CN118759245B - Power adapter-based current overload monitoring method and system - Google Patents

Abstract

The invention provides a current overload monitoring method and system based on a power adapter, and relates to the technical field of electronic equipment. The system comprises a historical data acquisition module, a feature extraction module, a model training module, a data prediction module and a current overload monitoring module. According to the invention, through comprehensively analyzing multiple parameters, an overload monitoring threshold value is accurately set and early warning is carried out on current overload, so that accurate monitoring and timely early warning on the current overload of the power adapter are realized, and by combining a sliding window analysis and difference analysis technology and a prediction model, the system can sensitively respond to abnormal conditions, provide enough reaction time, reduce false alarm, and ensure efficient utilization of system resources, thereby effectively ensuring safe operation of the power adapter.

Description

Power adapter-based current overload monitoring method and system

Technical Field

The present invention relates to the field of electronic devices, and in particular, to a method and a system for monitoring current overload based on a power adapter.

Background

The power adapter is an important component for supplying power to electronic equipment, and the reliability and the safety of the power adapter are particularly important. Current overload is one of the common failures of power adapters with the risk of causing damage to the electronic equipment. Therefore, current overload monitoring and early warning are important technologies for guaranteeing safe operation of the power adapter.

Most power adapters currently on the market use simple overload protection mechanisms, such as fuses or overcurrent protection circuits. Although these methods can protect equipment to some extent, they have certain drawbacks. For example, overload protection is carried out by simply relying on a fixed threshold value, so that the overload protection is easily influenced by instantaneous current fluctuation or external interference, the false alarm rate is higher, and the user experience is influenced; there are limitations to the current load characteristics in different environments (e.g., high temperature, humidity, etc.), if the monitoring strategy cannot be dynamically adjusted according to the environmental changes.

Disclosure of Invention

In order to solve the technical problems, the invention provides a current overload monitoring method and a system based on a power adapter, which are used for analyzing multidimensional data such as current, voltage, temperature and the like, adopting a sliding window technology to extract characteristics, combining a long-period memory network model to conduct data analysis and prediction, realizing dynamic adjustment of a monitoring threshold value and improving the monitoring precision of current overload.

To achieve the above object, a first aspect of the present invention provides a current overload monitoring method based on a power adapter, including:

acquiring historical working data of a power adapter, determining a plurality of abnormal events in the historical working data according to a preset monitoring threshold, and determining a first target period and a second target period according to event occurrence time of the abnormal events and a preset duration range;

Extracting first monitoring data corresponding to a first target period of each abnormal event and second monitoring data corresponding to a second target period, and associating the first monitoring data and the second monitoring data corresponding to each abnormal event to obtain reference characteristic data of each abnormal event;

Carrying out sliding window analysis and difference analysis on multiple groups of reference feature data to obtain reference difference features of each group of reference feature data, constructing a training data set based on the reference difference features of each group of reference feature data, and training through the training data set to obtain an operation parameter prediction model;

collecting current working monitoring data of a power adapter, screening target operation data from the working monitoring data based on a preset observation threshold value, carrying out sliding window analysis and difference analysis on the target operation data to generate target difference characteristics, and inputting the target difference characteristic data into an operation parameter prediction model to obtain predicted operation data;

splicing the target characteristic data with the predicted operation data to obtain operation reference data, analyzing the operation reference data, determining an observation reference node of the operation reference data, and extracting observation characteristics of the observation reference node to determine an overload monitoring threshold;

and carrying out current overload monitoring on the power adapter according to the overload monitoring threshold value to obtain a current overload monitoring result of the power adapter.

Preferably, sliding window analysis and difference analysis are performed on multiple sets of target feature data to obtain reference difference features of each set of reference feature data, including:

Sliding window analysis is carried out on multiple groups of first monitoring data and second monitoring data based on a preset sliding window and a preset sliding step length, multiple characteristic values in each preset sliding window are extracted, and a first characteristic matrix corresponding to each group of first monitoring data and a second characteristic matrix corresponding to each group of second monitoring data are constructed according to the multiple characteristic values corresponding to each preset sliding window;

And calculating the difference value and the change rate of the first feature matrix and the second feature matrix corresponding to each group of reference feature data on each feature to obtain the reference difference feature of each group of reference feature data, wherein the reference difference feature comprises the difference feature matrix corresponding to each group of reference feature data.

Preferably, the screening the target operation data from the operation monitoring data based on the preset observation threshold value includes:

Sliding window analysis is conducted on the working monitoring data based on preset sliding windows and preset sliding step sizes, current average values and temperature average values in each preset sliding window are calculated, a first current observation node is determined according to the preset sliding window corresponding to the current average value closest to the first current observation value, and a first temperature observation node is determined according to the preset sliding window corresponding to the temperature average value closest to the first temperature observation value;

And selecting an intermediate value of the first current observation node and the first temperature observation node as a first target observation node, and screening target operation data from the operation monitoring data according to a preset duration range and the first target observation node.

Preferably, the method for determining the overload monitoring threshold includes the steps of splicing target feature data with predicted operation data to obtain operation reference data, analyzing the operation reference data to determine an observation reference node of the operation reference data, and extracting observation features of the observation reference node to determine the overload monitoring threshold, including:

Judging whether an abnormal detection node exists in the operation reference data according to a preset monitoring threshold value, wherein if the current value at any moment exists in the operation reference data and is larger than the preset monitoring threshold value, the abnormal detection node exists in the operation reference data;

Selecting a time node with a current value larger than a preset monitoring threshold value in the operation reference data as an abnormality detection node, and determining an observation reference node of the operation reference data based on a preset observation time length and the abnormality detection node;

Taking the observation reference node as the center of a preset sliding window, and extracting the observation characteristic of the observation reference node based on operation reference data in the preset sliding window;

Extracting fluctuation reference data from operation reference data based on a preset duration range and an abnormality detection node, extracting reference fluctuation characteristics from the fluctuation reference data, correcting observation characteristics of an observation reference node based on the reference fluctuation characteristics to obtain target characteristics, and determining an overload monitoring threshold according to the target characteristics.

Preferably, extracting a reference fluctuation feature from the fluctuation reference data, correcting the observation feature of the observation reference node based on the reference fluctuation feature to obtain a target feature, and determining the overload monitoring threshold according to the target feature, including:

Sliding window analysis is carried out on the fluctuation reference data based on a preset sliding window and a preset sliding step length, a plurality of wavelet features corresponding to each preset sliding window are extracted, feature average values of each type of wavelet features are obtained, and reference fluctuation features corresponding to the fluctuation reference data are obtained;

And performing differential analysis on the observation characteristics of the observation reference nodes based on the reference fluctuation characteristics to obtain target characteristics after the observation characteristics of the observation reference nodes are corrected, and taking a plurality of characteristic values in the target characteristics as overload monitoring thresholds.

Preferably, the current overload monitoring of the power adapter according to the overload monitoring threshold comprises:

For the current work monitoring data of the power adapter obtained by monitoring, calculating to obtain an abnormal value corresponding to the k moment in the work monitoring data by adopting the following formula:

；

In the formula, For the outlier corresponding to time k in the operation monitoring data,For the k time point in the work monitoring dataThe value of the parameter of the item,Is the firstThe overload monitoring threshold value of the item parameter,Is the firstNormalized weight parameters of the term parameters,Monitoring the number of parameters in the data for the job;

And carrying out current overload early warning after detecting that the abnormal value corresponding to any moment in the current work monitoring data of the power adapter is larger than a preset abnormal threshold value.

Preferably, the operating parameter predictive model is a long and short term memory network model.

The second aspect of the present invention provides a current overload monitoring system based on a power adapter, configured to implement the above current overload monitoring method based on a power adapter, including:

the historical data acquisition module is used for acquiring historical working data of the power adapter, determining a plurality of abnormal events in the historical working data according to a preset monitoring threshold value, and determining a first target period and a second target period according to the event occurrence time of the abnormal events and a preset duration range;

the feature extraction module is used for extracting first monitoring data corresponding to a first target period of each abnormal event and second monitoring data corresponding to a second target period, and associating the first monitoring data and the second monitoring data corresponding to each abnormal event to obtain reference feature data of each abnormal event;

The model training module is used for carrying out sliding window analysis and difference analysis on multiple groups of reference feature data to obtain reference difference features of each group of reference feature data, constructing a training data set based on the reference difference features of each group of reference feature data, and training the training data set to obtain an operation parameter prediction model;

the data prediction module is used for collecting current working monitoring data of the power adapter, screening target operation data from the working monitoring data based on a preset observation threshold value, carrying out sliding window analysis and difference analysis on the target operation data to generate target difference characteristics, and inputting the target difference characteristic data into the operation parameter prediction model to obtain predicted operation data;

And the current overload monitoring module is used for splicing the target characteristic data with the predicted operation data to obtain operation reference data, analyzing the operation reference data, determining an observation reference node of the operation reference data, extracting the observation characteristic of the observation reference node to determine an overload monitoring threshold value, and carrying out current overload monitoring on the power adapter according to the overload monitoring threshold value to obtain a current overload monitoring result of the power adapter.

The invention has the following beneficial effects:

According to the invention, through comprehensively analyzing multiple parameters, an overload monitoring threshold value is accurately set and early warning is carried out on current overload, so that accurate monitoring and timely early warning on the current overload of the power adapter are realized, and by combining a sliding window analysis and difference analysis technology and a prediction model, the system can sensitively respond to abnormal conditions, provide enough reaction time, reduce false alarm, and ensure efficient utilization of system resources, thereby effectively ensuring safe operation of the power adapter.

Drawings

Fig. 1 is a schematic flow chart of a current overload monitoring method based on a power adapter according to an embodiment of the present invention.

Fig. 2 is a schematic structural diagram of a current overload monitoring system based on a power adapter according to an embodiment of the present invention.

Detailed Description

In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

Fig. 1 is a schematic flow chart of a current overload monitoring method based on a power adapter according to an embodiment of the present invention, please refer to fig. 1, the method includes the following steps:

step S1, acquiring historical working data of a power adapter, determining a plurality of abnormal events in the historical working data according to a preset monitoring threshold, and determining a first target period and a second target period according to event occurrence time of the abnormal events and a preset duration range;

In this embodiment, the historical working data of the power adapter includes time-series data of parameters (hereinafter simply referred to as current, voltage, temperature) such as output current, output voltage, device temperature, etc. of the power adapter in different operating states or different operating environments, and according to a plurality of abnormal events in the historical working data determined by a preset monitoring threshold, specifically, a moment of current overload existing in the historical working data, the preset monitoring threshold is a threshold for determining whether current overload occurs, for example, when the output current of the power adapter is greater than the preset monitoring threshold, the power adapter is considered to be current overload.

A first target period and a second target period determined according to the occurrence time of each abnormal event and a preset duration range, wherein the first target period refers to a certain period of time before the occurrence of the abnormal event, is used for capturing precursor data of the occurrence of the abnormal event, i.e. for mining some potential features before the occurrence of a current overload phenomenon, the second target period refers to a certain period of time before the first target period for extracting reference data for comparison, i.e. for characterizing normal data features in case no potential anomaly has occurred. The first target period is a period corresponding to the first 10 minutes of the moment when the current overload occurs, and the second target period is a period corresponding to the first 10 minutes of the first target period.

Step S2, extracting first monitoring data corresponding to a first target period of each abnormal event and second monitoring data corresponding to a second target period, and associating the first monitoring data and the second monitoring data corresponding to each abnormal event to obtain reference characteristic data of each abnormal event;

In this embodiment, the first monitoring data and the second monitoring data extracted from the historical working data include precursor feature data before occurrence of an abnormal event and normal operation data in a reference period. The characteristic change mode before the occurrence of the abnormal event can be identified by analyzing the reference characteristic data obtained by correlating the first monitoring data and the second monitoring data corresponding to each abnormal event.

S3, carrying out sliding window analysis and difference analysis on multiple groups of reference feature data to obtain reference difference features of each group of reference feature data, constructing a training data set based on the reference difference features of each group of reference feature data, and training through the training data set to obtain an operation parameter prediction model;

In this embodiment, the time sequence features included in the data are considered, the sliding window technology is adopted to analyze the reference feature data, then the feature data without abnormal information included in the second monitoring data are combined to perform differential analysis on the precursor feature data, interference information caused by factors such as noise in the precursor feature data is removed, reference differential features of each group of reference feature data are obtained, a training data set is built based on the reference differential features, an operation parameter prediction model is trained through the training data set, and the operation parameter prediction model has the function of predicting data in a future period according to the current voltage, current, temperature and other data.

S4, collecting current working monitoring data of the power adapter, screening target operation data from the working monitoring data based on a preset observation threshold value, carrying out sliding window analysis and difference analysis on the target operation data to generate target difference characteristics, and inputting the target difference characteristic data into an operation parameter prediction model to obtain predicted operation data;

In this embodiment, the current operation monitoring data of the power adapter may specifically be data corresponding to parameters such as current, voltage and temperature of the power adapter, which are collected in real time, and the preset observation threshold is used for determining minor anomalies occurring in the data, that is, identifying a time node corresponding to the fluctuation of the data exceeding a certain limit, screening out target operation data from the operation monitoring data according to the time node, extracting target difference features of the target operation data, and processing the target operation data through an operation parameter prediction model to generate predicted operation data.

Step S5, splicing the target characteristic data and the predicted operation data to obtain operation reference data, analyzing the operation reference data, determining an observation reference node of the operation reference data, and extracting observation characteristics of the observation reference node to determine an overload monitoring threshold;

In this embodiment, the target feature data is spliced with predicted operation data obtained through an operation parameter prediction model to generate operation reference data, the operation reference data includes predicted data and current monitored information in a period of time in the future, and an observation reference node of the operation reference data is determined by combining the predicted data. The observation reference node is determined based on the time of occurrence of the event, namely, it is determined that current overload may occur at a certain time in the future through predicted data, then the observation reference node is determined based on the time length of early warning, the predicted time of occurrence of the current overload is taken as a reference, for example, the time of the first 1 minute of the predicted time of occurrence of the current overload is recorded as the observation reference node, and an overload monitoring threshold value for current overload monitoring is determined according to the observation characteristics of the observation reference node.

And S6, carrying out current overload monitoring on the power adapter according to the overload monitoring threshold value to obtain a current overload monitoring result of the power adapter.

In the embodiment, the power adapter is monitored for current overload according to the overload monitoring threshold, so that accurate monitoring and early warning of the current overload of the power adapter can be realized, the working state of the power adapter can be monitored in real time through the steps, potential current overload risks can be recognized in advance through a data analysis means and a prediction model, and early warning signals can be sent out timely, thereby effectively guaranteeing safe operation of the power adapter.

As an exemplary implementation process, in step S2, sliding window analysis and difference analysis are performed on multiple sets of target feature data to obtain reference difference features of each set of reference feature data, which specifically includes:

Sliding window analysis is carried out on multiple groups of first monitoring data and second monitoring data based on a preset sliding window and a preset sliding step length, multiple characteristic values in each preset sliding window are extracted, and a first characteristic matrix corresponding to each group of first monitoring data and a second characteristic matrix corresponding to each group of second monitoring data are constructed according to the multiple characteristic values corresponding to each preset sliding window;

In this embodiment, the preset sliding window and the preset sliding step size may be determined according to the specific application scenario and the data characteristics, for example, 1 minute, and the preset sliding step size is a time interval of each movement of the sliding window, for example, 10 seconds. And carrying out sliding window analysis on multiple groups of first monitoring data and second monitoring data based on a preset sliding window and a preset sliding step length, and extracting multiple characteristic values in the preset sliding window in each sliding window operation so as to comprehensively reflect the change conditions of parameters such as current, voltage, temperature and the like. These characteristic values may include, but are not limited to: mean, the mean of the data in the sliding window, reflecting the central trend of the data; standard deviation (standard deviation), the fluctuation degree of the data in the sliding window reflects the discrete degree of the data; slope (slope), linear regression slope of data within sliding window, reflecting rate of change of data; peak (peak), the maximum and minimum of data within the sliding window, reflect the extreme cases of data.

After extracting a plurality of characteristic values in each preset sliding window, constructing a first characteristic matrix corresponding to each group of first monitoring data and a second characteristic matrix corresponding to each group of second monitoring data. Illustratively, the rows of the feature matrix represent different sliding windows and the columns represent different feature values extracted within each sliding window.

Calculating the difference value and the change rate of the first feature matrix and the second feature matrix corresponding to each group of reference feature data on each feature to obtain the reference difference feature of each group of reference feature data, wherein the reference difference feature comprises the difference feature matrix corresponding to each group of reference feature data;

In this embodiment, the reference difference feature includes the extracted feature difference value and the change rate in each sliding window, and the reference difference feature is expressed in the form of a difference feature matrix, and illustratively, rows of the difference feature matrix represent different sliding windows, and columns represent the feature difference value and the change rate extracted in each sliding window. The difference features can effectively reflect the change rule between the first monitoring data and the second monitoring data and help identify the feature change mode before the abnormal event occurs.

As an exemplary implementation process, in step S3, for the operation parameter prediction model, the operation parameter prediction model is built based on the long-short-term memory network architecture in this embodiment, for the training data set, the input feature and the output tag are created by using the sliding window in a manner of segmenting data, the input feature may be historical data of voltage, current, temperature, etc., the output tag may be data of voltage, current, temperature, etc., in a future period, the input feature and the output tag are divided into a training set and a verification set, the training of the LSTM model is performed through the training set, and the performance of the model on the verification set is monitored, so that the operation parameter prediction model capable of predicting relevant data in the future period according to the current data of voltage, current, temperature, etc. is obtained through training. The LSTM model is a technical means well known to those skilled in the art, and is not specifically limited in this embodiment.

As an exemplary implementation process, in step S4, the screening of the target operation data from the operation monitoring data based on the preset observation threshold specifically includes:

Sliding window analysis is conducted on the working monitoring data based on preset sliding windows and preset sliding step sizes, current average values and temperature average values in each preset sliding window are calculated, a first current observation node is determined according to the preset sliding window corresponding to the current average value closest to the first current observation value, and a first temperature observation node is determined according to the preset sliding window corresponding to the temperature average value closest to the first temperature observation value;

In this embodiment, in order to accurately screen out the target operation data with reference from the operation monitoring data, a preset observation threshold is taken as a reference, and data analysis is performed by a sliding window technology. Specifically, the sliding window analysis may be performed using the preset sliding window and the preset sliding step length mentioned in the foregoing, and the average value of the current and the temperature may be calculated in each sliding window. These averages reflect the overall level of current and temperature over the period of time, helping to identify abnormal fluctuations. The preset observation threshold value comprises a first current observation value and a first temperature observation value. The observations may be based on characteristic values in the historical data, such as the mean or median of current and temperature, and set in conjunction with a particular detection accuracy. And (3) finding a sliding window corresponding to the current average value closest to the first current observation value by analyzing the current average value in each sliding window, wherein the central point time of the sliding window is the first current observation node. Similarly, a sliding window corresponding to the temperature mean value closest to the first temperature observation value is found, and the central point time of the sliding window is the first temperature observation node.

Selecting an intermediate value of a first current observation node and a first temperature observation node as a first target observation node, and screening target operation data from the operation monitoring data according to a preset duration range and the first target observation node;

In this embodiment, the first current observation node and the first temperature observation node represent key observation time points of current and temperature, respectively. In order to comprehensively consider the influence of two parameters of current and temperature, the middle time point of the two observation nodes is selected as a first target observation node, and the time of the two observation nodes can be determined by calculating the average value of the time of the two observation nodes. And for screening the target operation data, taking the first target observation node as a reference center, and taking a preset duration range as a data screening range. For example, if the preset duration range is 3 minutes, the data of 1.5 minutes forward and 1.5 minutes backward of the first target observation node are taken into the target operation data, and all the data in the duration range including parameters such as current and temperature are extracted from the operation monitoring data, so that the target operation data is obtained.

It is worth to say that, the target operation data considers abnormal fluctuation of the data, and specifically exceeds the fluctuation existing in the data in a normal state, partial data with abnormal change trend can be extracted by reasonably setting a preset observation threshold, and then analysis is carried out by means of an operation parameter prediction model, so that the operation parameter prediction model needs to analyze and predict the data close to the abnormal occurrence in order to improve the prediction precision. If the data which deviate from the normal condition only slightly is predicted, the predicted data has a high probability that no node with current overload exists, or it can be understood that the current overload is not worried about for a long time, and in this case, the calculation resource is wasted easily by calling the operation parameter prediction model. The target operation data screened from the work monitoring data based on the preset observation threshold value can better improve the resource utilization rate.

As an exemplary implementation process, in step S5, splicing the target feature data with the predicted operation data to obtain operation reference data, analyzing the operation reference data, determining an observation reference node of the operation reference data, and extracting an observation feature of the observation reference node to determine an overload monitoring threshold, including:

Judging whether an abnormal detection node exists in the operation reference data according to a preset monitoring threshold value, wherein if the current value at any moment exists in the operation reference data and is larger than the preset monitoring threshold value, the abnormal detection node exists in the operation reference data;

Selecting a time node with a current value larger than a preset monitoring threshold value in the operation reference data as an abnormality detection node, and determining an observation reference node of the operation reference data based on a preset observation time length and the abnormality detection node;

in this embodiment, the target feature data is spliced with the predicted operation data obtained by the operation parameter prediction model, and the generated operation reference data includes actual monitoring data at the current moment and future data predicted based on the historical data and the model, so that the current and future operation states can be comprehensively considered by splicing, and more comprehensive information is provided for subsequent analysis.

In this embodiment, for each time in the operation reference data, it is checked whether the current value thereof is greater than a preset monitoring threshold, and if the current value at any time exceeds the monitoring threshold, it is determined that an abnormality detection node exists in the operation reference data. And for the abnormality detection node, selecting a time node of which the current value in the operation reference data exceeds a preset monitoring threshold value for the first time as the abnormality detection node.

Further, in order to improve the selected rationality of the anomaly detection node, the preset monitoring threshold may include a second current observation value and a second temperature observation value at the same time, and the change of the temperature data and the current data is comprehensively considered instead of the single consideration of the current data, and according to the second current observation value and the second temperature observation value, a time node in which the current value in the operation reference data exceeds the second current observation value for the first time is determined and is denoted as the second current observation node, and a time node in which the temperature value exceeds the second temperature observation value for the first time is denoted as the second temperature observation node, and an average value of the second current observation node and the second temperature observation node is selected as the anomaly detection node, which may also be referred to as the second target observation node, and the anomaly detection node is used for indicating that the current value or the temperature value exceeds a certain set threshold for the first time, so as to indicate that the overload risk may exist.

It should be noted that, although the conventional immediate alarm based on the threshold value, that is, the alarm is immediately sent out when a certain characteristic value is detected to exceed the preset threshold value, the implementation is simple, but if the threshold value is set unreasonably, a high false alarm rate may be caused, especially in a noisy environment. And the setting of the threshold value is also susceptible to temperature. For example, for a low-temperature environment and a high-temperature environment, the heat dissipation of the device is faster in the low-temperature environment, so that the temperature of the device increases slowly, the stability of the temperature of the device also makes the temperature change less interfere with the output current, and in this case, a temperature threshold or a current threshold which is slightly higher can be set in combination with the critical value of the temperature or the current. In contrast, heat dissipation of the device in a high-temperature environment is affected, so that the temperature of the device is easily too high, accurate control of output current is affected, the temperature of the device increases faster, the output current is greatly disturbed due to instability of the temperature, and in order to protect the device, a temperature threshold or a current threshold which is slightly lower is set by combining a temperature or a current critical value, and response to an abnormal condition is timely performed. Under the condition, the early warning mode based on time prediction can be adopted to better realize early warning.

In this embodiment, the preset observation time period is a preset pre-monitoring time period for executing an alarm, which may also be understood as an alarm at a certain time before an abnormal event occurs, for example, at a time of 30 seconds before an overload current occurs, so as to facilitate timely making an improvement policy to ensure the operation of the device temperature. The early warning based on time prediction can provide enough reaction time and reduce false alarms. After determining the abnormality detection node, an observation reference node for observing an abnormality event that may arrive in advance may be determined based on a preset observation period.

Taking the observation reference node as the center of a preset sliding window, and extracting the observation characteristic of the observation reference node based on operation reference data in the preset sliding window;

In this embodiment, the observation feature is specifically a feature value of data in a preset sliding window with an observation reference node as a center, the preset sliding window is applied to the observation reference node, and feature values such as a mean value, a standard deviation, a slope, a peak value and the like of parameters such as current and temperature are extracted, and the feature reflects an operation state of the power adapter at the moment.

Extracting fluctuation reference data from operation reference data based on a preset duration range and an abnormality detection node, extracting reference fluctuation characteristics from the fluctuation reference data, correcting observation characteristics of an observation reference node based on the reference fluctuation characteristics to obtain target characteristics, and determining an overload monitoring threshold according to the target characteristics.

In this embodiment, considering the influence of data fluctuation caused by factors such as noise data, fluctuation reference data is extracted from operation reference data, the fluctuation reference data is used for representing reference data in a period in which no abnormality occurs, and noise information contained in the fluctuation reference data can be used for correcting observation characteristics, so that interference caused by noise is reduced, and an overload monitoring threshold is determined according to target characteristics obtained by correction.

Extracting reference fluctuation characteristics from fluctuation reference data, correcting the observation characteristics of an observation reference node based on the reference fluctuation characteristics to obtain target characteristics, and determining an overload monitoring threshold according to the target characteristics, wherein the method specifically comprises the following steps:

Sliding window analysis is carried out on the fluctuation reference data based on a preset sliding window and a preset sliding step length, a plurality of wavelet features corresponding to each preset sliding window are extracted, feature average values of each type of wavelet features are obtained, and reference fluctuation features corresponding to the fluctuation reference data are obtained;

in this embodiment, the sliding window analysis process includes, but is not limited to, calculating a mean value of parameters such as current and temperature in the sliding window, and then obtaining a mean value among a plurality of sub-waveform features corresponding to each preset sliding window to form a reference waveform feature.

And performing differential analysis on the observation characteristics of the observation reference nodes based on the reference fluctuation characteristics to obtain target characteristics after the observation characteristics of the observation reference nodes are corrected, and taking a plurality of characteristic values in the target characteristics as overload monitoring thresholds.

In this embodiment, the differential analysis on the observation feature of the observation reference node based on the reference fluctuation feature may specifically be to calculate a difference value of each feature parameter between the observation feature of the observation reference node and the reference fluctuation feature, so as to eliminate deviation caused by data fluctuation, obtain a corrected target feature, and then use a plurality of feature values in the target feature as an overload monitoring threshold, where the target feature includes a temperature feature, a current feature and a voltage feature, and obtain overload monitoring thresholds corresponding to the temperature, the voltage and the current respectively.

As an exemplary implementation procedure, in step S6, current overload monitoring is performed on the power adapter according to an overload monitoring threshold, which specifically includes:

For the current work monitoring data of the power adapter obtained by monitoring, calculating to obtain an abnormal value corresponding to the k moment in the work monitoring data by adopting the following formula:

；

In the formula, For the outlier corresponding to time k in the operation monitoring data,For the k time point in the work monitoring dataThe value of the parameter of the item,Is the firstThe overload monitoring threshold value of the item parameter,Is the firstNormalized weight parameters of the term parameters,Monitoring the number of parameters in the data for the job;

And carrying out current overload early warning after detecting that the abnormal value corresponding to any moment in the current work monitoring data of the power adapter is larger than a preset abnormal threshold value.

It is worth to describe that, in the above scheme, a plurality of monitoring parameters (such as current, voltage, temperature, etc.) are comprehensively calculated, so that the running state of the power adapter can be comprehensively reflected, the contribution of each parameter to the abnormal value is considered, the monitoring blind area possibly brought by a single parameter is avoided, and the comprehensiveness and accuracy of monitoring are improved.

According to the current overload monitoring method based on the power adapter, accurate monitoring and early warning of current overload of the power adapter are achieved through comprehensive multi-parameter real-time monitoring and predictive analysis. Specifically, the invention comprehensively reflects the running state of the power adapter by monitoring and analyzing a plurality of parameters such as current, voltage, temperature and the like of the power adapter in real time. The characteristic data is extracted by adopting a sliding window technology and differential analysis, and the precursor characteristics before current overload can be accurately identified by comparing the data changes before and after the abnormal event, so that dead zones caused by single parameter monitoring are avoided, and the comprehensiveness and accuracy of monitoring are improved; the invention combines an instant alarm mechanism with a time prediction based early warning method, predicts data such as current, voltage and temperature in a period of time in the future by training an operation parameter prediction model, and sends early warning in advance before the predicted abnormal node, thereby improving the timeliness of early warning, effectively reducing false alarms caused by instant fluctuation or environmental noise and improving the accuracy and reliability of early warning.

Fig. 2 is a schematic structural diagram of a current overload monitoring system based on a power adapter according to an embodiment of the present invention, referring to fig. 2, the system includes:

the historical data acquisition module is used for acquiring historical working data of the power adapter, determining a plurality of abnormal events in the historical working data according to a preset monitoring threshold value, and determining a first target period and a second target period according to the event occurrence time of the abnormal events and a preset duration range;

the feature extraction module is used for extracting first monitoring data corresponding to a first target period of each abnormal event and second monitoring data corresponding to a second target period, and associating the first monitoring data and the second monitoring data corresponding to each abnormal event to obtain reference feature data of each abnormal event;

The model training module is used for carrying out sliding window analysis and difference analysis on multiple groups of reference feature data to obtain reference difference features of each group of reference feature data, constructing a training data set based on the reference difference features of each group of reference feature data, and training the training data set to obtain an operation parameter prediction model;

carrying out sliding window analysis and difference analysis on multiple groups of target feature data to obtain reference difference features of each group of reference feature data, wherein the method comprises the following steps:

Sliding window analysis is carried out on multiple groups of first monitoring data and second monitoring data based on a preset sliding window and a preset sliding step length, multiple characteristic values in each preset sliding window are extracted, and a first characteristic matrix corresponding to each group of first monitoring data and a second characteristic matrix corresponding to each group of second monitoring data are constructed according to the multiple characteristic values corresponding to each preset sliding window;

And calculating the difference value and the change rate of the first feature matrix and the second feature matrix corresponding to each group of reference feature data on each feature to obtain the reference difference feature of each group of reference feature data, wherein the reference difference feature comprises the difference feature matrix corresponding to each group of reference feature data.

The data prediction module is used for collecting current working monitoring data of the power adapter, screening target operation data from the working monitoring data based on a preset observation threshold value, carrying out sliding window analysis and difference analysis on the target operation data to generate target difference characteristics, and inputting the target difference characteristic data into the operation parameter prediction model to obtain predicted operation data;

the method for screening the target operation data from the operation monitoring data based on the preset observation threshold value comprises the following steps:

Sliding window analysis is conducted on the working monitoring data based on preset sliding windows and preset sliding step sizes, current average values and temperature average values in each preset sliding window are calculated, a first current observation node is determined according to the preset sliding window corresponding to the current average value closest to the first current observation value, and a first temperature observation node is determined according to the preset sliding window corresponding to the temperature average value closest to the first temperature observation value;

And selecting an intermediate value of the first current observation node and the first temperature observation node as a first target observation node, and screening target operation data from the operation monitoring data according to a preset duration range and the first target observation node.

And the current overload monitoring module is used for splicing the target characteristic data with the predicted operation data to obtain operation reference data, analyzing the operation reference data, determining an observation reference node of the operation reference data, extracting the observation characteristic of the observation reference node to determine an overload monitoring threshold value, and carrying out current overload monitoring on the power adapter according to the overload monitoring threshold value to obtain a current overload monitoring result of the power adapter.

The foregoing is merely exemplary of embodiments of the present invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art. Parts of the specification not described in detail belong to the prior art known to those skilled in the art.

Claims (8)

1. A power adapter-based current overload monitoring method, comprising:

acquiring historical working data of a power adapter, determining a plurality of abnormal events in the historical working data according to a preset monitoring threshold, and determining a first target period and a second target period according to event occurrence time of the abnormal events and a preset duration range;

Extracting first monitoring data corresponding to a first target period of each abnormal event and second monitoring data corresponding to a second target period, and associating the first monitoring data and the second monitoring data corresponding to each abnormal event to obtain reference characteristic data of each abnormal event;

Carrying out sliding window analysis and difference analysis on multiple groups of reference feature data to obtain reference difference features of each group of reference feature data, constructing a training data set based on the reference difference features of each group of reference feature data, and training through the training data set to obtain an operation parameter prediction model;

collecting current working monitoring data of a power adapter, screening target operation data from the working monitoring data based on a preset observation threshold value, carrying out sliding window analysis and difference analysis on the target operation data to generate target difference characteristics, and inputting the target difference characteristic data into an operation parameter prediction model to obtain predicted operation data;

splicing the target characteristic data with the predicted operation data to obtain operation reference data, analyzing the operation reference data, determining an observation reference node of the operation reference data, and extracting observation characteristics of the observation reference node to determine an overload monitoring threshold;

and carrying out current overload monitoring on the power adapter according to the overload monitoring threshold value to obtain a current overload monitoring result of the power adapter.

2. The power adapter-based current overload monitoring method of claim 1, wherein performing sliding window analysis and difference analysis on multiple sets of target feature data to obtain reference difference features of each set of reference feature data comprises:

Sliding window analysis is carried out on multiple groups of first monitoring data and second monitoring data based on a preset sliding window and a preset sliding step length, multiple characteristic values in each preset sliding window are extracted, and a first characteristic matrix corresponding to each group of first monitoring data and a second characteristic matrix corresponding to each group of second monitoring data are constructed according to the multiple characteristic values corresponding to each preset sliding window;

And calculating the difference value and the change rate of the first feature matrix and the second feature matrix corresponding to each group of reference feature data on each feature to obtain the reference difference feature of each group of reference feature data, wherein the reference difference feature comprises the difference feature matrix corresponding to each group of reference feature data.

3. The power adapter-based current overload monitoring method of claim 2, wherein screening the target operational data from the operational monitoring data based on a preset observation threshold comprises:

Sliding window analysis is conducted on the working monitoring data based on preset sliding windows and preset sliding step sizes, current average values and temperature average values in each preset sliding window are calculated, a first current observation node is determined according to the preset sliding window corresponding to the current average value closest to the first current observation value, and a first temperature observation node is determined according to the preset sliding window corresponding to the temperature average value closest to the first temperature observation value;

And selecting an intermediate value of the first current observation node and the first temperature observation node as a first target observation node, and screening target operation data from the operation monitoring data according to a preset duration range and the first target observation node.

4. A power adapter-based current overload monitoring method as claimed in claim 3 wherein the splicing of the target feature data with the predicted operational data to obtain operational reference data, analyzing the operational reference data to determine an observed reference node of the operational reference data, extracting the observed features of the observed reference node to determine an overload monitoring threshold comprises:

Judging whether an abnormal detection node exists in the operation reference data according to a preset monitoring threshold value, wherein if the current value at any moment exists in the operation reference data and is larger than the preset monitoring threshold value, the abnormal detection node exists in the operation reference data;

Selecting a time node with a current value larger than a preset monitoring threshold value in the operation reference data as an abnormality detection node, and determining an observation reference node of the operation reference data based on a preset observation time length and the abnormality detection node;

Taking the observation reference node as the center of a preset sliding window, and extracting the observation characteristic of the observation reference node based on operation reference data in the preset sliding window;

Extracting fluctuation reference data from operation reference data based on a preset duration range and an abnormality detection node, extracting reference fluctuation characteristics from the fluctuation reference data, correcting observation characteristics of an observation reference node based on the reference fluctuation characteristics to obtain target characteristics, and determining an overload monitoring threshold according to the target characteristics.

5. The power adapter-based current overload monitoring method of claim 4, wherein the extracting the reference fluctuation feature from the fluctuation reference data, correcting the observation feature of the observation reference node based on the reference fluctuation feature to obtain the target feature, and determining the overload monitoring threshold according to the target feature comprises:

Sliding window analysis is carried out on the fluctuation reference data based on a preset sliding window and a preset sliding step length, a plurality of wavelet features corresponding to each preset sliding window are extracted, feature average values of each type of wavelet features are obtained, and reference fluctuation features corresponding to the fluctuation reference data are obtained;

And performing differential analysis on the observation characteristics of the observation reference nodes based on the reference fluctuation characteristics to obtain target characteristics after the observation characteristics of the observation reference nodes are corrected, and taking a plurality of characteristic values in the target characteristics as overload monitoring thresholds.

6. The power adapter-based current overload monitoring method of claim 5, wherein the current overload monitoring of the power adapter according to the overload monitoring threshold comprises:

For the current work monitoring data of the power adapter obtained by monitoring, calculating to obtain an abnormal value corresponding to the k moment in the work monitoring data by adopting the following formula:

；

In the formula, For the outlier corresponding to time k in the operation monitoring data,For the k time point in the work monitoring dataThe value of the parameter of the item,Is the firstThe overload monitoring threshold value of the item parameter,Is the firstNormalized weight parameters of the term parameters,Monitoring the number of parameters in the data for the job;

And carrying out current overload early warning after detecting that the abnormal value corresponding to any moment in the current work monitoring data of the power adapter is larger than a preset abnormal threshold value.

7. The power adapter-based current overload monitoring method of claim 1, wherein the operating parameter prediction model is a long-term and short-term memory network model.

8. A power adapter based current overload monitoring system, characterized in that the system is adapted to implement the power adapter based current overload monitoring method of any one of claims 1-7, comprising:

the historical data acquisition module is used for acquiring historical working data of the power adapter, determining a plurality of abnormal events in the historical working data according to a preset monitoring threshold value, and determining a first target period and a second target period according to the event occurrence time of the abnormal events and a preset duration range;

the feature extraction module is used for extracting first monitoring data corresponding to a first target period of each abnormal event and second monitoring data corresponding to a second target period, and associating the first monitoring data and the second monitoring data corresponding to each abnormal event to obtain reference feature data of each abnormal event;

The model training module is used for carrying out sliding window analysis and difference analysis on multiple groups of reference feature data to obtain reference difference features of each group of reference feature data, constructing a training data set based on the reference difference features of each group of reference feature data, and training the training data set to obtain an operation parameter prediction model;

the data prediction module is used for collecting current working monitoring data of the power adapter, screening target operation data from the working monitoring data based on a preset observation threshold value, carrying out sliding window analysis and difference analysis on the target operation data to generate target difference characteristics, and inputting the target difference characteristic data into the operation parameter prediction model to obtain predicted operation data;

And the current overload monitoring module is used for splicing the target characteristic data with the predicted operation data to obtain operation reference data, analyzing the operation reference data, determining an observation reference node of the operation reference data, extracting the observation characteristic of the observation reference node to determine an overload monitoring threshold value, and carrying out current overload monitoring on the power adapter according to the overload monitoring threshold value to obtain a current overload monitoring result of the power adapter.