Method and system for intelligently switching and quickly detecting electrical parameters of load of electronic equipment

Abstract

The invention discloses a method and a system for intelligently switching and quickly detecting electrical parameters of a load of electronic equipment. In the method, through multi-channel broadband sampling and timestamp alignment data, by means of machine learning, a model is built for analysis, parameters are predicted, errors are calculated and compensated, and electrical parameter data are obtained. By implementing the technical scheme provided by the invention, multi-frequency-band voltage and current signals can be simultaneously acquired through multi-channel broadband sampling, the data acquisition range is greatly widened compared with a single channel, and frequency band signal omission during complex load switching is avoided. Timestamps are added to realize accurate time alignment of multiple paths of signals, and a foundation is built for subsequent analysis. A dynamic prediction model is constructed by combining historical load switching data and a machine learning algorithm, and real-time data prediction analysis is assisted by historical experience. The technologies work together, load switching signal information is comprehensively and accurately obtained, the problem of insufficient data coverage is effectively solved, the detection efficiency and accuracy are remarkably improved, and transient signal characteristics are completely extracted.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of load detection of electronic equipment, and in particular to a method and system for intelligent switching and rapid detection of electrical parameters of an electronic equipment load. BACKGROUND

[0002] With the widespread application of electronic equipment, the dynamic changes in load characteristics have an increasingly significant impact on the performance of the equipment and the stability of the system. In the fields of industrial automation, smart grid, etc., load switching of electronic equipment can cause dramatic fluctuations in voltage and current signals, thereby adversely affecting the normal operating state of the equipment. Therefore, rapidly and accurately detecting the electrical parameters of electronic equipment during load switching has become a key technical requirement for ensuring system stability and improving equipment performance.

[0003] In related technologies, a single-channel sampling scheme is usually used to collect voltage or current signals during load switching through a high-frequency sampling channel, and a low-pass filter is used to remove high-frequency noise to extract the main variation characteristics of the signals. To further analyze the signals, a frequency domain analysis method (such as Fourier transform) is often used to decompose the sampled data and extract the frequency spectrum characteristics of the voltage and current.

[0004] However, the single-channel sampling scheme has the problem of insufficient data coverage in complex load switching modes. Since single-channel sampling can only capture the signal characteristics of a single frequency band during load switching, when the load switching contains multiple frequency bands or high dynamic variation characteristics, the existing sampling method cannot fully cover the signal characteristics of these frequency bands. This limitation in data coverage range leads to the fact that in complex load switching scenarios, the existing technology may miss key transient signal characteristics, thereby affecting the accuracy and real-time performance of the detection. SUMMARY

[0005] The present application provides a method and system for intelligent switching and rapid detection of electrical parameters of an electronic equipment load, which solves the problem of incomplete transient signal characteristics caused by insufficient data coverage in complex load switching scenarios, and improves the efficiency and accuracy of detection.

[0006] In a first aspect, the application provides a method for intelligently switching and rapidly detecting electrical parameters of a load of an electronic device, comprising: acquiring, by a multi-channel broadband sampling method, voltage signals and current signals generated by the electronic device during a load switching process, to obtain multi-channel digital signal data; adding a time stamp to each channel of the multi-channel digital signal data, to obtain time-aligned multi-channel signal data; establishing a dynamic prediction model based on a load switching mode by a machine learning algorithm, according to historical load switching data and the time-aligned multi-channel signal data, wherein the historical load switching data includes typical modes of load switching and corresponding changes in electrical parameters; analyzing the time-aligned multi-channel signal data by the dynamic prediction model, to obtain transient change characteristics of the electrical parameters of the load of the electronic device; predicting an electrical parameter value at a future time according to the transient change characteristics; calculating a transient error between the electrical parameter value and the time-aligned multi-channel signal data; and compensating the time-aligned multi-channel signal data according to the transient error by the dynamic prediction model, to obtain electrical parameter data including a compensation result.

[0007] By using the above technical solution, multi-channel broadband sampling can simultaneously acquire voltage and current signals of multiple frequency bands, greatly widening the data acquisition range compared to single-channel sampling, and ensuring that signals of each frequency band are not missed during complex load switching. Adding a time stamp enables the multi-channel signal data to be accurately aligned in the time dimension, providing a reliable basis for subsequent analysis. The dynamic prediction model is established by combining historical load switching data and a machine learning algorithm, and real-time data is analyzed and predicted using historical experience. These technical means cooperate with each other to comprehensively and accurately acquire signal information during load switching, effectively solve the problem of insufficient data coverage, significantly improve detection efficiency and accuracy, and completely extract transient signal characteristics.

[0008] In some embodiments in combination with the first aspect, the dynamic prediction model based on the load switching mode is established by the machine learning algorithm according to the historical load switching data and the time-aligned multi-channel signal data, specifically comprising: pre-processing the acquired time-aligned multi-channel signal data and the historical load switching data to obtain pre-processed data, wherein the pre-processing includes denoising filtering, time synchronization correction, and normalization processing; extracting transient characteristics and feature labels from the pre-processed data, wherein the transient characteristics are transient characteristics of voltage and current during the load switching process extracted from the pre-processed time-aligned multi-channel signal data, including peak value, rise time, fall time, and oscillation frequency, and the feature labels are feature labels corresponding to the switching mode extracted from the pre-processed historical load switching data; training the dynamic prediction model by the machine learning algorithm using the transient characteristics and the feature labels, to obtain the dynamic prediction model based on the load switching mode.

[0009] By adopting the technical solution, the time-aligned multi-channel signal data and historical load switching data are preprocessed, the interference noise in the signal is removed through denoising filtering, time synchronization correction ensures the consistency of the data time, normalization processing makes the data comparable, and the data quality is improved. The transient characteristics and feature labels are extracted to provide key information for model training. A dynamic prediction model is trained based on the data using a machine learning algorithm, which enables the model to learn the internal relationship between the load switching mode and the change of the electrical parameter, enhances the adaptability of the model to the actual load switching scene, and makes the model more suitable for the actual situation, thereby laying a solid foundation for subsequent accurate analysis and prediction.

[0010] In combination with some embodiments of the first aspect, in some embodiments, the time-aligned multi-channel signal data is analyzed by the dynamic prediction model to obtain the transient change characteristics of the electrical parameter of the electronic device under load, specifically including: extracting the transient characteristics in the time-aligned multi-channel signal data by the dynamic prediction model, the transient characteristics including peak voltage, peak current, rise time, fall time and oscillation frequency; based on the transient characteristics and the features of different historical load switching modes, a feature template corresponding to different load switching modes is constructed; the real-time collected transient characteristics are matched and analyzed with the feature template to obtain a target load switching mode; under the target load switching mode, the transient characteristics are classified by the dynamic prediction model to obtain the transient change characteristics corresponding to the target load switching mode.

[0011] By adopting the technical solution, the dynamic prediction model extracts the transient characteristics of the multi-channel signal data, and constructs a feature template in combination with the features of the historical load switching mode to provide a reference standard for load switching mode identification. The real-time transient characteristics are matched with the template to quickly determine the target load switching mode, and then the transient characteristics are classified to obtain the transient change characteristics of the corresponding mode. This series of operations realizes accurate identification of the load switching mode and accurate analysis of the transient change characteristics, can timely grasp the change of the electrical parameter of the electronic device under different load switching modes, facilitates early discovery of potential problems, and takes corresponding measures to ensure stable operation of the device.

[0012] In combination with some embodiments of the first aspect, in some embodiments, according to the transient change characteristics, the electrical parameter value at a future time is predicted, specifically including: predicting the temporary electrical parameter value at a future time by the dynamic prediction model according to the transient characteristics and the known mode law in the historical load switching data, the mode law being a typical mode extracted from the historical load switching data, including transient behavior and electrical parameter change trend in the switching process; performing time series fitting on the temporary electrical parameter value to obtain the change trend of the electrical parameter in the load switching process; and predicting the electrical parameter value in a stable state after the load switching based on the change trend.

[0013] By employing this technical solution, the dynamic prediction model predicts temporary electrical parameter values ​​based on transient characteristics and historical patterns. It then uses time series fitting to determine the changing trends of the electrical parameters during load switching, and thus predicts the electrical parameter values ​​in the stable state. This process leverages historical experience and real-time data, taking into account the dynamic changes in load switching, and can predict the stable state of the electrical parameters after the load switching in advance. This provides a strong basis for system control and adjustment, allowing the system to be prepared in advance, effectively improving system stability and reliability, and reducing the adverse effects of load switching.

[0014] In combination with some embodiments of the first aspect, in some embodiments, the dynamic prediction model is used to perform compensation calculation on the time-aligned multi-channel signal data according to the transient error to obtain electrical parameter data containing the compensation result, specifically including: dynamically adjusting the weight parameters of the dynamic prediction model according to the distribution characteristics of the transient error to obtain the adjusted dynamic prediction model; using the adjusted dynamic prediction model to perform point-by-point calculation on the time-aligned multi-channel signal data to obtain the transient error of each data point; based on the transient error of each data point, using the real-time error compensation mechanism to correct each data point to obtain electrical parameter data containing the compensation result.

[0015] By adopting this technical solution, the weight parameters of the dynamic prediction model are dynamically adjusted based on the transient error distribution characteristics to optimize model performance. The adjusted model is used to calculate the transient error of each data point, and the data points are corrected through a real-time error compensation mechanism. This approach can promptly detect and correct data deviations, effectively reducing detection errors and improving data accuracy. The resulting electrical parameter data can more realistically and accurately reflect the actual load conditions, providing reliable data support for electronic equipment load electrical parameter testing and enhancing the credibility of the test results.

[0016] In combination with some embodiments of the first aspect, in some embodiments, after using the adjusted dynamic prediction model to perform point-by-point calculations on the time-aligned multi-channel signal data to obtain the transient error of each data point, the method also includes: fitting the transient error of each data point into a transient error sequence; based on the transient error sequence, constructing a transient error distribution model using normal distribution; judging whether the transient error exceeds a reasonable range through the transient error distribution model and the statistical process control method; if so, reallocating the weight parameters of the dynamic prediction model according to the time of the abnormal point and the corresponding transient characteristics to obtain a secondary adjusted dynamic prediction model, wherein the abnormal point is a point that exceeds the reasonable range; according to the secondary adjusted dynamic prediction model, performing point-by-point calculations on the time-aligned multi-channel signal data to obtain the transient error of each data point.

[0017] By adopting the technical solution, the transient error of the data points is fitted into a sequence to construct a transient error distribution model, and a statistical process control method is combined to determine whether the error is reasonable. When an abnormal point is found, the model weight parameters are redistributed according to the time and transient characteristics of the abnormal point to obtain a second-adjusted dynamic prediction model. This series of operations can timely find and handle abnormal errors, dynamically optimize the model, enhance the adaptability of the model to complex and variable load switching scenarios, ensure the accuracy and reliability of the detection results, and improve the robustness and stability of the entire detection system.

[0018] In combination with some embodiments of the first aspect, in some embodiments, the weight parameters of the dynamic prediction model are redistributed according to the time and corresponding transient characteristics of the abnormal point to obtain a second-adjusted dynamic prediction model, specifically including: calculating a transient error accumulation of the abnormal point in the time sequence according to the time of the abnormal point, the transient characteristics of the abnormal point and the distribution characteristics of the transient error, the accumulation reflecting the cumulative influence intensity of the abnormal point on the overall error; at the same time, the importance of the transient characteristics is obtained according to the influence degree of the transient characteristics on the load switching mode; the weight parameters of the dynamic prediction model are redistributed according to the accumulation and the importance to obtain a second-adjusted dynamic prediction model.

[0019] By adopting the technical solution, the accumulation is calculated according to the time, transient characteristics and distribution characteristics of the transient error of the abnormal point, the importance of the transient characteristics is determined according to the influence degree of the transient characteristics on the load switching mode, and the weight parameters of the model are redistributed based on the accumulation and the importance to obtain a second-adjusted dynamic prediction model. This method fully considers the influence of the abnormal point on the overall error and the role of the transient characteristics in the load switching mode, can more accurately adjust the model weight, makes the second-adjusted model more consistent with the actual situation, effectively reduces the influence of abnormal errors, significantly improves the prediction accuracy and stability of the model, guarantees the reliability of the detection results, and improves the overall performance of the detection system.

[0020] In the second aspect, the present application provides a detection system including one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code including computer instructions, and the one or more processors invoke the computer instructions to enable the detection system to perform the method described in the first aspect and any possible implementation manner of the first aspect.

[0021] In the third aspect, the present application provides an electronic device, when the electronic device runs on the detection system, enables the detection system to perform the method described in the first aspect and any possible implementation manner of the first aspect.

[0022] In a fourth aspect, the present application provides a computer readable storage medium storing computer instructions, which, when executed on a detection system, cause the detection system to perform the method as described in the first aspect and any possible implementation of the first aspect.

[0023] The one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. Since the technical means of multi-channel wideband parallel sampling, adding time stamps, and establishing a dynamic prediction model based on machine learning are adopted, the multi-channel wideband parallel sampling can simultaneously collect multiple multi-frequency signals, making up for the defect of insufficient data coverage of single-channel sampling; the addition of time stamps ensures the time alignment of multiple signals, providing protection for accurate analysis; the dynamic prediction model established in combination with historical data and machine learning algorithms can analyze real-time data using historical experience. The synergistic effect of the three effectively solves the problems of insufficient data coverage and inability to extract complete transient signal characteristics in the complex load switching scenario of the prior art, thereby achieving the technical effects of comprehensively acquiring signal information, improving detection efficiency and accuracy.

[0024] 2. Since the technical means of pre-processing data, extracting transient features and feature labels, and training a dynamic prediction model through a machine learning algorithm are adopted, the denoising filtering, time synchronization correction, and normalization processing in data preprocessing improve data quality; the extraction of transient features and feature labels provides core data for model training; the machine learning algorithm trains the model based on these data, enabling the model to learn the internal relationship between load switching patterns and changes in electrical parameters. These operations effectively solve the problems of low-quality model training data and difficulty in adapting the model to actual load switching conditions, thereby achieving the technical effects of enhancing model adaptability and accuracy and laying a reliable foundation for subsequent analysis and prediction.

[0025] 3. Since the technical means of adjusting model weight parameters according to transient error distribution characteristics, calculating transient errors point by point, and real-time compensation correction are adopted, adjusting model weight parameters according to transient error distribution characteristics can optimize model performance; calculating transient errors point by point can accurately locate data deviations; the real-time compensation correction mechanism timely corrects data points. The combination of the three effectively solves the problems of detection data errors and low accuracy, thereby achieving the technical effects of reducing detection errors, making electrical parameter data more realistically reflect the actual situation of the load, and improving the credibility of detection results. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 is a flowchart of a method for intelligently switching and rapidly detecting electrical parameters of a load of an electronic device using the embodiments of the present application; Figure 2is another flowchart of using the method for rapidly detecting electrical property parameters of an electronic device load by intelligent switching in the embodiments of the present application; Figure 3 is a hardware structure schematic diagram of an electronic device in the embodiments of the present application. DETAILED DESCRIPTION

[0027] The terms used in the following embodiments of the present application are only for the purpose of describing the specific embodiments and are not intended to be limiting of the present application. As used in the specification and the appended claims of the present application, the singular forms "a," "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "and / or," as used in the present application, refers to any or all possible combinations of one or more of the associated listed items.

[0028] Hereinafter, the terms "first" and "second" are only for the purpose of description and cannot be understood as implying or suggesting relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first" and "second" can explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, the meaning of "a plurality of" is two or more, unless otherwise specified.

[0029] For the convenience of understanding, the method for rapidly detecting electrical property parameters of an electronic device load by intelligent switching provided in the embodiments of the present application is described as follows: multi-channel wideband sampling is adopted to sample the voltage and current signals in parallel during the load switching of the electronic device to obtain multi-channel digital signals, and a time stamp is added to align the time. Then, a dynamic prediction model is established in combination with historical load switching data and machine learning algorithm, the time-aligned signals are analyzed by using the model, the transient change characteristics of the load electrical property parameters are obtained, and the future parameter values are predicted accordingly. By calculating the transient error between the predicted values and the actual signals, the signals are compensated and calculated by using the model, and finally the accurate electrical property parameter data is obtained.

[0030] Figure 1 is a flowchart of using the method for rapidly detecting electrical property parameters of an electronic device load by intelligent switching in the embodiments of the present application.

[0031] Please refer to Figure 1 , the specific description of the method for rapidly detecting electrical property parameters of an electronic device load by intelligent switching is as follows: 1. The voltage signals and current signals generated during the load switching of the electronic device are sampled in real time in parallel by using a multi-channel wideband sampling method to obtain multi-channel digital signal data; Multi-channel broadband sampling is mainly used for data acquisition, which is composed of multiple sampling channels with different frequency response ranges. In the field of industrial automation, electromagnetic interference generated by motor start and stop is reflected in different frequency bands of voltage and current signals. Multi-channel broadband sampling equipment can collect data through different channels. Low-frequency channels monitor the slow changes during motor start, and high-frequency channels capture the peak pulse during commutation. Through this method of sampling, the collected analog signals can be quickly converted into digital signals, providing digital format data for subsequent processing.

[0032] The collected data is mainly the voltage and current signals generated by electronic devices during load switching. The changes in these signals during load switching contain a wealth of information, such as the sharp rise in current and voltage drop during motor start, and the dynamic changes in current and voltage during the transition from standby to working state. Collecting these signals helps analyze the impact of load switching on devices and their operating status. Multi-channel broadband sampling is used for real-time parallel sampling. Multiple sampling channels with specific frequency response ranges are set up to cover various frequency bands of signals that may be generated during load switching. For example, in the industrial automation scenario, different frequency electromagnetic interference generated by motor operation is reflected in voltage and current signals, which can be collected by different channels for low and high frequency signals. Real-time sampling ensures that the data obtained during load switching is instantaneous, without significant time delay, and can truly reflect the dynamic changes in signals. Parallel sampling improves sampling efficiency and avoids signal information loss due to time differences. Each channel works independently to collect and convert analog signals into digital signals, ultimately obtaining multiple digital signal data.

[0033] 102. Adding a time stamp to each signal data of the multi-channel digital signal data to obtain time-aligned multi-channel signal data; After obtaining the multi-channel digital signal data, in order to ensure the accuracy and reliability of subsequent analysis, a time stamp needs to be added to each signal data to achieve time alignment. The addition of time stamp is based on an accurate clock source, which can be a high-precision crystal oscillator or an atomic clock. The clock source generates a stable clock signal to provide accurate time markers for each sampling time. When the multi-channel broadband sampling system collects signals, a time stamp is immediately added to each signal data according to the clock signal, ensuring that each data point has an accurate time record.

[0034] 103. Establishing a dynamic prediction model based on load switching mode through machine learning algorithm according to historical load switching data and the time-aligned multi-channel signal data, the historical load switching data including typical patterns of load switching and their corresponding electrical parameter changes; Firstly, remove the obvious error data caused by equipment failure or abnormal interference in historical data, and perform preliminary signal strength check on time-aligned multi-channel signal data to ensure data quality. Extract key features from the processed data. For voltage and current signals during load switching, pay attention to some basic features such as the maximum and minimum values of voltage and current in a certain period of time, and the average rate of change, etc. These features can roughly reflect the signal change trend during load switching. Mark different load switching modes from historical load switching data, such as common motor start mode, device standby to work switching mode, etc. Then, select a suitable machine learning algorithm, such as neural network, to train the dynamic prediction model using the extracted transient features and feature labels. Taking neural network as an example, it is composed of multiple neurons which are connected to form a complex network structure. The transient features are used as input data and the feature labels are used as output targets. By continuously adjusting the connection weights between neurons, the neural network learns the internal relationship between load switching mode and electrical parameter change.

[0035] Training and construction process of dynamic prediction model: the extracted transient features are used as input data and the corresponding feature labels are used as output targets, which are input into the selected machine learning algorithm for model training. Taking neural network as an example, during the training process, the data passes through the input layer, hidden layer and output layer in turn. The neurons in the hidden layer perform nonlinear transformation on the input data, and by adjusting the connection weights between neurons, the output of the model gradually approaches the true feature label. This process of adjusting weights usually uses the backpropagation algorithm to realize, which calculates the gradient of each neuron weight according to the error between the model output and the true label, and adjusts the weight according to the gradient, and iteratively trains until the error of the model converges to a small value, indicating that the model has learned the rules in the data. In the training process, in order to prevent the model from overfitting, some regularization techniques are usually used to constrain the weights of the model, to avoid the model overfitting to the noise and details in the training data, and to improve the generalization ability of the model, so that it can also have good performance on unknown data. At the same time, reasonably divide the training set, validation set and test set, use the training set to train the model, use the validation set to adjust the model parameters and monitor the training effect of the model to prevent overfitting, and finally use the test set to evaluate the final performance of the model.

[0036] After the training is completed, the constructed dynamic prediction model is evaluated using the test set data. Common evaluation indicators include accuracy, recall rate, mean square error (MSE), etc. Accuracy is used to measure the proportion of samples predicted correctly by the model, recall rate reflects the model's ability to identify positive samples, and mean square error evaluates the average error between the predicted value and the true value. For example, if the model has a high accuracy and small mean square error when predicting load switching mode, it indicates that the model can better identify different load switching modes, and the predicted electrical parameter value is close to the actual value. If the evaluation result is not ideal, analyze the problems of the model, such as whether there is overfitting or underfitting, whether the feature selection is reasonable, etc., and optimize and adjust the model according to these problems.

[0037] 104. Analyzing the time-aligned multi-channel signal data through the dynamic prediction model to obtain the transient variation characteristics of the electrical parameters of the electronic device load; First, input the time-aligned multi-channel signal data into the dynamic prediction model. When processing data, the model will extract and analyze the features of each time point in the multi-channel signal data. Using the long short-term memory network, information is memorized and updated through cell state and gating mechanism, which can capture long-term dependencies in data. In the analysis process, the model will mine the correlation between multi-channel signal data. For example, in an electronic device, there may be mutual influence between the power input voltage signal, the load current signal, and the internal temperature signal of the device. When the input voltage fluctuates, the load current may change accordingly, causing the internal temperature of the device to rise. The correlation between these signals is crucial to determining the transient variation characteristics of electrical parameters.

[0038] Through repeated iterative calculation and feature extraction of time-aligned multi-channel signal data, the dynamic prediction model can obtain the variation trend, rate, and amplitude of the electrical parameters of the electronic device load at different time points, etc. For example, in a smart charger device, by analyzing the input voltage signal, output current signal, and power signal of the charging device, etc. Multi-channel signal data, the model can determine how the transient variation characteristics of the current change from large current fast charging to small current trickle charging as the power gradually increases during the charging process, and the voltage fluctuation during this process, etc.

[0039] 105. Predicting the electrical parameter value at the future time according to the transient variation characteristics; Based on the transient change characteristics of the electrical parameters of the electronic device load, a dynamic prediction model is used to predict the electrical parameter values at future time. First, the model will identify the patterns and rules existing in the existing transient change characteristic data. For example, through analysis, it is found that the change of the electrical parameter presents periodic fluctuations, or there is a certain linear or nonlinear change trend. Taking periodic fluctuations as an example, suppose that when analyzing the power supply system of a certain server room, it is found that the load current presents obvious periodic change within 24 hours, with larger current during daytime and smaller current at night. The dynamic prediction model will extract the frequency, amplitude and other characteristic parameters of such periodic change through mathematical methods such as Fourier transform. Then, according to these characteristic parameters and the electrical parameter value at the current time, the electrical parameter value at the future time is predicted by combining the algorithm of the model itself. If the model identifies a linear change trend, for example, when analyzing the current of a DC motor with gradually increasing load, it is found that the current presents a linear growth trend with the increase of the load. In actual operation, as the load driven by the motor gradually increases, the motor needs to output more torque to maintain operation, which leads to the continuous rise of the current. After monitoring and analyzing the load and current data within a period of time, the model finds that there is a clear linear relationship between them, and based on this linear relationship, the current value at the future time is predicted.

[0040] 106、Calculate the transient error of the time-aligned multi-channel signal data of the electrical parameter value; After predicting the electrical parameter value at the future time, it needs to be compared with the actual time-aligned multi-channel signal data to calculate the transient error. First, the calculation method of the transient error is determined. Common calculation methods include mean square error (MSE), mean absolute error (MAE), etc. Taking mean square error as an example, suppose the predicted electrical parameter value is A, and the actual electrical parameter value is A i where i = 1, 2, …, n, and n is the number of data samples. The calculation formula of mean square error is: The formula measures the degree of deviation between the predicted results and the actual data by calculating the average of the squares of the differences between the predicted values and the actual values. The purpose of squaring is to amplify the impact of larger errors, so that the model pays more attention to those points with larger errors, thereby better optimizing the model. When calculating the transient error, the predicted electrical parameter value needs to be accurately matched with the actual electrical parameter value in the time-aligned multi-channel signal data. For example, when predicting and calculating the error of the voltage parameter of an electronic circuit, the predicted voltage value at a certain time needs to be compared with the actual voltage signal collected from the circuit at that time. Due to the noise interference and other problems in the multi-channel signal data, the actual signal data needs to be filtered before comparison. For example, Gaussian filtering, median filtering and other methods are used to remove high-frequency noise in the signal, so that the calculated transient error more accurately reflects the difference between the predicted value and the true value.

[0041] 107. Using the dynamic prediction model, the time-aligned multi-channel signal data is compensated and calculated based on the transient error, and electrical parameter data containing the compensation result is obtained.

[0042] Based on the obtained transient error, the dynamic prediction model is used to compensate and calculate the time-aligned multi-channel signal data. First, the dynamic prediction model will determine the adjustment strategy for the multi-channel signal data according to the size and direction of the transient error. If the transient error is large and positive, it means that the predicted value is greater than the actual value, and the model will accordingly reduce the prediction weight of the electrical parameter value at future time or adjust the current multi-channel signal data downward. For example, when predicting and compensating the current parameter of a motor, if it is found that the predicted current value is greater than the actual current value, the model will appropriately reduce the increment of the current prediction value in subsequent calculations, or reduce the current collected current signal by a certain proportion, so that the predicted value is closer to the actual value. On the contrary, if the transient error is large and negative, the model will increase the prediction weight of the electrical parameter value at future time or adjust the current multi-channel signal data upward. For example, when predicting the voltage parameter of a battery, if the predicted value is less than the actual value, the model will increase the increment of the voltage prediction value in subsequent calculations, or amplify the current voltage signal. In the compensation calculation, the dynamic prediction model can use various methods. One common method is based on the principle of feedback control, taking the transient error as the feedback signal and inputting it into the model to adjust the parameters of the model. For example, in a neural network-based dynamic prediction model, the transient error is propagated from the output layer to the hidden layer and the input layer through the backpropagation algorithm, and the weights in the network are updated according to the error, so that the model can better adapt to the changes of the actual data in subsequent prediction and calculation.

[0043] In addition, interpolation and fitting methods can also be used for compensation calculation. For points with large errors in the time-aligned multi-channel signal data, a more reasonable value can be estimated by interpolation method according to the values and change trend of adjacent points to replace. For example, when compensating the signal collected by a temperature sensor, if the temperature value at a certain time deviates greatly from the predicted value, linear interpolation or spline interpolation can be used to obtain a more accurate temperature value. For the overall signal data, a correction curve can be fitted according to the distribution of transient errors, and the original signal data can be adjusted as a whole. After the compensation calculation, the electrical parameter data containing the compensation result is obtained.

[0044] By using the intelligent switching rapid detection method in the embodiments of the present application, a series of technical means such as multi-channel wideband sampling, time stamp addition, construction of dynamic prediction model combining historical data and machine learning algorithm, and compensation calculation based on transient error are used to realize comprehensive, rapid and accurate detection and analysis of the electrical parameters of the electronic device load. Not only does it effectively solve the problem of insufficient data coverage and difficulty in extracting complete transient signal characteristics in complex load switching scenarios, but also significantly improves the detection efficiency and accuracy, providing strong support for ensuring stable operation of electronic devices and optimizing device performance. At the same time, through prediction of the electrical parameter value at future time and error compensation, the obtained electrical parameter data can more truly reflect the actual situation of the load, providing a reliable basis for the control and adjustment of related systems, and enhancing the stability and reliability of the entire electronic device system.

[0045] The following describes another method provided by the embodiments of the present application: multi-channel wideband sampling is used to obtain multi-channel digital signals, and time stamps are added for preprocessing and feature extraction, so as to train a dynamic prediction model. Then, the model is used to extract transient features, construct a feature template to determine the target load switching mode, obtain the transient change characteristics and predict the electrical parameter value. Then, the transient error is calculated, and the model weight is adjusted accordingly to construct an error distribution model to determine whether the error is reasonable. If there is an anomaly, the model weight is redistributed for secondary adjustment of the model. Finally, the adjusted model is used to calculate the error and correct the data to obtain accurate electrical parameter data.

[0046] Figure 2 is another flowchart of the method for intelligent switching rapid detection of electrical parameters of electronic device load in the embodiments of the present application.

[0047] Please refer to Figure 2 , another method for intelligent switching rapid detection of electrical parameters of electronic device load is described as follows: 201. Performing real-time parallel sampling of voltage and current signals generated by the electronic device during load switching using a multi-channel broadband sampling method to obtain multiple channels of digital signal data (this step has been described in 101 and will not be repeated here). 202. Adding a timestamp to each channel of the multiple channels of digital signal data to obtain time-aligned multiple channels of signal data (this step has been described in 102 and will not be repeated here). 203. Preprocess the acquired time-aligned multi-channel signal data and the historical load switching data to obtain preprocessed data, wherein the preprocessing includes denoising filtering, time synchronization correction, and normalization processing; After acquiring time-aligned multi-channel signal data and historical load switching data, preprocessing is performed to improve data quality. During actual acquisition, signals are susceptible to various noise interferences, such as electromagnetic noise in industrial environments and thermal noise from the equipment itself. This noise can obscure the true characteristics of the signal and affect the accuracy of the analysis results. Various denoising filtering methods can be used, such as mean filtering, median filtering, and wavelet filtering. For example, mean filtering calculates the average value of the data within a window of the signal and replaces the data at the center of the window, thereby smoothing the signal and removing random noise. For example, when acquiring the current signal during motor load switching, electromagnetic interference generates some glitches. After mean filtering, these glitches are effectively removed, resulting in a smoother current signal that truly reflects the current trend during motor load switching.

[0048] Time synchronization correction further ensures data consistency. Although timestamps have been added in the previous step, the data may still have slight time deviations due to factors such as differences in clock accuracy between devices and signal transmission delays. Using a time synchronization algorithm, the data timestamps are fine-tuned to ensure that all data is precisely aligned on the time axis. For example, in an electronic equipment monitoring network composed of multiple sensors, the clocks of each sensor may have slight errors. After time synchronization correction, the data collected by these sensors are completely synchronized in time, facilitating subsequent joint analysis of multiple parameters.

[0049] Normalization unifies data of different ranges and dimensions to the same scale. Historical load switching data and currently collected multi-channel signal data have different numerical ranges and dimensions for parameters such as voltage and current, which can affect the training effectiveness of machine learning algorithms. Normalization is used to map the data to the [0, 1] interval. For example, assuming the voltage range in the historical data is 0-500V and the current range is 0-20A, normalization converts both voltage and current data to the [0, 1] interval. This enhances data comparability and algorithm convergence speed, making subsequent model training more efficient and accurate.

[0050] 204、extracting transient features and feature labels from the preprocessed data, the transient features are extracted from the preprocessed time-aligned multi-channel signal data, including peak value, rise time, fall time and oscillation frequency, the feature labels are extracted from the preprocessed historical load switching data, corresponding to the switching mode; After completing the data preprocessing, key transient features and feature labels are extracted from the preprocessed data.

[0051] For the extraction of transient features, focus on the dynamic change characteristics of voltage and current during load switching. Take the motor starting process as an example, at the starting moment, the current will rise rapidly, by analyzing the preprocessed current signal, the peak value of the current, i.e. the maximum value of the current during the starting process, can be accurately captured; at the same time, the time taken for the current to reach the peak value from the start of the starting process is recorded, which is the rise time. When the motor stops, the current will gradually decrease, and the fall time of the current can also be obtained. In addition, in some circuits containing inductors, capacitors and other energy storage elements, load switching will cause oscillation of voltage and current. By using signal processing algorithms such as Fourier transform, the frequency components of the signal are analyzed to obtain the oscillation frequency. These transient features can directly reflect the change speed, amplitude and stability of voltage and current during load switching.

[0052] Feature labels are load switching mode identifiers extracted from preprocessed historical load switching data. Different load switching modes have different characteristics. According to the voltage, current change law and device operating state information in the historical data, the load switching mode is classified and given corresponding labels. For example, for the motor starting mode, the feature label can be defined as "motor starting"; for the switching mode from device standby to work, the label can be "standby to work switching". These feature labels provide clear targets for model training, so that machine learning algorithms can learn the corresponding relationship between different load switching modes and transient features.

[0053] 205、training a dynamic prediction model based on load switching mode by machine learning algorithm using the transient features and the feature labels; After obtaining the key transient features and feature labels, a dynamic prediction model based on load switching mode is constructed by machine learning algorithm.

[0054] During training, the model continuously feeds data samples, using transient features as input data and feature labels as desired outputs. Each data sample contains transient characteristics of voltage and current during load switching, such as peak value, rise time, and oscillation frequency, as well as the corresponding characteristic label of the load switching pattern. As the model processes this data, neurons calculate output values ​​based on the input data and current weight values. Initially, because the weights are randomly set, there is a significant difference between the model output and the true feature labels. This is achieved through the backpropagation algorithm, which backpropagates the error between the model output and the true label, starting from the output layer, and calculates the gradient of each neuron's weight. By iteratively adjusting the weights, the model's output gradually approaches the true feature label, meaning that the model can increasingly accurately identify load switching patterns.

[0055] To ensure the model's generalization ability and avoid overfitting, the training data is divided into a training set, a validation set, and a test set. The training set is used to train the model, allowing it to learn patterns and regularities in the data. The validation set is used to monitor the model's training process and evaluate its performance on the validation set to determine whether the model is overfitting. If the model performs well on the training set but degrades on the validation set, this indicates possible overfitting, and adjustments to the model's parameters or structure can be made. The test set is used for the final evaluation of the model's performance, ensuring that it performs well on unseen data.

[0056] 206. Extracting transient features from the time-aligned multi-channel signal data using the dynamic prediction model, the transient features including peak voltage, peak current, rise time, fall time, and oscillation frequency; When processing multiple signal channels, the model analyzes each signal point by point. For example, for the load switching signal of a transformer in power equipment, the model monitors changes in the voltage signal during the load switching process. When the transformer connects and disconnects a load, the voltage fluctuates. By monitoring and calculating the voltage signal in real time, the model accurately captures the peak voltage—the maximum value reached during the load switching process. It also records the time it takes for the voltage to reach its peak value, which is known as the voltage rise time. When the load switching ends and the voltage stabilizes, the model also accurately determines the voltage fall time. The same applies to current signals. During transformer load switching, the current changes are more pronounced. The model can quickly identify the current peak value and the time points when the current rises and falls, thereby calculating the rise and fall times.

[0057] In addition, in some complex power systems, load switching can cause oscillation of the circuit, resulting in periodic fluctuations in voltage and current. The model can accurately identify the oscillation frequency by performing spectral analysis on the signal, such as converting the time-domain signal to the frequency-domain signal using the Fourier transform. The oscillation frequency reflects the interaction between energy storage elements (such as inductance and capacitance) in the circuit and the load, which determines the stability of the circuit and the characteristics of the load.

[0058] 207、Based on the transient characteristics and the characteristics of different historical load switching modes, construct feature templates corresponding to different load switching modes; After extracting the transient characteristics in the time-aligned multi-channel signal data, combine the characteristics of different historical load switching modes to start constructing feature templates corresponding to different load switching modes.

[0059] First, deeply mine the historical load switching data. In these data, there are changes in voltage, current, and other electrical parameters under various different load switching modes. For example, for the common load switching mode of motor starting, the historical data shows that at the starting moment, the current will rise rapidly to a high peak value, and then gradually decrease to a stable operating value, while the voltage will drop to a certain extent. For the mode of motor stopping, the current will rapidly decrease from the stable operating value to zero, and the voltage will correspondingly rise. Organize and summarize these historical data to extract the typical characteristics of each load switching mode.

[0060] Then combine the extracted transient characteristics with the typical characteristics of historical load switching modes. Based on transient characteristics such as peak voltage, peak current, rise time, fall time, and oscillation frequency, construct feature templates for each load switching mode. For the feature template of the motor starting mode, the current peak value may be set within a certain range, the rise time within a certain time interval, and the voltage drop within a corresponding range. These characteristic values are not fixed, but are determined based on statistical analysis of historical data and actual application requirements to determine a reasonable fluctuation interval.

[0061] In the process of constructing feature templates, the similarities and differences between different load switching modes should also be considered. Some load switching modes may be similar in some characteristics, but different in other characteristics. For example, the load switching modes of motor starting and compressor starting may be similar in the trend of current rise, but different in current peak value, rise time, and oscillation frequency due to different device characteristics. By carefully analyzing these differences, the constructed feature templates can more accurately reflect the uniqueness of each load switching mode.

[0062] 208. Match and analyze the transient features collected in real time with the feature templates to obtain a target load switching mode. After constructing feature templates corresponding to different load switching modes, match and analyze the transient features collected in real time with these templates to determine the current target load switching mode.

[0063] The real-time transient features extracted by the dynamic prediction model from the time-aligned multi-channel signal data are obtained. These features include key information such as the peak value, rise time, fall time, and oscillation frequency of voltage and current. These real-time transient features are compared one by one with the pre-built feature templates. The comparison process adopts a similarity calculation algorithm. Taking the Euclidean distance algorithm as an example, it measures the similarity by calculating the distance between the real-time transient feature vector and each feature template vector. The closer the distance, the higher the similarity between the real-time feature and the template, which means that the current load switching mode is closer to the mode represented by the template. Assume that in the feature template of the motor starting mode, the current peak is I1, the rise time is T1, the voltage peak is V1, etc.; and in the transient features collected in real time, the current peak is I2, the rise time is T2, the voltage peak is V2, etc. The distance between the two is calculated using the Euclidean distance formula: The ellipsis here indicates the calculation of other transient features. After calculating the distances to each feature template, the load switching pattern corresponding to the template with the smallest distance is selected as the preliminary judgment result.

[0064] In real-world situations, multiple templates may share a high degree of similarity with real-time features. This necessitates further analysis and judgment, taking into account multiple factors, such as feature importance. Key features for determining load switching patterns, such as the current rise time during motor startup, are given higher weights, while less important features are given lower weights. By calculating weighted similarity, the target load switching pattern can be more accurately determined.

[0065] 209. Under the target load switching mode, classify the transient characteristics using a dynamic prediction model to obtain transient change characteristics corresponding to the target load switching mode; After determining the target load switching mode, the extracted transient features are carefully classified through the dynamic prediction model to explore the transient change characteristics corresponding to the target load switching mode.

[0066] When performing classification operations, the dynamic prediction model comprehensively considers the inherent connections and interactions between various transient characteristics. For example, when a power module load switching, a common occurrence in electronic devices, is used, the model classifies the involved transient characteristics, such as voltage and current, when the target load switching mode is switching from a low load to a high load. For voltage characteristics, the model classifies them based on factors such as voltage fluctuation and rate of change. These may be categorized as rapid voltage drop, slow voltage drop, or voltage drop followed by rise. Similarly, for current characteristics, the model classifies them based on factors such as the rate of rise and peak value, such as rapid current rise or step-by-step current rise.

[0067] During the classification process, the model leverages the knowledge and patterns learned during the training phase. By comparing and matching real-time transient features with the characteristic patterns in the training data, the model determines the category to which each transient feature belongs. For example, during the training phase, the model has learned that when a power module switches from a low load to a high load, if the current rises rapidly within a short period of time and has a large peak value, this characteristic pattern corresponds to a situation where the device needs to respond quickly to the load change. When the real-time current characteristics collected match this pattern, the model classifies it as a rapid current rise type.

[0068] By classifying all transient features, the model can summarize the complete transient change characteristics corresponding to the target load switching mode. These characteristics include key information such as the change pattern, change amplitude, and change time of parameters such as voltage and current during the load switching process.

[0069] 210. Predicting temporary electrical parameter values ​​at future moments using the dynamic prediction model based on the transient characteristics and known patterns in historical load switching data, where the patterns are typical patterns extracted from the historical load switching data, including transient behavior and electrical parameter change trends during the switching process; After obtaining the transient change characteristics corresponding to the target load switching pattern, the dynamic prediction model is used to combine these transient characteristics with the pattern rules extracted from the historical load switching data to predict the temporary electrical parameter values ​​at future moments.

[0070] During the prediction process, the dynamic prediction model fully utilizes the rich information contained in historical load switching data. This historical data records the changes in the electrical parameters of electronic devices under various load switching conditions. By analyzing and mining this data, typical patterns can be extracted. For example, in the historical data, it may be found that when a power module switches from a low load to a high load, the current will initially rise rapidly and then gradually stabilize over a period of time; the voltage will initially drop and then slowly return to a stable value. These patterns reflect the inherent operating mechanisms of electronic devices during load switching.

[0071] The model matches and integrates the transient characteristics of the current target load switching mode with historical mode rules. Taking the load switching of a power module as an example, it is known that the current transient characteristic is that the voltage drops rapidly at the moment of load switching, with a drop of 10%-15% of the rated voltage, and the current rises rapidly, with a rise speed of 10-20 amperes per second. The model will look for load switching cases with similar transient characteristics in historical data and analyze the subsequent changes of electrical parameters in these cases. If, in historical cases, after similar load switching, the voltage starts to rise within 5 milliseconds and recovers to 95% of the rated voltage after 10 milliseconds, and the current reaches a stable value within 8 milliseconds, then the model will predict the temporary electrical parameter values at future time points according to these historical rules and the current transient characteristics.

[0072] 211、Temporal fitting of the temporary electrical parameter values to obtain the trend of the electrical parameter changes during the load switching process; Select an appropriate time series fitting method. Common fitting methods include linear fitting, polynomial fitting, exponential fitting, etc. The choice of method depends on the distribution characteristics and change rules of the temporary electrical parameter values. Taking the voltage change in the load switching process of a power module as an example, if the temporary voltage values show an approximately linear change trend, then a linear fitting method can be selected. The basic principle of linear fitting is to find a straight line that minimizes the error between the straight line and the discrete temporary voltage values. Through algorithms such as least squares, the slope and intercept of the straight line can be determined, thus obtaining the linear fitting equation.

[0073] If the change of the temporary electrical parameter values is complex and shows nonlinear characteristics, methods such as polynomial fitting or exponential fitting may be needed. Polynomial fitting can better approximate complex curves by increasing the order of the polynomial. For example, for the change of current during load switching, it may first rise rapidly and then gradually stabilize, which may be more suitable for fitting with a quadratic or cubic polynomial. Exponential fitting is suitable for electrical parameter changes that show exponential growth or decay trends. For example, in some electronic devices, the voltage change during the charging or discharging process of a capacitor may follow an exponential law, so an exponential fitting method can be selected.

[0074] When performing time series fitting, the temporary electrical parameter values are preprocessed. This includes removing outliers, smoothing data, etc. Outliers may be due to measurement errors, interference, etc., and they can have a large impact on the fitting results. Through statistical analysis methods, such as methods based on standard deviation, outliers can be identified and removed. Smoothing data can reduce noise in the data, making the fitting results more accurate.

[0075] After the fitting is completed, the fitting results are evaluated. Evaluation indicators include goodness of fit, mean square error, etc. The goodness of fit reflects the fitting degree of the fitting curve and the original data, and the closer to 1 indicates the better the fitting effect. The mean square error measures the average error between the fitting curve and the original data, and the smaller the error indicates the better the fitting effect. If the evaluation result is not ideal, the fitting is re-performed by adjusting the fitting method or parameters. The variation trend curve of the electrical parameter obtained by time series fitting can directly show the variation of the electrical parameter with time during the load switching process.

[0076] 212、based on the variation trend, predicting the electrical parameter value in the stable state after the load switching; After obtaining the variation trend of the electrical parameter during the load switching process through time series fitting, the electrical parameter value in the stable state after the load switching is predicted based on the variation trend. Taking the voltage variation trend curve during the load switching process of the power module as an example, the curve may present a form of first falling and then rising, and finally tending to be flat, and the asymptote part of the curve is mainly concerned, that is, the stable value to which the curve gradually approaches. By analyzing the mathematical model of the curve, for example, limit calculation is performed on the equation obtained by polynomial fitting or exponential fitting, to determine the value of the curve when time tends to infinity. This value is the predicted voltage value in the stable state.

[0077] In the analysis process, errors and uncertainties that may exist in actual situations are also considered. Although the variation trend curve can roughly reflect the variation law of the electrical parameter, the stable state in the actual situation may be affected by various factors. For example, the heat dissipation of the electronic device, the interference of the external environment, etc. may cause a certain deviation between the electrical parameter value in the stable state and the theoretically predicted value. In order to cope with these uncertainties, statistical analysis method is used to analyze the stable state electrical parameter value under similar load switching conditions in the historical data, and calculate the statistical quantities such as mean value and standard deviation. Then, according to these statistical information, the predicted stable state electrical parameter value is corrected and adjusted.

[0078] 213、calculating the transient error of the multi-channel signal data aligned with the electrical parameter value and the time (this step has been described in 106, and will not be repeated here); 214、according to the distribution characteristics of the transient error, dynamically adjusting the weight parameter of the dynamic prediction model to obtain an adjusted dynamic prediction model; After obtaining the electrical parameter value in the stable state after the load switching and the transient error of each data point, the weight parameter of the dynamic prediction model needs to be dynamically adjusted according to the distribution characteristics of the transient error.

[0079] First, we need to analyze the distribution characteristics of transient errors in detail. Transient errors refer to the difference between the predicted electrical parameter values ​​and the actual measured values. Their distribution characteristics include information such as the magnitude, sign, and concentration of the error. Statistical analysis methods, such as plotting error histograms and calculating error statistics such as the mean, variance, skewness, and kurtosis, provide a comprehensive understanding of the distribution of transient errors. For example, if the mean of the error is not zero, it indicates that the model has systematic biases; if the variance is large, it indicates that the error fluctuates widely and the model is less stable.

[0080] Based on the analysis of the transient error distribution characteristics, an appropriate algorithm is used to dynamically adjust the model's weight parameters. Common algorithms include gradient descent and its variants, such as stochastic gradient descent and gradient descent with adaptive learning rates. Taking gradient descent as an example, its core idea is to update the model's weight parameters along the negative gradient of the error function, gradually decreasing the value of the error function. The error function is typically defined as the mean squared error (MSE) between the predicted and actual values, or another loss function. The gradient of the error function with respect to each weight parameter is calculated, and the weight parameter is then adjusted based on the magnitude and direction of the gradient.

[0081] 215. Using the adjusted dynamic prediction model, perform point-by-point calculation on the time-aligned multi-channel signal data to obtain a transient error of each data point; The time-aligned multi-channel signal data is input point by point into the adjusted dynamic prediction model. This multi-channel signal data includes real-time measurements of electrical parameters such as voltage and current during load switching. After the previous time alignment process, they are synchronized in the time dimension. Based on the input data, the model uses its internal weighting parameters and calculation rules to predict the electrical parameters corresponding to each data point.

[0082] For each data point, the model's predicted value is compared with the actual measured value, and the difference between them is calculated. This difference is the transient error for that data point. For example, for a voltage data point at a certain moment, if the model predicts a voltage value of V1 and the actual measured voltage value is V2, the transient error for that data point is: ΔV = V2 - V1. The same method can be applied to the error calculation of other electrical parameters such as current.

[0083] 216. Fit the transient error of each data point into a transient error sequence; After obtaining the transient error of each data point, these discrete transient error data are fitted into a continuous transient error sequence.

[0084] Select a suitable fitting method. Common fitting methods include polynomial fitting, spline fitting, Fourier series fitting, etc. Different fitting methods are suitable for different types of error data. For example, polynomial fitting is suitable for error data that shows a relatively simple polynomial trend; spline fitting can better handle error data with complex changes, which uses low-order polynomials for fitting in each small segment to ensure the smoothness and accuracy of the fitted curve; Fourier series fitting is suitable for error data with periodic changes, which can decompose error data into the superposition of sine and cosine functions of different frequencies.

[0085] When selecting a fitting method, consider the characteristics of the transient error data comprehensively. First, a preliminary visual analysis of the error data is performed, and a scatter plot of error versus time is drawn to observe the distribution and trend of the error data. If the error data roughly shows a linear or quadratic trend, polynomial fitting may be selected; if the error data has obvious fluctuations and local changes, spline fitting may be more appropriate; if the error data has periodic fluctuations, Fourier series fitting may be the best choice. After fitting the transient error of each data point into a transient error sequence, a continuous error curve can be obtained.

[0086] 217、Based on the transient error sequence, a normal distribution is used to construct a transient error distribution model; After fitting the transient error of each data point into a transient error sequence, a normal distribution is used to construct a transient error distribution model based on the sequence. Normal distribution is a common and important probability distribution, and in many practical problems, error data often approximately obeys normal distribution.

[0087] Statistical analysis is performed on the transient error sequence to calculate its mean μ and standard deviation σ. The mean μ reflects the average level of the transient error, and the standard deviation σ measures the dispersion of the transient error. According to the calculated mean μ and standard deviation σ, the parameters of the normal distribution can be determined. The probability density function of the normal distribution is where x represents the transient error value. This probability density function describes the probability distribution of the transient error at different values.

[0088] 218、Through the transient error distribution model and statistical process control method, it is judged whether the transient error exceeds a reasonable range; After the transient error distribution model is constructed, it is determined whether the transient error of each data point exceeds the pre-set reasonable range according to the model.

[0089] First, a reasonable range is determined. This reasonable range is usually determined according to the performance requirements of the electronic device, the actual application scenario, and past experience. Generally, the reasonable range is related to the parameters of the transient error distribution model, for example, it can be set as an interval of mean plus or minus several times the standard deviation. A common choice is mean plus or minus 2 or 3 times the standard deviation, because according to the characteristics of the normal distribution, about 95% of the data will fall within the interval of mean plus or minus 2 times the standard deviation, and about 99.7% of the data will fall within the interval of mean plus or minus 3 times the standard deviation. The reasonable range set in this way can effectively identify errors that may be caused by abnormal conditions while ensuring that most normal errors are included.

[0090] For the transient error of each data point, it is compared with the pre-set reasonable range. If the transient error of a certain data point does not exceed the reasonable range, it can be considered that the error is caused by normal fluctuations or measurement errors, which belongs to the acceptable range, and step 223 is continued; if the transient error of a certain data point exceeds the reasonable range, step 219 is executed.

[0091] 219、According to the time of the abnormal point, the transient characteristics of the abnormal point, and the distribution characteristics of the transient error, the cumulative amount of the transient error of the abnormal point in the time sequence is calculated, which reflects the cumulative influence strength of the abnormal point on the overall error; After determining that there are abnormal points whose transient errors exceed the reasonable range through the transient error distribution model and the statistical process control method, in-depth analysis is performed on these abnormal points, and the cumulative amount of the transient error of the abnormal point in the time sequence is calculated. First, according to the time when the abnormal point occurs, the position of the abnormal point in the entire time sequence data is accurately located.

[0092] Then, the transient characteristics corresponding to the abnormal point are extracted, which include key change information of voltage and current in the load switching process, such as peak value, rise time, fall time, and oscillation frequency. These transient characteristics not only reflect the instantaneous change of the signal at the abnormal point, but also provide clues for analyzing the causes of the abnormality. For example, if the current peak value at the abnormal point is abnormally high, it may indicate that the load has a short circuit fault or other abnormal conditions.

[0093] Then, the distribution characteristics of the transient error are combined to comprehensively consider the abnormal point. The distribution characteristics of the transient error include information reflected by statistical quantities such as mean, variance, skewness, and kurtosis. If the variance of the error distribution is large, the error fluctuation range is wide, and the influence of the abnormal point on the overall error may be more significant; skewness and kurtosis can further reveal the asymmetry and peak degree of the error distribution, helping to more comprehensively understand the characteristics of the error distribution.

[0094] When calculating the cumulative transient error, we typically start from the outlier point and trace back a certain time window along the time series, both forward and backward. Within this time window, a weighted sum is taken for the transient errors of each data point. The weighting takes into account the temporal distance between the outlier point and each data point, as well as the importance of each data point's transient characteristics. Data points closer to the outlier receive a higher weight, while time points corresponding to transient characteristics that significantly influence the load switching pattern receive a correspondingly higher weight. This weighted summation method more accurately measures the cumulative impact of the outlier point on the overall error within the time series.

[0095] 220. At the same time, the importance of the transient feature is obtained based on the degree of influence of the transient feature on the load switching mode; while calculating the cumulative amount of transient errors at abnormal points, the importance of the transient feature must also be determined. The historical load switching data is comprehensively sorted out and deeply analyzed. These historical data contain rich information on changes in electrical parameters such as voltage and current under various load switching modes, as well as the corresponding transient feature data. For each load switching mode, the relevant transient features are extracted. For example, in the motor starting mode, the rise time and peak value of the current are crucial for determining whether the starting process is normal; in the switching mode from standby to working, the voltage stabilization time and fluctuation amplitude are key features.

[0096] For each load switching pattern, statistical analysis is used to study the correlation between each transient feature and the load switching pattern. For example, the distribution differences of a transient feature across different load switching patterns are calculated. If the distribution of a transient feature varies significantly across different patterns, it indicates that it can effectively distinguish different load switching patterns, and thus the importance of this transient feature is relatively high. Conversely, if the distribution of a transient feature is relatively similar across various load switching patterns, its ability to distinguish different patterns is weaker and its importance is lower.

[0097] 221. Redistribute weight parameters of the dynamic prediction model according to the cumulative amount and the importance to obtain a secondary adjusted dynamic prediction model; After determining the cumulative transient error of the outlier and the importance of the transient features, we proceed to reallocate the weight parameters of the dynamic prediction model, thereby obtaining a secondary adjusted dynamic prediction model. Based on the values ​​of the cumulative transient error and transient feature importance, we construct a weight adjustment strategy. Because the cumulative transient error reflects the cumulative impact of the outlier on the overall error, and the transient feature importance reflects the criticality of each feature in determining the load switching mode, both factors must be considered when adjusting the weights.

[0098] Based on different weight adjustment strategies, the adjustment range and direction of each weight parameter are determined. Weight parameters associated with outliers that generate large cumulative transient errors and correspond to important transient features are adjusted more significantly. This is because these weight parameters significantly impact the accuracy of the model's predictions. Adjusting them can more effectively reduce the interference of outliers on the model, making the model more realistic. Weight parameters associated with less significant outliers or relatively minor transient features receive relatively smaller adjustments. After adjusting all weight parameters, a quadratically adjusted dynamic prediction model is obtained.

[0099] 222. Perform point-by-point calculation on the time-aligned multi-channel signal data according to the secondary adjusted dynamic prediction model to obtain a transient error of each data point; After obtaining the secondary adjustment dynamic prediction model, the model is used to perform point-by-point calculations on the time-aligned multi-channel signal data to obtain the transient error of each data point. The time-aligned multi-channel signal data is then input into the secondary adjustment dynamic prediction model in chronological order.

[0100] After receiving the signal data for each data point, the model predicts the corresponding electrical parameters based on its internal structure and weight parameters. The calculation process within the model involves complex mathematical operations and logical processing. For example, in a neural network model, data is processed by multiple layers of neurons. Each neuron performs a weighted sum based on the input data and its own weight, and then performs a nonlinear transformation through an activation function to ultimately obtain the prediction result.

[0101] Compare the model's predicted values ​​with the actual measured values. For each data point, calculate the difference between them; this difference represents the transient error for that data point. When calculating transient errors, ensure that the predicted and measured values ​​precisely correspond in terms of time and parameter type to ensure accurate calculations. When calculating transient errors for voltage data points, subtract the model-predicted voltage value from the actual measured voltage value at the same moment. After calculating the transient errors for all data points, organize and record them.

[0102] 223. Correct each data point using a real-time error compensation mechanism based on the transient error of each data point to obtain electrical parameter data including compensation results.

[0103] According to the transient error of each data point, the data points are corrected by means of a real-time error compensation mechanism, so as to obtain electrical parameter data containing compensation results. According to the size and direction of the transient error, a corresponding correction strategy is determined. For the case that the transient error is large, a more aggressive correction measure is taken; and for the data points with small error, a relatively mild adjustment is made to avoid excessive correction and cause new errors.

[0104] The data points are corrected by means of a real-time error compensation algorithm. Common compensation algorithms include a feedback control-based method, an interpolation method and a curve fitting method, etc. The feedback control-based algorithm takes the transient error as a feedback signal and inputs it into the data correction process. According to the size and direction of the error, the data points are adjusted correspondingly, so that the adjusted values are closer to the true values. The interpolation method estimates and corrects the data points with errors according to the values and variation trends of adjacent data points. If the transient error of a data point is large, a more reasonable value is calculated by linear interpolation or spline interpolation, etc., to replace the original data point, with reference to the values of adjacent data points before and after it. The curve fitting method fits an error correction curve by analyzing the transient errors of a series of data points. According to the curve, each data point is corrected so that the corrected data can more accurately reflect the actual electrical parameters in the load switching process of the electronic device. Finally, the corrected data is arranged into electrical parameter data containing compensation results.

[0105] By using the intelligent switching rapid detection method in the embodiments of the present application, a series of technical means such as acquiring multiple digital signals through multi-channel wideband sampling, adding time stamps to realize time alignment, pre-processing and feature extraction of data and training a dynamic prediction model, and model adjustment and data correction by using transient errors are adopted to comprehensively, accurately and efficiently detect and analyze the electrical parameters of the load of the electronic device. Not only the problems of insufficient data coverage and difficulty in extracting complete transient signal characteristics in the complex load switching scene are effectively solved, and the detection efficiency and accuracy are significantly improved, but also the electrical parameter data obtained can more truly reflect the actual situation of the load by predicting the electrical parameter values at future time and real-time compensation correction based on errors, thereby providing a strong guarantee for the stable operation and performance optimization of the electronic device and enhancing the stability and reliability of the entire electronic device system.

[0106] The embodiments of the intelligent switching rapid detection method for electrical parameter of load of electronic device are described above, and the electronic device in the embodiments of the present application is described from the perspective of hardware processing. Please refer to Figure 3 , which is a hardware structure schematic diagram of the electronic device in the embodiments of the present application.

[0107] It should be noted that, Figure 3The structure of the electronic device shown is merely an example and should not impose any limitation on the functions and use range of the embodiments of the present application.

[0108] As shown in Figure 3 The electronic device includes a central processing unit (CPU) 301 which can perform various appropriate actions and processes, such as executing the methods described in the above embodiments, according to programs stored in a read-only memory (ROM) 302 or programs loaded from a storage section 308 into a random access memory (RAM) 303. Various programs and data required for system operation are also stored in the random access memory (RAM) 303. The central processing unit (CPU) 301, the read-only memory (ROM) 302, and the random access memory (RAM) 303 are connected to each other through a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0109] The following components are connected to the input / output (I / O) interface 305: an input section 306 including an audio input device, a button switch, and the like; an output section 307 including a display, an audio output device, an indicator, and the like; the storage section 308 including a hard disk and the like; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, and the like. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the input / output (I / O) interface 305 as necessary. A removable medium 311 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like is mounted on the drive 310 as necessary, so that a computer program read therefrom is installed in the storage section 308 as necessary.

[0110] In particular, the processes described above with reference to the flowcharts can be implemented as a computer software program according to embodiments of the present application. For example, embodiments of the present application include a computer program product comprising a computer program carried on a computer readable medium, the computer program comprising computer program code for implementing the methods illustrated by the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via the communication section 309, and / or installed from the removable medium 311. When the computer program is executed by the central processing unit (CPU) 301, various functions defined in the present application are performed.

[0111] Note that specific examples of the computer readable storage medium can include one or more of the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present application, the computer readable storage medium can be any tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0112] The flowcharts and block diagrams in the attached drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowcharts or block diagrams can represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks can sometimes be executed in the reverse order, depending upon the functionality involved.

[0113] In particular, the electronic device of the present embodiment includes a processor and a memory coupled to the one or more processors, the memory storing computer program code comprising computer instructions to be invoked by the one or more processors to cause the electronic device to perform the method provided by the above-described embodiments.

[0114] As another aspect, the present application also provides a computer readable storage medium, which can be included in the electronic device described in the above embodiments, or can exist separately without being assembled into the electronic device. The storage medium carries one or more computer programs, which, when executed by a processor of the electronic device, enable the electronic device to implement the method provided in the above embodiments.

[0115] The above-described embodiments are merely used to illustrate the technical solutions of the present application, but not for limiting the present application; even though the present application has been described in detail with reference to the foregoing embodiments, those ordinarily skilled in the art should understand that: they can still make modifications to the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some of the technical features; and the modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0116] In the above embodiments, according to the context, the term "when" can be interpreted as meaning "if" or "after" or "in response to determining" or "in response to detecting". Similarly, according to the context, the phrase "upon determining" or "if detecting (the stated condition or event)" can be interpreted as meaning "if determining" or "in response to determining" or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)".

[0117] In the above embodiments, all or some of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or some of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or some of the processes or functions according to the embodiments of the present application are generated. The computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer readable storage medium, or transmitted from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line) or wireless (such as infrared, wireless, microwave, etc.) mode. The computer readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media. The available media can be magnetic media (such as floppy disk, hard disk, magnetic tape), optical media (such as DVD), or semiconductor media (such as solid state disk) and the like.

[0118] Those skilled in the art can understand that all or part of the processes in the above-mentioned method embodiments can be implemented by a computer program instructing relevant hardware to complete, the program can be stored in a computer-readable storage medium, and the program can include the processes of the above-mentioned method embodiments when executed. The aforementioned storage medium includes ROM or random storage memory RAM, magnetic disc or optical disc, and various storage medium capable of storing program codes.

Claims

1. A method for intelligent switching and rapid detection of load electrical parameters of electronic equipment, characterized in that: include: The voltage and current signals generated by the electronic device during load switching are sampled in real time and in parallel by using a multi-channel broadband sampling method to obtain multi-channel digital signal data; Adding a timestamp to each channel of signal data of the multiple channels of digital signal data to obtain time-aligned multiple channels of signal data; Establishing a dynamic prediction model based on load switching patterns using a machine learning algorithm based on historical load switching data and the time-aligned multi-channel signal data, wherein the historical load switching data includes typical load switching patterns and corresponding electrical parameter changes; Analyzing the time-aligned multi-channel signal data using the dynamic prediction model to obtain transient change characteristics of the electronic device load electrical parameters; Predicting electrical parameter values ​​at future times based on the transient change characteristics; Calculating a transient error between the electrical parameter value and the time-aligned multi-channel signal data; The dynamic prediction model is used to perform compensation calculation on the time-aligned multi-channel signal data according to the transient error to obtain electrical parameter data including compensation results.

2. The method according to claim 1, characterized in that Based on the historical load switching data and the time-aligned multi-channel signal data, a dynamic prediction model based on the load switching pattern is established through a machine learning algorithm, specifically including: Preprocessing the acquired time-aligned multi-channel signal data and the historical load switching data to obtain preprocessed data, wherein the preprocessing includes denoising filtering, time synchronization correction, and normalization processing; Extracting transient features and feature labels from the preprocessed data, wherein the transient features are transient features of voltage and current during the load switching process extracted from the preprocessed time-aligned multi-channel signal data, including peak value, rise time, fall time, and oscillation frequency; and the feature labels are feature labels corresponding to switching modes extracted from the preprocessed historical load switching data; The transient features and the feature labels are used to train a dynamic prediction model through a machine learning algorithm to obtain a dynamic prediction model based on the load switching mode.

3. The method according to claim 1, characterized in that The time-aligned multi-channel signal data is analyzed by the dynamic prediction model to obtain transient change characteristics of the electronic device load electrical parameters, specifically including: Extracting transient features from the time-aligned multi-channel signal data using the dynamic prediction model, wherein the transient features include peak voltage, peak current, rise time, fall time, and oscillation frequency; Based on the transient characteristics and the characteristics of different historical load switching modes, constructing feature templates corresponding to different load switching modes; Matching and analyzing the transient characteristics collected in real time with the characteristic template to obtain the target load switching mode; In the target load switching mode, the transient features are classified by a dynamic prediction model to obtain transient change characteristics corresponding to the target load switching mode.

4. The method according to claim 1, wherein Predicting electrical parameter values ​​at future times based on the transient change characteristics specifically includes: Predicting temporary electrical parameter values ​​at future moments using the dynamic prediction model based on the transient characteristics and known pattern rules in historical load switching data, wherein the pattern rules are typical patterns extracted from the historical load switching data, including transient behavior and electrical parameter change trends during the switching process; Performing time series fitting on the temporary electrical parameter values ​​to obtain a change trend of the electrical parameter during the load switching process; Based on the change trend, the electrical parameter value of the load in a stable state after switching is predicted.

5. The method according to claim 1, wherein Using the dynamic prediction model, compensation calculation is performed on the time-aligned multi-channel signal data according to the transient error to obtain electrical parameter data including compensation results, specifically including: Dynamically adjusting the weight parameters of the dynamic prediction model according to the distribution characteristics of the transient error to obtain an adjusted dynamic prediction model; Using the adjusted dynamic prediction model, performing point-by-point calculation on the time-aligned multi-channel signal data to obtain a transient error of each data point; According to the transient error of each data point, each data point is corrected using a real-time error compensation mechanism to obtain electrical parameter data including compensation results.

6. The method according to claim 5, characterized in that After the step of using the adjusted dynamic prediction model to perform point-by-point calculation on the time-aligned multi-channel signal data to obtain a transient error of each data point, the method further includes: Fitting the transient error of each data point into a transient error sequence; Based on the transient error sequence, a transient error distribution model is constructed using normal distribution; Determining whether the transient error exceeds a reasonable range by using the transient error distribution model and the statistical process control method; If so, reallocating the weight parameters of the dynamic prediction model according to the time of the abnormal point and the corresponding transient characteristics to obtain a secondary adjusted dynamic prediction model, wherein the abnormal point is a point outside the reasonable range; The time-aligned multi-channel signal data are calculated point by point based on the secondary adjusted dynamic prediction model to obtain a transient error of each data point.

7. The method according to claim 6, characterized in that Redistributing the weight parameters of the dynamic prediction model according to the time of the abnormal point and the corresponding transient characteristics to obtain a secondary adjusted dynamic prediction model, specifically including: Calculate the cumulative transient error of the abnormal point in the time series according to the time of the abnormal point, the transient characteristics of the abnormal point, and the distribution characteristics of the transient error, wherein the cumulative amount reflects the cumulative impact intensity of the abnormal point on the overall error; At the same time, according to the influence degree of the transient characteristics on the load switching mode, the importance of the transient characteristics is obtained; The weight parameters of the dynamic prediction model are redistributed according to the cumulative amount and the importance to obtain a secondary adjusted dynamic prediction model.

8. A detection system, characterized in that: including one or more processors and memory; The memory is coupled to the one or more processors, and is configured to store computer program codes, where the computer program codes include computer instructions. The one or more processors call the computer instructions to enable the detection system to execute the method according to any one of claims 1 to 7.

9. An electronic device, characterized in that: When the electronic device is run on the detection system, the detection system is caused to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing computer instructions, characterized in that: In due place When the computer instructions are executed on a detection system, the detection system is caused to perform the method according to any one of claims 1 to 7.

Images