Switching power supply high-side current detection method

Abstract

The invention relates to the field of switching power supply high-side current detection, in particular to a switching power supply high-side current detection method, which comprises the following steps of: acquiring current signals to form a sequence, acquiring an intrinsic mode by using empirical mode decomposition, analyzing each mode to construct a difference amplitude-frequency index, and calculating a noise index in combination with a similarity index and heterogeneity. According to the method, the de-noising window is intelligently adjusted, and the local outlier factor is utilized to detect the anomaly, so that accurate current anomaly identification is finally realized, the detection precision and the system intelligence level are remarkably improved, the resource utilization is optimized, and a powerful guarantee is provided for stable operation of a power system.

Description

Technical Field

[0001] The present application relates to the field of high-side current detection of switching power supplies, and in particular to a method for detecting high-side current of a switching power supply. Background Art

[0002] A switching power supply is an electronic power conversion device that converts input AC or DC power into the required stable DC output through high-frequency switching technology. Compared with traditional linear power supplies, switching power supplies have higher efficiency, smaller size and lighter weight, and are widely used in various electronic devices, such as computers, communication equipment, industrial control, medical equipment, etc. Its core principle is to use the switch tube to conduct and cut off at high frequency, transfer energy from the input end to the output end through a transformer or inductor, and then obtain a stable DC output through rectification and filtering.

[0003] Existing methods will generate some interference when detecting high-side current, such as the internal resistance, parasitic capacitance and inductance of the switch component. If a non-contact sensor is used, it will also be interfered by changes in the external magnetic field. There are large current changes and high-frequency noise in the switching power supply system. These may cause electromagnetic interference to the measurement circuit, and the interference signal may be superimposed on the current signal to be measured. Summary of the invention

[0004] In order to achieve high-precision and adaptive current anomaly detection, the present application provides a switching power supply high-side current detection method.

[0005] In a first aspect, the present application provides a method for detecting a high-side current of a switching power supply, comprising the following steps: S1: Collect the current signal of the high-side of the switching power supply to form an original high-side current sequence; S2: Perform empirical mode decomposition on the original high-side current sequence to obtain each eigenmode; S3: According to the maximum and minimum values ​​of each eigenmode, the upper and lower envelopes are constructed, and each eigenmode is divided according to the time sequence to obtain each subsequence. By analyzing the number of zero points and the upper and lower envelopes of each subsequence, the extrinsic difference amplitude-frequency index of each eigenmode subsequence is constructed; S4: construct the similarity index between the moment intrinsics of the subsequences and the moment intrinsic tuple heterogeneity, and calculate the high-side current potential noise index of each subsequence based on this; S5: according to the potential noise index of the high-side current of each subsequence, the window size of the moving average denoising algorithm for denoising each subsequence is obtained, and the denoised high-side current sequence is detected in combination with the local outlier factor detection algorithm to complete the detection of the high-side current of the switching power supply; S6: Input the denoised high-side current sequence into a local outlier factor detection algorithm, output the LOF value of each current data in the denoised high-side current sequence, and identify the current data with a LOF value higher than the abnormal threshold as abnormal current, otherwise it is normal current data.

[0006] Optionally, collecting the current signal of the high-side side of the switching power supply to form an original high-side current sequence includes the following steps: S11: Use a high-precision current sensor array to monitor the current changes on the high side of the switching power supply in real time; S12: Convert the analog current signal into a digital signal through a high-speed data acquisition card, and use edge computing technology to perform preliminary data processing; S13: Use deep learning models to pre-process digital signals, including intelligent filtering and adaptive denoising; S14: The preprocessed signal is stored as time series data to form an original high-side current sequence, and data compression technology is used to reduce storage space requirements.

[0007] Optionally, performing empirical mode decomposition on the original high-side current sequence to obtain each eigenmode comprises the following steps: S21: Apply the empirical mode decomposition algorithm and combine it with machine learning technology to automatically adjust the decomposition parameters; S22: Perform in-depth feature analysis on each IMF to ensure that it meets the definition of IMF and identify potential nonlinear patterns; S23: Through an iterative process, IMFs of different frequencies are gradually extracted from the original sequence, and the decomposition process is optimized using a genetic algorithm; S24: Record the characteristic parameters of each IMF, including frequency, amplitude and energy, and use neural network to predict the future trend of IMF.

[0008] Optionally, the method of constructing upper and lower envelopes according to the maximum and minimum values ​​of each eigenmode, dividing each eigenmode according to the time sequence, obtaining each subsequence, and constructing the extrinsic difference amplitude-frequency index of each eigenmode subsequence by analyzing the number of zero points and the upper and lower envelopes of each subsequence includes the following steps: S31: Use wavelet transform to enhance the envelope construction and improve the ability to capture signal details; S32: Each IMF is divided into multiple subsequences according to the time series, and the fuzzy clustering algorithm is used to optimize the division of subsequences; S33: Count the number of zero points in each subsequence and combine with wavelet coefficient analysis to enhance the recognition of signal features; S34: Based on the number of zero points and envelope characteristics, combined with the deep learning model, the extrinsic difference amplitude-frequency index of each subsequence is constructed to improve the intelligent level of signal analysis.

[0009] Optionally, constructing the similarity index between moment intrinsics and moment intrinsic tuple heterogeneity of the subsequences, and calculating the high-side current potential noise index of each subsequence based on the similarity index and moment intrinsic tuple heterogeneity, comprises the following steps: S41: Define a deep learning-based moment-to-moment eigensimilarity index to quantify the complex similarity between different IMF subsequences; S42: Calculate the moment eigentuple heterogeneity based on complex network theory to evaluate the dynamic diversity and complexity of subsequences; S43: Combining similarity index and heterogeneity, a machine learning model is used to evaluate the noise level of subsequences and improve the accuracy of the evaluation; S44: According to the evaluation result, combined with the adaptive threshold technology, the high-side current potential noise index of each subsequence is calculated for subsequent intelligent denoising processing.

[0010] Optionally, the step of obtaining the window size of each subsequence when the moving average denoising algorithm denoises each subsequence according to the high-side current potential noise index of each subsequence, and detecting the denoised high-side current sequence in combination with the local outlier factor detection algorithm to complete the detection of the high-side current of the switching power supply includes the following steps: S51: According to the noise index, combined with the reinforcement learning algorithm, the moving average window size is dynamically adjusted to achieve intelligent denoising; S52: Apply the local outlier factor (LOF) detection algorithm based on deep learning to intelligently identify outliers in the denoised sequence; S53: Set the LOF threshold based on statistical learning, combine domain knowledge and real-time data, and dynamically adjust the anomaly detection criteria; S54: Performing in-depth post-processing on the detection results, including intelligent smoothing processing.

[0011] Optionally, the step of inputting the denoised high-side current sequence into a local outlier factor detection algorithm, outputting the LOF value of each current data in the denoised high-side current sequence, and identifying the current data with a LOF value higher than the abnormal threshold as abnormal current, and vice versa as normal current data, comprises the following steps: S61: Using the denoised current sequence as input, executing the LOF algorithm based on deep learning to improve the intelligence level of anomaly detection; S62: Calculate the LOF value of each current data point, and combine it with the adaptive threshold technology to reflect its degree of outlier relative to the surrounding data; S63: Set abnormal thresholds based on machine learning to achieve dynamic adjustment; S64: Based on the comparison between the LOF value and the threshold value, combined with the intelligent decision-making system, the current data is marked as abnormal or normal, and the current detection process is completed.

[0012] In a second aspect, the present application provides a switching power supply high-side current detection system, which is used to perform the above-mentioned switching power supply high-side current detection method, including: The signal acquisition module is used to monitor the current changes on the high side of the switching power supply in real time, collect current signals using a high-precision current sensor array, and convert analog signals into digital signals through a data acquisition card; The signal processing and analysis module is used to pre-process the collected digital signals, including intelligent filtering, adaptive denoising and feature extraction. It also uses deep learning models and empirical mode decomposition techniques to decompose the original current sequence into multiple eigenmodes and further analyze the frequency, amplitude and energy characteristics of the signal. The noise assessment and anomaly detection module calculates the high-side current potential noise index of each subsequence based on the results of the signal processing and analysis module, dynamically adjusts the moving average window size for intelligent denoising, and then combines the local outlier factor (LOF) detection algorithm to intelligently identify outliers in the denoised sequence, set the LOF threshold based on statistical learning, and realize dynamic anomaly detection; The intelligent decision and output module, based on the output of the noise assessment and anomaly detection module, combines with the intelligent decision system to classify the current data and mark it as abnormal or normal.

[0013] In a third aspect, the present application provides an electronic device, which adopts the following technical solution: An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of a switching power supply high-side current detection method as described in any one of the above items when executing the computer program.

[0014] In a fourth aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of a switching power supply high-side current detection method as described in any one of the above items are implemented.

[0015] In summary, this application includes the following beneficial technical effects: 1. Through advanced technologies such as deep learning models and empirical mode decomposition, the current signal is intelligently preprocessed and feature analyzed, which effectively improves the accuracy of noise assessment and anomaly detection. In addition, combined with the dynamically adjusted moving average window size and the local outlier factor (LOF) detection algorithm based on deep learning, it can more accurately identify abnormal currents, reduce false alarms and missed alarms, and ensure the stable operation of the power system. 2. Using machine learning and adaptive threshold technology, the system can automatically adjust parameters such as noise assessment threshold, moving average window size and anomaly detection criteria according to real-time data and environmental changes, achieving intelligent decision-making and adaptive control; 3. Through intelligent signal preprocessing and data compression technology, the demand for data storage and transmission is reduced, and the system operating cost is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of a high-side current detection method of a switching power supply in an embodiment of the present application.

[0017] Figure 2 It is a flowchart of the sub-steps of step S1 in the embodiment of the present application.

[0018] Figure 3 It is a flowchart of the sub-steps of step S2 in the embodiment of the present application.

[0019] Figure 4 It is a flowchart of the sub-steps of step S3 in the embodiment of the present application.

[0020] Figure 5 It is a flowchart of the sub-steps of step S4 in the embodiment of the present application.

[0021] Figure 6 It is a flowchart of the sub-steps of step S5 in the embodiment of the present application.

[0022] Figure 7 It is a flowchart of the sub-steps of step S6 in the embodiment of the present application. DETAILED DESCRIPTION

[0023] The following is combined with Figure 1-Figure 6 This application is described in further detail.

[0024] Reference Figure 1 The present application discloses a method for detecting a high-side current of a switching power supply, comprising the following steps: S1: Collect the current signal of the high-side of the switching power supply to form an original high-side current sequence.

[0025] Reference Figure 2 , step S1 correspondingly includes the following sub-steps.

[0026] S11: Use a high-precision current sensor array to monitor the current changes on the high side of the switching power supply in real time.

[0027] S12: The analog current signal is converted into a digital signal through a high-speed data acquisition card, and preliminary data processing is performed using edge computing technology.

[0028] S13: Use deep learning models to pre-process digital signals, including intelligent filtering and adaptive denoising.

[0029] S14: The preprocessed signal is stored as time series data to form an original high-side current sequence, and data compression technology is used to reduce storage space requirements.

[0030] For steps S11-S14, the present application embodiment provides the following examples.

[0031] Suppose you are monitoring an industrial switching power supply, whose high-side current varies widely, from microamperes to several amperes. To ensure the accuracy and real-time nature of the data, a set of high-precision current sensor arrays are deployed, with each sensor having an accuracy of less than 0.1% and a sampling frequency of 10kHz. These sensors are closely arranged, covering the high-side of the power supply, and can capture small fluctuations and rapid changes in current.

[0032] Furthermore, the analog current signal captured by the sensor array is sent to a high-speed data acquisition card with a 16-bit resolution and a 200kHz sampling rate, which can convert the analog signal into a high-precision digital signal. Subsequently, the edge computing device immediately intervenes to perform preliminary processing on the digital signal, including signal integrity check, timestamp addition, and basic filtering to remove obvious noise interference.

[0033] At the same time, in the preprocessing stage, the deep learning model is used for intelligent filtering and adaptive denoising of the signal. By analyzing historical data and learning the normal mode and noise characteristics of the current signal, the model can intelligently identify and remove random noise and periodic interference in the signal while retaining the details and dynamic changes of the signal. This process improves the clarity of the signal and provides a high-quality data foundation for subsequent analysis and detection.

[0034] Finally, the preprocessed signal is converted into time series data to form the original high-side current sequence. For efficient storage and transmission, data compression technology is applied, such as predictive coding or wavelet transform, to compress the signal, significantly reducing the storage space requirements while maintaining the integrity of the signal. The compressed data is stored in the local database and transmitted to the central server via the network for further analysis and decision-making.

[0035] Replay Figure 1, the following steps are also included after step S1.

[0036] S2: Perform empirical mode decomposition on the original high-side current sequence to obtain each eigenmode.

[0037] Reference Figure 3 , step S2 correspondingly includes the following sub-steps.

[0038] S21: Apply the empirical mode decomposition algorithm and combine it with machine learning technology to automatically adjust the decomposition parameters.

[0039] S22: Perform in-depth feature analysis on each IMF to ensure that it meets the definition of IMF and identify potential nonlinear patterns.

[0040] S23: Through an iterative process, IMFs of different frequencies are gradually extracted from the original sequence, and the decomposition process is optimized using a genetic algorithm.

[0041] S24: Record the characteristic parameters of each IMF, including frequency, amplitude and energy, and use neural network to predict the future trend of IMF.

[0042] Through empirical mode decomposition (EMD) and machine learning technology, the original high-side current sequence is refined and analyzed to extract the intrinsic mode functions (IMF) of different frequency characteristics. Correspondingly, this process is described in detail through a specific example. For steps S21-S24, the embodiment of the present application is described as follows.

[0043] First, the EMD algorithm is applied to decompose the original high-side current sequence. The EMD algorithm can decompose complex signals into a series of IMFs, each of which represents fluctuations of different frequencies in the signal. In order to improve the accuracy and efficiency of the decomposition, machine learning technology is introduced to automatically adjust the parameters of the EMD, such as the smoothness of the Spline interpolation and the stop condition of the decomposition, by training the model. The model learns the characteristics of the signal based on historical data and can intelligently select the optimal parameters to ensure the reliability of the decomposition results.

[0044] Each decomposed IMF is subjected to deep feature analysis to verify whether it meets the definition of IMF, that is, at any point in time, the average value of the upper and lower envelopes of the IMF is zero. In addition, nonlinear analysis methods such as phase space reconstruction and Lyapunov exponent calculation are used to identify possible nonlinear patterns in the IMF, which helps to understand the complex dynamic characteristics of the current signal.

[0045] Correspondingly, the EMD algorithm is an iterative process that gradually separates fluctuations of different frequencies from the original sequence by continuously extracting IMFs. In order to optimize this process, a genetic algorithm is introduced to automatically adjust the parameters of the EMD by simulating the mechanism of natural selection and genetic variation to improve the efficiency and accuracy of decomposition. The genetic algorithm can quickly find the optimal parameter combination to ensure that the extraction of each IMF is optimal.

[0046] For each extracted IMF, its key characteristic parameters such as frequency, amplitude and energy are recorded, which can reflect the physical meaning and signal characteristics of IMF. In addition, a neural network model is used to predict the trend of future IMF based on historical IMF data, which helps to identify potential changes in current signals in advance and provide prediction information for subsequent anomaly detection and system optimization.

[0047] Replay Figure 1 , after step S2, the following steps are also included.

[0048] S3: According to the maximum and minimum values ​​of each eigenmode, the upper and lower envelopes are constructed, and each eigenmode is divided according to the time sequence to obtain each subsequence. By analyzing the number of zero points and the upper and lower envelopes of each subsequence, the extrinsic difference amplitude-frequency index of each eigenmode subsequence is constructed.

[0049] Reference Figure 4 , step S3 correspondingly includes the following sub-steps.

[0050] S31: Use wavelet transform to enhance the construction of envelope and improve the ability to capture signal details.

[0051] S32: Each IMF is divided into multiple subsequences according to the time series, and the fuzzy clustering algorithm is used to optimize the division of the subsequences.

[0052] S33: Count the number of zero points in each subsequence and combine it with wavelet coefficient analysis to enhance the recognition of signal features.

[0053] S34: Based on the number of zero points and envelope characteristics, combined with the deep learning model, the extrinsic difference amplitude-frequency index of each subsequence is constructed to improve the intelligent level of signal analysis.

[0054] For steps S31-S34, the following examples are given correspondingly.

[0055] First, wavelet transform is applied to each IMF to enhance the construction of envelope. Wavelet transform can capture the local characteristics of the signal and has good adaptability to both high-frequency and low-frequency signal details. By selecting appropriate wavelet basis functions, a more accurate envelope can be constructed, which helps to improve the ability to capture signal details, especially for those signal features that are short-lived in time but have an important impact on the system state.

[0056] Next, each IMF is divided into multiple subsequences according to the time series, and each subsequence represents the signal characteristics within a period of time. In order to optimize the division of subsequences, a fuzzy clustering algorithm is introduced. The algorithm can divide subsequences into different clusters according to the similarity of the signal, and the subsequences in each cluster have similar characteristics. Through fuzzy clustering, patterns and anomalies in the signal can be more accurately identified, providing a clearer perspective for subsequent analysis.

[0057] For each subsequence, the number of zero points is counted, that is, the number of zero crossings from positive to negative or from negative to positive. The number of zero points is an important indicator of signal activity and can reflect the dynamic characteristics of the signal. Combined with wavelet coefficient analysis, it is possible to have a deeper understanding of the local characteristics of the signal, such as the intensity, frequency and duration of the signal. This process enhances the recognition of signal characteristics and helps to analyze the nature of the signal more accurately.

[0058] Finally, based on the number of zero points and envelope characteristics, the deep learning model is used to construct the intrinsic mode function amplitude-frequency index (IMF-AFI) of each subsequence. IMF-AFI is a comprehensive indicator that can reflect the frequency characteristics and amplitude changes of the subsequence, while removing the inherent characteristics of the IMF itself, making the comparison between different subsequences more fair. Through the deep learning model, these features can be automatically learned and identified, and a more intelligent signal analysis model can be constructed to improve the intelligence level of signal analysis.

[0059] Replay Figure 1 , step S3 and the corresponding steps include the following steps.

[0060] S4: construct the similarity index between the moment intrinsics of the subsequences and the moment intrinsic tuple heterogeneity, and calculate the high-side current potential noise index of each subsequence based on this.

[0061] Reference Figure 5 , step S4 correspondingly includes the following sub-steps.

[0062] S41: Define a deep learning-based moment-to-moment eigenvalue similarity index to quantify the complex similarity between different IMF subsequences.

[0063] S42: Calculate the moment eigentuple heterogeneity based on complex network theory to evaluate the dynamic diversity and complexity of subsequences.

[0064] S43: Combining similarity index and heterogeneity, a machine learning model is used to evaluate the noise level of subsequences and improve the accuracy of the evaluation.

[0065] S44: According to the evaluation result, combined with the adaptive threshold technology, the high-side current potential noise index of each subsequence is calculated for subsequent intelligent denoising processing.

[0066] Correspondingly, in the S4 step of in-depth understanding of the high-side current detection method of the switching power supply, the IMF subsequences are evaluated for complex similarity and dynamic diversity through deep learning, complex network theory and machine learning technology to improve the accuracy of the noise level evaluation. The following is a detailed explanation of this process through a specific example: First, a deep learning-based intrinsic mode function temporal similarity index (IMF-TSI) is defined to quantify the complex similarity between different IMF subsequences. IMF-TSI uses a deep learning model to learn the feature representation of subsequences, including frequency, amplitude, and temporal pattern, and then calculates the similarity between these representations. This index not only considers the static characteristics of subsequences, but also their dynamic changes in time, which can more comprehensively reflect the complex similarity between subsequences.

[0067] Next, we use complex network theory to calculate the Intrinsic Mode Function Temporal Heterogeneity (IMF-TH) based on IMF subsequences. IMF-TH evaluates the dynamic diversity and complexity of subsequences by constructing a network between subsequences and analyzing the network's topological structure, such as degree distribution, clustering coefficient, and shortest path length. This indicator can reveal the interactions and dependencies between subsequences, and helps to understand the complex dynamic characteristics of the signal.

[0068] Then, IMF-TSI and IMF-TH are combined to evaluate the noise level of the subsequence using a machine learning model. By learning from historical data, the model can identify the impact pattern of noise on the similarity index and heterogeneity, thereby accurately evaluating the noise level of the subsequence. This process improves the accuracy of the evaluation and provides key information for subsequent intelligent denoising processing.

[0069] Finally, based on the evaluation results, combined with the adaptive threshold technology, the High-side Current Potential Noise Index (HCPN I) of each subsequence is calculated. HCPNI is a comprehensive indicator that can reflect the noise level of the subsequence while taking into account the dynamic characteristics and complexity of the signal. Through the adaptive threshold technology, the threshold can be dynamically adjusted according to the characteristics of the signal, and the noise can be more accurately identified and removed, thereby improving the clarity and reliability of the signal.

[0070] Replay Figure 1 , after step S4, the following steps are also included.

[0071] S5: According to the potential noise index of the high-side current of each subsequence, the window size of the moving average denoising algorithm for denoising each subsequence is obtained, and the denoised high-side current sequence is detected in combination with the local outlier factor detection algorithm to complete the detection of the high-side current of the switching power supply.

[0072] Reference Figure 6 , step S5 correspondingly includes the following sub-steps: S51: According to the noise index, combined with the reinforcement learning algorithm, the moving average window size is dynamically adjusted to achieve intelligent denoising.

[0073] S52: Apply the local outlier factor (LOF) detection algorithm based on deep learning to intelligently identify outliers in the denoised sequence.

[0074] S53: Set the LOF threshold based on statistical learning, combine domain knowledge and real-time data, and dynamically adjust the anomaly detection standard.

[0075] S54: Performing in-depth post-processing on the detection results, including intelligent smoothing processing.

[0076] Correspondingly, steps S51-S54 are described by way of example below.

[0077] First, based on the previously calculated high-side current potential noise index (HCPNI), combined with the reinforcement learning algorithm, the moving average window size is dynamically adjusted to achieve intelligent denoising. The reinforcement learning algorithm learns the optimal strategy to maximize long-term rewards by simulating the interaction between the environment and the agent. In this case, the agent (i.e., the algorithm) tries different moving average window sizes, observes the denoising effect (i.e., reward), and dynamically adjusts the window size based on the feedback of the noise index to achieve the best denoising effect. This process improves the intelligence level of denoising, can adapt to the dynamic changes of the signal, and improves the clarity of the signal.

[0078] Next, a local outlier factor (LOF) detection algorithm based on deep learning is applied to detect outliers in the denoised sequence. The LOF algorithm calculates the local density of each data point and identifies those data points that are significantly lower in local density than their neighbors, i.e., outliers. In this case, a deep learning model is used to learn the normal pattern of the signal, and then the LOF value of each data point is calculated to intelligently identify those data points that deviate from the normal pattern. This process improves the accuracy of outlier detection and can effectively identify anomalies in the signal.

[0079] Then, we set the LOF threshold based on statistical learning, combined with domain knowledge and real-time data, to dynamically adjust the anomaly detection criteria. The LOF threshold is a key parameter for distinguishing normal values ​​from abnormal values. Through statistical learning, we can dynamically adjust this threshold based on historical data and real-time signal characteristics to adapt to the dynamic changes of the signal. This process improves the flexibility and accuracy of anomaly detection and can more accurately identify anomalies in the signal.

[0080] Finally, the detection results are deeply post-processed, including intelligent smoothing. Intelligent smoothing can remove noise and fluctuations in the detection results and improve the smoothness and reliability of the signal. Using the deep learning model, the smoothing pattern of the signal is learned, and then the detection results are intelligently smoothed to remove false positives and false negatives in the outlier detection process. This process improves the intelligence level of signal processing and can more accurately reflect the true characteristics of the signal.

[0081] Replay Figure 1 , after step S5, the following steps are also included.

[0082] S6: Input the denoised high-side current sequence into a local outlier factor detection algorithm, output the LOF value of each current data in the denoised high-side current sequence, and identify the current data with a LOF value higher than the abnormal threshold as abnormal current, otherwise it is normal current data.

[0083] Reference Figure 7 , step S6 correspondingly includes the following sub-steps.

[0084] S61: The denoised current sequence is used as input to execute the LOF algorithm based on deep learning to improve the intelligence level of anomaly detection.

[0085] S62: Calculate the LOF value of each current data point, and combine it with the adaptive threshold technology to reflect its degree of outlier relative to the surrounding data.

[0086] S63: Set abnormal thresholds based on machine learning to achieve dynamic adjustment.

[0087] S64: Based on the comparison between the LOF value and the threshold value, combined with the intelligent decision-making system, the current data is marked as abnormal or normal, and the current detection process is completed.

[0088] For steps S61-S64, the following examples are given correspondingly.

[0089] First, the denoised current sequence is used as input to execute the local outlier factor (LOF) algorithm based on deep learning. The deep learning model can learn the normal pattern of the current sequence. By extracting the feature representation of the current data, the LOF algorithm can calculate the local density of each data point and identify those data points that are significantly lower in local density than their neighbors, i.e., outliers. This process improves the intelligence level of anomaly detection and can more accurately identify anomalies in the current sequence.

[0090] Next, the LOF value of each current data point is calculated and combined with the adaptive threshold technology to reflect its degree of outlier relative to the surrounding data. The LOF value is a key indicator to measure the degree of outlier of a data point. Through the adaptive threshold technology, the LOF threshold can be dynamically adjusted according to the characteristics of the current sequence to adapt to the dynamic changes of the signal. This process improves the flexibility and accuracy of anomaly detection and can more accurately identify anomalies in current data.

[0091] Then, set the abnormal threshold based on machine learning to achieve dynamic adjustment. By learning historical data, the machine learning model can identify the normal range and abnormal pattern of the current sequence, thereby setting a dynamic abnormal threshold. This threshold can be adjusted according to the real-time characteristics of the current sequence, improving the adaptability and accuracy of abnormal detection.

[0092] Finally, based on the comparison between the LOF value and the abnormal threshold, combined with the intelligent decision-making system, the current data is marked as abnormal or normal. The intelligent decision-making system can comprehensively consider the LOF value, threshold and domain knowledge to intelligently classify the current data. This process improves the intelligence level of current detection, can more accurately identify and mark abnormal situations of current data, and complete the current detection process.

[0093] This embodiment further provides a switching power supply high-side current detection system, which is used to execute the above-mentioned switching power supply high-side current detection method, including: The signal acquisition module is used to monitor the current changes on the high side of the switching power supply in real time. It uses a high-precision current sensor array to collect current signals and converts analog signals into digital signals through a data acquisition card.

[0094] The signal processing and analysis module is used to pre-process the collected digital signals, including intelligent filtering, adaptive denoising and feature extraction, and use deep learning models and empirical mode decomposition technologies to decompose the original current sequence into multiple intrinsic modes, and further analyze the frequency, amplitude and energy characteristics of the signal.

[0095] The noise assessment and anomaly detection module calculates the high-side current potential noise index of each subsequence based on the results of the signal processing and analysis module, dynamically adjusts the moving average window size for intelligent denoising, and then combines the local outlier factor (LOF) detection algorithm to intelligently identify outliers in the denoised sequence, set the LOF threshold based on statistical learning, and realize dynamic anomaly detection.

[0096] The intelligent decision and output module, based on the output of the noise assessment and anomaly detection module, combines with the intelligent decision system to classify the current data and mark it as abnormal or normal.

[0097] This embodiment also provides an electronic device. The electronic device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the electronic device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for detecting high-side current of a switching power supply is implemented: S1: Collect the current signal of the high-side of the switching power supply to form an original high-side current sequence.

[0098] S2: Perform empirical mode decomposition on the original high-side current sequence to obtain each eigenmode.

[0099] S3: According to the maximum and minimum values ​​of each eigenmode, the upper and lower envelopes are constructed, and each eigenmode is divided according to the time sequence to obtain each subsequence. By analyzing the number of zero points and the upper and lower envelopes of each subsequence, the extrinsic difference amplitude-frequency index of each eigenmode subsequence is constructed.

[0100] S4: construct the similarity index between the moment intrinsics of the subsequences and the moment intrinsic tuple heterogeneity, and calculate the high-side current potential noise index of each subsequence based on this.

[0101] S5: According to the potential noise index of the high-side current of each subsequence, the window size of the moving average denoising algorithm for denoising each subsequence is obtained, and the denoised high-side current sequence is detected in combination with the local outlier factor detection algorithm to complete the detection of the high-side current of the switching power supply.

[0102] S6: Input the denoised high-side current sequence into a local outlier factor detection algorithm, output the LOF value of each current data in the denoised high-side current sequence, and identify the current data with a LOF value higher than the abnormal threshold as abnormal current, otherwise it is normal current data.

[0103] When the computer program is executed by a processor, any one of the switching power supply high-side current detection methods in the above method embodiments can be implemented.

[0104] This embodiment further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: S1: Collect the current signal of the high-side of the switching power supply to form an original high-side current sequence.

[0105] S2: Perform empirical mode decomposition on the original high-side current sequence to obtain each eigenmode.

[0106] S3: According to the maximum and minimum values ​​of each eigenmode, the upper and lower envelopes are constructed, and each eigenmode is divided according to the time sequence to obtain each subsequence. By analyzing the number of zero points and the upper and lower envelopes of each subsequence, the extrinsic difference amplitude-frequency index of each eigenmode subsequence is constructed.

[0107] S4: construct the similarity index between the moment intrinsics of the subsequences and the moment intrinsic tuple heterogeneity, and calculate the high-side current potential noise index of each subsequence based on this.

[0108] S5: According to the potential noise index of the high-side current of each subsequence, the window size of the moving average denoising algorithm for denoising each subsequence is obtained, and the denoised high-side current sequence is detected in combination with the local outlier factor detection algorithm to complete the detection of the high-side current of the switching power supply.

[0109] S6: Input the denoised high-side current sequence into a local outlier factor detection algorithm, output the LOF value of each current data in the denoised high-side current sequence, and identify the current data with a LOF value higher than the abnormal threshold as abnormal current, otherwise it is normal current data.

[0110] When the computer program is executed by a processor, any one of the switching power supply high-side current detection methods in the above method embodiments can be implemented.

[0111] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0112] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above.

[0113] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application is described in detail in the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some of the technical features therein by equivalents. 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, and should be included in the protection scope of the present application.

Claims

1. A method for detecting high-side current of a switching power supply, characterized in that: The following steps are involved: S1: Collect the current signal of the high-side of the switching power supply to form an original high-side current sequence; S2: Perform empirical mode decomposition on the original high-side current sequence to obtain each eigenmode; S3: According to the maximum and minimum values ​​of each eigenmode, the upper and lower envelopes are constructed, and each eigenmode is divided according to the time sequence to obtain each subsequence. By analyzing the number of zero points and the upper and lower envelopes of each subsequence, the extrinsic difference amplitude-frequency index of each eigenmode subsequence is constructed; S4: construct the similarity index between the moment intrinsics of the subsequences and the moment intrinsic tuple heterogeneity, and calculate the high-side current potential noise index of each subsequence based on this; S5: according to the potential noise index of the high-side current of each subsequence, the window size of the moving average denoising algorithm for denoising each subsequence is obtained, and the denoised high-side current sequence is detected in combination with the local outlier factor detection algorithm to complete the detection of the high-side current of the switching power supply; S6: Input the denoised high-side current sequence into a local outlier factor detection algorithm, output the LOF value of each current data in the denoised high-side current sequence, and identify the current data with a LOF value higher than the abnormal threshold as abnormal current, otherwise it is normal current data.

2. A switching power supply high-side current detection method according to claim 1, characterized in that: The collecting of the current signal of the high-side side of the switching power supply to form an original high-side current sequence includes the following steps: S11: Use a high-precision current sensor array to monitor the current changes on the high side of the switching power supply in real time; S12: Convert the analog current signal into a digital signal through a high-speed data acquisition card, and use edge computing technology to perform preliminary data processing; S13: Use deep learning models to pre-process digital signals, including intelligent filtering and adaptive denoising; S14: The preprocessed signal is stored as time series data to form an original high-side current sequence, and data compression technology is used to reduce storage space requirements.

3. A switching power supply high-side current detection method according to claim 1, characterized in that: The process of performing empirical mode decomposition on the original high-side current sequence to obtain each eigenmode comprises the following steps: S21: Apply the empirical mode decomposition algorithm and combine it with machine learning technology to automatically adjust the decomposition parameters; S22: Perform in-depth feature analysis on each IMF to ensure that it meets the definition of IMF and identify potential nonlinear patterns; S23: Through an iterative process, IMFs of different frequencies are gradually extracted from the original sequence, and the decomposition process is optimized using a genetic algorithm; S24: Record the characteristic parameters of each IMF, including frequency, amplitude and energy, and use neural network to predict the future trend of IMF.

4. A switching power supply high-side current detection method according to claim 1, characterized in that: The method of constructing upper and lower envelopes according to the maximum and minimum values ​​of each eigenmode, dividing each eigenmode according to the time sequence, obtaining each subsequence, and constructing the extrinsic difference amplitude-frequency index of each eigenmode subsequence by analyzing the number of zero points and the upper and lower envelopes of each subsequence includes the following steps: S31: Use wavelet transform to enhance the envelope construction and improve the ability to capture signal details; S32: Each IMF is divided into multiple subsequences according to the time series, and the fuzzy clustering algorithm is used to optimize the division of subsequences; S33: Count the number of zero points in each subsequence and combine with wavelet coefficient analysis to enhance the recognition of signal features; S34: Based on the number of zero points and envelope characteristics, combined with the deep learning model, the extrinsic difference amplitude-frequency index of each subsequence is constructed to improve the intelligent level of signal analysis.

5. A switching power supply high-side current detection method according to claim 1, characterized in that: The constructing of the moment intrinsic similarity index and moment intrinsic tuple heterogeneity of the subsequences, and calculating the high-side current potential noise index of each subsequence based on the similarity index and moment intrinsic tuple heterogeneity, comprises the following steps: S41: Define a deep learning-based moment-to-moment eigensimilarity index to quantify the complex similarity between different IMF subsequences; S42: Calculate the moment eigentuple heterogeneity based on complex network theory to evaluate the dynamic diversity and complexity of subsequences; S43: Combining similarity index and heterogeneity, a machine learning model is used to evaluate the noise level of subsequences and improve the accuracy of the evaluation; S44: According to the evaluation result, combined with the adaptive threshold technology, the high-side current potential noise index of each subsequence is calculated for subsequent intelligent denoising processing.

6. A switching power supply high-side current detection method according to claim 1, characterized in that: The method of obtaining the window size of each subsequence denoised by the moving average denoising algorithm according to the high-side current potential noise index of each subsequence, combining the local outlier factor detection algorithm, detecting the denoised high-side current sequence, and completing the detection of the high-side current of the switching power supply includes the following steps: S51: According to the noise index, combined with the reinforcement learning algorithm, the moving average window size is dynamically adjusted to achieve intelligent denoising; S52: Apply the local outlier factor (LOF) detection algorithm based on deep learning to intelligently identify outliers in the denoised sequence; S53: Set the LOF threshold based on statistical learning, combine domain knowledge and real-time data, and dynamically adjust the anomaly detection criteria; S54: Performing in-depth post-processing on the detection results, including intelligent smoothing processing.

7. A switching power supply high-side current detection method according to claim 1, characterized in that: The method inputs the denoised high-side current sequence into a local outlier factor detection algorithm, outputs the LOF value of each current data in the denoised high-side current sequence, and identifies the current data with a LOF value higher than the abnormal threshold as abnormal current, and vice versa as normal current data, including the following steps: S61: Using the denoised current sequence as input, executing the LOF algorithm based on deep learning to improve the intelligence level of anomaly detection; S62: Calculate the LOF value of each current data point, and combine it with the adaptive threshold technology to reflect its degree of outlier relative to the surrounding data; S63: Set abnormal thresholds based on machine learning to achieve dynamic adjustment; S64: Based on the comparison between the LOF value and the threshold value, combined with the intelligent decision-making system, the current data is marked as abnormal or normal, and the current detection process is completed.

8. A switching power supply high-side current detection system, used to execute a switching power supply high-side current detection method as claimed in claims 1-7, characterized in that: include: The signal acquisition module is used to monitor the current changes on the high side of the switching power supply in real time, collect current signals using a high-precision current sensor array, and convert analog signals into digital signals through a data acquisition card; The signal processing and analysis module is used to pre-process the collected digital signals, including intelligent filtering, adaptive denoising and feature extraction. It also uses deep learning models and empirical mode decomposition techniques to decompose the original current sequence into multiple eigenmodes and further analyze the frequency, amplitude and energy characteristics of the signal. The noise assessment and anomaly detection module calculates the high-side current potential noise index of each subsequence based on the results of the signal processing and analysis module, dynamically adjusts the moving average window size for intelligent denoising, and then combines the local outlier factor (LOF) detection algorithm to intelligently identify outliers in the denoised sequence, set the LOF threshold based on statistical learning, and realize dynamic anomaly detection; The intelligent decision and output module, based on the output of the noise assessment and anomaly detection module, combines with the intelligent decision system to classify the current data and mark it as abnormal or normal.

9. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of a switching power supply high-side current detection method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of a switching power supply high-side current detection method as described in any one of claims 1 to 7 are implemented.

Images