Cable failure prediction processing method, device and electronic equipment

Abstract

The application discloses a kind of cable fault prediction processing method, device and electronic equipment.Therein, the method comprises: obtaining the partial discharge characteristic data of high-voltage cable;First training set data in partial discharge characteristic data is respectively input to initial radial basis function network model and initial long short-term memory network model for training;First test set data in partial discharge characteristic data is respectively input to the radial basis function network model after training and the long short-term memory network model after training, obtain the first output result output by the radial basis function network model after training, and the second output result output by the long short-term memory network model after training;Based on first output result and second output result, determine cable fault prediction model.The present application solves the technical problems of low cable fault prediction efficiency and poor prediction accuracy caused by poor performance of cable fault prediction model in related art.

Description

Technical Field

[0001] This invention relates to the field of smart grid safety monitoring, and more specifically, to a cable fault prediction and processing method, apparatus, and electronic equipment. Background Technology

[0002] High-voltage cables are crucial electrical equipment in power systems, and their operating status affects the safety and reliability of power supply. However, due to design flaws, manufacturing defects during installation, external damage, and water tree intrusion, insulation defects are inevitable in cable systems. Partial discharge (PD) is both a major cause of insulation degradation and an important indicator of cable insulation defects and aging. Timely and accurate identification of potential partial discharge faults in high-voltage cables plays a vital role in their safe and stable operation. Current technologies primarily rely on manual methods or neural network prediction for high-voltage cable partial discharge fault detection, which suffers from high detection costs and poor model performance, leading to low efficiency and accuracy in cable fault prediction and identification.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This invention provides a cable fault prediction processing method, apparatus, and electronic device to at least solve the technical problem of low cable fault prediction efficiency and poor prediction accuracy caused by the poor performance of cable fault prediction models in related technologies.

[0005] According to one aspect of the present invention, a cable fault prediction processing method is provided, comprising: acquiring partial discharge characteristic data of a high-voltage cable; inputting a first training set of data from the partial discharge characteristic data into an initial radial basis function network model and an initial long short-term memory network model for training, respectively, to obtain a trained radial basis function network model and a trained long short-term memory network model; inputting a first test set of data from the partial discharge characteristic data into the trained radial basis function network model and the trained long short-term memory network model, respectively, to obtain a first output result from the trained radial basis function network model and a second output result from the trained long short-term memory network model; and determining a cable fault prediction model based on the first output result and the second output result.

[0006] According to another aspect of the present invention, a cable fault prediction processing apparatus is also provided, comprising: a first acquisition module for acquiring partial discharge characteristic data of a high-voltage cable; a training module for inputting a first training set of data from the partial discharge characteristic data into an initial radial basis function network model and an initial long short-term memory network model for training, to obtain a trained radial basis function network model and a trained long short-term memory network model; a second acquisition module for inputting a first test set of data from the partial discharge characteristic data into the trained radial basis function network model and the trained long short-term memory network model, to obtain a first output result from the trained radial basis function network model and a second output result from the trained long short-term memory network model; and a determination module for determining a cable fault prediction model based on the first output result and the second output result.

[0007] According to another aspect of the present invention, a non-volatile storage medium is also provided, characterized in that the non-volatile storage medium stores a plurality of instructions, the instructions being adapted to be loaded by a processor and executed any one of the above-described cable fault prediction processing methods.

[0008] According to another aspect of the present invention, an electronic device is also provided, characterized in that it includes one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement any one of the above-described cable fault prediction processing methods.

[0009] In this embodiment of the invention, partial discharge characteristic data of high-voltage cables are acquired; the first training set data from the partial discharge characteristic data is input into an initial radial basis function network model and an initial long short-term memory network model for training, resulting in a trained radial basis function network model and a trained long short-term memory network model; the first test set data from the partial discharge characteristic data is input into the trained radial basis function network model and the trained long short-term memory network model, resulting in a first output result from the trained radial basis function network model and a second output result from the trained long short-term memory network model; based on the first and second output results, a cable fault prediction model is determined, achieving the goal of constructing a more accurate cable fault prediction model by integrating the features of multiple neural network models and improving model performance. This achieves the technical effect of improving the performance of the cable fault prediction model, thereby improving the efficiency and accuracy of cable fault prediction, and solving the technical problem of low cable fault prediction efficiency and poor prediction accuracy caused by the poor performance of cable fault prediction models in related technologies. Attached Figure Description

[0010] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0011] Figure 1 This is a schematic diagram of a cable fault prediction and processing method according to an embodiment of the present invention;

[0012] Figure 2 This is a schematic diagram of an optional cable fault prediction and processing method according to an embodiment of the present invention;

[0013] Figure 3 This is a schematic diagram of a cable fault prediction and processing device according to an embodiment of the present invention. Detailed Implementation

[0014] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0015] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0016] High-voltage cables are crucial electrical equipment in power systems, and their operating status affects the safety and reliability of power supply. However, due to design flaws, installation defects, external damage, and intrusion by water and trees, insulation defects inevitably occur in cable systems. Partial discharge (PD) is both a major cause of insulation degradation and an important indicator of cable insulation defects and aging.

[0017] Partial discharge refers to a discharge in an insulation system where only certain areas discharge under the influence of an electric field, without forming a through-discharge channel. The main cause of partial discharge is uneven electric field strength across the insulation system due to electrolyte inhomogeneity. In some areas, the electric field strength reaches the breakdown strength, causing discharge, while other areas retain their insulating properties. Large electrical equipment has complex insulation structures using various materials, resulting in uneven electric field distribution throughout the insulation system. Imperfections in design or manufacturing processes can lead to air gaps in the insulation system, or moisture absorption during long-term operation can cause moisture to decompose under the influence of an electric field, forming bubbles. Because the dielectric constant of air is lower than that of the insulating material, even under relatively low electric fields, the electric field strength of the air gap bubbles can be high, leading to partial discharge when the field strength reaches a certain value. Furthermore, internal defects or impurities within the insulation, or poor electrical connections within the insulation structure, can cause localized electric field concentration. At these concentrated areas, surface discharge and floating potential discharge can occur. Therefore, partial discharge can be broadly classified into four types: air gap discharge, surface discharge, corona discharge, and suspension discharge. Related technologies primarily rely on manual methods or neural network prediction for high-voltage cable partial discharge fault detection. However, this approach suffers from high detection costs and poor model performance, leading to low efficiency and accuracy in cable fault prediction and identification.

[0018] Based on the above problems, this invention provides a method embodiment for cable fault prediction processing. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0019] Figure 1 This is a flowchart of a cable fault prediction and processing method according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0020] Step S102: Obtain partial discharge characteristic data of the high-voltage cable.

[0021] In one optional embodiment, the acquisition of partial discharge characteristic data of high-voltage cables includes: acquiring historical partial discharge fault data of high-voltage cables; performing noise reduction processing on the historical partial discharge fault data to obtain noise-reduced historical partial discharge fault data; and performing feature extraction processing on the noise-reduced historical partial discharge fault data to obtain the partial discharge characteristic data.

[0022] Optionally, but not limited to, during the actual operation of the high-voltage cable, historical data on partial discharge faults of the high-voltage cable may be obtained, i.e., historical data on partial discharge faults includes operating data of the high-voltage cable prior to the current operating time.

[0023] Optionally, the above-mentioned feature extraction processing of the noise-reduced historical partial discharge fault data to obtain the above-mentioned partial discharge feature data includes: performing feature extraction processing on the noise-reduced historical partial discharge fault data to obtain feature sample data; and performing normalization processing on the above-mentioned feature sample data to obtain the above-mentioned partial discharge feature data. The cable partial discharge (PD) features included in the above-mentioned partial discharge feature data may include, but are not limited to, the positive half-cycle skewness of the maximum discharge quantity distribution. Maximum discharge distribution negative half-cycle skew Maximum discharge distribution positive half-cycle spurs Maximum discharge distribution negative half-cycle spuriousness Maximum discharge quantity distribution asymmetry Q m Correlation of maximum discharge distribution (CC) m skewness of the positive half-cycle of the average discharge quantity distribution Average discharge distribution negative half-cycle skew Average discharge distribution positive half-cycle spurs Average discharge distribution negative half-cycle spuriousness Average discharge quantity distribution asymmetry Q a Correlation of average discharge quantity distribution (CC) a Discharge frequency distribution positive half-cycle skewness Discharge frequency distribution negative half-cycle skew Discharge frequency distribution positive half-cycle spurs Discharge frequency distribution negative half-cycle spurious Discharge frequency distribution asymmetry Q n Correlation of discharge number distribution (CC) n 18 types.

[0024] By using the above methods, the historical data of partial discharge faults in high-voltage cables are noise-reduced, thereby improving the signal-to-noise ratio. Based on this, the partial discharge feature data obtained by feature extraction has higher data quality.

[0025] In one optional embodiment, the above-mentioned noise reduction processing of the above-mentioned partial discharge fault historical data to obtain noise-reduced partial discharge fault historical data includes: determining whether there is missing data in the above-mentioned partial discharge fault historical data; if there is missing data in the above-mentioned partial discharge fault historical data, using interpolation to fill the missing data to obtain filled partial discharge fault historical data; and performing noise reduction processing on the filled partial discharge fault historical data to obtain the noise-reduced partial discharge fault historical data.

[0026] By using the above method, when missing data is detected in the historical data of partial discharge faults, interpolation is used to obtain filler data based on other historical data with the same characteristics as the missing data in the historical data of partial discharge faults. The missing data is then filled with the filler data to obtain the filled historical data of partial discharge faults, thereby ensuring the integrity of the training data.

[0027] Step S104: Input the first training set data from the above partial discharge feature data into the initial radial basis function network model and the initial long short-term memory network model for training, respectively, to obtain the trained radial basis function network model and the trained long short-term memory network model.

[0028] It should be noted that Radial Basis Function (RBF) neural networks are high-performance feedforward networks, possessing advantages such as optimal approximation, simple training, fast learning convergence, and the ability to overcome local minima. It has been proven that RBF networks can approximate any continuous function with arbitrary precision. Therefore, they have been widely used in pattern recognition, nonlinear control, and image processing. Long Short-Term Memory (LSTM) networks are an improved recurrent neural network that addresses the problem of RNNs' inability to handle long-range dependencies, thus avoiding the gradient vanishing problem of RNNs. In this embodiment of the invention, the above two models are combined when constructing the cable fault prediction model, resulting in a cable fault prediction model with better performance.

[0029] Optionally, the partial discharge feature data is first divided into a first training set and a first test set. Taking the partial discharge feature data as an example, which includes 6000 sample data sets, the data is randomly shuffled and split in an 8:2 ratio, with the first 4800 sets forming the first training set and the last 1200 sets forming the first test set. It should be noted that the number of sample data sets used in this embodiment is not fixed. During subsequent operation, cable partial discharge fault data is remotely acquired from a server for prediction, and the collected data is continuously written into the original data, achieving self-updating of the sample data. This ensures the richness of the sample data and gradually improves the prediction accuracy.

[0030] In an optional embodiment, before inputting the first training set data from the partial discharge feature data into the initial radial basis function network model and the initial long short-term memory network model for training to obtain the trained radial basis function network model and the trained long short-term memory network model, the method further includes: determining the feature types corresponding to the partial discharge feature data of the high-voltage cable; and determining the initialization hyperparameters corresponding to the initial radial basis function network model and the initial long short-term memory network model based on the feature types, wherein the initialization hyperparameters include at least the initial number of nodes and the initial learning rate corresponding to the initial radial basis function network model and the initial long short-term memory network model, respectively.

[0031] It should be noted that before training the model, it is necessary to first determine the initial number of neurons, initial number of nodes, and initial learning rate of the initial radial basis function network model and the initial long short-term memory network model based on the characteristics of the partial discharge feature data. For example, if the cable PD feature type is used as the input node of the input layer, the number of nodes is 18. If the cable partial discharge fault type (corona discharge, suspension discharge, air gap discharge, surface discharge) is used as the output layer, the number of neurons is 4, and the learning rate is set to 0.002, and so on.

[0032] Optionally, but not limited to, the following formula can be used as the activation function for the initial radial basis function network model (i.e., the RBF model):

[0033]

[0034] In the formula, μ t With the center point, σ tThe radial basis function width determines the rate of descent of the radial basis functions. In the initial Long Short-Term Memory (LSTM) network model, the activation function of the gate network is defined as the sigmoid function, and the activation function of the output layer is the tanh function. The sigmoid function resembles a mathematical function with an S-shaped curve; it can be understood as a squeezing function, limiting its output to between 0 and 1, making it very useful in probabilistic prediction. The tanh function is one of the hyperbolic functions; tanh() represents the hyperbolic tangent. In mathematics, the hyperbolic tangent "tanh" is derived from the basic hyperbolic functions hyperbolic sine and hyperbolic cosine.

[0035] Step S106: Input the first test set data from the partial discharge feature data into the trained radial basis function network model and the trained long short-term memory network model respectively to obtain the first output result of the trained radial basis function network model and the second output result of the trained long short-term memory network model.

[0036] Using the above methods, the trained radial basis function network model and the trained long short-term memory network model are tested based on the first test set data in the partial discharge feature data, respectively. The output results of the two models are obtained and used to evaluate the model performance and construct the cable fault prediction model by comparing the output results of the two models.

[0037] Step S108: Based on the first output result and the second output result, determine the cable fault prediction model.

[0038] In an optional embodiment, when the first test set data includes multiple sets of test data and both the first output result and the second output result are multiple, determining the cable fault prediction model based on the first output result and the second output result includes: determining the comparison results between the multiple first output results and their corresponding second output results, wherein the comparison results are: first comparison results where the first output results are consistent with their corresponding second output results, or second comparison results where the first output results are inconsistent with their corresponding second output results; determining the total number of first results among the comparison results between the multiple first output results and their corresponding second output results, and the number of the first comparison results; determining the first proportion of the number of the first comparison results to the total number of the first results; and, if the first proportion is greater than a preset proportion threshold, constructing the cable fault prediction model based on the trained radial basis function network model and the trained long short-term memory network model.

[0039] The above method is used to determine whether the output results (i.e., the first output result and the second output result) are consistent. If the output results are consistent, the prediction is correct; if the output results are inconsistent, the prediction is incorrect. Then, the accuracy of the model is calculated based on the first proportion of the number of correctly predicted results to the total number of results. Further, it is determined whether the accuracy meets the requirements, i.e., whether the accuracy is greater than a preset accuracy threshold (or whether the aforementioned first proportion is greater than a preset proportion threshold). If it does, a cable fault prediction model is constructed by combining the trained RBF network model and the trained LSTM network model.

[0040] In an optional embodiment, the method further includes: when the first ratio is not greater than a preset ratio threshold, using a gradient descent algorithm to update the trained radial basis function network model and the trained long short-term memory network model respectively, to obtain a new radial basis function network model and a new long short-term memory network model; inputting the second training set data from the partial discharge feature data into the new radial basis function network model and the new long short-term memory network model respectively for training, to obtain a new trained radial basis function network model and a new trained long short-term memory network model; inputting the second test set data from the partial discharge feature data into the new trained radial basis function network model and the new trained long short-term memory network model respectively, to obtain a third output result from the new trained radial basis function network model and the new trained long short-term memory network model. The model outputs a fourth output result; in the case that the second test set data includes multiple sets of test data, and both the third and fourth output results are multiple, the comparison results between the multiple third output results and their corresponding fourth output results are determined, wherein the comparison results are: third comparison results where the third output results are consistent with their corresponding fourth output results, or fourth comparison results where the third output results are inconsistent with their corresponding fourth output results; the total number of second results among the comparison results between the multiple third output results and their corresponding fourth output results, and the number of third comparison results are determined; the second proportion of the number of third comparison results to the total number of second results is determined; if the second proportion is greater than the preset proportion threshold, the cable fault prediction model is constructed based on the new trained radial basis function network model and the new trained long short-term memory network model.

[0041] If, using the above method, the first proportion is not greater than a preset proportion threshold, it indicates that the trained radial basis function network (RBF) model and the trained long short-term memory (LSTM) network model do not meet the accuracy requirements. At this point, gradient descent is introduced to update the initial weights and initial neuron biases in the initial LSTM and RBF network models, resulting in new RBF and LSM models. The models are then retrained and tested. If the outputs of the new RBF and LSM models meet the accuracy requirements, a cable fault prediction model is constructed based on these models. If the accuracy requirements are still not met, gradient descent is used again to update the weights and neuron biases in both models. This effectively improves the model accuracy, thereby enhancing the accuracy of cable fault prediction.

[0042] In an optional embodiment, after determining the cable fault prediction model based on the first output result and the second output result, the method further includes: acquiring the test feature data of the cable under test; inputting the test feature data into the cable fault prediction model for testing, and obtaining the discharge type corresponding to the cable under test.

[0043] By using the above methods, the cable fault prediction model is deployed to the field operation environment to monitor the cable under test. After obtaining the test feature data of the cable under test, the test feature data is input into the cable fault prediction model for testing, and the discharge type of the cable under test can be obtained, realizing real-time monitoring of partial discharge of high-voltage cables.

[0044] Through steps S102 to S108, the training set data from the cable partial discharge feature data is input into the prediction model composed of a combination of RBF neural network and LSTM neural network for training. The trained model is then tested using the test set data from the cable partial discharge feature data. Passing the test yields a cable fault prediction model that meets the required accuracy. This achieves the goal of constructing a more accurate cable fault prediction model by integrating features from multiple neural network models, thereby improving model performance. This improves the performance of the cable fault prediction model, thus enhancing the efficiency and accuracy of cable fault prediction. Ultimately, it solves the technical problem of low efficiency and poor accuracy in cable fault prediction caused by the poor performance of existing cable fault prediction models in related technologies.

[0045] It should be noted that this embodiment of the invention utilizes a combined RBF and LSTM neural network to process and train cable partial discharge fault data. This improves upon existing technologies that suffer from unstable predictions, low accuracy, and the need for on-site support, which is both labor-intensive and cumbersome. The invention automatically updates cable partial discharge fault data and outputs prediction results. Furthermore, a gradient descent algorithm is introduced to learn and update the weights of the network prediction model. This effectively improves prediction accuracy, enhances the ability of personnel to predict transformer faults, and increases the intelligence level of substation equipment. It avoids the risks associated with on-site operations and the high rate of human error, saves significant manpower, accelerates identification speed, and improves identification accuracy. Ultimately, it effectively ensures the safe operation of high-voltage cables and related equipment.

[0046] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation method. Figure 2 This is a flowchart of an optional cable fault prediction processing method according to an embodiment of the present invention, such as... Figure 2 As shown, the method includes:

[0047] Step S1 involves acquiring historical data of partial discharge faults in the cable as raw data, performing noise reduction processing, extracting partial discharge feature data, and then performing normalization preprocessing. This includes the following sub-steps:

[0048] Step (1.1): Obtain historical data on partial discharge defects in high-voltage cables (i.e., historical data on partial discharge faults) through previous monitoring, as the raw data. As the cable continues to operate, the raw data will be automatically updated and gradually increased. The continuously enriched historical data on partial discharge faults can gradually improve the accuracy of prediction.

[0049] Step (1.2): Check the completeness of the historical partial discharge fault data. If there are missing data, fill them in using interpolation. For example, take the median of all values ​​of the same characteristic gas in the historical partial discharge fault data that are in the missing data to fill in the missing data, and obtain the filled historical partial discharge fault data.

[0050] Step (1.3): Perform wavelet denoising on the padded partial discharge fault historical data to obtain denoised partial discharge fault historical data, thereby improving the data signal-to-noise ratio and obtaining higher quality model training input data. For example, perform Fourier denoising on the partial discharge fault historical data to improve the signal-to-noise ratio. Fourier transform formula:

[0051]

[0052] In the formula, ω represents frequency, t represents time, and e -iωt It is a complex function.

[0053] Step (1.4): Perform feature extraction processing on the noise-reduced historical partial discharge fault data to extract feature data from the historical partial discharge fault data, obtaining feature sample data of the high-voltage cable. The cable partial discharge (PD) features in the feature sample data may include, but are not limited to, the positive half-cycle skewness of the maximum discharge quantity distribution. Maximum discharge distribution negative half-cycle skew Maximum discharge distribution positive half-cycle spurs Maximum discharge distribution negative half-cycle spuriousness Maximum discharge quantity distribution asymmetry Q m Correlation of maximum discharge distribution (CC) m skewness of the positive half-cycle of the average discharge quantity distribution Average discharge distribution negative half-cycle skew Average discharge distribution positive half-cycle spurs Average discharge distribution negative half-cycle spuriousness Average discharge quantity distribution asymmetry Q a Correlation of average discharge quantity distribution (CC) a Discharge frequency distribution positive half-cycle skewness Discharge frequency distribution negative half-cycle skew Discharge frequency distribution positive half-cycle spurs Discharge frequency distribution negative half-cycle spurious Discharge frequency distribution asymmetry Q n Correlation of discharge number distribution (CC) n The 18 extracted PD sample data will be used as sample feature data.

[0054] Step (1.5): Normalize the sample feature data to obtain the partial discharge feature data of the high-voltage cable, which can accelerate the training speed. The normalization formula is:

[0055]

[0056] Where, x i For sample feature data, y i This is the normalized result corresponding to the sample feature data (i.e., the partial discharge feature data).

[0057] Step S2: Determine the initial hyperparameters (hidden layers, number of nodes, learning rate) corresponding to the initial radial basis function network model and the initial long short-term memory network model respectively based on the partial discharge characteristic data; initialize the weights and neuron biases.

[0058] Step S3: Input the first training set data from the partial discharge feature data, along with the set hyperparameters, initial weights, and initial neuron biases, into the RBF network model and the LSTM network model simultaneously to train the model, resulting in the trained RBF network model and the trained LSTM network model. Input the first test set data from the partial discharge feature data into the trained RBF network model and the trained LSTM network model respectively to obtain the first output result of the trained RBF network model and the second output result of the trained LSTM network model.

[0059] Step S4: Determine whether the two outputs (i.e., the first output and the second output) are consistent. If the outputs are consistent, the prediction is correct; if the outputs are inconsistent, the prediction is incorrect. Then, calculate the model's accuracy based on the proportion of correctly predicted results to the total number of results.

[0060] Step S5: Determine if the accuracy meets the requirements, i.e., whether the accuracy is greater than the preset accuracy threshold. If it does, construct a cable fault prediction model based on the trained RBF network model and the trained LSTM network model. If it does not, use the gradient descent algorithm to update the initial weights and initial neuron biases in the initial LSTM network model and the initial RBF network model, respectively, and then feed them into the RBF network model and the LSTM network model along paths 1 and 2, respectively. It should be noted that paths 1 and 2 are performed simultaneously.

[0061] Step S6: Repeat steps S3 to S5 until the prediction results meet the accuracy requirements. Based on the newly obtained trained radial basis function network model and the newly trained long short-term memory network model, construct a cable fault prediction model.

[0062] Step S7: Deploy the cable fault prediction model to the field operating environment to monitor the cable under test and obtain the characteristic data of the cable under test. Input the characteristic data into the cable fault prediction model for testing to obtain the discharge type corresponding to the cable under test, thereby realizing real-time monitoring of partial discharge of high-voltage cables. It should be noted that the prediction results need to be inversely normalized before outputting the prediction results to obtain the discharge type corresponding to the cable under test.

[0063] It should be noted that this invention is applied to the field of large power grid prediction and diagnosis. A new prediction model is obtained by combining an RBF network model with an LSTM model. Historical data on partial discharge faults in cables is collected and updated automatically, and this data is used as input to predict the type of partial discharge fault. Compared to traditional neural network models, this model can adjust the number of PD feature types according to actual conditions, reducing the large amount of complex data processing workload. Furthermore, it avoids the problems of low accuracy and stability of prediction results associated with traditional neural networks, providing power plant personnel with a stable and reliable basis for assessing transformer operating status and ensuring the safety and stability of the entire power grid.

[0064] This embodiment also provides a cable fault prediction and processing device, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as already described. As used below, the terms "module" and "device" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0065] According to an embodiment of the present invention, an apparatus embodiment for implementing the above-described cable fault prediction and processing method is also provided. Figure 3 This is a schematic diagram of the structure of a cable fault prediction and processing device according to an embodiment of the present invention, as shown below. Figure 3 As shown, the above-mentioned cable fault prediction and processing device includes: a first acquisition module 300, a training module 302, a second acquisition module 304, and a determination module 306, wherein:

[0066] The first acquisition module 300 mentioned above is used to acquire partial discharge characteristic data of the high-voltage cable;

[0067] The training module 302 is connected to the first acquisition module 300 and is used to input the first training set data in the partial discharge feature data into the initial radial basis function network model and the initial long short-term memory network model for training, so as to obtain the trained radial basis function network model and the trained long short-term memory network model.

[0068] The second acquisition module 304 is connected to the training module 302 and is used to input the first test set data in the partial discharge feature data into the trained radial basis function network model and the trained long short-term memory network model respectively, to obtain the first output result of the trained radial basis function network model and the second output result of the trained long short-term memory network model.

[0069] The aforementioned determining module 306 is connected to the aforementioned second acquiring module 304 and is used to determine the cable fault prediction model based on the aforementioned first output result and the aforementioned second output result.

[0070] In this embodiment of the invention, the first acquisition module 300 is used to acquire partial discharge characteristic data of the high-voltage cable; the training module 302, connected to the first acquisition module 300, is used to input the first training set data from the partial discharge characteristic data into the initial radial basis function network model and the initial long short-term memory network model for training, to obtain the trained radial basis function network model and the trained long short-term memory network model; the second acquisition module 304, connected to the training module 302, is used to input the first test set data from the partial discharge characteristic data into the trained radial basis function network model and the trained long short-term memory network model, to obtain the trained radial basis function network model and the trained long short-term memory network model. The first output result of the trained radial basis function network model and the second output result of the trained long short-term memory network model are used to determine the cable fault prediction model based on the first and second output results. This achieves the goal of constructing a more accurate cable fault prediction model by integrating the features of multiple neural network models and improving model performance. This improves the performance of the cable fault prediction model, thereby enhancing the efficiency and accuracy of cable fault prediction. It also solves the technical problem of low efficiency and poor accuracy of cable fault prediction caused by the poor performance of cable fault prediction models in related technologies.

[0071] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0072] It should be noted that the first acquisition module 300, training module 302, second acquisition module 304, and determination module 306 mentioned above correspond to steps S102 to S108 in the embodiments. The instances and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments. It should be noted that the above modules, as part of the device, can run in a computer terminal.

[0073] It should be noted that the optional or preferred implementation methods of this embodiment can be found in the relevant descriptions in the embodiments, and will not be repeated here.

[0074] The aforementioned cable fault prediction and processing device may further include a processor and a memory. The first acquisition module 300, training module 302, second acquisition module 304, determination module 306, etc., are all stored in the memory as program modules, and the processor executes the aforementioned program modules stored in the memory to realize the corresponding functions.

[0075] The processor contains a core that retrieves the corresponding program modules from memory. One or more cores may be configured. Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory includes at least one memory chip.

[0076] According to an embodiment of this application, an embodiment of a non-volatile storage medium is also provided. Optionally, in this embodiment, the non-volatile storage medium includes a stored program, wherein, when the program is running, it controls the device containing the non-volatile storage medium to execute any of the cable fault prediction processing methods described above.

[0077] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals, and the non-volatile storage medium includes stored programs.

[0078] Optionally, during program execution, the device containing the non-volatile storage medium performs the following functions: acquiring partial discharge characteristic data of the high-voltage cable; inputting the first training set data from the partial discharge characteristic data into the initial radial basis function network model and the initial long short-term memory network model for training, respectively, to obtain the trained radial basis function network model and the trained long short-term memory network model; inputting the first test set data from the partial discharge characteristic data into the trained radial basis function network model and the trained long short-term memory network model, respectively, to obtain the first output result of the trained radial basis function network model and the second output result of the trained long short-term memory network model; and determining the cable fault prediction model based on the first output result and the second output result.

[0079] According to an embodiment of this application, an embodiment of a processor is also provided. Optionally, in this embodiment, the processor is used to run a program, wherein the program executes any of the cable fault prediction processing methods described above.

[0080] According to an embodiment of this application, an embodiment of a computer program product is also provided, which, when executed on a data processing device, is adapted to execute a program that initializes the cable fault prediction and processing method steps described above.

[0081] Optionally, when the aforementioned computer program product is executed on a data processing device, it is suitable to execute an initialization program with the following steps: acquiring partial discharge characteristic data of a high-voltage cable; inputting the first training set data from the partial discharge characteristic data into an initial radial basis function network model and an initial long short-term memory network model for training, to obtain a trained radial basis function network model and a trained long short-term memory network model; inputting the first test set data from the partial discharge characteristic data into the trained radial basis function network model and the trained long short-term memory network model, to obtain a first output result from the trained radial basis function network model and a second output result from the trained long short-term memory network model; and determining a cable fault prediction model based on the first output result and the second output result.

[0082] This invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs the following steps: acquiring partial discharge characteristic data of a high-voltage cable; inputting a first training set of the partial discharge characteristic data into an initial radial basis function network model and an initial long short-term memory network model for training, respectively, to obtain a trained radial basis function network model and a trained long short-term memory network model; inputting a first test set of the partial discharge characteristic data into the trained radial basis function network model and the trained long short-term memory network model, respectively, to obtain a first output result from the trained radial basis function network model and a second output result from the trained long short-term memory network model; and determining a cable fault prediction model based on the first and second output results.

[0083] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0084] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0085] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of modules described above can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between modules, and may be electrical or other forms.

[0086] The modules described above as separate components may or may not be physically separate. Similarly, the components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple modules. Some or all of the modules can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0087] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0088] If the aforementioned integrated modules are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable non-volatile storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a non-volatile storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned non-volatile storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0089] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for predicting and processing cable faults, characterized in that, include: Obtain partial discharge characteristic data of high-voltage cables; The first training set data in the partial discharge feature data is input into the initial radial basis function network model and the initial long short-term memory network model respectively for training, to obtain the trained radial basis function network model and the trained long short-term memory network model. The first test set data in the partial discharge feature data is input into the trained radial basis function network model and the trained long short-term memory network model respectively to obtain the first output result of the trained radial basis function network model and the second output result of the trained long short-term memory network model. Based on the first output result and the second output result, a cable fault prediction model is determined; Where the first test set data includes multiple sets of test data, and both the first output result and the second output result are multiple, the step of determining the cable fault prediction model based on the first output result and the second output result includes: determining the comparison results between the multiple first output results and their corresponding second output results, wherein the comparison results are: first comparison results where the first output results are consistent with their corresponding second output results, or second comparison results where the first output results are inconsistent with their corresponding second output results; determining the total number of first results among the comparison results between the multiple first output results and their corresponding second output results, and the number of first comparison results; determining a first proportion of the number of first comparison results to the total number of first results; and, if the first proportion is greater than a preset proportion threshold, constructing the cable fault prediction model based on the trained radial basis function network model and the trained long short-term memory network model.

2. The method according to claim 1, characterized in that, The method further includes: If the first ratio is not greater than a preset ratio threshold, the gradient descent algorithm is used to update the trained radial basis function network model and the trained long short-term memory network model respectively, to obtain a new radial basis function network model and a new long short-term memory network model. The second training set data in the partial discharge feature data is input into the new radial basis function network model and the new long short-term memory network model respectively for training, to obtain the new trained radial basis function network model and the new trained long short-term memory network model. The second test set data in the partial discharge feature data is input into the new trained radial basis function network model and the new trained long short-term memory network model respectively, to obtain the third output result of the new trained radial basis function network model and the fourth output result of the new trained long short-term memory network model. In the case where the second test set data includes multiple sets of test data and both the third output result and the fourth output result are multiple, the comparison results between the multiple third output results and the corresponding fourth output results are determined. The comparison results are: third comparison results where the third output results are consistent with the corresponding fourth output results, or fourth comparison results where the third output results are inconsistent with the corresponding fourth output results. Determine the total number of second results in the comparison results between the plurality of third output results and their corresponding fourth output results, and the number of third comparison results; Determine the second proportion of the number of the third comparison results to the total number of the second results; When the second ratio is greater than the preset ratio threshold, the cable fault prediction model is constructed based on the new trained radial basis function network model and the new trained long short-term memory network model.

3. The method according to claim 1, characterized in that, Before inputting the first training set data from the partial discharge feature data into the initial radial basis function network model and the initial long short-term memory network model for training, respectively, to obtain the trained radial basis function network model and the trained long short-term memory network model, the method further includes: Determine the feature type corresponding to the partial discharge characteristic data of the high-voltage cable; Based on the aforementioned feature types, the initialization hyperparameters corresponding to the initial radial basis function network model and the initial long short-term memory network model are determined respectively. The initialization hyperparameters include at least the initial number of nodes and the initial learning rate corresponding to the initial radial basis function network model and the initial long short-term memory network model respectively.

4. The method according to claim 1, characterized in that, After determining the cable fault prediction model based on the first output result and the second output result, the method further includes: Obtain the test characteristic data of the cable under test; The test feature data is input into the cable fault prediction model for testing to obtain the discharge type corresponding to the cable under test.

5. The method according to any one of claims 1 to 4, characterized in that, The acquisition of partial discharge characteristic data of high-voltage cables includes: Obtain historical data on partial discharge faults in high-voltage cables; The partial discharge fault historical data is denoised to obtain denoised partial discharge fault historical data. The partial discharge fault history data after noise reduction is processed by feature extraction to obtain the partial discharge feature data.

6. The method according to claim 5, characterized in that, The step of denoising the historical partial discharge fault data to obtain denoised historical partial discharge fault data includes: Determine whether there are any missing data in the historical data of partial discharge faults; If the missing data exists in the partial discharge fault history data, the missing data is filled by interpolation to obtain the filled partial discharge fault history data. The partially discharge fault history data after filling is subjected to noise reduction processing to obtain the noise-reduced partially discharge fault history data.

7. A cable fault prediction and processing device, characterized in that, include: The first acquisition module is used to acquire partial discharge characteristic data of the high-voltage cable; The training module is used to input the first training set data in the partial discharge feature data into the initial radial basis function network model and the initial long short-term memory network model for training, so as to obtain the trained radial basis function network model and the trained long short-term memory network model. The second acquisition module is used to input the first test set data in the partial discharge feature data into the trained radial basis function network model and the trained long short-term memory network model respectively, and obtain the first output result of the trained radial basis function network model and the second output result of the trained long short-term memory network model. The determination module is used to determine a cable fault prediction model based on the first output result and the second output result; Where the first test set data includes multiple sets of test data, and both the first output result and the second output result are multiple, the determining module is further configured to determine the comparison results between the multiple first output results and their corresponding second output results, wherein the comparison results are: a first comparison result where the first output result is consistent with the corresponding second output result, or a second comparison result where the first output result is inconsistent with the corresponding second output result; determine the total number of first results among the comparison results between the multiple first output results and their corresponding second output results, and the number of first comparison results; determine the first proportion of the number of first comparison results to the total number of first results; and, if the first proportion is greater than a preset proportion threshold, construct the cable fault prediction model based on the trained radial basis function network model and the trained long short-term memory network model.

8. A non-volatile storage medium, characterized in that, The non-volatile storage medium stores multiple instructions, which are adapted to be loaded by a processor and executed by the cable fault prediction processing method according to any one of claims 1 to 6.

9. An electronic device, characterized in that, It includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the cable fault prediction processing method according to any one of claims 1 to 6.

Images