A direct current method, device, equipment and medium for denoising time-domain data

Abstract

The present invention provides a method, device, equipment and medium for denoising time-domain data of direct current electricity method, which relates to the field of data processing technology and aims to alleviate the problem of limited denoising effect in the existing technology. The present invention generates a highly simulated and diversified direct current electricity method time-domain potential difference data set by combining the actual ground electric field noise characteristics; normalizes the data set and uses a learning curve to determine the number of samples used for network training; determines the hyperparameter combination of the network through the Bayesian optimization method, and constructs a width learning network model for denoising time-domain potential difference data, thereby achieving effective denoising of direct current electricity method time-domain potential difference data. By integrating the constructed denoising model with time-domain potential difference data acquisition, a new mode of direct current electricity method data acquisition and processing is formed to improve the quality of observation data, thereby enhancing the application effect of direct current electricity method in near-surface exploration.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a direct current method, device, equipment and medium for denoising time-domain data. Background Art

[0002] The DC resistivity method is one of the important technologies in near-surface engineering surveys, resource exploration, and geological disaster monitoring. However, with the increase and widespread distribution of human electromagnetic interference, its observation data is facing increasingly serious noise interference, which reduces the detection accuracy.

[0003] In the existing technology, denoising is mainly performed on apparent resistivity data after DC electrical preprocessing. Since the noise characteristics in the apparent resistivity data are relatively few, the denoising effect is limited. At the same time, commonly used methods such as median filtering and wavelet transform mainly rely on manual experience in parameter setting. If the parameters are not selected properly, it is easy to cause signal distortion, affecting the accuracy and stability of the denoised data. Summary of the Invention

[0004] In view of this, the object of the present invention is to provide a method, device, equipment and medium for denoising time-domain data of direct current electrical method, so as to improve the quality of time-domain data of direct current electrical method, thereby enhancing the exploration effect of direct current electrical method.

[0005] In a first aspect, the present application provides a direct current method for time domain data denoising, comprising:

[0006] Obtain noisy time-domain potential difference data of the DC method;

[0007] Based on the noisy time-domain potential difference data, the time-domain data denoising model is used to denoise the noisy time-domain potential difference data to obtain the target noise-free time-domain potential difference data, wherein the time-domain data denoising model extracts features from the noisy time-domain potential difference data through the feature node layer to obtain mapping feature data; performs nonlinear transformation on the mapping feature data through the enhancement node layer to obtain enhanced feature data; and obtains the target noise-free time-domain potential difference data based on the mapping feature data, the enhanced feature data and the network connection weight through the output layer.

[0008] Optionally, the DC time-domain data denoising method provided in this application further includes:

[0009] Acquire a training data set; wherein the training data set includes a plurality of training sample data; each training sample data includes noisy time-domain potential difference data and noise-free time-domain potential difference data corresponding to the noisy time-domain potential difference data;

[0010] Based on the training data set, a training operation is performed on the time domain data denoising model, wherein the training operation includes:

[0011] Select target training sample data from the training data set;

[0012] The noisy time-domain potential difference data in the target training sample data is input into the time-domain data denoising model, so that the time-domain data denoising model extracts features from the noisy time-domain potential difference data through the feature node layer to obtain mapping feature data; the mapping feature data is nonlinearly transformed through the enhancement node layer to obtain enhanced feature data; the mapping feature data and the enhanced feature data are simultaneously connected to the output layer, and the network connection weights of the network are solved through pseudo-inverse.

[0013] Optionally, the DC time-domain data denoising method provided in this application further includes:

[0014] During the training operation of the time domain data denoising model, various hyperparameters of the time domain data denoising model are optimized; wherein the hyperparameters include the number of feature nodes, the number of mapping feature groups, the number of enhancement nodes, and the regularization parameter.

[0015] Optionally, the hyperparameters of the time domain data denoising model are optimized, including:

[0016] Set the value range of each hyperparameter and construct the search space for Bayesian optimization;

[0017] Based on the search space, multiple sets of hyperparameter combinations are generated through random sampling to obtain the initial sample set;

[0018] Based on the initial sample set, an iterative optimization operation is performed on the proxy model corresponding to the time-domain data denoising model until it is determined that the iterative optimization termination condition is met, and then a training operation is performed based on the hyperparameters of the time-domain data denoising model optimized in the last iterative optimization operation; wherein the iterative optimization operation includes:

[0019] Select the target hyperparameter combination from the initial sample set;

[0020] Inputting the target hyperparameter combination into the proxy model so that the proxy model obtains an error value corresponding to the target hyperparameter combination based on the target hyperparameter combination;

[0021] Based on the sampling function, update the target hyperparameter combination;

[0022] Based on the updated target hyperparameter combination, the error value of the proxy model is optimized to obtain the optimized hyperparameter combination.

[0023] Optionally, obtain a training dataset, including:

[0024] The noise-free time-domain potential difference data is obtained by randomly generating square wave signals of different amplitudes;

[0025] Noisy time-domain potential difference data are obtained by superimposing various types of noise on noise-free time-domain potential difference data;

[0026] A training data set is obtained based on the noise-free time-domain potential difference data and the noisy time-domain potential difference data.

[0027] Optionally, after obtaining the training data set, the following steps are also included:

[0028] The training sample data in the training data set is normalized by the normalization function, where the normalization function is:

[0029]

[0030] Where, is the normalized training sample data, is the training sample data, is a constant.

[0031] Optionally, target training sample data is selected from the training data set, including:

[0032] Based on the mapping relationship between the number of training samples and the absolute error of the validation set, the number of training sample data in the target training sample data is determined.

[0033] In a second aspect, the present application provides a direct current method time domain data denoising device, comprising:

[0034] A data acquisition module is used to obtain the noisy time-domain potential difference data of the direct current method;

[0035] The data denoising module is used to denoise the noisy time-domain potential difference data using a time-domain data denoising model to obtain target noise-free time-domain potential difference data. The time-domain data denoising model extracts features from the noisy time-domain potential difference data through a feature node layer to obtain mapping feature data; performs nonlinear transformation on the mapping feature data through an enhancement node layer to obtain enhanced feature data; and obtains target noise-free time-domain potential difference data based on the mapping feature data, the enhanced feature data, and the network connection weights through an output layer.

[0036] In a third aspect, the present application also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above-mentioned DC electrical method time domain data denoising method is implemented.

[0037] In a fourth aspect, the present application also provides a computer-readable storage medium, which stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to execute the above-mentioned DC time-domain data denoising method.

[0038] The embodiments of the present invention provide a direct current (DC) electrical method time domain data denoising method, apparatus, device and medium. The method obtains noisy time domain potential difference data of the DC method; uses a time domain data denoising model to denoise the noisy time domain potential difference data to obtain target noise-free time domain potential difference data. The time domain data denoising model extracts features from the noisy time domain potential difference data through a feature node layer to obtain mapping feature data; performs nonlinear transformation on the mapping feature data through an enhanced node layer to obtain enhanced feature data; and obtains target noise-free time domain potential difference data based on the mapping feature data, enhanced feature data and network connection weights through an output layer. Compared with the denoising of apparent resistivity data, the time domain potential difference data of the DC method receiving electrode contains effective signals and rich noise information, which provides the possibility for more effective denoising, thereby improving the quality of DC method observation data and thus improving the exploration effect of the DC method.

[0039] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 A flow chart of a direct current method time domain data denoising method provided by an embodiment of the present invention is shown;

[0042] Figure 2 A schematic diagram of noise-free data provided by an embodiment of the present invention is shown;

[0043] Figure 3 A schematic diagram of Gaussian noise provided by an embodiment of the present invention is shown;

[0044] Figure 4 A schematic diagram of noisy time domain data provided by an embodiment of the present invention is shown;

[0045] Figure 5 A learning curve diagram provided by an embodiment of the present invention is shown;

[0046] Figure 6 A schematic diagram illustrating a process flow of optimizing a time domain data denoising model provided by an embodiment of the present invention is shown;

[0047] Figure 7 A schematic structural diagram of a power supply control device provided by an embodiment of the present invention is shown;

[0048] Figure 8 A schematic diagram showing the structure of an acquisition control device provided by an embodiment of the present invention is shown;

[0049] Figure 9 It shows a schematic structural diagram of a water tank provided by an embodiment of the present invention;

[0050] Figure 10 A schematic diagram of an indoor experimental observation system provided by an embodiment of the present invention is shown;

[0051] Figure 11 A schematic diagram of time domain data before denoising of measurement points provided by an embodiment of the present invention is shown;

[0052] Figure 12 A schematic diagram of time domain data after denoising of measurement points provided by an embodiment of the present invention is shown;

[0053] Figure 13 A schematic diagram of the inversion result of the time domain data without denoising provided by an embodiment of the present invention is shown;

[0054] Figure 14 A schematic diagram showing the inversion results of denoised data using the time domain data denoising model provided by an embodiment of the present invention is shown;

[0055] Figure 15 A schematic diagram illustrating a process of constructing and testing a time domain data denoising model according to an embodiment of the present invention is shown;

[0056] Figure 16 A schematic structural diagram of a direct current method time domain data denoising device provided by an embodiment of the present invention is shown;

[0057] Figure 17 A schematic structural diagram of an electronic device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.

[0059] Currently, denoising techniques are used to denoise apparent resistivity data from the direct current method to improve the quality of the observed data. Denoising techniques include removing mutation points, median spatial filtering, and wavelet transforms. While removing mutation points is simple, removing too many data points in areas of strong interference can lead to exploration gaps, increasing the difficulty of interpretation. Median spatial filtering has a good denoising effect under specific noise conditions, but due to the potential presence of multiple types of noise in field data, denoising data collected in the field is not very effective. Wavelet transform denoising is easily influenced by personal experience and parameter selection. If the mother wavelet is not chosen appropriately, it can easily lead to signal distortion, affecting the accuracy and reliability of the denoised data.

[0060] To this end, in this application, a time domain data denoising model is established based on the nonlinear mapping relationship between noisy time domain potential difference data and noise-free time domain potential difference data. The time domain data denoising model is used to denoise the time domain potential difference data of the direct current method to obtain the target time domain potential difference data, so as to improve the quality of the time domain potential difference data and thus improve the exploration effect of the direct current method.

[0061] After introducing the application scenarios and design concepts of the present invention, the technical solutions provided by the present invention are described in detail below.

[0062] The present invention provides a method for training a time domain data denoising model using a direct current method. Figure 1 As shown, the DC time-domain data denoising model training method provided in the embodiment of the present application includes at least the following steps:

[0063] Step 110: Obtain noisy time-domain potential difference data of the direct current method.

[0064] In the embodiment of the present application, the following methods may be used to obtain the noisy time-domain potential difference data of the direct current method, but are not limited to:

[0065] The time series data of the injected current and potential difference of the DC electrical method are collected by a DC electrical method instrument. The time series data is the potential difference or current signal recorded continuously in the time dimension and is used to invert the underground resistivity distribution.

[0066] Step 120: Based on the noisy time-domain potential difference data, a time-domain data denoising model is used to denoise the noisy time-domain potential difference data to obtain target noise-free time-domain potential difference data, wherein the time-domain data denoising model extracts features from the noisy time-domain potential difference data through a feature node layer to obtain mapping feature data; performs nonlinear transformation on the mapping feature data through an enhancement node layer to obtain enhanced feature data; and obtains target noise-free time-domain potential difference data based on the mapping feature data, the enhanced feature data, and the network connection weight through an output layer.

[0067] In an embodiment of the present application, the acquired noisy time-domain potential difference data of the direct current method is input into a time-domain data denoising model, so that the time-domain data denoising model performs feature extraction on the noisy time-domain potential difference data through a feature node layer to obtain mapping feature data; performs nonlinear transformation on the mapping feature data through an enhanced node layer to obtain enhanced feature data; obtains target noise-free time-domain potential difference data based on the mapping feature data, enhanced feature data and network connection weights through an output layer to improve the quality of the observation data, thereby improving the application effect of the direct current method in pre-surface exploration. By denoising the noisy time-domain potential difference data of the direct current method, it helps to improve the application effect of the direct current method in a strong interference environment, and provide technical support for urban underground structure exploration and geological disaster monitoring.

[0068] The following describes in detail the training process of the time domain data denoising model provided in the embodiment of the present application. The training process of the time domain data denoising model provided in the embodiment of the present application is as follows:

[0069] Step 210 , obtaining a training data set; wherein the training data set includes a plurality of training sample data; each training sample data includes noisy time-domain potential difference data and noise-free time-domain potential difference data corresponding to the noisy time-domain potential difference data.

[0070] In the embodiments of the present application, when obtaining a training data set, the following methods are included but not limited to:

[0071] Acquire noise-free time-domain potential difference data and noisy time-domain potential difference data; wherein the noise-free time-domain potential difference data is obtained by randomly generating square wave signals of different amplitudes; and the noisy time-domain potential difference data is obtained by superimposing multiple types of noise on the noise-free time-domain potential difference data;

[0072] A training data set is obtained based on the noise-free time-domain potential difference data and the noisy time-domain potential difference data.

[0073] In a specific embodiment, square wave signals of different amplitudes are randomly generated as training data for noise-free time-domain potential difference data. Figure 2 Noise-free data shown;

[0074] In order to simulate the complex noise in the actual observation environment, in this application, various types of noisy time-domain potential difference data training samples are constructed to enhance the diversity of training samples. When obtaining noisy time-domain potential difference data, various types of noise can be superimposed on the noise-free time-domain potential difference data to obtain noisy time-domain potential difference data, and different types of noise can be superimposed on the noise-free time-domain potential difference data to obtain noisy time-domain potential difference data.

[0075] Furthermore, the electrical noise of the device and the environmental background noise can be simulated by Gaussian random signals (e.g., with a mean of 0 and a variance of 1), as shown in the following example: Figure 3The Gaussian noise shown in the figure is added to the noise-free time domain potential difference data to obtain the following Figure 4 The noisy time-domain potential difference data shown;

[0076] You can also use a pulse signal with random amplitude and interference time to simulate the interference caused by sudden events such as equipment failure and lightning to obtain mixed noise; and use a sine wave signal with random period and phase to simulate the noise caused by industrial electricity and equipment grounding. The amplitude of the sine noise is random data between [0, 1] and the phase is [0, 2 ] with a period of [0, 4T] seconds, where T is the power supply period of the DC electrical instrument. The synthesized typical noise is superimposed on the noise-free time-domain potential difference data as the noisy time-domain potential difference data training sample.

[0077] This application enriches the types of training samples in the training sample set by adjusting the amplitude, noise amplitude, frequency and other parameters of the noise-free time-domain potential difference data; and in order to simulate the influence of electrode extremes on the observed data, a randomly generated constant is introduced into each noisy time-domain potential difference data, so that the characteristics of the noisy time-domain potential difference data are closer to the actual observed data.

[0078] In an embodiment of the present application, after obtaining the training data set, the training sample data in the training data set is further normalized.

[0079] In order to improve the stability and performance of the time domain data denoising model, in this application, the training sample data in the training data set is normalized. Specifically, the noisy time domain potential difference data and the noise-free time domain potential difference data are normalized to the range of [0, 1] by a normalization function to obtain a normalized training data set, where the expression of the normalization function is:

[0080]

[0081] Where, is the normalized training sample data, is the training sample data, is a constant, and the maximum value of the noisy time-domain potential difference data is taken.

[0082] In the embodiment of the present application, since the time domain potential difference signal in the direct current method is usually in the form of a square wave with clear polarity and amplitude characteristics, the maximum value normalization avoids the nonlinear disturbance of the extreme point on the overall normalization ratio while maintaining the dominant structure of the signal, which helps to improve the robustness and adaptability of the time domain data denoising model to data under different acquisition conditions, thereby enhancing the stability and universality of the denoising performance.

[0083] Step 220: Based on the training data set, a training operation is performed on the time domain data denoising model, wherein the training operation includes: selecting target training sample data from the training data set; inputting the noisy time domain potential difference data in the target training sample data into the time domain data denoising model, so that the time domain data denoising model extracts features from the noisy time domain potential difference data through the feature node layer to obtain mapping feature data; performing a nonlinear transformation on the mapping feature data through the enhancement node layer to obtain enhanced feature data; connecting the mapping feature data and the enhanced feature data to the output layer at the same time, and solving the network connection weights of the network through pseudo-inverse solution.

[0084] In order to improve the training efficiency of the time domain data denoising model, in this application, the number of training sample data in the target training sample data is determined based on the mapping relationship between the number of training samples and the absolute error of the validation set, such as Figure 5 As shown in the learning curve, as the number of training samples increases, the absolute error of the validation set first decreases and then gradually stabilizes.

[0085] In a specific embodiment, the inflection point of the learning curve is used to approximately determine the optimal number of training sample data in the target training sample data. This can optimize the allocation of training resources while ensuring effective learning of the time domain data denoising model. In this application, the specific process of determining the inflection point of the learning curve based on the gain value is as follows:

[0086] Based on the newly added training samples, calculate the absolute error of the validation set corresponding to the newly added training samples to obtain the gain value;

[0087] It is determined whether the gain value exceeds a preset gain threshold, and if not, the gain value is determined to be an inflection point.

[0088] The absolute error of the validation set can be calculated using the following formula:

[0089]

[0090] Where, is the absolute error of the validation set, is the predicted value of the validation set, is the expected value of the validation set.

[0091] In the embodiment of the present application, the input data is noisy time-domain potential difference data; the output data is target noise-free time-domain potential difference data; the time-domain data denoising model extracts features from the noisy time-domain potential difference data through the feature node layer to obtain mapping feature data; the mapping feature data is nonlinearly transformed through the enhancement node layer to obtain enhanced feature data; the mapping feature data and the enhanced feature data are simultaneously connected to the output layer, and the network connection weights of the network are solved through pseudo-inverse. In this way, by establishing a time-domain data denoising model based on the nonlinear mapping relationship between the noisy time-domain potential difference data and the noise-free time-domain potential difference data, the noisy time-domain potential difference data is denoised to obtain the target noise-free time-domain potential difference data, thereby improving the observation data instructions and thus improving the exploration effect of the direct current method;

[0092] By combining the actual geoelectric field noise characteristics, a highly simulated and diverse training dataset of DC electrical time-domain potential difference data is generated; the training dataset is normalized, and the number of samples used for network training is determined using a learning curve; the constructed denoising model is integrated with the time-domain potential difference data acquisition to form a new DC electrical data acquisition and processing mode to improve the quality of observation data, thereby enhancing the application effect of DC electrical method in pre-surface exploration.

[0093] In order to improve the efficiency and rationality of the hyperparameter selection of the time domain data denoising model, in the embodiment of the present application, in the process of training the time domain data denoising model, the hyperparameters of the time domain data denoising model can also be optimized; wherein the hyperparameters include the number of feature nodes, the number of mapping feature groups, the number of enhancement nodes and the regularization parameter, specifically, Figure 6 As shown, the optimization process of the time domain data denoising model provided in the embodiment of the present application is as follows:

[0094] Step 310: Set the value range of each hyperparameter and construct a Bayesian optimization search space.

[0095] In a specific embodiment, a reasonable value range (such as minimum value, maximum value) is set for each hyperparameter to construct a hyperparameter search space, i.e., a Bayesian optimization search space. Specifically, a reasonable value range is set for each hyperparameter, which is: the value range of the number of feature nodes is set to ; The value range of the number of feature node groups is set to ; The value range of the enhanced node number is set to ; The range of the regularization parameter is ; and set the maximum number of iterations of the optimization process; in this application, the Gaussian process is selected as the proxy model to fit the complex nonlinear relationship between hyperparameters and the performance of the time domain data denoising model; the expected improvement is used as the acquisition function to guide the search process to converge to the global optimal solution.

[0096] Step 320: Based on the search space, generate multiple sets of hyperparameter combinations through random sampling to obtain an initial sample set.

[0097] In a specific embodiment, within the hyperparameter search space, several sets of hyperparameter combinations are generated by random sampling. Where, is a hyperparameter, , as the initial sample set for Bayesian optimization.

[0098] Step 330: Select a target hyperparameter combination from the initial sample set;

[0099] Step 340: Input the target hyperparameter combination into the proxy model so that the proxy model obtains an error value corresponding to the target hyperparameter combination based on the target hyperparameter combination; and update the target hyperparameter combination based on the sampling function.

[0100] In a specific embodiment, each set of hyperparameter combinations is used to train the time domain data denoising model, and the error of the time domain data denoising model on the validation set is calculated. , the error on the validation set As the performance evaluation result of the current hyperparameter combination, the initial dataset is finally formed. , providing data support for the construction of agent models;

[0101] Based on the above initial dataset , building a Gaussian process agent model ,in, is the mean function; is the covariance function, g represents the kernel function, and RBF is used as the kernel function. and Represents two different hyperparameter combinations, each of which includes the number of feature nodes, the number of feature groups, the number of augmented nodes, and the regularization parameter. This Gaussian process surrogate model provides probabilistic predictions for subsequent hyperparameter selection by fitting the complex nonlinear relationship between hyperparameters and the performance of the time-domain data denoising model.

[0102] Using the acquisition function Select the next set of hyperparameter combinations from the surrogate model that may improve model performance , using expected improvement as the acquisition function:

[0103]

[0104] Where, Represents the best model performance index among existing samples, represents the predicted mean and standard deviation of the surrogate model in processing, Represents the normal distribution function and probability density function, maximized in the search space , you can get the current optimal hyperparameter combination Then, the time domain data denoising model is trained using the newly selected hyperparameter combination and its error on the validation set is calculated. .

[0105] Step 350, determine whether the iterative optimization termination condition is met; if so, execute step 360; if not, return to step 330; wherein, the iterative optimization termination condition is that the number of iterations is not less than the number threshold, or the error value is not higher than the preset threshold.

[0106] In a specific embodiment, it is determined whether the current optimization process meets the iterative optimization termination condition (such as reaching the maximum number of iterations or the validation set error is less than a preset threshold). If the iterative optimization termination condition is met, the optimal hyperparameter combination is output; otherwise, the new hyperparameter combination and validation set error are added to the existing data set. , and return to the above to continue optimization until the stopping condition is met.

[0107] Step 360: Based on the updated target hyperparameter combination, optimize the error value of the proxy model to obtain an optimized hyperparameter combination.

[0108] In order to verify the denoising accuracy of the constructed and optimized time domain data denoising model, in the embodiment of the present application, the denoising accuracy of the constructed and optimized time domain data denoising model is tested in an indoor laboratory. Specifically, the time domain data denoising model provided in the present application is tested in an indoor laboratory, and the specific content is as follows:

[0109] The laboratory collection device consists of two parts, one of which is Figure 7 The power supply control device shown is connected to the power supply motor and can control parameters such as power supply cycle and amplitude; the other part is Figure 8 The acquisition control device shown is connected to the receiving motor and can synchronously acquire the time domain potential difference data of all receiving electrodes, and the sampling frequency can be adjusted as needed; wherein, the experiment is carried out in a water tank, the photo of the water tank is shown in Figure 9 As shown, the water tank size is 1 m (length) × 0.64 m (width) × 0.55 m (height).

[0110] Figure 10This is a schematic diagram of the indoor experimental observation system. The red dots represent power electrodes. The electrodes used in this experiment were copper electrodes with a diameter of 0.005 meters and a length of 0.02 meters. There were seven pairs of power electrodes (A1-B1, ..., A7-B7). Each power supply cycle was 4 seconds, and the supply voltage amplitude was 10 volts. The black dots represent measuring electrodes, totaling 42 measuring electrodes. During each power supply cycle, all measuring electrodes acquired data at a sampling rate of 200 Hz. The difference between any two pairs of measuring electrodes was used as the measurement point data. Over seven power cycles, the 42 measuring electrodes collected data at 6027 measurement points. To simulate the effects of interference sources in field measurements, a pair of interference electrodes (in the pink five-pointed star area) were placed in the water tank. The interference signal was supplied to the water through the interference electrodes. The interference signal waveform was a 5-second sinusoidal wave with an amplitude of 3 volts. To increase the complexity of the interference signal, a sliding rheostat was connected in series between the interference power source and the interference electrodes. During the measurement, the sliding rheostat continuously varied the interference source waveform. In addition, two anomalies were placed in the water tank. The blue matrix represents the projection position of the anomaly on the horizontal plane. The anomaly is a high-resistance insulating rubber block with a size of 0.04 m (length) × 0.04 m (width) × 0.04 m (height). The distance between the top interface of the two rubber blocks and the water surface is 0.02 m.

[0111] The time domain potential difference data measured in the indoor experiment is input into the time domain data denoising model for denoising. Figure 11 and Figure 12 The time-domain potential difference data of the measuring point before and after denoising are shown. The black dotted line represents the measured time-domain potential difference data; the cyan dashed line represents the time-domain potential difference data after denoising by the time-domain data denoising model, and the blue dotted line represents the noise removed by the time-domain data denoising model. It can be seen from the figure that the use of the time-domain data denoising model can suppress the noise in the time-domain potential difference data of the DC method.

[0112] In order to further test the denoising effect of the time domain data denoising model, the data measured in the indoor experiment were inverted. Figure 13 and Figure 14 is the inversion result of indoor experimental data, where Figure 13 This is the inversion result of the time-domain potential difference data without denoising. Due to the influence of noise, many false anomalies appear in the inversion result, making it impossible to accurately identify and determine the position of the rubber block. Figure 14 This is the inversion result of the denoised data using the time domain data denoising model. The inversion result in the rubber block area shows high resistance, which is consistent with the actual situation. Figure 14 and Figure 13 Compared with the results of the time domain data denoising model, the false structure in the background is significantly reduced. Compared with the inversion results of noisy data, the denoising data of the time domain data denoising model can improve the accuracy of the inversion results.

[0113] The following is a detailed description of the process of constructing and testing the time domain data denoising model provided in the embodiment of the present application. Figure 15 As shown, the process of constructing and testing the time domain data denoising model provided in the embodiment of the present application is as follows:

[0114] Step a: Generate training data samples:

[0115] Randomly generate and analyze measured data, and add noise types to the measured data to obtain typical noise; the measured data is measured data by randomly generating square wave signals of different amplitudes as noise-free time-domain potential difference data; Gaussian noise is obtained by simulating the electrical noise of the equipment and the environmental background noise through Gaussian random signals (such as mean 0 and variance 1), and the Gaussian noise is superimposed on the noise-free time-domain potential difference data to obtain noisy time-domain potential difference data; pulse signals with random amplitude and interference time can also be used to simulate the interference caused by sudden events such as equipment failure and lightning to obtain mixed noise; and sinusoidal wave signals with random period and phase are used to simulate the noise caused by industrial electricity and equipment grounding. The amplitude of the sinusoidal noise is random data between [0, 1] and the phase is [0, 2 ] with a period of [0, 4T] seconds, where T is the power supply period of the DC electrical instrument. The synthesized typical noise is superimposed on the noise-free time-domain potential difference data as the noisy time-domain potential difference data training sample.

[0116] Step b: Build a time domain data denoising model

[0117] Perform normalized sample preprocessing on the training sample data in the training data set;

[0118] The inflection point of the learning curve is used to approximately determine the optimal number of training sample data in the target training sample data;

[0119] A Bayesian optimization grid model is used to optimize the grid hyperparameters; wherein, the value range of each hyperparameter is set, and a Bayesian optimization search space is constructed; based on the search space, multiple sets of hyperparameter combinations are generated by random sampling to obtain an initial sample set; based on the initial sample set, an iterative optimization operation is performed on the proxy model corresponding to the time domain data denoising model, until it is determined that the iterative optimization termination condition is met, and a training operation is performed based on each hyperparameter of the time domain data denoising model optimized during the last iterative optimization operation; wherein, the iterative optimization operation includes: selecting a target hyperparameter combination from the initial sample set; inputting the target hyperparameter combination into the proxy model, so that the proxy model obtains an error value corresponding to the target hyperparameter combination based on the target hyperparameter combination; based on the sampling function, the target hyperparameter combination is updated; based on the updated target hyperparameter combination, the error value of the proxy model is optimized to obtain the target hyperparameter combination.

[0120] Based on the training data set, a training operation is performed on the time domain data denoising model, wherein the training operation includes: selecting target training sample data from the training data set; inputting the noisy time domain potential difference data in the target training sample data into the time domain data denoising model, so that the time domain data denoising model extracts features from the noisy time domain potential difference data through the feature node layer to obtain mapping feature data; performing nonlinear transformation on the mapping feature data through the enhancement node layer to obtain enhanced feature data; connecting the mapping feature data and the enhanced feature data to the output layer at the same time, and solving the network connection weights of the network through pseudo-inverse solution.

[0121] Step c: obtain the experimentally measured noisy time-domain potential difference data through the laboratory's power supply control device and acquisition control device, and input the noisy time-domain potential difference data into the time-domain data denoising model for denoising to obtain experimental noise-free time-domain potential difference data, and perform inversion interpretation and result evaluation on the indoor experimental measurement data.

[0122] Based on the above embodiments, the present application provides a DC electrical method time domain data denoising device, see Figure 16 As shown, the DC time domain data denoising device provided in the embodiment of the present application includes at least:

[0123] The data acquisition module 410 is used to acquire the noisy time-domain potential difference data of the direct current method;

[0124] The data denoising module 420 is used to denoise the noisy time-domain potential difference data using a time-domain data denoising model to obtain target noise-free time-domain potential difference data, wherein the time-domain data denoising model extracts features from the noisy time-domain potential difference data through a feature node layer to obtain mapping feature data; performs nonlinear transformation on the mapping feature data through an enhancement node layer to obtain enhanced feature data; and obtains target noise-free time-domain potential difference data based on the mapping feature data, the enhanced feature data and the network connection weight through an output layer.

[0125] In an optional embodiment, the DC electrical method time-domain potential difference data denoising device also includes a model training module 430, which is used to obtain a training data set; wherein the training data set includes multiple training sample data; each training sample data includes noisy time-domain potential difference data and noise-free time-domain potential difference data corresponding to the noisy time-domain potential difference data; based on the training data set, a training operation is performed on the time-domain data denoising model, wherein the training operation includes: selecting target training sample data from the training data set; inputting the noisy time-domain potential difference data in the target training sample data into the time-domain data denoising model, so that the time-domain data denoising model extracts features from the noisy time-domain potential difference data through the feature node layer to obtain mapping feature data; performing nonlinear transformation on the mapping feature data through the enhancement node layer to obtain enhanced feature data; connecting the mapping feature data and the enhanced feature data to the output layer at the same time, and solving the network connection weights of the network through pseudo-inverse solution.

[0126] In an optional embodiment, the model training module 430 is used to:

[0127] During the training operation of the time domain data denoising model, various hyperparameters of the time domain data denoising model are optimized; wherein the hyperparameters include the number of feature nodes, the number of mapping feature groups, the number of enhancement nodes, and the regularization parameter.

[0128] In an optional embodiment, the model training module 430 is used to:

[0129] Optimizing the hyperparameters of the time-domain data denoising model includes: setting the value range of each hyperparameter and constructing a Bayesian optimization search space; based on the search space, generating multiple sets of hyperparameter combinations by random sampling to obtain an initial sample set; based on the initial sample set, performing an iterative optimization operation on the proxy model corresponding to the time-domain data denoising model, until it is determined that the iterative optimization termination condition is met, and performing a training operation based on the hyperparameters of the time-domain data denoising model optimized during the last iterative optimization operation; wherein the iterative optimization operation includes: selecting a target hyperparameter combination from the initial sample set; inputting the target hyperparameter combination into the proxy model, so that the proxy model obtains an error value corresponding to the target hyperparameter combination based on the target hyperparameter combination; based on the sampling function, updating the target hyperparameter combination; and optimizing the error value of the proxy model based on the updated target hyperparameter combination to obtain an optimized hyperparameter combination.

[0130] In an optional embodiment, the model training module 430 is used to:

[0131] Noise-free time-domain potential difference data is obtained by randomly generating square wave signals of different amplitudes; noisy time-domain potential difference data is obtained by superimposing multiple types of noise on the noise-free time-domain potential difference data; and a training data set is obtained based on the noise-free time-domain potential difference data and the noisy time-domain potential difference data.

[0132] In an optional embodiment, the model training module 430 is used to:

[0133] The training sample data in the training data set is normalized by a normalization function, wherein the normalization function is:

[0134]

[0135] Where, is the normalized training sample data, is the training sample data, is a constant.

[0136] In an optional embodiment, the model training module 430 is used to:

[0137] Selecting target training sample data from the training data set also includes determining the number of training sample data in the target training sample data based on a mapping relationship between the number of training samples and the absolute error of the validation set.

[0138] It should be noted that the principle of solving the technical problem of the DC electrical time domain data denoising device provided in the embodiment of the present application is similar to the DC electrical time domain data denoising method provided in the embodiment of the present application. Therefore, the implementation of the DC electrical time domain data denoising device provided in the embodiment of the present application can refer to the implementation of the DC electrical time domain data denoising method provided in the embodiment of the present application, and the repeated parts will not be repeated.

[0139] After introducing the direct current time domain data denoising method and device provided in the embodiments of the present application, the electronic device provided in the embodiments of the present application is briefly introduced next.

[0140] See Figure 17 As shown, the electronic device 500 provided in the embodiment of the present application includes at least a processor 501, a memory 502, and a computer program stored in the memory 502 and executable on the processor 501. When the processor 501 executes the computer program, the DC electrical method time domain data denoising method provided in the embodiment of the present application is implemented.

[0141] The electronic device 500 provided in the embodiment of the present application may further include a bus 503 connecting different components (including the processor 501 and the memory 502). The bus 503 represents one or more of several types of bus structures, including a memory bus, a peripheral bus, a local bus, etc.

[0142] Memory 502 may include a readable storage medium in the form of volatile memory, such as random access memory (RAM) 5021 and / or cache memory 5022, and may further include read-only memory (ROM) 5023. Memory 502 may also include a program tool 5025 having a set (at least one) of program modules 5024. Program modules 5024 include, but are not limited to, an operating subsystem, one or more application programs, other program modules, and program data. Each of these examples, or some combination thereof, may include an implementation of a network environment.

[0143] Processor 501 can be a single processing element or a collective term for multiple processing elements. For example, processor 501 can be a central processing unit (CPU) or one or more integrated circuits configured to implement the DC electrical method time-domain data denoising method provided in the embodiments of the present application. Specifically, processor 501 can be a general-purpose processor, including but not limited to a CPU, an application-specific integrated circuit (ASIC), an off-the-shelf field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, etc.

[0144] The electronic device 500 can communicate with one or more external devices 504 (e.g., keyboard, remote control, etc.), and can also communicate with one or more devices that enable a user to interact with the electronic device 500 (e.g., mobile phone, computer, etc.), and / or communicate with a device that enables the electronic device 500 to communicate with one or more other electronic devices 500 (e.g., router, modem, etc.). Such communication can be performed through an input / output (I / O) interface 505. In addition, the electronic device 500 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN) and / or public network, such as the Internet) through a network adapter 506. Figure 17 As shown, the network adapter 506 communicates with other modules of the electronic device 500 via the bus 503. Figure 17Not shown, other hardware and / or software modules may be used in conjunction with the electronic device 500, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, disk arrays (Redundant Arrays of Independent Disks, RAID) subsystems, tape drives, and data backup storage subsystems.

[0145] It should be noted that Figure 17 The electronic device 500 shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0146] The following describes the computer-readable storage medium provided in the embodiments of the present application. The computer-readable storage medium provided in the embodiments of the present application stores computer instructions that, when executed by a processor, implement the DC time-domain data denoising method provided in the embodiments of the present application. Specifically, the computer instructions may be built into or installed within the processor. Thus, the processor can implement the DC time-domain data denoising method provided in the embodiments of the present application by executing the built-in or installed computer instructions.

[0147] In addition, the DC electrical time-domain data denoising method provided in the embodiment of the present application can also be implemented as a computer program product, which includes program code. When the program code is run on a processor, it implements the DC electrical time-domain data denoising method provided in the embodiment of the present application.

[0148] The computer program product provided in the embodiments of the present application may adopt one or more computer-readable storage media, and the computer-readable storage medium may be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any suitable combination of the above. Specifically, more specific examples of computer-readable storage media (a non-exhaustive list) include an electrical connection with one or more wires, a portable disk, a hard disk, RAM, ROM, Erasable Programmable Read Only Memory (EPROM), optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.

[0149] The computer program product provided in the embodiments of the present application may be a CD-ROM and include program code, and may also be run on an electronic device such as a computer. However, the computer program product provided in the embodiments of the present application is not limited thereto. In the embodiments of the present application, the computer-readable storage medium may be any tangible medium that contains or stores program code, and the program code may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0150] It should be noted that although several units or subunits of the device are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, depending on the embodiment of the application, the features and functions of two or more units described above can be embodied in a single unit. Conversely, the features and functions of a single unit described above can be further divided and embodied by multiple units.

[0151] Furthermore, although the operations of the method of the present application are described in a particular order in the accompanying drawings, this does not require or imply that the operations must be performed in this particular order, or that all illustrated operations must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0152] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they become aware of the basic creative concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0153] Obviously, those skilled in the art may make various changes and modifications to the embodiments of the present application without departing from the spirit and scope of the embodiments of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalents, the present application also intends to include such modifications and variations.

Claims

1. A direct current method for time domain data denoising, characterized in that: include: Obtain noisy time-domain potential difference data of the DC method; Based on the noisy time-domain potential difference data, a time-domain data denoising model is used to denoise the noisy time-domain potential difference data to obtain target noise-free time-domain potential difference data, wherein the time-domain data denoising model extracts features from the noisy time-domain potential difference data through a feature node layer to obtain mapping feature data; performs nonlinear transformation on the mapping feature data through an enhanced node layer to obtain enhanced feature data; obtains target noise-free time-domain potential difference data based on the mapping feature data, the enhanced feature data and the network connection weight through an output layer; wherein, in the process of training the time-domain data denoising model, each hyperparameter of the time-domain data denoising model is optimized; wherein, each hyperparameter of the time-domain data denoising model is optimized, including: updating the target hyperparameter combination based on a sampling function; wherein, utilizing an acquisition function Select the next set of hyperparameter combinations from the surrogate model to improve model performance , using expected improvement as the acquisition function : Where, Represents the best model performance index among existing samples, represents the predicted mean and standard deviation of the surrogate model in processing, Represents the normal distribution function and probability density function, maximized in the search space , you can get the current optimal hyperparameter combination Then, the time domain data denoising model is trained using the newly selected hyperparameter combination and its error on the validation set is calculated. , is the initial data set, and X is the hyperparameter combination.

2. The DC electrical method time domain data denoising method according to claim 1, characterized in that: Also includes: Acquire a training data set; wherein the training data set includes a plurality of training sample data; each of the training sample data includes noisy time-domain potential difference data and noise-free time-domain potential difference data corresponding to the noisy time-domain potential difference data; Based on the training data set, a training operation is performed on the time domain data denoising model, wherein the training operation includes: Selecting target training sample data from the training data set; The noisy time-domain potential difference data in the target training sample data is input into the time-domain data denoising model, so that the time-domain data denoising model extracts features from the noisy time-domain potential difference data through a feature node layer to obtain mapping feature data; performs nonlinear transformation on the mapping feature data through an enhancement node layer to obtain enhanced feature data; and connects the mapping feature data and the enhanced feature data to an output layer at the same time, and solves the network connection weights of the network through pseudo-inverse solution.

3. The DC electrical method time domain data denoising method according to claim 2, characterized in that: The hyperparameters include the number of feature nodes, the number of mapping feature groups, the number of enhancement nodes and the regularization parameter.

4. The DC electrical method time domain data denoising method according to claim 3, characterized in that: Optimizing the hyperparameters of the time domain data denoising model includes: Setting the value range of each hyperparameter and constructing a Bayesian optimization search space; Based on the search space, generating multiple sets of hyperparameter combinations by random sampling to obtain an initial sample set; Based on the initial sample set, an iterative optimization operation is performed on the proxy model corresponding to the time-domain data denoising model until it is determined that an iterative optimization termination condition is satisfied, and then the training operation is performed based on each hyperparameter of the time-domain data denoising model optimized during the last execution of the iterative optimization operation; wherein the iterative optimization operation includes: Selecting a target hyperparameter combination from the initial sample set; Inputting the target hyperparameter combination into the proxy model, so that the proxy model obtains an error value corresponding to the target hyperparameter combination based on the target hyperparameter combination; Based on the sampling function, updating the target hyperparameter combination; Based on the updated target hyperparameter combination, the error value of the proxy model is optimized to obtain an optimized hyperparameter combination.

5. The direct current method time domain data denoising method according to any one of claims 2 to 4, characterized in that: Get the training dataset, including: The noise-free time-domain potential difference data is obtained by randomly generating square wave signals of different amplitudes; Obtaining noisy time-domain potential difference data by superimposing multiple types of noise on the noise-free time-domain potential difference data; The training data set is obtained based on the noise-free time-domain potential difference data and the noisy time-domain potential difference data.

6. The DC electrical method time domain data denoising method according to claim 5, characterized in that: After obtaining the training data set, it also includes: The training sample data in the training data set is normalized by a normalization function, wherein the normalization function is: Where, is the normalized training sample data, is the training sample data, is a constant.

7. The method for denoising time-domain data using a direct current method according to claim 2, wherein: Selecting target training sample data from the training data set includes: The number of training sample data in the target training sample data is determined based on a mapping relationship between the number of training samples and the absolute error of the validation set.

8. A direct current method time domain data denoising device, characterized in that: include: A data acquisition module is used to obtain the noisy time-domain potential difference data of the direct current method; A data denoising module is used to denoise the noisy time-domain potential difference data using a time-domain data denoising model to obtain target noise-free time-domain potential difference data, wherein the time-domain data denoising model extracts features from the noisy time-domain potential difference data through a feature node layer to obtain mapping feature data; performs nonlinear transformation on the mapping feature data through an enhanced node layer to obtain enhanced feature data; obtains target noise-free time-domain potential difference data based on the mapping feature data, the enhanced feature data and network connection weights through an output layer; wherein, in the process of training the time-domain data denoising model, each hyperparameter of the time-domain data denoising model is optimized; wherein, the optimization of each hyperparameter of the time-domain data denoising model includes: updating the target hyperparameter combination based on a sampling function; wherein, utilizing an acquisition function Select the next set of hyperparameter combinations from the surrogate model to improve model performance , using expected improvement as the acquisition function : Where, Represents the best model performance index among existing samples, represents the predicted mean and standard deviation of the surrogate model in processing, Represents the normal distribution function and probability density function, maximized in the search space , you can get the current optimal hyperparameter combination Then, the time domain data denoising model is trained using the newly selected hyperparameter combination and its error on the validation set is calculated. , is the initial data set, and X is the hyperparameter combination.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the direct current method time domain data denoising method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by the processor, the machine-executable instructions prompt the processor to execute the direct current time-domain data denoising method according to any one of claims 1 to 7.

Images