Self-adaptive overcurrent protection method, device, equipment and storage medium

Abstract

The invention discloses a self-adaptive overcurrent protection method and device, equipment and a storage medium. The method comprises the following steps: acquiring target multi-dimensional data through a multi-sensor acquisition system; determining fusion data corresponding to the target multi-dimensional data according to a sensor fusion algorithm; extracting abnormal signal features existing in the fused data; classifying the abnormal signal features through a machine learning algorithm, and judging whether an overcurrent event occurs in the target equipment or not; and if the target device has the overcurrent event, performing overcurrent protection on the target device according to a preset protection logic. Through the implementation of the scheme of the invention, target multi-dimensional data is acquired according to a multi-sensor acquisition system, and technical means such as a sensor fusion algorithm, an abnormal signal feature extraction algorithm and a machine learning algorithm are comprehensively utilized to realize accurate prediction and dynamic protection of an overcurrent event.

Description

Technical Field

[0001] The present application relates to the field of electronics, and in particular to an adaptive overcurrent protection method, device, equipment and storage medium. Background Art

[0002] With the acceleration of industrialization and the development of smart grids, various electrical devices play a vital role in power systems. However, current overload, as one of the main factors affecting the safe operation of equipment, has been a major concern in the industry due to its potential for serious consequences such as equipment damage, system failures, and even fires.

[0003] Existing technologies primarily rely on fixed current thresholds and simple protection logic to implement overcurrent protection. This approach is inadequate for handling complex and changing load characteristics and environmental conditions. On the one hand, fixed current thresholds often fail to accurately reflect the actual overcurrent risk under different loads, potentially leading to over- or underprotection. On the other hand, existing technologies lack the ability to comprehensively analyze data from different sensors and extract abnormal signal characteristics, making it difficult to accurately predict and promptly respond to overcurrent events. Furthermore, existing overcurrent protection mechanisms often lack adaptability and learning capabilities, making it impossible to dynamically adjust protection strategies based on historical data and the current environment, thus limiting the effectiveness of protection. Summary of the Invention

[0004] The present application provides an adaptive overcurrent protection method, apparatus, device and storage medium, which are used to solve the problem in the related art that the existing overcurrent protection mechanism has limited protection effect due to lack of adaptability and learning ability.

[0005] A first aspect of the present application provides an adaptive overcurrent protection method, the adaptive overcurrent protection method comprising: Acquire target multi-dimensional data through a multi-sensor acquisition system; Determining fusion data corresponding to the target multidimensional data according to a sensor fusion algorithm; Extracting abnormal signal features present in the fused data; Classifying the abnormal signal features through a machine learning algorithm to determine whether an overcurrent event occurs in the target device; If an overcurrent event occurs in the target device, overcurrent protection is performed on the target device according to a preset protection logic.

[0006] Optionally, in a first implementation of the first aspect of the present application, the step of acquiring target multi-dimensional data through a multi-sensor acquisition system includes: generating a multidimensional time series data set based on the multidimensional data collected by the multi-sensor acquisition system; Performing time synchronization processing on all data of the multidimensional time series data set; The target multidimensional data is determined by performing denoising processing on the multidimensional data after the time synchronization processing is completed.

[0007] Optionally, in a second implementation of the first aspect of the present application, the sensor fusion algorithm includes a weighted Kalman filter formula, and the step of determining the fused data corresponding to the target multidimensional data according to the sensor fusion algorithm includes: The fusion data is calculated by the following formula: , in, Represented as the fused data, Represented as the target multidimensional data, is the Kalman gain matrix, is the observation vector, is the observation matrix, Indicated as the corresponding time.

[0008] Optionally, in a third implementation of the first aspect of the present application, the step of extracting abnormal signal features present in the fused data includes: extracting a current signal from the fused data; Obtaining a frequency domain feature vector of the current signal by performing a fast Fourier transform on the current signal; Determining the time domain characteristic vector and dynamic characteristics of the current signal; generating a comprehensive feature vector according to the frequency domain feature vector, the time domain feature vector and the dynamic feature; The abnormal signal feature is determined according to the Mahalanobis distance formula and the comprehensive feature vector.

[0009] Optionally, in a fourth implementation of the first aspect of the present application, the machine learning algorithm includes a support vector machine, and the step of classifying the abnormal signal features using the machine learning algorithm to determine whether an overcurrent event occurs in the target device includes: Training a support vector machine classifier based on historical normal data and fault data of the target device; The abnormal signal feature is input into the support vector machine classifier to determine whether an overcurrent event occurs in the target device.

[0010] Optionally, in a fifth implementation of the first aspect of the present application, the method further includes: Dynamically updating the overcurrent protection threshold according to the real-time load characteristics and environmental parameters of the target device; Modifying the overcurrent protection threshold according to the judgment result of the support vector machine classifier on the overcurrent event; It is determined whether an overcurrent event occurs in the target device according to the corrected real-time overcurrent protection threshold.

[0011] Optionally, in a sixth implementation of the first aspect of the present application, after the step of performing overcurrent protection on the target device according to a preset protection logic if an overcurrent event occurs in the target device, the method further includes: When the overcurrent protection is completed, the event parameters corresponding to the overcurrent protection event are obtained; updating the training set of the support vector machine classifier according to the event parameters; The support vector machine classifier is optimized according to the training set.

[0012] A second aspect of the present application provides an adaptive overcurrent protection device, the adaptive overcurrent protection device comprising: An acquisition module is used to acquire target multi-dimensional data through a multi-sensor acquisition system; A determination module, configured to determine fusion data corresponding to the target multidimensional data according to a sensor fusion algorithm; An extraction module, configured to extract abnormal signal features present in the fused data; A judgment module, configured to classify the abnormal signal features using a machine learning algorithm to determine whether an overcurrent event occurs in the target device; The protection module is used to perform overcurrent protection on the target device according to a preset protection logic if an overcurrent event occurs in the target device.

[0013] A third aspect of an embodiment of the present application provides an electronic device, comprising a memory and a processor, wherein the processor is used to execute a computer program stored on the memory. When the processor executes the computer program, it implements the steps of the adaptive overcurrent protection method provided in the first aspect of the embodiment of the present application.

[0014] The fourth aspect of the embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the adaptive overcurrent protection method provided in the first aspect of the embodiment of the present application are implemented.

[0015] In summary, according to the adaptive overcurrent protection method, device, equipment and storage medium provided by the present application, target multi-dimensional data is acquired through a multi-sensor acquisition system; fusion data corresponding to the target multi-dimensional data is determined according to a sensor fusion algorithm; abnormal signal features present in the fusion data are extracted; the abnormal signal features are classified by a machine learning algorithm to determine whether an overcurrent event has occurred in the target device; if an overcurrent event has occurred in the target device, overcurrent protection is performed on the target device according to a preset protection logic. Through the implementation of the present application, target multi-dimensional data is acquired according to a multi-sensor acquisition system, and technical means such as sensor fusion algorithms, abnormal signal feature extraction and machine learning algorithms are comprehensively used to achieve accurate prediction and dynamic protection of overcurrent events. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flow chart of an adaptive overcurrent protection method provided in an embodiment of the present application; Figure 2 A schematic diagram of a program module of an adaptive overcurrent protection device provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] In order to make the purpose, features, and advantages of the invention of this application more obvious and easy to understand, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of this application.

[0018] In order to solve the problem that the existing overcurrent protection mechanism in the related art limits the protection effect due to the lack of adaptability and learning ability, the embodiment of the present application provides an adaptive overcurrent protection method, such as Figure 1 This is a flow chart of the adaptive overcurrent protection method provided in this embodiment. The adaptive overcurrent protection method includes the following steps: Step 110: Acquire target multi-dimensional data through a multi-sensor acquisition system.

[0019] Specifically, in this embodiment, appropriate sensors, such as current sensors, temperature sensors, and voltage sensors, are selected based on the characteristics and requirements of the target device. These sensors should possess high precision, high stability, and high reliability to ensure accurate and reliable collected data. The selected sensors are placed in appropriate locations on the target device to ensure comprehensive coverage of all key components. For example, on a power transformer, multiple current sensors can be deployed to monitor the current flow of different phases; on a motor, temperature sensors can be deployed to monitor temperature changes. The multi-sensor acquisition system is activated to collect multidimensional data from the target device in real time. This data can include various physical quantities such as current, voltage, temperature, and vibration. The collected multidimensional data is preprocessed, including steps such as data cleaning, denoising, and calibration, to eliminate outliers and noise in the data, thereby obtaining the final target multidimensional data and improving the accuracy and efficiency of subsequent data processing.

[0020] Step 120: Determine fusion data corresponding to the target multi-dimensional data according to a sensor fusion algorithm.

[0021] Specifically, in this embodiment, according to the characteristics and requirements of the target multidimensional data, an appropriate sensor fusion algorithm is selected, including but not limited to the Kalman filter method. The sensor fusion algorithm can be selected and adjusted according to different data types and requirements. The preprocessed multidimensional data is input into the selected fusion algorithm for data fusion processing. The fusion algorithm will perform a comprehensive analysis of the data from different sensors to eliminate redundancy and contradictions between the data and improve the accuracy and reliability of the data.

[0022] Step 130: Extract abnormal signal features in the fused data.

[0023] Specifically, abnormal signal features are extracted from the fused data for subsequent machine learning classification and judgment. In this embodiment, the fused data is input into the selected feature extraction method for feature extraction processing, wherein the feature extraction method includes but is not limited to time domain analysis, frequency domain analysis, wavelet transform, etc., and key features related to overcurrent events can be extracted from the fused data. The feature extraction method will conduct an in-depth analysis of the fused data, extract abnormal signal features related to overcurrent events, and screen and optimize the extracted features to ensure the validity and accuracy of the features.

[0024] Step 140: Classify the abnormal signal features through a machine learning algorithm to determine whether an overcurrent event occurs in the target device.

[0025] Specifically, in this embodiment, a machine learning algorithm is used to classify and determine the extracted abnormal signal features. These features are then fed into a selected machine learning algorithm to train the model. During the training process, the algorithm learns and identifies the characteristic patterns of overcurrent events based on known overcurrent event data and non-overcurrent event data. The trained model is then validated to ensure it can accurately and reliably determine whether an overcurrent event has occurred in the target device. The abnormal signal features of the target device to be determined are then fed into the trained model for classification and determination. The model then identifies whether an overcurrent event has occurred in the target device based on the characteristic patterns.

[0026] Step 150: If an overcurrent event occurs in the target device, overcurrent protection is performed on the target device according to a preset protection logic.

[0027] Specifically, in this embodiment, based on the classification judgment results of the machine learning algorithm, it is determined whether an overcurrent event occurs in the target device. If an overcurrent event occurs in the target device, the corresponding protection measures are triggered according to the preset protection logic, including but not limited to cutting off the power supply, reducing the load, and starting the alarm. The protection logic should be set and adjusted according to the characteristics and requirements of the target device. At the same time, the results and process of the protection execution are recorded to provide a reference for subsequent analysis and optimization.

[0028] In an optional implementation of this embodiment, the step of acquiring target multidimensional data through a multi-sensor acquisition system includes: generating a multidimensional time series data set based on the multidimensional data collected by the multi-sensor acquisition system; performing time synchronization processing on all data in the multidimensional time series data set; and determining the target multidimensional data by denoising the multidimensional data after the time synchronization processing is completed.

[0029] Specifically, in this embodiment, current, voltage, and temperature signals are collected by a current sensor, a temperature sensor, and a voltage sensor, respectively, to form an initial multi-dimensional time series data set: , in, Expressed as the current value at time t, Expressed as the temperature value at time t, Expressed as the voltage value at time t, the collected data of different sensors are time-synchronized, and the timestamp of each sensor sampling time is recorded. Based on a reference time axis (such as the system master clock), other sensor data are mapped to the reference time axis. If the sampling time points of different sensors do not completely overlap, interpolation methods can be used to fill in the data of unsampled points. Among them, interpolation methods include but are not limited to linear difference, spline interpolation, etc. The multi-dimensional sensor data set after time synchronization can ensure that the data of each dimension is consistent in the time dimension. Afterwards, the window function is used Denoise the data to obtain smoothed multidimensional data: , in, It is a Gaussian window function, which assigns Gaussian distribution weights to the data points in the window and defines a window size w , calculate the Gaussian weight array, and for each data point, calculate the weighted average value within the window, It is the time delay correction parameter. The target multidimensional data obtained after denoising will provide input for the subsequent fusion algorithm.

[0030] In an optional implementation of this embodiment, the sensor fusion algorithm includes but is not limited to the weighted Kalman filter formula: , in, Represents the fused multidimensional state vector, which is the optimal estimate of the Kalman filter output. It is directly calculated using the weighted Kalman filter formula. Represents the smoothed multidimensional data vector, which is the sensor data that has been preprocessed (such as denoising, synchronization), and is obtained through the preprocessing of multidimensional data. Expressed as the Kalman gain matrix, it determines the weight of new observations in the state estimation and is calculated by the following formula: , is the state covariance matrix, which represents the uncertainty of the estimate and is updated by the Kalman filter recursive formula: , A is the state transfer matrix, which describes the dynamic changes of the system state. Q is the process noise covariance matrix, which represents the uncertainty of the process model. is the observation matrix, which represents the observation model and maps the state vector to the observation space. It is determined according to the system modeling and is generally a matrix known or obtained through system calibration. is the observation noise covariance matrix, which represents the uncertainty of the observation. The observation vector is the sensor observation data at the current moment. It is obtained directly from the data collected by the sensor after preprocessing. The multidimensional information of different data sources is fused through the Kalman filtering algorithm to improve robustness and consistency.

[0031] In an optional implementation of this embodiment, the step of extracting abnormal signal features present in the fused data includes: extracting a current signal from the fused data; obtaining a frequency domain feature vector of the current signal by performing a fast Fourier transform on the current signal; determining a time domain feature vector and a dynamic feature of the current signal; generating a comprehensive feature vector based on the frequency domain feature vector, the time domain feature vector, and the dynamic feature; and determining the abnormal signal features based on the Mahalanobis distance formula and the comprehensive feature vector.

[0032] Specifically, in this embodiment, basic time-domain statistical features are extracted from the fused data, including the average amplitude, standard deviation, maximum value, minimum value, etc. of the current. It can be understood that the time-domain features are an intuitive description of the signal strength and fluctuation range, which can be used as a preliminary reference for anomaly detection. The time-domain statistical formula is: , , in, is the average amplitude of the current signal, is the standard deviation of the current signal, indicating the degree of fluctuation, is the current signal at time t, To analyze the time window, that is, in the process of time domain feature extraction, the current signal in the input fusion data , calculate the average amplitude of the current and standard deviation , and calculate the minimum value of the current signal and maximum value , and finally output the time domain feature vector : , Secondly, frequency domain analysis tools (such as fast Fourier transform (FFT)) are used to extract the frequency characteristics of the signal and obtain the harmonic components and amplitudes of the current signal. Frequency domain characteristics can reflect the periodicity and high-frequency components of the signal, which is particularly important for anomaly detection. For example, high-order harmonics may indicate equipment failure. The fast Fourier transform formula is: , in, is the frequency domain representation of the current signal, is the angular frequency, is the time domain current signal, is a complex exponential, representing the basis function of frequency domain transformation. It can be understood that the current signal is extracted from the fused data. , the current signal Perform fast Fourier transform to convert it to the frequency domain and get the spectrum , where the key to extracting frequency domain features is to obtain the amplitude of the frequency component And the corresponding energy distribution: , in, is the amplitude of frequency component ii, and are the real and imaginary parts of the frequency domain signal, To obtain the corresponding frequency components, after converting the time domain current signal to the frequency domain, the amplitudes of the main frequency components (such as the amplitudes of the fundamental wave, second harmonic, and third harmonic) are extracted. , and calculate the total harmonic distortion THD: , And output the final frequency domain feature vector .

[0033] Furthermore, the dynamic changes of the current signal in the time and frequency domains are analyzed, including the change trend (increase or decrease speed) and mutation point detection. The dynamic characteristics can capture transient abnormal behaviors, such as short-term overcurrent or sudden high-frequency noise. The dynamic trend formula is: , , in, is the instantaneous rate of change of the signal, For the time interval The current increment on the , and the mutation detection formula (detecting the change point by the second-order derivative) is: , in, is a mutation point detection marker, is the second derivative of the signal, which is used to capture the acceleration of the rate of change, is the threshold for mutation point detection, which is usually set based on historical data. In the corresponding formula, it can be understood that the instantaneous rate of change is obtained by taking the first-order derivative of the current signal. ,right Then take the derivative and get the detection of significant change points , which is an abnormal signal, marks all mutation points, records the time position and amplitude changes, and finally outputs the dynamic characteristics: .

[0034] The features extracted in each step are integrated to form a unified feature vector to provide input for subsequent machine learning classification. The features are normalized and standardized using statistical methods to further analyze the degree of signal abnormality. 、 Integrate to form a comprehensive feature vector , and then use the Mahalanobis distance to calculate the degree to which the current signal characteristics deviate from the normal distribution to determine whether there is an anomaly. The Mahalanobis distance formula is: , in, is the Mahalanobis distance of the feature, is the mean vector of normal working conditions. Finally, the above features are integrated again to output abnormal signal features including time domain features, frequency domain features, dynamic features and abnormal signal mutation points: , For subsequent machine learning classification.

[0035] In an optional implementation of this embodiment, the step of classifying abnormal signal features through a machine learning algorithm to determine whether an overcurrent event has occurred in the target device includes: training a support vector machine classifier based on the historical normal data and fault data of the target device; inputting the abnormal signal features into the support vector machine classifier to determine whether an overcurrent event has occurred in the target device.

[0036] Specifically, in this embodiment, historical normal data and fault data (i.e., current overcurrent data) are used to train a support vector machine (SVM) classifier. The constrained SVM function is: , The constraints of this function are: , in, is the normal vector of the hyperplane, b is the bias of the hyperplane, is a slack variable used to process inseparable samples, C is a regularization parameter that controls the influence of the slack variable on the objective function, is the sample label, that is, the collected historical normal data and fault data, is the sample feature vector. It can be understood that the constrained SVM function can be converted into an equivalent unconstrained optimization form by introducing the hinge loss function: , in, Hinge loss ,when When , the loss is 0, otherwise the loss increases linearly.

[0037] The abnormal signal characteristics will be determined Input into the trained support vector machine classifier to determine the overcurrent event result: , Overcurrent judgment result , for dynamic threshold adjustment.

[0038] In an optional implementation of this embodiment, the overcurrent protection threshold is dynamically updated according to the real-time load characteristics and environmental parameters of the target device. The dynamically updated threshold is: , in, is the real-time protection threshold, is the initial threshold, and is the adjustment coefficient, and are the variance of load characteristics and the mean of environmental parameters respectively. Then, the overcurrent classification results are used to Correction threshold: , in, It is the correction coefficient. After the correction is completed, the corrected real-time overcurrent protection threshold is output. , and according to Determine whether an overcurrent event occurs in the target device.

[0039] In an optional implementation of this embodiment, if the real-time current signal When greater than the overcurrent protection threshold , it is determined to be an overcurrent event, triggering the protection logic to execute protection strategies such as cutting off the load, reducing the output power, or issuing an alarm signal, and collecting the event parameters corresponding to the overcurrent protection event. The model training set of the support vector machine classifier is updated according to the event parameters, and the support vector machine classifier is continuously optimized based on the model training set, that is, the objective function corresponding to the support vector machine classifier is updated: , in, is the regularization parameter, is the loss function, and the target device is over-protected according to the continuously optimized support vector machine classifier.

[0040] According to the adaptive overcurrent protection method provided by the present application, target multidimensional data is acquired through a multi-sensor acquisition system; fusion data corresponding to the target multidimensional data is determined based on a sensor fusion algorithm; abnormal signal features present in the fusion data are extracted; the abnormal signal features are classified using a machine learning algorithm to determine whether an overcurrent event has occurred in the target device; if an overcurrent event has occurred in the target device, overcurrent protection is performed on the target device based on preset protection logic. Through the implementation of the present application, target multidimensional data is acquired based on a multi-sensor acquisition system, and technical means such as sensor fusion algorithms, abnormal signal feature extraction, and machine learning algorithms are comprehensively utilized to achieve accurate prediction and dynamic protection of overcurrent events.

[0041] Figure 2 An adaptive overcurrent protection device is provided in an embodiment of the present application. The adaptive overcurrent protection device can be used to implement the adaptive overcurrent protection method in the above embodiment. Figure 2 As shown, the adaptive overcurrent protection device mainly includes: An acquisition module 10 is used to acquire target multi-dimensional data through a multi-sensor acquisition system; A determination module 20 is configured to determine fusion data corresponding to the target multidimensional data according to a sensor fusion algorithm; Extraction module 30, used to extract abnormal signal features present in the fused data; A judgment module 40 is used to classify abnormal signal features through a machine learning algorithm to determine whether an overcurrent event occurs in the target device; The protection module 50 is configured to perform overcurrent protection on the target device according to a preset protection logic if an overcurrent event occurs in the target device.

[0042] In an optional implementation of this embodiment, the acquisition module is specifically used to: generate a multidimensional time series data set based on the multidimensional data collected by the multi-sensor acquisition system; perform time synchronization processing on all data in the multidimensional time series data set; and determine the target multidimensional data by denoising the multidimensional data after the time synchronization processing is completed.

[0043] In an optional implementation of this embodiment, the extraction module is specifically used to: extract the current signal from the fused data; obtain the frequency domain feature vector of the current signal by performing a fast Fourier transform on the current signal; determine the time domain feature vector and dynamic characteristics of the current signal; generate a comprehensive feature vector based on the frequency domain feature vector, the time domain feature vector and the dynamic characteristics; and determine the abnormal signal characteristics based on the Mahalanobis distance formula and the comprehensive feature vector.

[0044] In an optional implementation of this embodiment, the determination module is specifically used to: train a support vector machine classifier based on historical normal data and fault data of the target device; input abnormal signal features into the support vector machine classifier to determine whether an overcurrent event occurs in the target device.

[0045] Furthermore, in an optional implementation of this embodiment, the adaptive overcurrent protection device further includes an update module and a correction module. The update module is configured to dynamically update the overcurrent protection threshold based on the real-time load characteristics and environmental parameters of the target device. The correction module is configured to correct the overcurrent protection threshold based on the support vector machine classifier's judgment of the overcurrent event. The judgment module is further configured to determine whether an overcurrent event has occurred in the target device based on the corrected real-time overcurrent protection threshold.

[0046] Furthermore, in another optional implementation of this embodiment, the adaptive overcurrent protection device further includes an optimization module. The acquisition module is further configured to: obtain event parameters corresponding to the overcurrent protection event after the overcurrent protection is completed. The update module is further configured to: update the training set of the support vector machine classifier based on the event parameters. The optimization module is configured to: optimize the support vector machine classifier based on the training set.

[0047] According to the application plan Figure 3 An electronic device provided in an embodiment of the present application can be used to implement the adaptive overcurrent protection method in the aforementioned embodiment, mainly comprising: Memory 301, processor 302, and computer program 303 stored in memory 301 and executable on processor 302. Memory 301 and processor 302 are connected via communication. When processor 302 executes computer program 303, the adaptive overcurrent protection method described in the aforementioned embodiment is implemented. The number of processors may be one or more.

[0048] The memory 301 can be a high-speed random access memory (RAM) memory or a non-volatile memory such as a disk drive. The memory 301 is used to store executable program code. The processor 302 is coupled to the memory 301 .

[0049] Furthermore, the embodiment of the present application also provides a computer-readable storage medium, which can be provided in the electronic device in the above embodiments. The computer-readable storage medium can be the above Figure 3 Memory in the illustrated embodiment.

[0050] The computer-readable storage medium stores a computer program that, when executed by a processor, implements the adaptive overcurrent protection method described in the aforementioned embodiment. Furthermore, the computer-readable storage medium may be any medium capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), RAM, a magnetic disk, or an optical disk.

[0051] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0052] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.

[0053] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An adaptive overcurrent protection method, characterized in that: include: Acquire multi-dimensional target data through a multi-sensor acquisition system; Determine fusion data corresponding to the target multidimensional data according to a sensor fusion algorithm; Extracting abnormal signal features present in the fused data; Classifying the abnormal signal features through a machine learning algorithm to determine whether an overcurrent event occurs in the target device; If an overcurrent event occurs in the target device, overcurrent protection is performed on the target device according to a preset protection logic.

2. The adaptive overcurrent protection method according to claim 1, characterized in that: The step of acquiring target multi-dimensional data through a multi-sensor acquisition system includes: Generate a multidimensional time series data set based on the multidimensional data collected by the multi-sensor acquisition system; Performing time synchronization processing on all data of the multidimensional time series data set; The target multi-dimensional data is determined by performing denoising processing on the multi-dimensional data after the time synchronization processing is completed.

3. The adaptive overcurrent protection method according to claim 1, characterized in that: The sensor fusion algorithm includes a weighted Kalman filter formula, and the step of determining the fusion data corresponding to the target multidimensional data according to the sensor fusion algorithm includes: The fused data is calculated by the following formula: ， in, Represented as the fused data, Represented as the target multidimensional data, is the Kalman gain matrix, is the observation vector, is the observation matrix, Expressed as the corresponding time.

4. The adaptive overcurrent protection method according to claim 1, characterized in that: The step of extracting abnormal signal features present in the fused data comprises: extracting a current signal from the fused data; Obtaining a frequency domain feature vector of the current signal by performing a fast Fourier transform on the current signal; Determining the time domain characteristic vector and dynamic characteristics of the current signal; Generate a comprehensive feature vector according to the frequency domain feature vector, the time domain feature vector and the dynamic feature; The abnormal signal feature is determined according to the Mahalanobis distance formula and the comprehensive feature vector.

5. The adaptive overcurrent protection method according to claim 1, characterized in that: The machine learning algorithm includes a support vector machine, and the step of classifying the abnormal signal features by the machine learning algorithm to determine whether an overcurrent event occurs in the target device includes: Training a support vector machine classifier according to historical normal data and fault data of the target device; The abnormal signal feature is input into the support vector machine classifier to determine whether an overcurrent event occurs in the target device.

6. The adaptive overcurrent protection method according to claim 5, characterized in that: The method further comprises: Dynamically updating the overcurrent protection threshold according to the real-time load characteristics and environmental parameters of the target device; According to the judgment result of the support vector machine classifier on the overcurrent event, the overcurrent protection threshold is modified; It is determined whether an overcurrent event occurs in the target device according to the corrected real-time overcurrent protection threshold.

7. The adaptive overcurrent protection method according to claim 5, characterized in that: After the step of performing overcurrent protection on the target device according to a preset protection logic if an overcurrent event occurs in the target device, the method further includes: When the overcurrent protection is completed, the event parameters corresponding to the overcurrent protection event are obtained; Updating the training set of the support vector machine classifier according to the event parameters; The support vector machine classifier is optimized according to the training set.

8. An adaptive overcurrent protection device, characterized in that: The adaptive overcurrent protection device comprises: An acquisition module, used to acquire target multi-dimensional data through a multi-sensor acquisition system; A determination module, used to determine fusion data corresponding to the target multidimensional data according to a sensor fusion algorithm; An extraction module, used to extract abnormal signal features present in the fused data; A judgment module, used to classify the abnormal signal features through a machine learning algorithm to determine whether an overcurrent event occurs in the target device; The protection module is used to perform overcurrent protection on the target device according to a preset protection logic if an overcurrent event occurs in the target device.

9. An electronic device, characterized in that: The device comprises a memory and a processor, wherein: The processor is used to execute the computer program stored in the memory; When the processor executes the computer program, the steps in the adaptive overcurrent protection method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps in the adaptive overcurrent protection method according to any one of claims 1 to 7 are implemented.

Images