ extremely poor. Direct numerical reconstruction will result in extremely low robustness and numerical stability. In addition, according to the formula for reconstructing the entire physical field above, it can be seen that when there is no sensing data for a certain physical field, numerical reconstruction cannot be achieved. Therefore, in the next step of this application, a neural network based on time series recursion - shallow decoder is constructed and combined with a reduced - order model to reconstruct the temperature field and velocity field.

[0094] It should be noted that when the sensor positions are not fixed, the DEIM or QR - DEIM algorithm can be used to optimize the sensor positions and improve the condition number of Θ. However, under the first - order constraints of the sensor positions targeted by this application, simply using these two algorithms will not result in a significant improvement in the condition number of Θ.

[0095] In step S104, based on the final sensing data, basis, and time evolution coefficients, a neural network surrogate model for GIL multi - physical - field reconstruction is constructed.

[0096] It can be understood that the basic structure of the field reconstruction surrogate model in the embodiments of this application is as Figure 3 shown.

[0097] In the actual execution process, the embodiments of this application can construct a neural network surrogate model for GIL multi - physical - field reconstruction based on the final sensing data, basis, and time evolution coefficients. The basic structure is a time series recursion network followed by a decoder. The input of the network is where k is the number of time series, and the network output is which respectively represent the time evolution coefficients of the temperature field and velocity field.

[0098] Among them, in an embodiment of this application, the loss function uses MSE, and the construction formula of the neural network surrogate model for field reconstruction is:

[0099]

[0100] where N u is the batch_size, is the vector of evolution coefficients corresponding to the Ψ φ POD modes fitted by the neural network, is the corresponding Ground_Truth.

[0101] Randomly sample the data in the above step S102 and organize it into a time series form, and split it into a training set, a validation set, and a test set. During the training process, monitor the Loss of the validation set and the training set to prevent the model from overfitting. When the Loss of the validation set no longer decreases within 20 epochs, perform early stopping processing.

[0102] After the training is completed, the robustness of the model can be tested by adding different degrees of noise to all the sensing data:

[0103]

[0104] The reconstructed field can be expressed as:

[0105]

[0106] The reconstructed field at time t j The relative error at the moment can be calculated by the following formula:

[0107]

[0108] The overall error of the model can be characterized by taking the average of the above formula in the time dimension:

[0109]

[0110] Since both the time series network and the decoder adopt shallow structures, the training efficiency of the model is very fast. The comparison with the traditional reconstruction method (15 random experiments are conducted under each noise level) is as follows. Table 1 is the comparison table of the anti-noise performance of the method of the present application and the traditional method. As shown in Table 1:

[0111] Table 1

[0112] Method / Noise level 0.05 0.1 0.15 0.2 POD(DEIM) 0.024 0.047 0.071 0.094 POD(QR-DEIM) 0.023 0.046 0.069 0.092 Shallow decoder 0.022 0.022 0.023 0.026 Temporal recursive - Shallow decoder 0.0025 0.0043 0.0054 0.0069

[0113] The present application can efficiently realize the high-accuracy and robust reconstruction of the GIL temperature field - velocity field under the sparse sensing constraint of the outer shell.

[0114] Figure 4 Provide a GIL multi-physical field reconstruction method under the constraint of sparse sensing data of the outer shell, which is applied to the model application stage and includes the following steps:

[0115] In step S401, obtain the sensing data corresponding to the actual sensor positions and the basis and time evolution coefficients.

[0116] In the actual execution process, the embodiments of the present application can obtain the sensing data corresponding to the actual sensor positions and the basis and time evolution coefficients, providing support for the subsequent construction of the neural network proxy model for field reconstruction.

[0117] In step S402, the sensing data, the substrate, and the time evolution coefficient are input into a pre-constructed neural network surrogate model for field reconstruction to reconstruct the GIL according to the neural network surrogate model for field reconstruction, where the neural network surrogate model for field reconstruction is constructed from the sensing data, the substrate, and the time evolution coefficient.

[0118] Specifically, in the embodiments of the present application, the sensing data, the substrate, and the time evolution coefficient can be input into a pre-constructed neural network surrogate model for field reconstruction to reconstruct the GIL according to the neural network surrogate model for field reconstruction, which can efficiently achieve high-accuracy and robust reconstruction of the GIL temperature field-velocity field under the sparse sensing constraint of the enclosure. The neural network surrogate model for field reconstruction is constructed from the sensing data, the substrate, and the time evolution coefficient.

[0119] According to the GIL multi-physical field reconstruction method under the sparse sensing data constraint of the enclosure proposed in the embodiments of the present application, a machine learning algorithm can be used to solve the "double constraint" dilemma faced by high-voltage equipment such as GIL in field-level state detection, that is, the problem of degradation of the stability and accuracy of traditional reconstruction algorithms under the conditions of single sensor type and limited spatial layout. And an online monitoring of the shell temperature data and an offline reduced-order model co-driven high-robustness reconstruction method for the GIL temperature field-flow velocity field is proposed to meet the basic requirements for constructing a GIL digital twin. Thus, the problems in the related art are solved, such as the limited layout position of sensors for high-voltage enclosed equipment such as gas-insulated metal-enclosed transmission lines, and the intrusion sensors may affect the normal and stable operation of the equipment, which results in a poor condition number of the inversion matrix and the inability to monitor the internal SF6 gas flow velocity.

[0120] Next, a GIL multi-physical field reconstruction device under the sparse sensing data constraint of the enclosure proposed in the embodiments of the present application will be described with reference to the accompanying drawings.

[0121] Figure 5 FIG. is a schematic structural diagram of the GIL multi-physical field reconstruction device under the sparse sensing data constraint of the enclosure of the embodiments of the present application applied to the model construction stage.

[0122] As Figure 5 shown, the GIL multi-physical field reconstruction device 10 under the sparse sensing data constraint of the enclosure includes: a first construction module 100, a second construction module 200, a calculation module 300, and a third construction module 400.

[0123] Specifically, the first construction module 100 is configured to construct a high-fidelity multi-physical field numerical simulation model of the electromagnetic field-temperature field-flow field based on the typical working conditions of the gas-insulated metal-enclosed transmission line GIL, and generate simulation data of the GIL according to the high-fidelity multi-physical field numerical simulation model.

[0124] The second construction module 200 is used to construct a reduced-order model of the temperature field - flow velocity field based on simulation data, and generate a basis for the low-rank space and time evolution coefficients according to the reduced-order model.

[0125] The calculation module 300 is used to calculate the relationship between the sampling matrix, sensor data, and the screen snapshot according to the actual sensor positions, so as to generate the final sensing data based on the relationship.

[0126] The third construction module 400 is used to construct a neural network surrogate model for GIL multi-physical field reconstruction based on the final sensing data, basis, and time evolution coefficients.

[0127] Optionally, in an embodiment of the present application, the internal SF6 flow pattern formula under typical operating conditions of a gas-insulated metal-enclosed transmission line GIL is:

[0128]

[0129] where Q 11 is the convective heat transfer between the shell and the outside air, Q7 is the radiative heat transfer between the shell and the outside air, Q 21 is the convective heat transfer between the conductor through the interlayer SF6 and the shell, Q 22 is the radiative heat transfer between the conductor and the shell.

[0130] Optionally, in an embodiment of the present application, the second construction module 200 includes: an acquisition unit, a preprocessing unit, and a generation unit.

[0131] Among them, the acquisition unit is used to collect screen snapshots of the temperature field and flow velocity field based on simulation data at a preset period, so as to generate a snapshot matrix according to the screen snapshots.

[0132] The preprocessing unit is used to preprocess the snapshot matrix to perform normalization processing on the physical fields of each spatial point, and generate a processed snapshot matrix.

[0133] The generation unit is used to perform singular value decomposition on the processed snapshot matrix to generate decomposition data, and linearly scale the time evolution coefficient matrix according to the decomposition data, so as to generate a basis for the low-rank space and time evolution coefficients.

[0134] Optionally, in an embodiment of the present application, the scaling formula of the time evolution coefficient matrix is:

[0135]

[0136] where is the time evolution coefficient corresponding to the POD mode at time j t , is The continuous form of the evolution coefficient corresponding to the POD mode, where T represents the temperature field, u represents the x-component of the velocity field, and v represents the y-component of the velocity field.

[0137] Optionally, in an embodiment of the present application, the relationship formula between the sampling matrix, sensor data, and screen snapshot is:

[0138]

[0139] where C is the sampling matrix, ∈ represents the error between the measured value and the full-order model data, Ψ = [ψ1,..., ψ r is the POD principal mode corresponding to the reduced-order model, and a i is the time evolution coefficient vector at time t i moment.

[0140] Optionally, in an embodiment of the present application, the construction formula of the neural network surrogate model for field reconstruction is:

[0141]

[0142] where N u is the batch_size, is the vector of evolution coefficients corresponding to the Ψ φ POD mode fitted by the neural network, and is the corresponding Ground_Truth.

[0143] Figure 6 There is provided a GIL multi-physical field reconstruction device 20 under the constraint of shell sparse sensing data, which is applied to the model application stage and includes: an acquisition module 500 and a reconstruction module 600.

[0144] Specifically, the acquisition module 500 is used to acquire the sensing data corresponding to the actual sensor positions, the basis, and the time evolution coefficients.

[0145] The reconstruction module 600 is used to input the sensing data, the basis, and the time evolution coefficients into a pre-constructed neural network surrogate model for field reconstruction, so as to reconstruct the GIL according to the neural network surrogate model for field reconstruction, where the neural network surrogate model for field reconstruction is constructed from the sensing data, the basis, and the time evolution coefficients.

[0146] It should be noted that the foregoing explanation of the embodiment of the GIL multi-physical field reconstruction method under the constraint of shell sparse sensing data also applies to the GIL multi-physical field reconstruction device under the constraint of shell sparse sensing data in this embodiment, and will not be elaborated here.

[0147] The GIL multi-physical field reconstruction device under the constraint of sparse shell sensing data proposed according to the embodiments of the present application can solve the "double constraint" dilemma faced by high-voltage equipment such as GIL in field-level state detection through machine learning algorithms, that is, under the conditions of single sensor type and limited spatial layout, the problem of degradation of the stability and accuracy of traditional reconstruction algorithms, and propose a highly robust reconstruction method for the GIL temperature field-flow velocity field driven by online monitoring of shell temperature data and offline reduced-order models, meeting the basic requirements for building a GIL digital twin. Thus, it solves the problems in the related art that the layout positions of sensors for high-voltage enclosed equipment such as gas-insulated metal-enclosed transmission lines are limited, and intrusive sensors may affect the normal and stable operation of the equipment, resulting in a poor condition number of the inversion matrix and the inability to monitor the internal SF6 gas flow velocity.

[0148] Figure 7 The structural schematic diagram of the electronic device provided by the embodiment of the present application. The electronic device may include:

[0149] A memory 701, a processor 702, and a computer program stored on the memory 701 and executable on the processor 702.

[0150] When the processor 702 executes the program, it implements the GIL multi-physical field reconstruction method under the constraint of sparse shell sensing data provided in the above embodiment.

[0151] Further, the electronic device further includes:

[0152] A communication interface 703 for communication between the memory 701 and the processor 702.

[0153] The memory 701 is used to store a computer program executable on the processor 702.

[0154] The memory 701 may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory.

[0155] If the memory 701, the processor 702, and the communication interface 703 are implemented independently, the communication interface 703, the memory 701, and the processor 702 can be interconnected through a bus and communicate with each other. The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation,Figure 7 It is represented by only one thick line, but it does not mean that there is only one bus or one type of bus.

[0156] Optionally, in a specific implementation, if the memory 701, the processor 702, and the communication interface 703 are integrated on a single chip, the memory 701, the processor 702, and the communication interface 703 can communicate with each other through an internal interface.

[0157] The processor 702 may be a central processing unit (CPU for short), or an application specific integrated circuit (ASIC for short), or one or more integrated circuits configured to implement the embodiments of the present application.

[0158] In the description of this specification, the descriptions with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0159] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "N" is at least two, such as two, three, etc., unless otherwise specifically defined.

[0160] Any process or method description shown in the flowchart or described in other ways herein can be understood as representing a module, segment, or part of code including one or N executable instructions for implementing a customized logic function or process, and the scope of the preferred embodiments of the present application includes additional implementations, where the functions may be executed in a manner that is not shown or discussed, including in a substantially simultaneous manner according to the involved functions or in a reverse order, which should be understood by those skilled in the art to which the embodiments of the present application belong.

[0161] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definable sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or used in combination with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection part (electronic device) having one or N wirings, a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically by optically scanning the paper or other media, followed by editing, interpretation, or otherwise processing as appropriate, and then stored in a computer memory.

[0162] It should be understood that various parts of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above-described embodiments, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits having suitable combinational logic gate circuits, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0163] Those of ordinary skill in the art of this technology can understand that all or part of the steps carried by the method of implementing the above embodiments can be completed by a program instructing relevant hardware, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0164] In addition, each functional unit in various embodiments of the present application may be integrated into a processing module, may exist physically alone for each unit, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0165] The above-mentioned storage medium may be a read-only memory, a magnetic disk, an optical disc, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A GIL multi-physical field reconstruction method under the constraint of sparse shell sensing data, characterized in that, Applied to the model construction stage, including the following steps: Based on the typical operating conditions of the gas-insulated metal-enclosed transmission line (GIL), construct a high-fidelity multi-physics field numerical simulation model of the electromagnetic field - temperature field - flow field, and generate simulation data of the GIL according to the high-fidelity multi-physics field numerical simulation model; Based on the simulation data, construct a reduced-order model of the temperature field - flow velocity field, and generate the basis of the low-rank space and the time evolution coefficient according to the reduced-order model; Calculate the relationship between the sampling matrix, sensor data, and screen snapshot according to the actual sensor position, so as to generate the final sensing data according to the relationship; Based on the final sensing data and the basis and time evolution coefficient, construct a neural network surrogate model for the multi-physics field reconstruction of the GIL.

2. The method according to claim 1, wherein The internal SF6 flow pattern formula under the typical operating conditions of the gas-insulated metal-enclosed transmission line (GIL) is: Among them, Q 11 is the convective heat transfer between the housing and the outside air, Q7 is the radiative heat transfer between the housing and the outside air, Q 21 is the convective heat transfer between the conductor and the housing through the intermediate layer of SF6, Q 22 is the radiative heat transfer between the conductor and the housing.

3. The method according to claim 1, characterized in that, The step of constructing a reduced-order model of the temperature field - flow velocity field based on the simulation data and generating the basis of the low-rank space and the time evolution coefficient according to the reduced-order model includes: Based on the simulation data, collect screen snapshots of the temperature field and flow velocity field at a preset period, so as to generate a snapshot matrix according to the screen snapshots; Preprocess the snapshot matrix to standardize the physical field of each spatial point and generate a processed snapshot matrix; Perform singular value decomposition on the processed snapshot matrix to generate decomposition data, and linearly scale the time evolution coefficient matrix according to the decomposition data to generate the basis of the low-rank space and the time evolution coefficient.

4. The method according to claim 3, characterized in that, The scaling formula of the time evolution coefficient matrix is: Among them, is t j moment The time evolution coefficient corresponding to the POD mode, is The continuous form of the evolution coefficient corresponding to the POD mode, T represents the temperature field, u represents the x-direction component of the velocity field, and v is the y-direction component of the velocity field.

5. The method according to claim 1, wherein The relationship formula between the sampling matrix, the sensor data, and the screen snapshot is: where C is the sampling matrix, ∈ represents the error between the measured value and the full-order model data, Ψ = [ψ1,..., ψ r is the POD principal mode corresponding to the reduced-order model, a i is the time evolution coefficient vector at time t i instant.

6. The method according to claim 1, wherein The construction formula of the neural network surrogate model for field reconstruction is: Among them, N u is the batch_size, is Ψ fitted by the neural network φ the evolution coefficient vector corresponding to the POD mode, is the corresponding Ground_Truth.

7. A GIL multi-physical field reconstruction method under the constraint of sparse shell sensing data, characterized in that, Applied to the model application stage, including the following steps: Obtain the sensing data corresponding to the actual sensor position and the basis and time evolution coefficient; Input the sensing data and the basis and time evolution coefficient into a pre-constructed neural network surrogate model for field reconstruction, so as to reconstruct the GIL according to the neural network surrogate model for field reconstruction, where the neural network surrogate model for field reconstruction is constructed from the sensing data and the basis and time evolution coefficient.

8. A GIL multi-physical field reconstruction device under the constraint of sparse sensing data of the shell, characterized in that Applied to the model construction stage, including: The first construction module is used to construct a high-fidelity multi-physics field numerical simulation model of the electromagnetic field - temperature field - flow field based on the typical operating conditions of the gas-insulated metal-enclosed transmission line (GIL), and generate simulation data of the GIL according to the high-fidelity multi-physics field numerical simulation model; The second construction module is used to construct a reduced-order model of the temperature field - flow velocity field based on the simulation data, and generate the basis of the low-rank space and the time evolution coefficient according to the reduced-order model; The calculation module is used to calculate the relationship between the sampling matrix, sensor data, and screen snapshot according to the actual sensor position, so as to generate the final sensing data according to the relationship; The third construction module is used to construct a neural network surrogate model for the multi-physics field reconstruction of the GIL based on the final sensing data and the basis and time evolution coefficient.

9. A GIL multi-physical field reconstruction device under the constraint of sparse sensing data of the shell, characterized in that, Applied to the model application stage, including: An acquisition module, configured to acquire sensing data corresponding to the actual sensor position and the substrate and the time evolution coefficient; A reconstruction module, configured to input the sensing data and the substrate and the time evolution coefficient into a pre-constructed neural network proxy model for field reconstruction, so as to reconstruct the GIL according to the neural network proxy model for field reconstruction, wherein the neural network proxy model for field reconstruction is constructed from the sensing data and the substrate and the time evolution coefficient.

10. An electronic device, characterized in that, Including: A memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the program to implement the GIL multi-physical field reconstruction method under the constraint of shell sparse sensing data as described in any one of claims 1-6 or the GIL multi-physical field reconstruction method under the constraint of shell sparse sensing data as described in claim 7.

Images