Error compensation method and device, computer device, readable storage medium and program product

Abstract

This application relates to an error compensation method, apparatus, computer device, computer-readable storage medium, and computer program product. The method includes: acquiring a source domain dataset and a target domain dataset, wherein the source domain dataset includes simulated non-contact measurement signals, and the target domain dataset includes actual non-contact measurement signals; determining a source domain compensation model based on the source domain dataset; the source domain compensation model includes a physical approximation layer and a residual compensation layer, wherein the physical approximation layer is used to establish a mapping relationship between the simulated non-contact measurement signals and the actual non-contact measurement signals, and the residual compensation layer is used to learn the residual characteristics between the output of the physical approximation layer and the actual non-contact measurement signals; determining a target domain compensation model based on the source domain compensation model and the target domain dataset using a transfer learning strategy; and using the target domain compensation model to perform error compensation on non-contact measurement signals acquired under field operating conditions. This method can effectively reduce costs.

Description

Technical Field

[0001] This application relates to the field of non-contact strategy technology, and in particular to an error compensation method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Technology

[0002] Non-contact voltage and current measurement is typically based on physical principles such as electric field coupling and magnetic field coupling. It indirectly inverts the voltage or current value of a conductor by measuring the electric or magnetic field signals around the conductor, offering advantages such as safety, convenience, and non-destructive operation of the circuit under test. However, complex field operating environments can lead to performance degradation of ideal electromagnetic coupling models, thereby affecting measurement accuracy.

[0003] In existing technologies, most methods employ extensive field-contact data to calibrate electromagnetic coupling models in order to improve measurement accuracy. However, this approach is costly. Summary of the Invention

[0004] Therefore, it is necessary to provide a low-cost error compensation method, apparatus, computer equipment, computer-readable storage medium, and computer program product to address the aforementioned technical problems.

[0005] Firstly, this application provides an error compensation method, including:

[0006] Acquire source domain datasets and target domain datasets. The source domain dataset includes simulated non-contact measurement signals, and the target domain dataset includes actual non-contact measurement signals.

[0007] The source domain compensation model is determined based on the source domain dataset. The source domain compensation model includes a physical approximation layer and a residual compensation layer. The physical approximation layer is used to establish the mapping relationship between the simulated non-contact measurement signal and the actual non-contact measurement signal. The residual compensation layer is used to learn the residual characteristics between the output of the physical approximation layer and the actual non-contact measurement signal.

[0008] Using a transfer learning strategy, a target domain compensation model is determined based on the source domain compensation model and the target domain dataset. The target domain compensation model is used to compensate for errors in non-contact measurement signals collected in the field operating environment.

[0009] In one embodiment, determining the source domain compensation model based on the source domain dataset includes: constructing a physical approximation layer based on electromagnetic coupling relationships, and determining residual data based on the output of the physical approximation layer and the actual non-contact measurement signal; constructing a residual compensation layer based on the residual data, and determining the source domain compensation model based on the physical approximation layer and the residual compensation layer.

[0010] In one embodiment, a target domain compensation model is determined based on a source domain compensation model and a target domain dataset using a transfer learning strategy. This includes: determining an initial target domain compensation model based on the model parameters of the source domain compensation model; determining a first target domain compensation model based on the target domain dataset, the source domain dataset, and the initial target domain compensation model; determining a second target domain compensation model based on the first target domain compensation model, the source domain compensation model, and the target domain dataset; obtaining a source domain coupling matrix and a target domain coupling matrix, and determining the target domain compensation model based on the source domain coupling matrix, the target domain coupling matrix, and the second target domain compensation model.

[0011] In one embodiment, determining a first target domain compensation model based on a target domain dataset, a source domain dataset, and an initial target domain compensation model includes: performing feature extraction processing on the source domain dataset and the target domain dataset to obtain a first feature corresponding to the source domain dataset and a second feature corresponding to the target domain dataset; performing alignment processing on the first feature based on the second feature, the alignment processing including amplitude alignment, phase alignment, and harmonic alignment; and determining the first target domain compensation model based on the second feature, the aligned first feature, and the initial target domain compensation model.

[0012] In one embodiment, determining a second target domain compensation model based on a first target domain compensation model, a source domain compensation model, and a target domain dataset includes: determining source domain residual features corresponding to the source domain dataset based on the source domain compensation model, and determining target domain residual features corresponding to the target domain dataset based on the first target domain compensation model; determining the distribution difference between the source domain residual features and the target domain residual features, and performing a linear transformation on the source domain residual features based on the distribution difference; and determining the second target domain compensation model based on the linearly transformed source domain residual features, the target domain residual features, and the first target domain compensation model.

[0013] In one embodiment, determining the target domain compensation model based on the source domain coupling matrix, the target domain coupling matrix, and the second target domain compensation model includes: performing matrix alignment processing on the source domain coupling matrix and the target domain coupling matrix to obtain the target coupling matrix; and using a loss function to determine the target domain compensation model based on the target coupling matrix and the second target domain compensation model.

[0014] Secondly, this application also provides an error compensation device, comprising:

[0015] The acquisition module is used to acquire source domain datasets and target domain datasets. The source domain dataset includes simulated non-contact measurement signals, and the target domain dataset includes actual non-contact measurement signals.

[0016] The determination module is used to determine the source domain compensation model based on the source domain dataset. The source domain compensation model includes a physical approximation layer and a residual compensation layer. The physical approximation layer is used to establish the mapping relationship between the simulated non-contact measurement signal and the actual non-contact measurement signal. The residual compensation layer is used to learn the residual characteristics between the output of the physical approximation layer and the actual non-contact measurement signal.

[0017] The execution module is used to determine the target domain compensation model based on the source domain compensation model and the target domain dataset through a transfer learning strategy; the target domain compensation model is used to perform error compensation on non-contact measurement signals collected in the field operation environment.

[0018] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in any of the embodiments of the first aspect above.

[0019] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any of the embodiments of the first aspect above.

[0020] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any of the embodiments of the first aspect above.

[0021] The aforementioned error compensation method, apparatus, computer equipment, computer-readable storage medium, and computer program product first acquire a source domain dataset and a target domain dataset. The source domain dataset includes simulated non-contact measurement signals, and the target domain dataset includes actual non-contact measurement signals. Then, a source domain compensation model is determined based on the source domain dataset. The source domain compensation model includes a physical approximation layer and a residual compensation layer. The physical approximation layer is used to establish the mapping relationship between the simulated non-contact measurement signals and the actual non-contact measurement signals, and the residual compensation layer is used to learn the residual characteristics between the output of the physical approximation layer and the actual non-contact measurement signals. Finally, a target domain compensation model is determined based on the source domain compensation model and the target domain dataset using a transfer learning strategy. The target domain compensation model is used to perform error compensation on non-contact measurement signals collected in the field operating environment. The error compensation method provided in this application replaces a large amount of traditional field contact data with simulated non-contact measurement signals and combines transfer learning to reuse the source domain compensation model to determine the target domain compensation model, thereby using the target domain compensation model to perform error compensation on non-contact measurement signals, effectively reducing the cost of data acquisition and model training. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating an error compensation method in one embodiment;

[0024] Figure 2 This is a flowchart illustrating a method for determining a source domain compensation model in one embodiment;

[0025] Figure 3 This is a flowchart illustrating a method for determining a target domain compensation model in one embodiment;

[0026] Figure 4 This is a flowchart illustrating a method for determining a compensation model for a first target domain in one embodiment;

[0027] Figure 5 This is a flowchart illustrating a method for determining a second target domain compensation model in one embodiment;

[0028] Figure 6 This is a flowchart illustrating a method for determining a target domain compensation model in one embodiment;

[0029] Figure 7 This is a flowchart illustrating the error compensation method in another embodiment;

[0030] Figure 8 This is a structural block diagram of an error compensation device in one embodiment;

[0031] Figure 9 This is an internal structural diagram of a computer device in one embodiment;

[0032] Figure 10 This is a diagram of the internal structure of a computer device in another embodiment. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0034] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0035] Non-contact voltage and current measurement is typically based on physical principles such as electric field coupling and magnetic field coupling. It indirectly inverts the voltage or current value of a conductor by measuring the electric or magnetic field signals around the conductor, offering advantages such as safety, convenience, and non-destructive operation of the circuit under test. However, complex field operating environments can lead to performance degradation of ideal electromagnetic coupling models, thereby affecting measurement accuracy.

[0036] In existing technologies, most methods employ a large amount of on-site contact data to calibrate electromagnetic coupling models in order to improve measurement accuracy.

[0037] However, the need to rely on a large amount of on-site contact data necessitates the investment in contact equipment, and the complex data collection process results in high costs.

[0038] In view of this, this application provides an error compensation method. First, a source domain dataset and a target domain dataset are acquired. The source domain dataset includes simulated non-contact measurement signals, and the target domain dataset includes actual non-contact measurement signals. Then, a source domain compensation model is determined based on the source domain dataset. The source domain compensation model includes a physical approximation layer and a residual compensation layer. The physical approximation layer is used to establish the mapping relationship between the simulated and actual non-contact measurement signals, and the residual compensation layer is used to learn the residual characteristics between the output of the physical approximation layer and the actual non-contact measurement signals. Finally, a target domain compensation model is determined based on the source domain compensation model and the target domain dataset using a transfer learning strategy. The target domain compensation model is used to perform error compensation on non-contact measurement signals collected in the field operating environment. The error compensation method provided in this application replaces a large amount of traditional field contact data with simulated non-contact measurement signals and combines transfer learning to reuse the source domain compensation model to determine the target domain compensation model, thereby using the target domain compensation model to perform error compensation on non-contact measurement signals, effectively reducing the cost of data acquisition and model training.

[0039] The error compensation method provided in this application can be implemented by a computer device, which can be a terminal or a server.

[0040] In one exemplary embodiment, such as Figure 1 As shown, an error compensation method is provided, which includes the following steps:

[0041] Step 101: Obtain the source domain dataset and the target domain dataset.

[0042] The source domain dataset may include simulated non-contact measurement signals, and the target domain dataset may include actual non-contact measurement signals.

[0043] For example, simulated non-contact measurement signals can be non-contact voltage / current analog signals. Actual non-contact measurement signals can be voltage / current sensing signals directly acquired from real-world operating scenarios.

[0044] In some exemplary embodiments, the computer device may acquire a source domain dataset and a target domain dataset.

[0045] Specifically, for source domain datasets, computer equipment can construct a non-contact measurement simulation environment using finite element simulation software. Based on the principle of electromagnetic coupling, it can simulate different load conditions, sensor installation angles, ambient temperature and humidity, and electromagnetic interference scenarios to generate simulated non-contact measurement signals in batches and simultaneously preset the corresponding actual measurement values ​​as labels to obtain source domain datasets.

[0046] For the target domain dataset, computer equipment can acquire non-contact measurement signals from actual operating scenarios such as power distribution networks by using non-contact sensors deployed on-site.

[0047] Step 102: Determine the source domain compensation model based on the source domain dataset.

[0048] Optionally, the source domain compensation model may include a physical approximation layer and a residual compensation layer.

[0049] For example, the physical approximation layer is used to establish the mapping relationship between the simulated non-contact measurement signal and the actual non-contact measurement signal. The residual compensation layer is used to learn the residual characteristics between the output of the physical approximation layer and the actual non-contact measurement signal.

[0050] The specific network structures of the residual compensation layer include, but are not limited to, fully connected neural networks, convolutional neural networks, recurrent neural networks, and Transformers.

[0051] In some exemplary embodiments, after acquiring the source domain dataset and the target domain dataset, the computer device can determine the source domain compensation model based on the source domain dataset.

[0052] Specifically, the computer equipment can construct a physical approximation layer based on simulated non-contact measurement signals and actual non-contact measurement signals, and construct a residual compensation layer based on the output of the physical approximation layer and the actual non-contact measurement signals to obtain a source domain compensation model.

[0053] Step 103: Determine the target domain compensation model based on the source domain compensation model and the target domain dataset using a transfer learning strategy.

[0054] Among them, the target domain compensation model can be used to compensate for errors in non-contact measurement signals collected in the field operating environment.

[0055] In some exemplary embodiments, after determining the source domain compensation model based on the source domain dataset, the computer device can determine the target domain compensation model based on the source domain compensation model and the target domain dataset through a transfer learning strategy.

[0056] Specifically, computer devices can use transfer learning strategies to take the model parameters of the source domain compensation model as the initial model parameters of the target domain compensation model, and then use the target domain dataset to optimize and adjust the initial model parameters of the target domain compensation model to obtain the target domain compensation model.

[0057] The aforementioned error compensation method first acquires a source domain dataset and a target domain dataset. The source domain dataset includes simulated non-contact measurement signals, and the target domain dataset includes actual non-contact measurement signals. Then, a source domain compensation model is determined based on the source domain dataset. The source domain compensation model includes a physical approximation layer and a residual compensation layer. The physical approximation layer is used to establish the mapping relationship between the simulated non-contact measurement signals and the actual non-contact measurement signals, and the residual compensation layer is used to learn the residual characteristics between the output of the physical approximation layer and the actual non-contact measurement signals. Finally, a target domain compensation model is determined based on the source domain compensation model and the target domain dataset using a transfer learning strategy. The target domain compensation model is used to perform error compensation on non-contact measurement signals collected in the field operating environment. The error compensation method provided in this application replaces a large amount of traditional field contact data with simulated non-contact measurement signals and combines transfer learning to reuse the source domain compensation model to determine the target domain compensation model, thereby using the target domain compensation model to perform error compensation on non-contact measurement signals, effectively reducing the cost of data acquisition and model training.

[0058] In one exemplary embodiment, such as Figure 2 As shown, determining the source domain compensation model based on the source domain dataset includes the following steps:

[0059] Step 201: Construct a physical approximation layer based on the electromagnetic coupling relationship, and determine the residual data based on the output of the physical approximation layer and the actual non-contact measurement signal.

[0060] Alternatively, the electromagnetic coupling relationship can be based on the principle of electrostatic induction, Faraday's law of electromagnetic induction, etc.

[0061] In some exemplary embodiments, the computer device may construct a physical approximation layer based on electromagnetic coupling relationships.

[0062] Specifically, the computer equipment can select the corresponding electromagnetic coupling relationship according to the measurement type. For non-contact voltage measurement, a mapping function U=k1×S+b2 can be constructed based on the principle of electrostatic induction, where S is the amplitude of the simulated non-contact voltage signal in the source domain dataset, k1 is the relevant parameter of the sensor induction coefficient, and b is the signal offset. For non-contact current measurement, a mapping function I=k2×(dx / dt)+b2 can be constructed based on Faraday's law of electromagnetic induction, where dx / dt is the rate of change of magnetic flux corresponding to the simulated non-contact current signal, k2 is the relevant proportional parameter of the number of coil turns, and b2 is the baseline offset.

[0063] Then, the computer device can use the simulated non-contact measurement signals and corresponding actual measurement values ​​(labels) in the source domain dataset to fit and solve for the optimal values ​​of k1, b1 or k2, b2 through the least squares method or gradient descent method, so as to minimize the initial error between the output of the physical approximation layer and the actual measurement value, and complete the construction of the physical approximation layer.

[0064] Furthermore, after constructing a physical approximation layer based on electromagnetic coupling, the computer equipment can determine the residual data based on the output of the physical approximation layer and the actual non-contact measurement signal.

[0065] Specifically, the computer device can input all simulated non-contact measurement signals from the source domain dataset one by one into the constructed physical approximation layer to obtain the corresponding approximate output values. Then, it calls the actual non-contact measurement signals associated with the simulated signals from the source domain dataset to calculate the difference, which is the residual data. This residual dataset contains nonlinear errors that the physical approximation layer failed to fit, such as deviations caused by the nonlinear characteristics of the sensor itself, residual errors caused by environmental temperature and humidity interference in the simulated scenario, and systematic errors caused by installation angle deviations.

[0066] Step 202: Construct a residual compensation layer based on the residual data, and determine the source domain compensation model based on the physical approximation layer and the residual compensation layer.

[0067] In some exemplary embodiments, after determining the residual data based on the output of the physical approximation layer and the actual non-contact measurement signal, the computer device can construct a residual compensation layer based on the residual data.

[0068] Specifically, the computer equipment can first perform feature extraction processing on the residual data to construct a residual feature vector. For example, key features such as the rate of change of residual amplitude, time-series fluctuation characteristics, and frequency domain harmonic components can be extracted. Then, a residual compensation layer is built using a deep learning network. The input of this network is the residual feature vector, and the output is the compensation value for the residual.

[0069] Taking a convolutional neural network as an example, a convolutional neural network can be set with 2 convolutional layers, 1 pooling layer and 2 fully connected layers. The training objective is to minimize the mean square error between the compensation value and the real residual data. The network weight parameters are updated iteratively using the stochastic gradient descent algorithm. The learning rate is set to 0.001 and the number of iterations is 100. When the loss function value is lower than the preset threshold, the model converges to obtain the residual compensation layer.

[0070] Furthermore, after constructing a residual compensation layer based on residual data, the computer device can determine the source domain compensation model based on the physical approximation layer and the residual compensation layer.

[0071] Specifically, computer equipment can connect and integrate the physical approximation layer and the residual compensation layer in series to obtain the source domain compensation model.

[0072] In one exemplary embodiment, such as Figure 3 As shown, the target domain compensation model is determined based on the source domain compensation model and the target domain dataset using a transfer learning strategy, including the following steps:

[0073] Step 301: Determine the initial target domain compensation model based on the model parameters of the source domain compensation model.

[0074] In some exemplary embodiments, the computer device can determine the initial target domain compensation model based on the model parameters of the source domain compensation model.

[0075] Specifically, computer equipment can directly reuse the model parameters of the source domain compensation model into the target domain compensation model framework to obtain the initial target domain compensation model.

[0076] Step 302: Determine the first target domain compensation model based on the target domain dataset, the source domain dataset, and the initial target domain compensation model.

[0077] In some exemplary embodiments, after determining the initial target domain compensation model based on the model parameters of the source domain compensation model, the computer device can determine the first target domain compensation model based on the target domain dataset, the source domain dataset, and the initial target domain compensation model.

[0078] Specifically, the computer device can input the source domain dataset and the target domain dataset into the initial target domain compensation model, use the source domain data labels as supervision, adjust the model parameters, and iterate until the model error converges to obtain the first target domain compensation model.

[0079] Step 303: Determine the second target domain compensation model based on the first target domain compensation model, the source domain compensation model, and the target domain dataset.

[0080] In some exemplary embodiments, after determining a first target domain compensation model based on the target domain dataset, the source domain dataset, and the initial target domain compensation model, the computer device may determine a second target domain compensation model based on the first target domain compensation model, the source domain compensation model, and the target domain dataset.

[0081] Specifically, the computer device can extract source domain residual features through the source domain compensation model, extract the residual features corresponding to the target domain dataset through the first target domain compensation model, adjust the residual compensation layer parameters of the first target domain compensation model with the goal of matching the distributions of the two, and iterate until the feature distribution difference converges to obtain the second target domain compensation model.

[0082] Step 304: Obtain the source domain coupling matrix and the target domain coupling matrix, and determine the target domain compensation model based on the source domain coupling matrix, the target domain coupling matrix, and the second target domain compensation model.

[0083] In some exemplary embodiments, after determining the second target domain compensation model based on the first target domain compensation model, the source domain compensation model, and the target domain dataset, the computer device can obtain the source domain coupling relationship matrix and the target domain coupling relationship matrix.

[0084] Specifically, computer equipment can generate a source domain coupling matrix based on parameters such as network topology and line impedance in the source domain simulation environment, and extract a target domain coupling matrix from field operation data.

[0085] Furthermore, after obtaining the source domain coupling matrix and the target domain coupling matrix, the computer device can determine the target domain compensation model based on the source domain coupling matrix, the target domain coupling matrix, and the second target domain compensation model.

[0086] Specifically, the computer equipment can adapt and align the source domain coupling matrix and the target domain coupling matrix. Using the alignment result as a physical constraint, the relevant parameters of the second target domain compensation model are adjusted to ensure that the model output conforms to the on-site power grid operation rules. After iterating until the constraints are satisfied and the error converges, the target domain compensation model is obtained.

[0087] In one exemplary embodiment, such as Figure 4 As shown, the first target domain compensation model is determined based on the target domain dataset, the source domain dataset, and the initial target domain compensation model, including the following steps:

[0088] Step 401: Perform feature extraction processing on the source domain dataset and the target domain dataset to obtain the first feature corresponding to the source domain dataset and the second feature corresponding to the target domain dataset.

[0089] In some exemplary embodiments, a computer device may perform feature extraction processing on a source domain dataset and a target domain dataset to obtain a first feature corresponding to the source domain dataset and a second feature corresponding to the target domain dataset.

[0090] Specifically, the computer equipment can first preprocess the simulated non-contact measurement signals of the source domain dataset and the actual non-contact measurement signals of the target domain dataset using wavelet transform or moving average algorithms to remove random noise and spike interference from the signals. Then, signal processing algorithms can simultaneously extract the time-domain and frequency-domain features of the two datasets. The time-domain features can include the signal's peak value, effective value (root mean square value), average value, peak factor, waveform distortion rate, rising / falling slope, pulse width, and other dynamic features. The frequency-domain features are extracted after converting the preprocessed signal from the time domain to the frequency domain using Fast Fourier Transform (FFT), covering the fundamental frequency, fundamental amplitude, amplitude and phase of the 3rd, 5th, and 7th harmonics, total harmonic distortion rate (and frequency domain energy distribution entropy, etc.). Finally, the extracted time-domain and frequency-domain features are concatenated in a preset order to form a source domain first feature vector and a target domain second feature vector with completely consistent dimensions and types.

[0091] Step 402: Align the first feature with the second feature as a reference.

[0092] Alignment processing can include amplitude alignment, phase alignment, and harmonic alignment.

[0093] In some exemplary embodiments, after obtaining the first feature corresponding to the source domain dataset and the second feature corresponding to the target domain dataset, the computer device can perform alignment processing on the first feature based on the second feature.

[0094] Specifically, the computer equipment can use the second feature of the target domain as a reference benchmark. First, by calculating the mean and standard deviation of the amplitude of the second feature, the amplitude of the first feature of the source domain is linearly scaled or translated to achieve a uniform amplitude distribution between the two. Then, for the fundamental wave and the 3rd, 5th, and 7th harmonic phases in the frequency domain features, the phase deviation between the source domain and the target domain is calculated, and the corresponding phases of the first feature are corrected to eliminate phase shift and complete phase alignment. Finally, the amplitude proportion of each harmonic in the first feature is adjusted so that the energy distribution of the first harmonic, each harmonic, and the total harmonic distortion rate are consistent with the second feature, thus achieving harmonic alignment.

[0095] Step 403: Determine the first target domain compensation model based on the second feature, the first feature after alignment processing, and the initial target domain compensation model.

[0096] In some exemplary embodiments, after aligning the first feature, the computer device can determine the first target domain compensation model based on the second feature, the aligned first feature, and the initial target domain compensation model.

[0097] Specifically, the computer device can input the second feature and the aligned first feature into the initial target domain compensation model, use the actual measurement value in the source domain dataset as the supervision signal, and use a small learning rate to adjust the model parameters. During the process, the core parameters of the physical approximation layer are fixed to preserve the physical laws of electromagnetic coupling, and only the weight parameters of the residual compensation layer are iteratively updated. When the loss function of the predicted value output by the model and the supervision signal drops below the preset threshold and tends to stabilize, the iteration stops, and the first target domain compensation model is obtained.

[0098] In one exemplary embodiment, such as Figure 5 As shown, the second target domain compensation model is determined based on the first target domain compensation model, the source domain compensation model, and the target domain dataset, including the following steps:

[0099] Step 501: Determine the source domain residual features corresponding to the source domain dataset based on the source domain compensation model, and determine the target domain residual features corresponding to the target domain dataset based on the first target domain compensation model.

[0100] In some exemplary embodiments, a computer device can determine the source domain residual characteristics corresponding to the source domain dataset based on a source domain compensation model.

[0101] Specifically, the computer equipment can input the simulated non-contact measurement signal of the source domain dataset into the trained source domain compensation model, pass through the physical approximation layer to obtain the basic approximate output value, and then pass through the residual compensation layer to obtain the compensation correction value. The difference between the final output value of the model and the preset actual measurement value in the source domain dataset is calculated, and key attributes such as the amplitude change rate, time-series fluctuation characteristics and frequency domain harmonic distribution of the difference are extracted to obtain the source domain residual characteristics.

[0102] Furthermore, the computer device can also determine the target domain residual characteristics corresponding to the target domain dataset based on the first target domain compensation model.

[0103] Specifically, the computer equipment can input the actual non-contact measurement signal of the target domain dataset into the first target domain compensation model. After processing by the physical approximation layer and the residual compensation layer, the model output value is obtained. The difference between the output value and the original non-contact measurement signal of the target domain is calculated, and the time domain and frequency domain features corresponding to the difference are extracted to obtain the target domain residual features.

[0104] Step 502: Determine the distribution difference between the source domain residual features and the target domain residual features, and perform a linear transformation on the source domain residual features based on the distribution difference.

[0105] In some exemplary embodiments, the computer device can determine the source domain residual characteristics and the target domain residual characteristics after determining the source domain residual characteristics and the target domain residual characteristics.

[0106] Specifically, computer equipment can calculate the statistical distribution parameters such as the mean, variance, and covariance of the source domain residual features and the target domain residual features, and use indicators such as KL divergence or mean square error to quantify the distribution differences of the two types of residual features in the time domain and frequency domain, thereby determining the degree of distribution shift between the source domain and the target domain residual features.

[0107] Furthermore, after determining the distributional differences between the source domain residual features and the target domain residual features, the computer device can perform a linear transformation on the source domain residual features based on the distributional differences.

[0108] Specifically, computer equipment can determine linear transformation coefficients, also known as scaling and translation factors, based on distribution differences. Using the distribution parameters of the target domain residual features as a benchmark, a linear transformation operation is performed on the source domain residual features. By adjusting the amplitude scaling ratio and baseline offset of the source domain residual features, the statistical distribution of the transformed source domain residual features is made consistent with that of the target domain residual features.

[0109] Step 503: Determine the second target domain compensation model based on the source domain residual characteristics, target domain residual characteristics, and the first target domain compensation model after linear transformation.

[0110] In some exemplary embodiments, after performing a linear transformation on the source domain residual features based on the distribution differences, the computer device can determine a second target domain compensation model based on the linearly transformed source domain residual features, the target domain residual features, and the first target domain compensation model.

[0111] Specifically, the computer equipment can use the source domain residual features and target domain residual features after linear transformation as joint training samples and input them into the first target domain compensation model. During the training process, the core parameters of the physical approximation layer are fixed to maintain the consistency of the electromagnetic coupling physical laws. The weight parameters of the residual compensation layer are iteratively updated in small increments with the goal of achieving the optimal matching degree of the residual feature distribution. The gradient descent algorithm is used to gradually reduce the model output error after the input of the two types of residual features. When the loss function value drops to a preset threshold and tends to stabilize, the iteration stops to obtain the second target domain compensation model.

[0112] In one exemplary embodiment, such as Figure 6 As shown, the target domain compensation model is determined based on the source domain coupling matrix, the target domain coupling matrix, and the second target domain compensation model, including the following steps:

[0113] Step 601: Perform matrix alignment on the source domain coupling matrix and the target domain coupling matrix to obtain the target coupling matrix.

[0114] In some exemplary embodiments, a computer device may perform matrix alignment processing on the source domain coupling matrix and the target domain coupling matrix to obtain the target coupling matrix.

[0115] Specifically, the computer equipment can first unify the dimensions of the source domain coupling relationship matrix and the target domain coupling relationship matrix, remove redundant topological nodes specific to the simulation scenario in the source domain matrix, and supplement the parameters of the branch nodes that actually exist in the target domain to match the dimensions; then, through the eigenvalue matching algorithm, based on the topological structure and coupling coefficient distribution of the target domain coupling relationship matrix, the corresponding element values ​​of the source domain matrix are corrected to eliminate the deviation between the simulation and the field in terms of electrical coupling strength and node connection relationship, and complete the matrix alignment process to obtain a target coupling relationship matrix that is consistent with the actual electrical characteristics of the field.

[0116] Step 602: Using the loss function, determine the target domain compensation model based on the target coupling relationship matrix and the second target domain compensation model.

[0117] In some exemplary embodiments, the computer device may obtain the target coupling relationship matrix after obtaining the target coupling relationship matrix.

[0118] Specifically, the computer equipment can construct a loss function that incorporates physical constraints. This function includes a model prediction error term and a coupling constraint term. The model prediction error term is the mean square error between the output value of the second target domain compensation model calculated based on the target domain dataset and the actual non-contact measurement signal. The coupling constraint term, based on the target coupling matrix, quantifies the deviation between the model output and the on-site electrical coupling law, ensuring that the model output conforms to the coupling logic corresponding to the impedance and mutual inductance coefficient of the distribution network nodes. Then, the target coupling matrix and the target domain dataset are input into the second target domain compensation model. With the goal of minimizing the loss function value, the parameters of the model residual compensation layer are adjusted. The parameters are iteratively updated using a gradient descent algorithm until the loss function value drops to a preset threshold and converges stably, thus obtaining the target domain compensation model.

[0119] In one exemplary embodiment, such as Figure 7 As shown, another error compensation method is provided, which includes the following steps:

[0120] Step 701: Obtain the source domain dataset and the target domain dataset. The source domain dataset includes simulated non-contact measurement signals, and the target domain dataset includes actual non-contact measurement signals. Construct a physical approximation layer based on the electromagnetic coupling relationship, and determine the residual data based on the output of the physical approximation layer and the actual non-contact measurement signals. Construct a residual compensation layer based on the residual data, and determine the source domain compensation model based on the physical approximation layer and the residual compensation layer. The source domain compensation model includes a physical approximation layer and a residual compensation layer. The physical approximation layer is used to establish the mapping relationship between the simulated non-contact measurement signals and the actual non-contact measurement signals, and the residual compensation layer is used to learn the residual characteristics between the output of the physical approximation layer and the actual non-contact measurement signals.

[0121] Step 702: Determine the initial target domain compensation model based on the model parameters of the source domain compensation model; perform feature extraction processing on the source domain dataset and the target domain dataset to obtain the first feature corresponding to the source domain dataset and the second feature corresponding to the target domain dataset; use the second feature as a reference to perform alignment processing on the first feature, including amplitude alignment, phase alignment, and harmonic alignment; determine the first target domain compensation model based on the second feature, the aligned first feature, and the initial target domain compensation model.

[0122] Step 703: Determine the source domain residual features corresponding to the source domain dataset based on the source domain compensation model, and determine the target domain residual features corresponding to the target domain dataset based on the first target domain compensation model; determine the distribution difference between the source domain residual features and the target domain residual features, and perform a linear transformation on the source domain residual features based on the distribution difference; determine the second target domain compensation model based on the linearly transformed source domain residual features, target domain residual features, and the first target domain compensation model.

[0123] Step 704: Obtain the source domain coupling matrix and the target domain coupling matrix, perform matrix alignment on the source domain coupling matrix and the target domain coupling matrix to obtain the target coupling matrix; use the loss function to determine the target domain compensation model based on the target coupling matrix and the second target domain compensation model; the target domain compensation model is used to compensate for errors in non-contact measurement signals collected in the field operating environment.

[0124] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0125] Based on the same inventive concept, this application also provides an error compensation device for implementing the error compensation method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more error compensation device embodiments provided below can be found in the limitations of the error compensation method described above, and will not be repeated here.

[0126] In one exemplary embodiment, such as Figure 8 As shown, an error compensation device 800 is provided, including: an acquisition module 801, a determination module 802, and an execution module 803, wherein:

[0127] The acquisition module 801 is used to acquire a source domain dataset and a target domain dataset. The source domain dataset includes simulated non-contact measurement signals, and the target domain dataset includes actual non-contact measurement signals.

[0128] The determination module 802 is used to determine the source domain compensation model based on the source domain dataset. The source domain compensation model includes a physical approximation layer and a residual compensation layer. The physical approximation layer is used to establish the mapping relationship between the simulated non-contact measurement signal and the actual non-contact measurement signal. The residual compensation layer is used to learn the residual characteristics between the output of the physical approximation layer and the actual non-contact measurement signal.

[0129] The execution module 803 is used to determine the target domain compensation model based on the source domain compensation model and the target domain dataset through a transfer learning strategy; the target domain compensation model is used to perform error compensation on non-contact measurement signals collected in the field operating environment.

[0130] In one embodiment, the determining module 802 is specifically used to construct a physical approximation layer based on the electromagnetic coupling relationship, and to determine residual data based on the output of the physical approximation layer and the actual non-contact measurement signal; to construct a residual compensation layer based on the residual data, and to determine a source domain compensation model based on the physical approximation layer and the residual compensation layer.

[0131] In one embodiment, the execution module 803 is specifically used to determine an initial target domain compensation model based on the model parameters of the source domain compensation model; determine a first target domain compensation model based on the target domain dataset, the source domain dataset, and the initial target domain compensation model; determine a second target domain compensation model based on the first target domain compensation model, the source domain compensation model, and the target domain dataset; obtain the source domain coupling relationship matrix and the target domain coupling relationship matrix, and determine the target domain compensation model based on the source domain coupling relationship matrix, the target domain coupling relationship matrix, and the second target domain compensation model.

[0132] In one embodiment, the execution module 803 is specifically used to perform feature extraction processing on the source domain dataset and the target domain dataset to obtain a first feature corresponding to the source domain dataset and a second feature corresponding to the target domain dataset; based on the second feature, perform alignment processing on the first feature, the alignment processing including amplitude alignment, phase alignment and harmonic alignment; and determine a first target domain compensation model based on the second feature, the aligned first feature and the initial target domain compensation model.

[0133] In one embodiment, the execution module 803 is specifically used to determine the source domain residual features corresponding to the source domain dataset based on the source domain compensation model, and to determine the target domain residual features corresponding to the target domain dataset based on the first target domain compensation model; to determine the distribution difference between the source domain residual features and the target domain residual features, and to perform a linear transformation on the source domain residual features based on the distribution difference; and to determine a second target domain compensation model based on the linearly transformed source domain residual features, the target domain residual features, and the first target domain compensation model.

[0134] In one embodiment, the execution module 803 is specifically used to perform matrix alignment processing on the source domain coupling relationship matrix and the target domain coupling relationship matrix to obtain the target coupling relationship matrix; and to determine the target domain compensation model based on the target coupling relationship matrix and the second target domain compensation model using a loss function.

[0135] Each module in the aforementioned error compensation device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0136] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9 As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements an error compensation method.

[0137] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10 As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements an error compensation method.

[0138] Those skilled in the art will understand that Figure 9 and Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0139] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0140] Acquire source domain datasets and target domain datasets. The source domain dataset includes simulated non-contact measurement signals, and the target domain dataset includes actual non-contact measurement signals.

[0141] The source domain compensation model is determined based on the source domain dataset. The source domain compensation model includes a physical approximation layer and a residual compensation layer. The physical approximation layer is used to establish the mapping relationship between the simulated non-contact measurement signal and the actual non-contact measurement signal. The residual compensation layer is used to learn the residual characteristics between the output of the physical approximation layer and the actual non-contact measurement signal.

[0142] Using a transfer learning strategy, a target domain compensation model is determined based on the source domain compensation model and the target domain dataset. The target domain compensation model is used to compensate for errors in non-contact measurement signals collected in the field operating environment.

[0143] In one embodiment, when the processor executes the computer program, it further performs the following steps: constructing a physical approximation layer based on the electromagnetic coupling relationship, and determining residual data based on the output of the physical approximation layer and the actual non-contact measurement signal; constructing a residual compensation layer based on the residual data, and determining a source domain compensation model based on the physical approximation layer and the residual compensation layer.

[0144] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining an initial target domain compensation model based on the model parameters of the source domain compensation model; determining a first target domain compensation model based on the target domain dataset, the source domain dataset, and the initial target domain compensation model; determining a second target domain compensation model based on the first target domain compensation model, the source domain compensation model, and the target domain dataset; obtaining the source domain coupling matrix and the target domain coupling matrix, and determining the target domain compensation model based on the source domain coupling matrix, the target domain coupling matrix, and the second target domain compensation model.

[0145] In one embodiment, when the processor executes the computer program, it further performs the following steps: performing feature extraction processing on the source domain dataset and the target domain dataset to obtain a first feature corresponding to the source domain dataset and a second feature corresponding to the target domain dataset; performing alignment processing on the first feature based on the second feature, the alignment processing including amplitude alignment, phase alignment and harmonic alignment; and determining a first target domain compensation model based on the second feature, the aligned first feature and the initial target domain compensation model.

[0146] In one embodiment, when the processor executes the computer program, it further performs the following steps: determining the source domain residual features corresponding to the source domain dataset based on the source domain compensation model, and determining the target domain residual features corresponding to the target domain dataset based on the first target domain compensation model; determining the distribution difference between the source domain residual features and the target domain residual features, and performing a linear transformation on the source domain residual features based on the distribution difference; and determining a second target domain compensation model based on the linearly transformed source domain residual features, the target domain residual features, and the first target domain compensation model.

[0147] In one embodiment, when the processor executes the computer program, it further performs the following steps: performing matrix alignment processing on the source domain coupling matrix and the target domain coupling matrix to obtain the target coupling matrix; and using a loss function to determine the target domain compensation model based on the target coupling matrix and the second target domain compensation model.

[0148] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0149] Acquire source domain datasets and target domain datasets. The source domain dataset includes simulated non-contact measurement signals, and the target domain dataset includes actual non-contact measurement signals.

[0150] The source domain compensation model is determined based on the source domain dataset. The source domain compensation model includes a physical approximation layer and a residual compensation layer. The physical approximation layer is used to establish the mapping relationship between the simulated non-contact measurement signal and the actual non-contact measurement signal. The residual compensation layer is used to learn the residual characteristics between the output of the physical approximation layer and the actual non-contact measurement signal.

[0151] Using a transfer learning strategy, a target domain compensation model is determined based on the source domain compensation model and the target domain dataset. The target domain compensation model is used to compensate for errors in non-contact measurement signals collected in the field operating environment.

[0152] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: constructing a physical approximation layer based on the electromagnetic coupling relationship, and determining residual data based on the output of the physical approximation layer and the actual non-contact measurement signal; constructing a residual compensation layer based on the residual data, and determining a source domain compensation model based on the physical approximation layer and the residual compensation layer.

[0153] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining an initial target domain compensation model based on the model parameters of the source domain compensation model; determining a first target domain compensation model based on the target domain dataset, the source domain dataset, and the initial target domain compensation model; determining a second target domain compensation model based on the first target domain compensation model, the source domain compensation model, and the target domain dataset; obtaining the source domain coupling relationship matrix and the target domain coupling relationship matrix, and determining the target domain compensation model based on the source domain coupling relationship matrix, the target domain coupling relationship matrix, and the second target domain compensation model.

[0154] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: performing feature extraction processing on the source domain dataset and the target domain dataset to obtain a first feature corresponding to the source domain dataset and a second feature corresponding to the target domain dataset; performing alignment processing on the first feature based on the second feature, the alignment processing including amplitude alignment, phase alignment and harmonic alignment; and determining a first target domain compensation model based on the second feature, the aligned first feature and the initial target domain compensation model.

[0155] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining the source domain residual features corresponding to the source domain dataset based on the source domain compensation model, and determining the target domain residual features corresponding to the target domain dataset based on the first target domain compensation model; determining the distribution difference between the source domain residual features and the target domain residual features, and performing a linear transformation on the source domain residual features based on the distribution difference; and determining a second target domain compensation model based on the linearly transformed source domain residual features, the target domain residual features, and the first target domain compensation model.

[0156] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: performing matrix alignment processing on the source domain coupling matrix and the target domain coupling matrix to obtain the target coupling matrix; and using a loss function to determine the target domain compensation model based on the target coupling matrix and the second target domain compensation model.

[0157] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0158] Acquire source domain datasets and target domain datasets. The source domain dataset includes simulated non-contact measurement signals, and the target domain dataset includes actual non-contact measurement signals.

[0159] The source domain compensation model is determined based on the source domain dataset. The source domain compensation model includes a physical approximation layer and a residual compensation layer. The physical approximation layer is used to establish the mapping relationship between the simulated non-contact measurement signal and the actual non-contact measurement signal. The residual compensation layer is used to learn the residual characteristics between the output of the physical approximation layer and the actual non-contact measurement signal.

[0160] Using a transfer learning strategy, a target domain compensation model is determined based on the source domain compensation model and the target domain dataset. The target domain compensation model is used to compensate for errors in non-contact measurement signals collected in the field operating environment.

[0161] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: constructing a physical approximation layer based on the electromagnetic coupling relationship, and determining residual data based on the output of the physical approximation layer and the actual non-contact measurement signal; constructing a residual compensation layer based on the residual data, and determining a source domain compensation model based on the physical approximation layer and the residual compensation layer.

[0162] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining an initial target domain compensation model based on the model parameters of the source domain compensation model; determining a first target domain compensation model based on the target domain dataset, the source domain dataset, and the initial target domain compensation model; determining a second target domain compensation model based on the first target domain compensation model, the source domain compensation model, and the target domain dataset; obtaining the source domain coupling relationship matrix and the target domain coupling relationship matrix, and determining the target domain compensation model based on the source domain coupling relationship matrix, the target domain coupling relationship matrix, and the second target domain compensation model.

[0163] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: performing feature extraction processing on the source domain dataset and the target domain dataset to obtain a first feature corresponding to the source domain dataset and a second feature corresponding to the target domain dataset; performing alignment processing on the first feature based on the second feature, the alignment processing including amplitude alignment, phase alignment and harmonic alignment; and determining a first target domain compensation model based on the second feature, the aligned first feature and the initial target domain compensation model.

[0164] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: determining the source domain residual features corresponding to the source domain dataset based on the source domain compensation model, and determining the target domain residual features corresponding to the target domain dataset based on the first target domain compensation model; determining the distribution difference between the source domain residual features and the target domain residual features, and performing a linear transformation on the source domain residual features based on the distribution difference; and determining a second target domain compensation model based on the linearly transformed source domain residual features, the target domain residual features, and the first target domain compensation model.

[0165] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: performing matrix alignment processing on the source domain coupling matrix and the target domain coupling matrix to obtain the target coupling matrix; and using a loss function to determine the target domain compensation model based on the target coupling matrix and the second target domain compensation model.

[0166] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0167] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0168] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An error compensation method, characterized in that, The method includes: Obtain a source domain dataset and a target domain dataset, wherein the source domain dataset includes simulated non-contact measurement signals and the target domain dataset includes actual non-contact measurement signals; A source domain compensation model is determined based on the source domain dataset. The source domain compensation model includes a physical approximation layer and a residual compensation layer. The physical approximation layer is used to establish a mapping relationship between the simulated non-contact measurement signal and the actual non-contact measurement signal. The residual compensation layer is used to learn the residual characteristics between the output of the physical approximation layer and the actual non-contact measurement signal. Using a transfer learning strategy, a target domain compensation model is determined based on the source domain compensation model and the target domain dataset; the target domain compensation model is used to compensate for errors in non-contact measurement signals collected in the field operating environment.

2. The method according to claim 1, characterized in that, The step of determining the source domain compensation model based on the source domain dataset includes: The physical approximation layer is constructed based on the electromagnetic coupling relationship, and the residual data is determined based on the output of the physical approximation layer and the actual non-contact measurement signal. The residual compensation layer is constructed based on the residual data, and the source domain compensation model is determined based on the physical approximation layer and the residual compensation layer.

3. The method according to claim 1 or 2, characterized in that, The step of determining the target domain compensation model based on the source domain compensation model and the target domain dataset using a transfer learning strategy includes: The initial target domain compensation model is determined based on the model parameters of the source domain compensation model; A first target domain compensation model is determined based on the target domain dataset, the source domain dataset, and the initial target domain compensation model. A second target domain compensation model is determined based on the first target domain compensation model, the source domain compensation model, and the target domain dataset. Obtain the source domain coupling matrix and the target domain coupling matrix, and determine the target domain compensation model based on the source domain coupling matrix, the target domain coupling matrix, and the second target domain compensation model.

4. The method according to claim 3, characterized in that, The step of determining the first target domain compensation model based on the target domain dataset, the source domain dataset, and the initial target domain compensation model includes: Feature extraction processing is performed on the source domain dataset and the target domain dataset to obtain a first feature corresponding to the source domain dataset and a second feature corresponding to the target domain dataset; Based on the second feature, the first feature is aligned, and the alignment process includes amplitude alignment, phase alignment, and harmonic alignment. The first target domain compensation model is determined based on the second feature, the first feature after alignment processing, and the initial target domain compensation model.

5. The method according to claim 3, characterized in that, The step of determining the second target domain compensation model based on the first target domain compensation model, the source domain compensation model, and the target domain dataset includes: Based on the source domain compensation model, the source domain residual features corresponding to the source domain dataset are determined, and based on the first target domain compensation model, the target domain residual features corresponding to the target domain dataset are determined. Determine the distribution difference between the source domain residual features and the target domain residual features, and perform a linear transformation on the source domain residual features based on the distribution difference; The second target domain compensation model is determined based on the source domain residual characteristics after linear transformation, the target domain residual characteristics, and the first target domain compensation model.

6. The method according to claim 3, characterized in that, The step of determining the target domain compensation model based on the source domain coupling matrix, the target domain coupling matrix, and the second target domain compensation model includes: The source domain coupling matrix and the target domain coupling matrix are aligned to obtain the target coupling matrix. The target domain compensation model is determined using the loss function, based on the target coupling relationship matrix and the second target domain compensation model.

7. An error compensation device, characterized in that, The device includes: The acquisition module is used to acquire a source domain dataset and a target domain dataset, wherein the source domain dataset includes simulated non-contact measurement signals and the target domain dataset includes actual non-contact measurement signals. A determination module is used to determine a source domain compensation model based on the source domain dataset; the source domain compensation model includes a physical approximation layer and a residual compensation layer, the physical approximation layer is used to establish a mapping relationship between the simulated non-contact measurement signal and the actual non-contact measurement signal, and the residual compensation layer is used to learn the residual characteristics between the output of the physical approximation layer and the actual non-contact measurement signal; The execution module is used to determine the target domain compensation model based on the source domain compensation model and the target domain dataset through a transfer learning strategy; the target domain compensation model is used to perform error compensation on non-contact measurement signals collected in the field operation environment.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images