10kV cable single-phase earth fault detection method based on adaptive algorithm

Abstract

The invention relates to the technical field of power system fault detection, and discloses a 10kV cable single-phase earth fault detection method based on an adaptive algorithm, and the method comprises the following steps: S1, employing a voltage collector and a current collector to collect the power frequency and transient signals of a switching station integrated intelligent terminal strip; s2, filtering and de-noising the collected signals through a self-adaptive digital signal processing algorithm; s3, adaptively determining a single-phase earth fault signal component, and identifying the most significant frequency band; s4, extracting features of the fault signal by using a multi-band multi-criterion adaptive algorithm; and S5, based on the extracted features, selecting the most significant criterion for single-phase earth fault detection. Through an adaptive algorithm and a deep learning model, the detection and recognition precision of a complex single-phase earth fault is improved, real-time intelligent processing is realized through an integrated monitoring module, misjudgment is reduced, the operation and maintenance efficiency is improved, the system adapts to multi-scene application, and the intelligentization and power supply reliability of the system are enhanced.

Description

Technical Field

[0001] The invention relates to the technical field of power system fault detection, and in particular to a 10kV cable single-phase grounding fault detection method based on an adaptive algorithm. Background Art

[0002] With the development of power systems and the gradual complexity of urban power grids, the role of 10kV cable distribution networks in power transmission and distribution has become increasingly significant. However, single-phase grounding faults in cable distribution networks have always been an important issue affecting the safe operation of power grids. The occurrence of single-phase grounding faults will not only lead to a decline in power supply quality, but may also cause more serious electrical accidents. Especially in low-current grounding systems, where the fault current is small, traditional detection methods are difficult to accurately identify and locate fault points, and often rely on manual inspections and experience judgments, which are inefficient and have the risk of missed and misjudgment.

[0003] Current detection technologies mainly rely on fixed-frequency fault signal analysis methods, which cannot adapt to the diversity and complexity of fault signals. For example, the power frequency steady-state signal detection method is easily affected by system noise and load fluctuations, while the transient signal detection method performs poorly under complex fault conditions such as high-resistance grounding and arc grounding. In addition, existing systems generally lack the ability to deeply analyze the nature of faults, making it difficult to distinguish different types of single-phase grounding faults (such as high resistance, low resistance, arc, intermittent, insulation discharge, etc.), thereby limiting the accuracy of fault location and processing.

[0004] Therefore, how to achieve high-precision single-phase grounding fault detection in a complex cable distribution network environment and accurately identify different fault types has become an urgent problem to be solved in the current technical field. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a 10kV cable single-phase grounding fault detection method based on an adaptive algorithm, which solves the problem that it is difficult to detect and identify different types of single-phase grounding faults with high accuracy in a complex cable distribution network environment.

[0006] To achieve the above objectives, the present invention is implemented by the following technical scheme: A 10kV cable single-phase grounding fault detection method based on an adaptive algorithm comprises the following steps: S1. Use voltage and current collectors to collect power frequency and transient signals from the terminal block of the integrated intelligent terminal of the switchgear; S2, filtering and denoising the collected signal by using an adaptive digital signal processing algorithm; S3, adaptively determine the single-phase ground fault signal component and identify the most significant frequency band; S4, using multi-band multi-criteria adaptive algorithm to extract the characteristics of fault signals; S5. Based on the extracted features, the most significant criterion is selected for the detection of single-phase grounding faults; S6. Use the labeled field recording data set to train a deep learning neural network to identify different types of single-phase grounding faults; S7. Through the fault monitoring module, the detected fault information is connected to the integrated intelligent terminal of the switchgear through the remote signal communication node, and the detailed data is uploaded to the distribution automation platform.

[0007] Preferably, in step S2, the adaptive filter minimizes the error by adjusting the weights in real time, and the specific operation formula is: The filter weight vector at the nth iteration, is the learning rate, which controls the step size of weight adjustment, e(n) is the current error, that is, the difference between the filter output and the expected signal, and x(n) is the input signal vector.

[0008] Preferably, the expression of the adaptive noise suppression algorithm in step S2 is: , where F(f) is the frequency response function of the filter, which is used to suppress the noise frequency band and retain the signal frequency band.

[0009] Preferably, when identifying and locking the most significant frequency band in the fault signal in step S3, the formula used is as follows: ,in, is the power spectral density of the signal, and Represent the characteristic factors related to fault signals and noise respectively.

[0010] Preferably, the step S4 specifically includes the following steps: S4.1. Use a multi-band algorithm to analyze the characteristics of multiple frequency bands of the signal; S4.2, applying multiple criteria to evaluate the signal characteristics of each frequency band; S4.3. Comprehensively evaluate the results and extract the most significant fault signal features.

[0011] Preferably, the step S4 involves an adaptive weight allocation mechanism, which is used to assign different weights to each criterion in the multi-judgment evaluation process, and the formula is: ,in, represents the adaptive weight of criterion k in frequency band f, is the characteristic factor of criterion k in frequency band f, is the power spectrum or other characteristic value of the criterion k in the frequency band f.

[0012] Preferably, in the step S4.3, when extracting comprehensive features by combining multiple criterion evaluation results, the following adaptive comprehensive feature extraction formula is used: ,in, is the comprehensive eigenvalue on the frequency band f, is the feature score for judging k in frequency band f, is the above adaptive weight.

[0013] Preferably, the step S6 specifically includes the following steps: S6.1. Use the labeled massive field recording data set as the training set; S6.2, training a stacked denoising autoencoder neural network model; S6.3. Use the trained model to identify different types of single-phase grounding faults, including high-resistance faults, low-resistance faults, arc faults, intermittent faults, and insulation discharge faults.

[0014] Preferably, in the training process of the autoencoder model in step S6.2, an adaptive mechanism is used to adjust the weight of each layer of encoder, so that the model can better process fault signals of different complexities. The specific formula is as follows: · ,in, is the weight matrix of the lth layer, is the learning rate, which is dynamically adjusted according to different training stages or fault signal types. is the gradient of the loss function with respect to the weight.

[0015] Preferably, the fault monitoring module in step S7 includes an integrated cable current and voltage collector and processing device, which is used to perform real-time data processing and fault analysis on site.

[0016] The present invention provides a 10kV cable single-phase grounding fault detection method based on an adaptive algorithm. It has the following beneficial effects: 1. The present invention introduces a multi-band multi-criteria adaptive algorithm to more accurately identify and judge the characteristics of single-phase grounding faults, especially when facing complex types such as high resistance, low resistance, arc light, intermittent faults, etc., which significantly improves the accuracy of fault detection and solves the problems of insufficient detection accuracy, missed judgment or misjudgment in traditional methods when detecting complex types of single-phase grounding faults such as high resistance, low resistance, arc light, intermittent faults, etc.

[0017] 2. The present invention utilizes a large amount of labeled field recording data sets and combines it with a deep learning neural network (SDAE model) for intensive training, which can more accurately identify and analyze the properties of various single-phase grounding faults. This capability greatly improves the system's ability to identify and judge the nature of grounding faults.

[0018] 3. The 10kV cable distribution network grounding fault monitoring module developed by the present invention integrates current and voltage collectors and fault processing devices, and has edge computing functions. It can collect and analyze fault signals in real time, and upload data to the distribution automation platform through remote signaling nodes. This integrated design not only improves the real-time performance of fault monitoring, but also enhances the intelligence level of the system.

[0019] 4. Through accurate fault detection and analysis, the present invention can significantly reduce misjudgments and missed judgments, improve the efficiency of cable operation and maintenance, reduce unnecessary maintenance time and costs, thereby improving power supply reliability and reducing power outage losses.

[0020] 5. The design of the present invention enables the system to not only be applicable to different types of single-phase grounding faults, but also be flexibly deployed in different environments and application scenarios, such as switchgear, ring network cabinets, etc., providing broad space for the promotion and application of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0022] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] Example: Please see attached Figure 1 The embodiment of the present invention provides a 10kV cable single-phase grounding fault detection method based on an adaptive algorithm, comprising the following steps: S1. Use voltage and current collectors to collect power frequency and transient signals from the terminal block of the integrated intelligent terminal of the switchgear; S2, filtering and denoising the collected signal by using an adaptive digital signal processing algorithm; In step S2, the adaptive filter minimizes the error by adjusting the weights in real time. The specific operation formula is: The filter weight vector at the nth iteration, is the learning rate, which controls the step size of weight adjustment, e(n) is the current error, that is, the difference between the filter output and the expected signal, and x(n) is the input signal vector.

[0024] The expression of the adaptive noise suppression algorithm in step S2 is: , where F(f) is the frequency response function of the filter, which is used to suppress the noise frequency band and retain the signal frequency band.

[0025] Specifically, in step S2, the present invention proposes a filtering and denoising processing method based on an adaptive digital signal processing algorithm for detecting a single-phase grounding fault in a 10kV cable. The method adjusts the weight of the filter in real time to minimize the error in the signal processing process, and suppresses the noise frequency band through an adaptive noise suppression algorithm while retaining the key frequency band of the fault signal.

[0026] Adaptive filter operation: The adaptive filter minimizes the error in the signal processing process by adjusting the weights in real time. The specific operation formula is as follows: The filter weight vector at the nth iteration, is the learning rate, which controls the step size of weight adjustment, e(n) is the current error, that is, the difference between the filter output and the expected signal, and x(n) is the input signal vector.

[0027] This adaptive filter can dynamically adjust its weights to ensure that the signal error is always reduced in the best way during the processing process, thereby accurately filtering and retaining the key characteristics of the fault signal.

[0028] The expression of adaptive noise suppression algorithm is: In the adaptive noise suppression process, the frequency response function F(f) of the filter is determined using the following expression: , where F(f) is the frequency response function of the filter, which is used to suppress the noise frequency band and retain the signal frequency band, S(f) is the power spectral density of the signal frequency band, and N(f) is the power spectral density of the noise frequency band.

[0029] This formula adaptively adjusts the frequency response of the filter to ensure that noise can be effectively suppressed at different frequencies while retaining the key fault signal frequency band, further improving the accuracy of fault detection.

[0030] Through the above-mentioned adaptive digital signal processing algorithm, the present invention achieves the following beneficial effects: Improve the processing accuracy of fault signals: By adjusting the filter weights and the adaptive noise suppression algorithm in real time, the present invention can effectively reduce the error in signal processing and ensure the accurate capture of fault signals, especially in complex cable fault environments.

[0031] Enhanced noise suppression: An adaptive frequency response adjustment mechanism is adopted to effectively suppress various noise frequency bands while retaining the sensitivity to fault signals, thereby significantly improving the accuracy of fault detection.

[0032] Adapt to different types of fault signals: The adaptive filtering and noise suppression algorithms can be dynamically adjusted according to the characteristics of the actual fault signal and are applicable to different types of single-phase to ground faults, which makes the system more widely applicable.

[0033] The present invention is based on an adaptive digital signal processing algorithm, and through real-time dynamic adjustment of the filter weights and noise suppression mechanism, effectively solves the problem of inaccurate fault detection caused by noise interference and fixed filters in traditional signal processing methods. The core of the invention is that through the adaptive algorithm, the filter can be adjusted according to the actual signal situation, which maximizes the retention of the characteristics of the fault signal and suppresses irrelevant noise.

[0034] This adaptive processing method not only improves the sensitivity and accuracy of the detection system, but also enables the system to flexibly respond to different cable fault scenarios, significantly improving the reliability of fault detection.

[0035] S3, adaptively determine the single-phase ground fault signal component and identify the most significant frequency band; When identifying and locking the most significant frequency band in the fault signal in step S3, the formula used is as follows: ,in, is the power spectral density of the signal, and Represent the characteristic factors related to fault signals and noise respectively.

[0036] Specifically, in step S3, the present invention proposes an adaptive algorithm for determining and identifying single-phase ground fault signal components, which can automatically identify and lock the most significant frequency band in the fault signal to facilitate subsequent fault analysis and processing.

[0037] Adaptive identification and locking of fault signal frequency band: When identifying and locking the most significant frequency band in the fault signal, the following formula is used to determine the optimal frequency band: ,in, is the power spectral density of the signal, and Represent the characteristic factors related to fault signals and noise respectively.

[0038] Through this formula, the present invention can automatically select the most representative frequency band among various frequency components, so as to more accurately identify the key features of the fault signal, especially in an environment with strong noise interference.

[0039] Through the above adaptive algorithm, the present invention achieves the following beneficial effects: Improve the accuracy of fault signal identification: By automatically identifying and locking the most significant fault frequency band, the sensitivity and accuracy of fault detection are significantly improved, especially in a complex noise background, and the fault signal can be effectively identified and separated.

[0040] Enhance the system's noise immunity: The adaptive algorithm can dynamically adjust the analysis strategy according to the real-time signal and noise characteristics, thereby suppressing irrelevant noise components and retaining only the core frequency band of the fault signal. This capability greatly improves the system's noise resistance performance in practical applications.

[0041] Adapt to various types of single-phase grounding faults: The adaptive frequency band identification algorithm of the present invention is applicable to various types of single-phase grounding faults, including high-resistance faults, low-resistance faults, arc faults, intermittent faults, etc., ensuring the stability and reliability of the system under different fault scenarios.

[0042] The present invention can more accurately lock and analyze these key frequency bands by adaptively determining the significant frequency bands of single-phase ground fault signals. Its principle is based on real-time analysis of fault signals and noise characteristics, and determines the optimal frequency band by dynamically adjusting the combination of power spectrum and characteristic factors.

[0043] The core of this method is its adaptability, which can automatically adjust the analysis strategy according to different fault types and environmental conditions. This not only improves the detection accuracy of the system, but also enables the system to maintain good performance in a wider range of application scenarios.

[0044] S4, using multi-band multi-criteria adaptive algorithm to extract the characteristics of fault signals; The S4 step specifically includes the following steps: S4.1. Use a multi-band algorithm to analyze the characteristics of multiple frequency bands of the signal; S4.2, applying multiple criteria to evaluate the signal characteristics of each frequency band; S4.3. Comprehensively evaluate the results and extract the most significant fault signal features.

[0045] Step S4 involves an adaptive weight allocation mechanism, which is used to assign different weights to each criterion in the multi-judgment evaluation process. The formula is: ,in, represents the adaptive weight of criterion k in frequency band f, is the characteristic factor of criterion k in frequency band f, is the power spectrum or other characteristic value of the criterion k in the frequency band f.

[0046] In step S4.3, when extracting comprehensive features by combining multiple criterion evaluation results, the following adaptive comprehensive feature extraction formula is used: ,in, is the comprehensive eigenvalue on the frequency band f, is the feature score for judging k in frequency band f, is the above adaptive weight.

[0047] Specifically, in step S4, the present invention adopts a multi-band multi-criteria adaptive algorithm to extract the characteristics of the fault signal. The algorithm extracts the most significant fault signal characteristics by performing multi-band analysis on the signal, applying multiple criteria for evaluation, and combining these evaluation results.

[0048] In step S4.1, the present invention first performs multi-band analysis on the signal. By decomposing the signal into multiple frequency bands, the characteristics of each frequency band can be extracted respectively, ensuring a comprehensive analysis of different frequency components.

[0049] In step S4.2, the present invention applies multiple criteria to evaluate the signal characteristics of each frequency band. These criteria may include the power spectrum, energy distribution, phase characteristics, etc. of the signal. Different weights are assigned to each criterion to reflect its importance in different frequency bands. The specific adaptive weight allocation formula is as follows: ,in, represents the adaptive weight of criterion k in frequency band f, is the characteristic factor of criterion k in frequency band f, is the power spectrum or other characteristic value of the criterion k in the frequency band f.

[0050] This formula combines the characteristic factors and power spectra of each criterion and dynamically adjusts the criterion weights, thereby achieving more accurate fault signal evaluation in different frequency bands.

[0051] In step S4.3, the present invention integrates the evaluation results of multiple criteria and uses an adaptive comprehensive feature extraction formula to extract the most significant fault signal feature. The specific formula is as follows: ,in, is the comprehensive eigenvalue on the frequency band f, is the feature score for judging k in frequency band f, is the above adaptive weight.

[0052] In this way, the present invention can effectively extract the most representative features in fault diagnosis and ensure accurate identification of fault signals.

[0053] Through the multi-band multi-criteria adaptive algorithm in step S4, the present invention achieves the following beneficial effects: Enhance the accuracy of feature extraction: The adaptive weight allocation mechanism ensures that each criterion can obtain appropriate weights in different frequency bands, so as to better reflect the key characteristics of the fault signal during comprehensive feature extraction.

[0054] Improve the robustness of fault diagnosis: Combining the multi-criteria evaluation results, fault signals can be identified more comprehensively and accurately, especially in complex signal environments, which can effectively improve the system's fault diagnosis capabilities.

[0055] Adapt to various fault types and complex signal environments: The adaptive algorithm of the present invention can not only handle different types of single-phase grounding faults, but also maintain high-precision feature extraction capabilities in complex noise environments, thereby improving the adaptability and wide applicability of the system.

[0056] The present invention uses a multi-band multi-criteria adaptive algorithm, an adaptive weight allocation mechanism and a comprehensive feature extraction formula in the signal analysis process to achieve accurate identification of fault signals. Its core principle is to adaptively adjust the weights of each criterion to adapt to the changes in real-time signals and ensure the accuracy and robustness of feature extraction. This adaptive mechanism enables the system to dynamically respond to different fault signal characteristics, significantly improving the overall performance of fault diagnosis.

[0057] S5. Based on the extracted features, the most significant criterion is selected for the detection of single-phase grounding faults; Specifically, in step S5, the present invention further analyzes the previously extracted fault signal features and selects the most significant criteria to detect single-phase grounding faults. This step is the key to ensuring fault detection accuracy because it determines which key criteria the system will use to judge and identify faults.

[0058] The process of selecting the most significant criterion: Feature analysis and criteria evaluation: The fault signal features extracted in the previous steps are input into multiple criteria for evaluation. Each criterion is quantitatively scored based on its performance in different frequency bands and its relevance to the fault type.

[0059] The comprehensive scoring formula is:

[0060] in, is the comprehensive score of criterion k, is the adaptive weight on frequency band f, is the feature score of criterion k in frequency band f; M represents the total number of frequency bands analyzed.

[0061] Priority order of criteria: By ranking the combined scores of the criteria, the system can identify the most significant criteria in the current signal environment. This process may involve some form of scoring or ranking algorithm to ensure that the most critical criteria are selected.

[0062] Final criterion selection: Based on the comprehensive evaluation results, the system will select one or more criteria with the highest scores as the main basis for fault detection. These criteria will be used in the final fault judgment process to ensure that the system can accurately identify single-phase grounding faults.

[0063] The process formula for selecting the most significant criterion is as follows:

[0064] in, is the most significant criterion selected, is the comprehensive score of criterion k.

[0065] By selecting the most significant criterion in step S5, the present invention achieves the following beneficial effects: Improve the accuracy of fault detection: By selecting the most representative criterion, the present invention can ensure that the discrimination criteria in the fault detection process are optimized, thereby greatly improving the accuracy of fault detection.

[0066] Optimize system response speed: Selecting the most significant criterion can simplify the detection process, avoid processing too many irrelevant signal features, and thus improve the response speed and real-time performance of the system.

[0067] Enhance system adaptability: Under different fault types and signal environments, the system can adaptively select the most appropriate criteria to ensure that a high level of fault detection performance can be maintained in a variety of application scenarios.

[0068] The core principle of the present invention in step S5 is to dynamically select the most significant criteria for fault detection based on feature analysis and criterion priority sorting. This mechanism enables the system to automatically adjust the fault detection criteria according to real-time signal characteristics, ensuring the reliability and accuracy of the detection results in different fault scenarios.

[0069] In this way, the system can effectively distinguish fault types in a complex power network environment, avoid missed or misjudgment problems caused by improper selection of judgment criteria, and thus improve the stability and operation efficiency of the power system.

[0070] S6. Use the labeled field recording data set to train a deep learning neural network to identify different types of single-phase grounding faults; Step S6 specifically includes the following steps: S6.1. Use the labeled massive field recording data set as the training set; S6.2, training a stacked denoising autoencoder neural network model; S6.3. Use the trained model to identify different types of single-phase grounding faults, including high-resistance faults, low-resistance faults, arc faults, intermittent faults, and insulation discharge faults.

[0071] In the training process of the autoencoder model in step S6.2, an adaptive mechanism is used to adjust the weight of each layer of encoder so that the model can better handle fault signals of different complexities. The specific formula is as follows: · ,in, is the weight matrix of the lth layer, is the learning rate, which is dynamically adjusted according to different training stages or fault signal types. is the gradient of the loss function with respect to the weight.

[0072] Specifically, in step S6, the present invention uses the labeled field recording data set to train the deep learning neural network to identify different types of single-phase grounding faults. Specifically, the present invention adopts a stacked denoising autoencoder (SDAE) neural network model and introduces an adaptive mechanism during the training process to dynamically adjust the weights of each layer of encoders, thereby optimizing the performance of the model so that it can better process fault signals of different complexities.

[0073] In step S6.1, the present invention uses a large number of labeled field recording data sets as training sets. These data sets contain a large number of actual fault signals and non-fault signals, which can accurately reflect various types of single-phase grounding faults after being labeled. By using these data sets for training, it can be ensured that the model has a high generalization ability and can accurately identify different types of faults.

[0074] In step S6.2, the stacked denoising autoencoder neural network model is trained through an adaptive mechanism. Specifically, during the training process of each layer of encoder, the following adaptive mechanism is used to adjust the weight matrix: · ,in, is the weight matrix of the lth layer, is the learning rate, which is dynamically adjusted according to different training stages or fault signal types. is the gradient of the loss function with respect to the weight.

[0075] This adaptive adjustment mechanism ensures that each layer of encoder can be optimized according to the characteristics of the input data, so that the model can more effectively extract and represent the characteristics of the fault signal. Especially when processing complex fault signals, the model can better capture key features, thereby improving the accuracy of fault identification.

[0076] In step S6.3, the trained SDAE model is used to classify and identify the actual fault signal. The model can identify multiple types of single-phase grounding faults, including high-resistance faults, low-resistance faults, arc faults, intermittent faults, and insulation discharge faults. Since the model has adapted to various complex fault signal types during the training process, it can provide high-accuracy fault identification in practical applications.

[0077] By adopting an adaptive mechanism to train the SDAE model in step S6, the present invention achieves the following beneficial effects: Enhance the learning ability of the model: By adaptively adjusting the encoder weights of each layer, the model can better learn and extract the features of complex fault signals, thereby improving the overall performance of the model.

[0078] Improve the accuracy of fault identification: Since the model dynamically adjusts the weights and learning rate during the training process, it can better adapt to different types of fault signals, so that the model can more accurately identify various single-phase grounding faults in practical applications.

[0079] Adapt to various fault types: The SDAE model of the present invention can adapt to and identify various types of single-phase grounding faults, including high resistance, low resistance, arc, intermittent faults and insulation discharge faults, thereby enhancing the versatility and applicability of the system.

[0080] The present invention trains the SDAE neural network model by using a large amount of labeled field recording data sets, and combines the adaptive weight adjustment mechanism to enable the model to better cope with various complex fault signals. During the training process, each layer of the model encoder optimizes the feature extraction capability of different types of fault signals by dynamically adjusting the weights, thereby significantly improving the accuracy of fault identification. This training method based on the adaptive mechanism not only improves the learning ability of the model, but also ensures the high performance and high reliability of the system in actual fault detection.

[0081] S7. Through the fault monitoring module, the detected fault information is connected to the integrated intelligent terminal of the switchgear through the remote signal communication node, and the detailed data is uploaded to the distribution automation platform.

[0082] The fault monitoring module in step S7 includes an integrated cable current and voltage collector and processing device, which is used to perform real-time data processing and fault analysis on site.

[0083] Specifically, in step S7, the present invention describes a fault monitoring module, which can connect the detected fault information to the switchgear integrated intelligent terminal (DTU) through the remote signaling node, and upload the detailed data to the distribution automation platform. The design of the fault monitoring module enables it to perform real-time data processing and fault analysis on site, thereby providing a fast and accurate fault response for the power system.

[0084] Composition of fault monitoring module The fault monitoring module includes the following components: Cable current collector: used to collect current signals in cables.

[0085] Voltage collector: used to collect voltage signals in cables.

[0086] Signal processing device: An integrated processing device used to perform real-time data processing and fault analysis on the collected current and voltage signals.

[0087] These components are designed as a whole and integrated into the fault monitoring module, so that the system can respond immediately when a fault occurs and process and analyze the fault information.

[0088] Data processing and fault analysis On site, the fault monitoring module performs data processing and fault analysis through the following steps: Real-time data acquisition: The current and voltage collectors continuously monitor the current and voltage signals in the cable and transmit these data to the signal processing device.

[0089] Data processing: The signal processing device filters, removes noise and extracts features from the collected data, and analyzes the fault characteristics in the signal in real time.

[0090] Fault judgment: Based on the extracted features and the system preset criteria, the signal processing device determines in real time whether there is a fault and identifies the fault type.

[0091] Data upload and platform integration Once a fault is identified, the fault monitoring module will connect the fault information to the integrated intelligent terminal (DTU) of the switchgear through the remote signaling node. The DTU will then upload detailed data (including fault current, voltage waveform, fault type, etc.) to the distribution automation platform. The platform is used for further fault analysis, recording and response decision-making, thereby realizing comprehensive automated management of the power system.

[0092] Through the fault monitoring module designed in step S7, the present invention achieves the following beneficial effects: Real-time fault response capability: The fault monitoring module can process and analyze data in real time at the moment of fault occurrence, so as to make a quick fault judgment. This real-time response capability significantly reduces the time of fault detection and processing, and improves the operational stability of the power system.

[0093] Efficiency of integrated design: The module integrates current and voltage collectors and signal processing devices, enabling the system to complete the entire process from data collection to fault diagnosis in a single device. This integrated design improves the efficiency and reliability of the system while reducing the complexity and maintenance costs of the equipment.

[0094] Data upload and system linkage: By uploading fault information to the distribution automation platform, the present invention realizes centralized management and analysis of fault data, which not only improves the speed and accuracy of fault response, but also makes the management of the entire power system more intelligent and efficient.

[0095] The fault monitoring module of the present invention integrates multiple functional components to achieve full process automation from data collection to fault analysis. When a fault occurs, the module can quickly process and analyze real-time data and respond in a linked manner through intelligent terminals and automation platforms. This design significantly improves the fault detection and processing capabilities of the power system, ensuring the stable operation of the system and power supply reliability.

[0096] Experimental Design: Exploring the Performance of 10kV Cable Single-Phase Ground Fault Detection System Objective: To verify the detection accuracy and reliability of a 10kV cable single-phase grounding fault detection system based on an adaptive algorithm under different fault types and complex signal environments.

[0097] Experimental materials and equipment: 10kV cable distribution system simulation device Fault monitoring module (including current, voltage collector and processing device) A large amount of annotated field recording data sets Distribution Automation Platform Data logger Current injectors and voltage controllers to simulate different types of faults Experimental steps: Experimental preparation: Set up a 10kV cable distribution system simulation device and connect the fault monitoring module to the simulation device.

[0098] The labeled field recording dataset is loaded into the detection system for training the deep learning neural network model (SDAE model).

[0099] Model training: Use the loaded waveform dataset to train the SDAE model, adjust the adaptive weights and learning rate, and optimize the model to adapt to fault signals of different complexities.

[0100] Record the weight adjustment and loss function changes during model training.

[0101] Experiment 1: Fault type identification and performance testing under multiple fault conditions Purpose: Verify the system's recognition accuracy under various single-phase grounding fault conditions, as well as the system's processing capabilities when multiple faults exist simultaneously.

[0102] step: Set up a 10kV cable simulator to simulate the following single and combined fault types: High resistance fault Low resistance fault Arc Fault Intermittent failure Insulation discharge fault High resistance and arc combined fault Low resistance and intermittent combined fault The fault monitoring module collects current and voltage signals in real time and performs data processing and fault analysis.

[0103] Record the detection results of single and combined faults and evaluate the performance of the system under complex fault conditions.

[0104] Table 1: Fault type identification and multiple fault condition test results Experimental summary: The results show that the system has a high detection accuracy under single fault conditions, especially for high resistance, arc and insulation discharge faults, with an accuracy of 90%. Under compound fault conditions, the detection accuracy drops slightly, but still remains above 75%. This shows that the system still has a strong recognition ability in complex fault scenarios, although further optimization is needed to improve the processing accuracy of compound faults.

[0105] Experiment 2: Noise interference environment and multi-source interference test Purpose: Evaluate the fault detection accuracy of the system under noise interference of different types and intensities, and when multiple noise sources exist simultaneously.

[0106] step: The following noise disturbances were introduced into the simulation device, and high-resistance and low-resistance faults were simulated: High frequency interference (30dB, 50dB) Power frequency interference (30dB, 50dB) Random noise (40dB) High frequency and power frequency composite noise (40dB) Record the system's detection results, especially the fault identification accuracy under different interferences.

[0107] Table 2: Noise interference environment and multi-source interference test results Experimental summary: The experimental results show that the system has strong anti-interference ability under single noise interference, especially in a 30dB noise environment, it still maintains a high detection accuracy (80-90%). However, when the noise intensity increases to 50dB or when complex noise sources exist at the same time, the detection accuracy decreases (60-70%). This shows that the robustness of the system in extreme noise environments needs to be further enhanced, especially the ability to handle multi-source noise.

[0108] Experiment 3: Real-time response and load change test Purpose: Measure the system's real-time response capability to sudden faults under different load conditions and evaluate the impact of load changes on system response time.

[0109] step: Set up load simulation to introduce sudden faults under light load, full load, and overload conditions.

[0110] The measurement system response time from fault occurrence to fault detection, data processing and uploading.

[0111] Compare the response time differences under different load conditions.

[0112] Table 3: Real-time response and load change test results Experimental summary: The system response time under light load and full load conditions is short, maintained between 0.4-0.5 seconds, showing good real-time response capabilities. However, under overload conditions, the total response time increases to 0.8 seconds, mainly due to the extension of processing and upload time. This shows that the system's processing capacity under high load or overload conditions needs to be optimized to ensure that it can still maintain a fast response under extreme conditions.

[0113] Experiment 4: System stability and dynamic environment testing Purpose: Verify the stability of the system during long-term operation, evaluate the impact of dynamic external environmental changes such as temperature and humidity on fault detection performance, and emphasize the stability of the system after proper optimization.

[0114] step: The system was allowed to run continuously for 48 hours under normal conditions, during which various fault types were introduced regularly and changes in temperature and humidity were simulated.

[0115] In a high temperature and high humidity environment, use an optimized heat dissipation and moisture resistance design, and observe the system's fault detection accuracy, response time, and any abnormalities under these conditions.

[0116] Evaluate the detection performance and stability of the optimized system under environmental changes.

[0117] Table 4: System stability and dynamic environment test results Experimental summary: By improving the system's heat dissipation design and anti-humidity treatment, the optimized system shows better stability and reliability in high temperature and high humidity environments. Although the detection performance is slightly affected under extreme conditions (such as 40°C high temperature and 75% humidity), the overall detection accuracy remains above 80%, and there is no obvious signal loss or detection delay. This shows that after proper optimization, the robustness and stability of the system in extreme environments have been significantly improved.

[0118] Adjusted summary Experimental summary and overall conclusion: 1. Fault identification capability: The system shows a high detection accuracy when dealing with single fault types, especially for high resistance, arc and insulation discharge faults. Under compound fault conditions, despite the increased complexity, the detection accuracy remains above 75%.

[0119] 2. Noise immunity and robustness: The system performs well under low- to medium-level single noise interference, but the detection accuracy decreases in high-level noise or compound noise environments. However, the optimized system can effectively cope with most noise interference and has improved robustness in extreme cases.

[0120] 3. Real-time responsiveness: The system responds quickly under light load and full load conditions, with a total response time between 0.4 and 0.5 seconds, showing good real-time responsiveness. Under overload conditions, the response time increases, but after optimization it remains within an acceptable range (0.7 to 0.8 seconds).

[0121] 4. Stability and environmental adaptability: The optimized system shows high stability in long-term operation and extreme environmental conditions, especially in high temperature and high humidity environments, with detection accuracy maintained at more than 80% without obvious delay or signal loss. This shows that after the improvement, the system can operate stably under various harsh conditions.

[0122] Overall conclusion: Through optimized design, the 10kV cable single-phase ground fault detection system based on the adaptive algorithm has shown excellent performance and can maintain a high level of detection accuracy and stability in most application scenarios. In the face of complex fault scenarios and extreme environmental conditions, the system's anti-interference ability and real-time response performance have been significantly enhanced. After these improvements, the system has shown good application prospects and provided a solid foundation for future technology applications and market promotion.

[0123] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A 10kV cable single-phase grounding fault detection method based on an adaptive algorithm is characterized in that: The following steps are involved: S1. Use voltage and current collectors to collect power frequency and transient signals from the terminal block of the integrated intelligent terminal of the switchgear; S2, filtering and denoising the collected signal by using an adaptive digital signal processing algorithm; S3, adaptively determine the single-phase ground fault signal component and identify the most significant frequency band; S4, using multi-band multi-criteria adaptive algorithm to extract the characteristics of fault signals; S5. Based on the extracted features, the most significant criterion is selected for the detection of single-phase grounding faults; S6. Use the labeled field recording data set to train a deep learning neural network to identify different types of single-phase grounding faults; S7. Through the fault monitoring module, the detected fault information is connected to the integrated intelligent terminal of the switchgear through the remote signal communication node, and the detailed data is uploaded to the distribution automation platform.

2. The 10kV cable single-phase grounding fault detection method based on adaptive algorithm according to claim 1 is characterized in that: In the step S2, the adaptive filter minimizes the error by adjusting the weights in real time. The specific operation formula is: The filter weight vector at the nth iteration, is the learning rate, which controls the step size of weight adjustment, e(n) is the current error, that is, the difference between the filter output and the expected signal, and x(n) is the input signal vector.

3. The 10kV cable single-phase grounding fault detection method based on adaptive algorithm according to claim 1 is characterized in that: The expression of the adaptive noise suppression algorithm in step S2 is: , where F(f) is the frequency response function of the filter, which is used to suppress the noise frequency band and retain the signal frequency band, S(f) is the power spectral density of the signal frequency band, and N(f) is the power spectral density of the noise frequency band.

4. The 10kV cable single-phase grounding fault detection method based on adaptive algorithm according to claim 1 is characterized in that: When the most significant frequency band in the fault signal is identified and locked in step S3, the formula used is as follows: ,in, is the power spectral density of the signal, and Represent the characteristic factors related to fault signals and noise respectively.

5. The 10kV cable single-phase grounding fault detection method based on adaptive algorithm according to claim 1 is characterized in that: The S4 step specifically includes the following steps: S4.

1. Use a multi-band algorithm to analyze the characteristics of multiple frequency bands of the signal; S4.2, applying multiple criteria to evaluate the signal characteristics of each frequency band; S4.

3. Comprehensively evaluate the results and extract the most significant fault signal features.

6. The 10kV cable single-phase grounding fault detection method based on adaptive algorithm according to claim 1 is characterized in that: The step S4 involves an adaptive weight allocation mechanism, which is used to assign different weights to each criterion in the multi-judgment evaluation process, and its formula is: ,in, represents the adaptive weight of criterion k in frequency band f, is the characteristic factor of criterion k in frequency band f, is the power spectrum or other characteristic value of the criterion k in the frequency band f.

7. The 10kV cable single-phase grounding fault detection method based on adaptive algorithm according to claim 5 is characterized in that: In the step S4.3, when extracting comprehensive features by combining multiple criteria evaluation results, the following adaptive comprehensive feature extraction formula is used: ,in, is the comprehensive eigenvalue on the frequency band f, is the feature score for judging k in frequency band f, is the above adaptive weight.

8. The 10kV cable single-phase grounding fault detection method based on adaptive algorithm according to claim 1 is characterized in that: The step S6 specifically includes the following steps: S6.

1. Use the labeled massive field recording data set as the training set; S6.2, training a stacked denoising autoencoder neural network model; S6.

3. Use the trained model to identify different types of single-phase grounding faults, including high-resistance faults, low-resistance faults, arc faults, intermittent faults, and insulation discharge faults.

9. The 10kV cable single-phase grounding fault detection method based on adaptive algorithm according to claim 8 is characterized in that: In the training process of the autoencoder model in step S6.2, an adaptive mechanism is used to adjust the weight of each layer of encoder so that the model can better process fault signals of different complexities. The specific formula is as follows: · ,in, is the weight matrix of the lth layer, is the learning rate, which is dynamically adjusted according to different training stages or fault signal types. is the gradient of the loss function with respect to the weight.

10. The 10kV cable single-phase grounding fault detection method based on adaptive algorithm according to claim 1, characterized in that: The fault monitoring module in step S7 includes an integrated cable current and voltage collector and processing device, which is used to perform real-time data processing and fault analysis on site.

Images