Ring main unit fault processing method, system and device based on Mamba model and medium
By using a deep learning method based on the Mamba model, the problem of low efficiency in traditional ring main unit fault detection is solved, enabling real-time monitoring and early warning of ring main unit faults, thereby improving the accuracy of fault detection and the intelligence level of the power system.
Patent Information
- Application Number
- CN202510932775.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-07
- Publication Date
- 2025-11-07
AI Technical Summary
Traditional methods for detecting faults in distribution network ring main units rely on manual inspections, which are inefficient and make it difficult to achieve real-time monitoring and early warning. Data collection and analysis are also limited, making it difficult to deal with complex faults.
A deep learning method based on the Mamba model is adopted. Real-time data of the ring main unit is collected, preprocessed, feature extracted and FFT transformed to build a deep learning model, and supervised training is carried out using the power system confidence loss function to realize the identification and early warning of fault types.
It improves the accuracy and efficiency of fault detection, enables real-time monitoring and early warning of various types of faults, and enhances the intelligence level of the power system.
Smart Images
Figure CN120910641A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of power system intelligence, and particularly relates to a ring main unit fault processing method and system based on a Mamba model, an electronic device and a computer readable storage medium. BACKGROUND
[0002] As a core component of urban power systems, the stability and reliability of ring main units in distribution networks are crucial for ensuring the continuity and safety of power supply. With the rapid development of smart grid technology, there is an increasing demand for real-time monitoring and fault warning of ring main units in distribution networks. Traditional fault detection methods rely on manual inspection and regular maintenance, which are not only inefficient but also difficult to respond in a timely manner when faced with complex faults. In addition, these methods have limitations in data collection and analysis, making it difficult to achieve real-time monitoring and warning of faults. SUMMARY
[0003] The present application aims to provide a ring main unit fault processing method and system based on a Mamba model, an electronic device and a computer readable storage medium, which aims to achieve real-time monitoring and warning of faults in ring main units in distribution networks.
[0004] To achieve the above-mentioned purpose, in a first aspect, the present application provides a ring main unit fault processing method based on a Mamba model, which comprises: Collecting raw data of specified parameters during the operation of a ring main unit in a distribution network; Preprocessing the raw data and labeling the preprocessed raw data as normal type samples and fault type samples; and using one-hot encoding to encode different fault types, wherein the different fault types include partial discharge fault, cable joint fault, high temperature fault and / or equipment aging fault; According to the labeled and encoded raw data, extracting the specified time domain features of the data signal, and performing FFT transformation on the extracted data to obtain specified frequency domain features; Concatenating the specified time domain features and the specified frequency domain features to obtain high-dimensional information features of the data signal; Based on the high-dimensional information features of the data signal, training in the Mamba module to obtain a deep learning model; and using a pre-set power system confidence loss function to supervise the trained deep learning model to obtain a ring main unit fault prediction model based on Mamba; Obtaining real-time monitoring data of the ring main unit in the distribution network, and identifying different types of faults of the ring main unit in the distribution network based on the ring main unit fault prediction model, to perform corresponding warning for the identified fault type.
[0005] Preferably, the step of extracting specified time domain features of the data signal from the labeled and encoded original data, and performing FFT transformation on the extracted data to obtain specified frequency domain features comprises: For the current, voltage, ultra-high frequency electromagnetic wave signal and / or temperature data signal in the original data, the amplitude average, standard deviation, maximum value and minimum value of the signal are calculated in the time domain, and the kurtosis, number of peaks, maximum height of peaks, minimum height of peaks, average height of peaks, maximum width of peaks, minimum width of peaks, average width of peaks are extracted; For each group of extracted data, the signal amplitude energy, maximum energy, minimum energy, standard deviation, variance, skewness, root mean square and rectified average value of the data converted into frequency domain signal are calculated in the frequency domain.
[0006] Preferably, the high-dimensional information feature based on the data signal is trained in the Mamba module to obtain a deep learning model, which specifically comprises: The original tensor format of the high-dimensional information feature data to be input into the Mamba module is set as , wherein B is the batch of training, and C is the dimension of the data feature. For A dimension is added to the specified position of the tensor to obtain , and the formula is as follows: , The 1D convolution is used for preliminary processing of , and the tensor format after convolution adjustment is , and the formula is as follows: , After entering the Mamba module, the format is adjusted first to obtain , and the formula is as follows: , After the long sequence modeling feature extraction of Mamba, the formula is as follows: , The format of the tensor is adjusted to the two-dimensional input of Softmax , and the formula is as follows: , The probability distribution of each fault is obtained after the Softmax processing , and the formula is as follows: .
[0007] The power system confidence loss function comprises an exponential cross-entropy enhancement loss function and a power system focal loss function , , wherein, is a probability distribution of model output, is a one-hot encoding of true labels, is the number of classes, is a hyperparameter for controlling the influence of exponential power; , wherein, is the confidence of the model prediction for the i-th sample, is a confidence adjustment factor for the i-th sample, configured to dynamically adjust according to the confidence of the sample; is a regulation factor for controlling the shape of the loss function; The exponential cross-entropy enhancement loss function and the power system focal loss function are combined to obtain a power system confidence loss function for supervision of the model: , , wherein, and are weight factors for balancing the contributions of the two loss functions.
[0008] After the step of training the deep learning model in the Mamba module based on the high-dimensional information features of the data signal, and using the preset power system confidence loss function to supervise the trained deep learning model to obtain the Mamba-based ring main unit fault prediction model, the method further comprises: Evaluating the proportion of correct prediction of the fault category of the ring main unit fault prediction model by the accuracy Accurac based on the Mamba model: , wherein, TP is the number of samples correctly predicted as positive class by the model; TN is the number of samples correctly predicted as negative class by the model; FP is the number of samples incorrectly predicted as positive class by the model; FN is the number of samples incorrectly predicted as negative class by the model; Evaluating the proportion of identifying all actual faults by the ring main unit fault prediction model by the recall rate Recall: , Evaluating the prediction performance of the ring main unit fault prediction model in the case of imbalance of fault categories by the F1 score: , wherein, the F1 score is a harmonic mean of the precision and the recall, for comprehensive evaluation of the precision and the recall; the precision is defined as: .
[0009] Preferably, the step of collecting the original data of the specified parameters in the operation process of the power distribution network ring main unit comprises: Real-time collection of voltage, current, temperature, humidity and / or ultra-high frequency electromagnetic wave signal related data through sensors installed at specified positions of the power distribution network ring main unit.
[0010] Preferably, the preprocessing of the original data specifically comprises: Cleaning processing of missing values, outliers and noises on the original data; wherein, for the transient characteristic signals in the original data, wavelet function denoising and normalization processing are adopted.
[0011] In a second aspect, the present application provides a ring main unit fault processing system based on a Mamba model, which comprises: A collection module for collecting original data of specified parameters in the operation process of the power distribution network ring main unit; A preprocessing module for preprocessing the original data, labeling the preprocessed original data into normal type samples and fault type samples, and using one-hot encoding to encode different fault types, wherein the different fault types include partial discharge fault, cable joint fault, high temperature fault and / or equipment aging fault; An extraction module for extracting specified time domain features of data signals according to the labeled and encoded original data, and performing FFT transformation on the extracted data to obtain specified frequency domain features; A splicing module for splicing the specified time domain features and the specified frequency domain features to obtain high-dimensional information features of the data signals; A training module for training a deep learning model in the Mamba module based on the high-dimensional information features of the data signals, and supervising the trained deep learning model using a preset power system confidence loss function to obtain a ring main unit fault prediction model based on Mamba; An identification module for obtaining real-time monitoring data of the power distribution network ring main unit, and identifying different types of faults of the power distribution network ring main unit based on the ring main unit fault prediction model to perform corresponding early warning for the identified fault types.
[0012] Thirdly, the present invention provides an electronic device, the electronic device including a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the computer program, when executed by the processor, implements the steps of the ring main unit fault handling method based on the Mamba model as described above.
[0013] Fourthly, the present invention provides a computer-readable storage medium storing a ring main unit fault handling system based on the Mamba model, the ring main unit fault handling system based on the Mamba model being executable by at least one processor to cause the at least one processor to perform the steps of the ring main unit fault handling method based on the Mamba model as described above.
[0014] The beneficial effects of using the above embodiments are: High-quality real-time monitoring raw data is acquired from distribution network ring main units, including data on normal operation and fault conditions. After data preprocessing, features helpful for fault detection are extracted from continuous monitoring data, including time-domain and frequency-domain features. A deep learning model is then trained in the Mamba module using the data to effectively extract feature information from the data, and to classify and predict it. This not only improves the accuracy of fault detection but also enables real-time monitoring and early warning of various types of faults. Attached Figure Description
[0015] Figure 1 A flowchart illustrating an embodiment of the ring main unit fault handling method based on the Mamba model provided in this application; Figure 2 A schematic diagram of the overall architecture of the Mamba model in one embodiment of the ring main unit fault handling method based on the Mamba model provided in this application; Figure 3 A schematic diagram of an embodiment of the ring main unit fault handling system based on the Mamba model provided in this application; Figure 4 A schematic block diagram of an embodiment of the electronic device provided in this application. Detailed Implementation
[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.
[0017] It should be noted that the descriptions involving "first", "second" and the like in the present application are only for the purpose of description, and cannot be understood as indicating or implying the relative importance of the indicated technical features or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first", "second" can be explicitly or implicitly included at least one of the features. In addition, the technical solutions of various embodiments can be combined with each other, but it must be based on the realization of ordinary skilled in the art, when the combination of technical solutions appears contradictory or cannot be realized, it should be considered that the combination of technical solutions does not exist, also not within the protection scope required by the present application.
[0018] The inventors of the present application found that with the development of deep learning, especially the development of convolutional neural network (CNN), recurrent neural network (RNN) and long short-term memory network (LSTM), powerful tools are provided for image and sequence data processing. The successful application of these technologies in image recognition, natural language processing and other fields provides a new idea for fault detection of distribution network ring network cabinet. Therefore, it is of great significance to study a distribution network ring network cabinet fault classification and automatic warning system based on deep learning for improving the intelligent level of power system.
[0019] Further, the partial discharge signal is a key signal for monitoring and control in the power system. Abnormal partial discharge signals may indicate potential fault risks. Therefore, developing a deep learning algorithm that can accurately analyze partial discharge signals and accurately identify abnormalities is of great importance to improve the accuracy and efficiency of fault detection. By collecting partial discharge signal data under normal operation and fault conditions of the ring network cabinet, a deep learning model can be trained to effectively extract feature information from the partial discharge signal and classify and predict it. This method not only improves the accuracy of fault detection, but also realizes real-time monitoring and early warning of faults.
[0020] Further, the present application provides a ring network cabinet fault processing method and system based on Mamba model, electronic device and storage medium, which are described in detail below.
[0021] Reference Figure 1 The method flowchart of one embodiment of the ring network cabinet fault processing method based on the Mamba model provided by the present application comprises the following steps: S1, collecting original data of specified parameters in the operation process of the distribution network ring network cabinet; S2, preprocessing the original data, labeling the preprocessed original data into normal type samples and fault type samples; and using one-hot encoding to encode different fault types, wherein the different fault types include partial discharge fault, cable joint fault, high temperature fault and / or equipment aging fault; The ring main unit of the power distribution network is a key device for carrying and distributing high voltage and large current in the power system. The cable joint inside is the most prone to failure due to the generation of a large amount of heat for a long time. If these failures are not detected and handled in time, they may seriously affect the stable operation of the power system, and even cause safety accidents. Therefore, in this embodiment, the cable joint failure is taken as a type of fault to be identified.
[0022] S3, according to the labeled and encoded original data, extracting the specified time domain features of the data signal, and performing FFT transformation on the extracted data to obtain the specified frequency domain features; S4, splicing the specified time domain features and the specified frequency domain features to obtain the high-dimensional information features of the data signal; S5, based on the high-dimensional information features of the data signal, training in the Mamba module to obtain a deep learning model; and using a preset power system confidence loss function to supervise the trained deep learning model to obtain a ring main unit fault prediction model based on Mamba; S6, obtaining real-time monitoring data of the ring main unit of the power distribution network, and identifying different types of faults of the ring main unit based on the ring main unit fault prediction model, to make corresponding early warning for the identified fault type.
[0023] This embodiment obtains high-quality real-time monitoring original data from the ring main unit of the power distribution network, including normal operation data and fault condition data. After data preprocessing, features that are helpful for fault detection are extracted from continuous monitoring data, including time domain features and frequency domain features. A deep learning model is trained in the Mamba module using the data to effectively extract feature information from the data and classify and predict it. Not only can it improve the accuracy of fault detection, but also can realize real-time monitoring and early warning of multiple types of faults.
[0024] In this embodiment, high-quality real-time monitoring data needs to be obtained from the ring main unit, including voltage, current, temperature and other parameters, combined with historical data. The acquisition of these data relies on sensor technology, and the preprocessing of data is crucial for subsequent analysis. Then, features that are helpful for fault detection are extracted from continuous monitoring data, which may include statistical features, frequency domain features and time domain features. Effective feature extraction is the key to accurate fault classification. Through deep learning method, the historical data is learned to realize the classification and timely prediction of each fault.
[0025] Specifically, in an alternative embodiment, a deep learning-based power distribution network ring main unit fault classification and automatic early warning method is provided, which can classify faults and give timely early warning, and specifically includes the following steps: Step 1: Data Acquisition. Sensors are installed at key locations in the distribution network ring main unit to monitor critical parameters such as voltage, current, temperature, humidity, and UHF electromagnetic wave signals in real time. The data types are shown in Table 1 below. Sensor selection should be based on their accuracy, durability, and suitability for specific environments. Raw data (real-time monitoring data) is collected from the ring main unit during operation using sensors. Raw data may also include partial discharge signal data.
[0026] Table 1
[0027] Step 2: Dataset Construction; (1) Data preprocessing The raw data is cleaned to remove missing values, outliers, and noise. This step may include wavelet denoising, data normalization, filtering, and other operations. For transient signals such as voltage and current, the Stationary wavelet function with a vanishing moment of 4 and a decomposition level of 3 can be used. The wavelet-denoised sample signals are then normalized to the range [-1, 1].
[0028] (2) Data labeling For supervised learning tasks, data annotation is required, especially for fault detection, where samples of normal and various fault states need to be labeled. In this embodiment, the distribution cabinet may encounter faults such as partial discharge, abnormal voltage, and abnormal temperature.
[0029] (3) Data encoding In the fault classification process, one-hot encoding is used to encode the fault type. One-hot encoding converts each category label into a binary vector, where only one position is 1 and the rest are 0, thus achieving the effect of classification encoding. In this embodiment, the faults that need to be predicted are partial discharge, cable joint faults, high temperature faults, and equipment aging faults.
[0030] Step 3: As Figure 2 As shown, the Mamba-based ring main unit fault prediction model is constructed as follows: (1) Construction of backbone network In the field of deep learning, 1D convolutions based on CNNs are commonly used for modeling and predicting long sequences. Considering that this embodiment is applied to a task with historical data, falling into the category of long sequences, Mamba is used for long sequence modeling to achieve better modeling capabilities and results. The specific data inflow and model output are as follows: At the data acquisition end, M pieces of voltage, current, ultra-high frequency electromagnetic wave signals and temperature information at transient moments after wavelet processing and normalization are collected by high-sensitivity sensors. The M pieces of data are divided into N groups, and each batch B pieces of data.
[0031] In terms of feature dimension construction, the amplitude average, standard deviation, maximum value and minimum value of the signals are calculated in the time domain for the current, voltage, ultra-high frequency electromagnetic wave signal and temperature data signal. In addition, in order to add more dimensional feature information, kurtosis, number of peaks, maximum height of peaks, minimum height of peaks, average height of peaks, maximum width of peaks, minimum width of peaks, average width of peaks are extracted. FFT transform is performed on each group of data to convert the data into frequency domain signals. The energy sum, maximum energy, minimum energy, standard deviation, variance, skewness, root mean square and rectified average of the signal amplitude of the data converted into frequency domain signals are calculated in the frequency domain. For the ultra-high frequency electromagnetic wave signal, the energy sum, maximum energy, minimum energy, standard deviation, variance, skewness, root mean square and rectified average of the signal amplitude of the data converted into frequency domain signals are calculated in the frequency domain; finally, the feature quantities extracted in the time domain and the feature quantities extracted in the frequency domain are sequentially spliced to obtain a 71-dimensional feature vector, and the value of C is 71. In this way, high-dimensional information features of the data signal can be obtained.
[0032] The tensor format of the data in the model is (B, C), where B is the batch for training, and C is the dimension of the data features. In this embodiment, 100,000 pieces of data at transient moments are collected, the batch B is set to 32, and the original tensor format of the data is Before the tensor enters the model, unsqueeze(1) is performed to increase a dimension to adapt to the needs of the network. At this time, the tensor is Then, 1D convolution is performed for preliminary processing, and the tensor format after convolution adjustment is Next, the Mamba module is entered, and the format is adjusted first to obtain the tensor format. After the long sequence modeling feature extraction of Mamba, the tensor format is The format of the tensor is adjusted to the two-dimensional input of Softmax: (batch_size, classes), where batch_size is the number of samples in the batch, and classes is the number of classes. Here, batch_size is 32, and the number of classes is 4. Finally, the probability distribution of each fault is obtained after Softmax processing: .
[0033] The specific format conversion can be represented by the following formula: .
[0034] (2) PowerSystem Confidence Loss. To supervise the network's learning more effectively in the task of multi-fault classification, this embodiment proposes a PowerSystem Confidence Loss, i.e. . This loss function contains an Exponential cross-entropy enhancement Loss and a PowerSys Focal Loss.
[0035] Exponential cross-entropy enhancement Loss, abbreviated as .
[0036]
[0037] Here, is the probability distribution of the model output, is the one-hot encoding of the true label, is the number of classes. is a hyperparameter that controls the influence of the exponential power. This modified loss function can better handle predictions that the model is very confident in (i.e. close to 0 or 1), as it amplifies the loss value of these predictions.
[0038] PowerSys Focal Loss, abbreviated as .
[0039]
[0040] where, is the confidence of the model's prediction for the th sample, which is a function of , specifically: 1-softmax . is the confidence adjustment factor of the th sample, which can be dynamically adjusted according to the confidence of the sample. is the adjustment factor, which is used to control the shape of the loss function.
[0041] Finally, these two loss functions are combined to obtain the total loss function, which is used for model supervision. and are weight factors used to balance the contributions of the two loss functions: .
[0042] Step 4: Model Evaluation (1) Evaluation Metrics: In this embodiment, a deep learning-based fault classification and automatic warning method for distribution network ring network cabinets is proposed. This method aims to analyze and process real-time data of distribution network ring network cabinets, use deep learning models to identify and classify possible fault types, and then realize automatic warning of faults. Therefore, the selection of evaluation metrics should reflect the performance and reliability of the system in the field of power system intelligent technology.
[0043] The following evaluation metrics are mainly selected: Accuracy: Measures the proportion of correctly predicted fault categories by the model, which is a basic indicator for evaluating classification performance.
[0044]
[0045] Among them, TP is the true positive (True Positives), that is, the number of samples correctly predicted as positive by the model. TN is the true negative (True Negatives), that is, the number of samples correctly predicted as negative by the model. FP is the false positive (False Positives), that is, the number of samples incorrectly predicted as positive by the model. FN is the false negative (False Negatives), that is, the number of samples incorrectly predicted as negative by the model.
[0046] Recall: Measures the proportion of all actual faults identified by the model. For a fault warning system, a high recall rate means fewer missed reports of faults. Recall measures the ability of the model to identify all actual positives (faults).
[0047]
[0048] F1 Score: The harmonic mean of precision and recall, used to consider both precision and recall. F1 score is the harmonic mean of precision and recall, which balances between precision and recall, especially suitable for class imbalance situations.
[0049]
[0050] Among them, precision (Precision) is defined as:
[0051] (2) Model Testing: In order to verify the effectiveness and accuracy of the proposed deep learning-based fault classification and automatic warning method for distribution network ring network cabinets, the following are the specific steps for model testing: Collect real-time monitoring data such as voltage, current, temperature, etc. from the distribution network ring main unit, including data under normal operation and fault conditions. Run the trained model on the test set, collect the prediction results, evaluate the model's prediction results according to the defined evaluation indicators, and analyze the faults that the model failed to predict correctly to identify the model's shortcomings.
[0052] Finally, after the model training and testing are completed, real-time monitoring data of the distribution network ring main unit are obtained, and different types of faults of the distribution network ring main unit are identified in real time based on the ring main unit fault prediction model, so as to timely warn against the identified fault types.
[0053] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.
[0054] The embodiment also provides a ring main unit fault processing system based on a Mamba model, which corresponds to the ring main unit fault processing method based on the Mamba model in the above embodiment. As shown in the figure, the ring main unit fault processing system based on the Mamba model comprises: Figure 3 The acquisition module 100 is configured to acquire original data of specified parameters during operation of the distribution network ring main unit. The preprocessing module 110 is configured to preprocess the original data, label the preprocessed original data into normal type samples and fault type samples, and use one-hot encoding to encode different fault types, wherein the different fault types include local discharge fault, cable joint fault, high temperature fault and / or equipment aging fault. The extraction module 120 is configured to extract specified time domain features of data signals according to the labeled and encoded original data, and perform FFT transformation on the extracted data to obtain specified frequency domain features. The splicing module 130 is configured to splice the specified time domain features and the specified frequency domain features to obtain high-dimensional information features of the data signals. The training module 140 is configured to train a deep learning model in the Mamba module based on the high-dimensional information features of the data signals, and supervise the trained deep learning model using a preset power system confidence loss function to obtain a ring main unit fault prediction model based on Mamba. The identification module 150 is configured to obtain real-time monitoring data of the distribution network ring main unit, and identify different types of faults of the distribution network ring main unit based on the ring main unit fault prediction model, so as to timely warn against the identified fault types.
[0055] The specific definition of each module of the Mamba model-based ring main unit fault processing system can refer to the definition of the Mamba model-based ring main unit fault processing method in the foregoing, which will not be repeated here. Each module in the Mamba model-based ring main unit fault processing system described above can be realized by software, hardware, and a combination thereof in whole or in part. The above-mentioned modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to each module.
[0056] Referring to Figure 4 is a schematic diagram of the operating environment of the preferred embodiment of the Mamba model-based ring main unit fault processing system 10 of the present application.
[0057] In this embodiment, the Mamba model-based ring main unit fault processing system 10 is installed and run in the electronic device 1. The electronic device 1 is a device capable of automatically performing numerical calculation and / or information processing according to the instructions set or stored in advance. The electronic device 1 can be a computer, a single network server, a server group composed of multiple network servers, or a cloud composed of a large number of hosts or network servers based on cloud computing, wherein cloud computing is a kind of distributed computing, which is a super virtual computer composed of a group of loosely coupled computer clusters. (The electronic device 1 can be a server, a smart phone, a tablet computer, a portable computer, a desktop computer, etc. Terminal equipment with storage and operation functions. In an embodiment, when the electronic device 1 is a server, the server can be one or more of rack-mounted servers, blade servers, tower servers, or cabinet servers, etc.) In this embodiment, the electronic device 1 can include, but is not limited to, a memory 11, a processor 12, and a network interface 13 that can communicate with each other through a system bus, and the memory 11 stores the Mamba model-based ring main unit fault processing system 10 that can run on the processor 12. It should be noted that, Figure 1 Only the electronic device 1 with components 11-13 is shown, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented instead.
[0058] The memory 11 includes an internal memory and at least one type of readable storage medium. The internal memory provides a cache for the operation of the electronic device 1; the readable storage medium can be a nonvolatile storage medium such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the readable storage medium can be an internal storage unit of the electronic device 1, such as a hard disk of the electronic device 1; in other embodiments, the nonvolatile storage medium can also be an external storage device of the electronic device 1, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. In this embodiment, the readable storage medium of the memory 11 is generally used to store an operating system and various application software installed in the electronic device 1, such as the ring main unit fault processing system 10 based on the Mamba model in an embodiment of the present application. In addition, the memory 11 can also be used to temporarily store various data that have been output or will be output.
[0059] The processor 12 can be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor 12 is generally used to control the overall operation of the electronic device 1, such as performing control and processing related to data interaction or communication with other devices, etc. In this embodiment, the processor 12 is used to run program codes or process data stored in the memory 11, such as the ring main unit fault processing system 10 based on the Mamba model.
[0060] The network interface 13 can include a wireless network interface or a wired network interface, and the network interface 13 is generally used to establish a communication connection between the electronic device 1 and other electronic devices.
[0061] The ring main unit fault processing system 10 based on the Mamba model includes at least one computer readable instruction stored in the memory 11, and the at least one computer readable instruction can be executed by the processor 12 to implement various embodiments of the present application.
[0062] In addition, the application further provides a computer readable storage medium, which stores a ring net cabinet fault processing system based on a Mamba model, and the ring net cabinet fault processing system based on the Mamba model can be executed by at least one processor to enable the at least one processor to execute steps of each of the above embodiments. The computer readable storage medium embodiment of the application is basically the same as the above-mentioned electronic device 1 and method embodiments, and will not be described here.
[0063] It should be noted that in this paper, the term "including", "containing" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of other identical elements in the process, method, article or device including the element.
[0064] Through the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by means of software and necessary general hardware platform, of course, it can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the application can be embodied in the form of software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a plurality of instructions for making a terminal device (which can be a mobile phone, computer, server, air conditioner or network device) execute the method described in each embodiment of the application.
[0065] The preferred embodiments of the application are described above with reference to the accompanying drawings, which do not limit the scope of the application. The above-mentioned embodiment serial numbers of the application are only for description, not representing the advantages and disadvantages of the embodiments. In addition, although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that here.
Claims
1. A method for handling faults in a ring main unit based on a Mamba model, characterized by, The ring main unit fault processing method based on the Mamba model comprises: Collecting original data of specified parameters during operation of a ring main unit of a power distribution network; Preprocessing the original data, labeling the preprocessed original data into normal type samples and fault type samples, and using one-hot encoding to encode different fault types, wherein the different fault types include partial discharge fault, cable joint fault, high temperature fault and / or equipment aging fault; According to the labeled and encoded original data, extracting specified time domain features of data signals thereof, and performing FFT transformation on the extracted data to obtain specified frequency domain features; Splicing the specified time domain features and the specified frequency domain features to obtain high-dimensional information features of the data signals; Training a deep learning model in a Mamba module based on the high-dimensional information features of the data signals, and supervising the trained deep learning model using a preset power system confidence loss function to obtain a ring main unit fault prediction model based on Mamba; Obtaining real-time monitoring data of the ring main unit of the power distribution network, and identifying different types of faults of the ring main unit of the power distribution network based on the ring main unit fault prediction model to perform corresponding early warning for the identified fault types.
2. The Mamba model-based ring main unit fault processing method of claim 1, wherein, The step of extracting specified time domain features of data signals thereof according to the labeled and encoded original data, and performing FFT transformation on the extracted data to obtain specified frequency domain features comprises: For current, voltage, ultra-high frequency electromagnetic wave signals and / or temperature data signals in the original data, calculating the amplitude average, standard deviation, maximum value and minimum value of the signals in the time domain, and performing feature extraction of kurtosis, number of peaks, maximum height of peaks, minimum height of peaks, average height of peaks, maximum width of peaks, minimum width of peaks, and average width of peaks; For each group of extracted data, calculating the energy sum, maximum energy, minimum energy, standard deviation, variance, skewness, root mean square, and rectified average value of the signal amplitude of the data converted into frequency domain signals.
3. The Mamba model-based ring main unit fault processing method of claim 1, wherein, The step of training a deep learning model in a Mamba module based on the high-dimensional information features of the data signals, and supervising the trained deep learning model using a preset power system confidence loss function to obtain a ring main unit fault prediction model based on Mamba comprises: The original tensor format of the high-dimensional information feature data to be input into the Mamba module is set as where B is a batch of training, and C is the dimension of data features. For Adding one dimension to the specified position of the tensor, resulting in , the formula is as follows: ， Pre-processed by 1D convolution The tensor format after convolution adjustment is , and the formula is as follows: ， After entering the Mamba module, the format is adjusted first to get , and the formula is as follows: ， After the long sequence modeling feature extraction of Mamba, we get , the formula is as follows: ， Adjusting the format of a tensor to be a two-dimensional input for Softmax , as follows: ， The probability distribution of each fault is obtained after Softmax processing The formula is as follows: 。 4. The Mamba model-based ring main unit fault processing method of claim 1, wherein, The power system confidence loss function includes an exponential cross-entropy boosting loss function and a power system focal loss function , ， where, is the probability distribution of the model output, is the one-hot encoding of the true label, is the number of classes, is a hyperparameter that controls the influence of the exponential power; ， in, The model is for the first The confidence level of the prediction for each sample. It is the first The confidence adjustment factor for each sample is configured to be dynamically adjusted based on the confidence level of the sample. It is an adjustment factor used to control the shape of the loss function; combining the exponential cross-entropy boosting loss function and the power system focal loss function to obtain a power system confidence loss function for supervision of the model: ， wherein, and are weight factors to balance the contribution of the two loss functions.
5. The Mamba model-based ring main unit fault processing method according to any one of claims 1-4, characterized in that, After the step of training a deep learning model in a Mamba module based on the high-dimensional information features of the data signals, and supervising the trained deep learning model using a preset power system confidence loss function to obtain a ring main unit fault prediction model based on Mamba, further comprising: Evaluating the proportion of correct prediction of fault categories of the ring main unit fault prediction model based on Mamba model through accuracy Accurac: ， Wherein TP is the number of samples correctly predicted as positive class by the model; TN is the number of samples correctly predicted as negative class by the model; FP is the number of samples incorrectly predicted as positive class by the model; FN is the number of samples incorrectly predicted as negative class by the model; Evaluating the proportion of identification of all actual faults by the ring main unit fault prediction model through recall rate Recall: ， Evaluating the prediction performance of the ring main unit fault prediction model in the case of unbalanced fault categories through F1 score: ， Wherein, the F1 score is a harmonic mean of the precision and the recall, used to evaluate the precision and the recall comprehensively; the precision is defined as: 。 6. The Mamba model based ring main unit fault handling method according to any one of claims 1-4, characterized in that, The step of collecting the original data of the specified parameters in the operation process of the power distribution network ring main unit comprises: The sensor installed at the specified position of the power distribution network ring main unit collects the relevant data of voltage, current, temperature, humidity and / or ultra-high frequency electromagnetic wave signals in real time.
7. The Mamba model based ring main unit fault handling method according to any one of claims 1-4, characterized in that, The preprocessing of the original data specifically comprises: The original data is cleaned of missing values, outliers and noise; wherein, for the transient characteristic signals in the original data, wavelet function denoising and normalization processing are performed.
8. A Mamba model-based ring main unit fault handling system, characterized in that, The system comprises: a collection module configured to collect original data of specified parameters in an operation process of a power distribution network ring main unit; a preprocessing module configured to preprocess the original data, label the preprocessed original data into normal type samples and fault type samples, and use one-hot encoding to encode different fault types, wherein the different fault types include local discharge fault, cable joint fault, high temperature fault and / or equipment aging fault; an extraction module configured to extract specified time domain features of data signals according to the labeled and encoded original data, and perform FFT transformation on the extracted data to obtain specified frequency domain features; a splicing module configured to splice the specified time domain features and the specified frequency domain features to obtain high-dimensional information features of the data signals; a training module configured to train a deep learning model in a Mamba module based on the high-dimensional information features of the data signals, and supervise the trained deep learning model using a preset power system confidence loss function to obtain a ring main unit fault prediction model based on Mamba; an identification module configured to obtain real-time monitoring data of the power distribution network ring main unit, and identify different types of faults of the power distribution network ring main unit based on the ring main unit fault prediction model to perform corresponding early warning for the identified fault types.
9. An electronic device, comprising: The electronic device comprises a memory and a processor, the memory stores a computer program executable on the processor, and the computer program is executed by the processor to implement the steps of the ring main unit fault processing method based on the Mamba model according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the ring main unit fault processing method based on the Mamba model according to any one of claims 1 to 7.