CN-120596819-B - Comprehensive power supply testing method and system for integrated fault prediction

CN120596819BCN 120596819 BCN120596819 BCN 120596819BCN-120596819-B

Abstract

The invention discloses a comprehensive power supply testing method and system for integrated fault prediction, and relates to the technical field of power supply testing. A comprehensive power supply test system integrating fault prediction comprises a test parameter acquisition module, a reference state updating module, a data fusion module, a fault prediction module and a retest recording module. According to the invention, the model paths are flexibly selected by setting the first prediction channel and the second prediction channel and combining initial abnormal judgment results, the prediction accuracy is improved while the prediction efficiency is ensured, the operation requirements under different complexity scenes are met, and the intelligent verification and self-adaptive adjustment of the consistency of the model results and the stability and the robustness of the fault judgment process are enhanced by setting the circulation analysis mechanism and the dynamic circulation limit value and the weight increasing operation based on the maximum influence data item.

Inventors

WU ZHIGANG
ZHANG QI
GAN TAO
LIANG JUN
XIONG TAO

Assignees

江西智矿自动化技术有限公司

Dates

Publication Date: 20260508
Application Date: 20250512

Claims (7)

1. The integrated fault prediction comprehensive power supply testing method is characterized by comprising the following steps of: acquiring multi-mode data of power equipment in a power supply test process, wherein the multi-mode data comprises electrical parameter information, running environment information, thermal imaging information and operation log information; Constructing a reference state transition diagram corresponding to the power supply equipment according to the historical multi-mode data, wherein the reference state transition diagram comprises a fault evolution path from a normal state to an abnormal trend to a fault state, and a fault knowledge map corresponding to a fault type-parameter representation-processing suggestion relation of each standard fault case; Calculating the stable fluctuation period of each item of data in the electrical parameter information, taking the weighted average value of the stable fluctuation period of each item of data as a fusion period, acquiring all multi-mode data in the fusion period, carrying out data fusion to obtain a period fusion characteristic, and acquiring a multi-dimensional health state image of the power supply equipment through characteristic engineering processing; Inputting the period fusion characteristics into a fault prediction model for fault prediction, wherein the fault prediction model selects a prediction channel according to an initial abnormality judgment result, the prediction channel comprises two independent prediction channels, the first prediction channel is used for carrying out rapid analysis based on a lightweight model, the second prediction channel is used for carrying out deep analysis based on a convolutional neural network, and the result of the prediction channel is used for carrying out circulation analysis in the fault prediction model to obtain an initial measurement result; Matching a fault knowledge graph according to the initial test result, acquiring a re-inspection weight according to the corresponding parameter expression, testing the power supply equipment again, weighting the periodic fusion characteristic by using the re-inspection weight to obtain a re-inspected periodic fusion characteristic, carrying out fault prediction by a fault prediction model, and determining whether to output a correction result or a final result according to the number of times of circulation analysis exceeds a circulation limit value; if the output result is a correction result, matching the fault knowledge graph by using the correction result, acquiring a re-inspection weight, and testing again until a final result is acquired or the number of re-tests exceeds a set threshold; The re-verification weights are fused into personalized equipment test samples after weighting the cycle fusion characteristics each time, wherein the personalized equipment test samples are used for weighting in the data fusion process; the method comprises the steps of carrying out cluster analysis on a personalized equipment test sample, and constructing personalized equipment characteristics in a twin digital model of equipment according to a cluster analysis result, wherein the personalized equipment characteristics are used for correcting a fault evolution path.
2. The integrated fault prediction integrated power supply testing method according to claim 1, wherein the fault prediction model uses a first prediction channel for analysis when the initial abnormality judgment result is normal; If the analysis result of the prediction channel is consistent with the initial abnormal judgment result, outputting the analysis result as a primary measurement result, otherwise, performing flow analysis, namely switching the prediction channel for analysis, and if the analysis result of the prediction channel after flow analysis is inconsistent with the analysis result of the prediction channel before flow analysis, continuing to perform flow analysis until the result is consistent or the frequency of flow analysis exceeds a flow limit value, and outputting the primary measurement result.
3. The integrated fault prediction comprehensive power supply testing method according to claim 2, wherein the circulation analysis is characterized in that before the confirmation is carried out, the fault prediction model increases the weight of the input period fusion feature according to the maximum influence data item corresponding to the result of the prediction channel, namely increases the set weight value of the maximum influence data item in the corresponding part of the period fusion feature.
4. The integrated fault-predicted comprehensive power supply testing method according to claim 2, wherein the circulation limit value is a dynamic value which is dynamically adjusted by setting a mathematical function by combining an initial set value with the number of retests as an independent variable.
5. The integrated fault-predicted integrated power test method of claim 1 wherein the data fusion comprises: In the fusion period, respectively extracting the characteristics of the electrical parameter information, the running environment information, the thermal imaging information and the operation log information to form a structured characteristic set; giving initial fusion weight to each feature by a feature importance evaluation method, fusing a structural feature set to generate a high-dimensional fusion vector with a unified format, and performing dimension reduction processing by using a t-SNE method to obtain a final periodic fusion feature; and performing similarity matching on the periodic fusion characteristics and the equipment characteristics in the personalized equipment test sample, and dynamically adjusting fusion weights according to similarity threshold values to realize self-adaptive characteristic reconstruction aiming at different equipment states.
6. The integrated fault prediction integrated power supply testing method as claimed in claim 1, wherein said initial anomaly determination comprises: Constructing a time sequence feature track diagram based on multi-mode data, performing trend fitting on electrical parameter information and operation environment information by utilizing a sliding window mechanism, and extracting statistical features; The statistical characteristics are subjected to discriminant analysis by using an abnormal trend detection model, wherein the abnormal trend detection model is an unsupervised abnormal detection model constructed based on an isolated forest algorithm and is used for identifying data fragments which have not triggered a hard fault threshold but have potential abnormal trends; and judging and outputting an initial abnormality judgment result according to the hard fault threshold and the results of the abnormality trend detection model, wherein the judgment mode is that when only 2 results are normal, the initial abnormality judgment result is output as normal, and otherwise, the initial abnormality judgment result is abnormal.
7. An integrated fault prediction integrated power supply testing system, characterized in that the system applies the method of any one of claims 1 to 6, comprising: The test parameter acquisition module is used for acquiring multi-mode data of the power supply equipment in the power supply test process; The reference state updating module is used for acquiring and updating a reference state transition diagram corresponding to the power supply equipment; The data fusion module is used for fusing all the multi-mode data in the fusion period and carrying out data fusion; the fault prediction module is used for calculating the period fusion data according to the fault prediction model to obtain a primary test result, a correction result and a final test result of the fault prediction; The retest recording module is used for acquiring and recording retest weight, personalized equipment test sample and personalized equipment characteristics in the retest process of the power supply equipment.

Description

Comprehensive power supply testing method and system for integrated fault prediction Technical Field The invention relates to the technical field of power supply testing, in particular to a comprehensive power supply testing method and system for integrated fault prediction. Background Power supply devices are used as a key infrastructure in modern industrial and information systems, and their operating status is directly related to the stability and safety of the system. With the rapid development of intelligent manufacturing, data centers, and communication equipment, power supply systems face higher operating loads and stability requirements. However, the conventional power supply test and fault diagnosis method relies on regular inspection and rule-based abnormality judgment, so that it is difficult to find potential fault trends in time or capture multifactor abnormalities in complex scenes. The existing power supply testing method mainly focuses on measurement of electrical parameters, such as real-time monitoring of indexes of voltage, current, frequency and the like, and part of the system introduces auxiliary means such as infrared thermal imaging and the like for image identification after faults occur. However, the method has the technical defects of single data dimension, lack of deep fusion analysis, lack of effective prediction capability, weak model generalization capability, poor individual equipment difference adaptability, knowledge accumulation fracture and lack of closed loop learning mechanism. Therefore, a power supply test system and method capable of integrating multi-mode data, having self-adaptive capability and fault prediction capability and supporting knowledge accumulation and dynamic update are needed to improve the recognition and response capability of complex faults and realize the transition from passive response to active prediction. Disclosure of Invention A comprehensive power supply testing method for integrated fault prediction comprises the following steps: acquiring multi-mode data of power equipment in a power supply test process, wherein the multi-mode data comprises electrical parameter information, running environment information, thermal imaging information and operation log information; Constructing a reference state transition diagram corresponding to the power supply equipment according to the historical multi-mode data, wherein the reference state transition diagram comprises a fault evolution path from a normal state to an abnormal trend to a fault state, and a fault knowledge map corresponding to a fault type-parameter representation-processing suggestion relation of each standard fault case; Calculating the stable fluctuation period of each item of data in the electrical parameter information, taking the weighted average value of the stable fluctuation period of each item of data as a fusion period, acquiring all multi-mode data in the fusion period, carrying out data fusion to obtain a period fusion characteristic, and acquiring a multi-dimensional health state image of the power supply equipment through characteristic engineering processing; The method comprises the steps that a period fusion characteristic is input into a fault prediction model for fault prediction, the fault prediction model selects a prediction channel according to an initial abnormality judgment result, the prediction channel comprises two independent prediction channels, a first prediction channel is used for carrying out rapid analysis based on a lightweight model, a second prediction channel is used for carrying out deep analysis based on a convolutional neural network, and a result of the prediction channel is used for carrying out circulation analysis in the fault prediction model to obtain an initial measurement result. As a preferred technical scheme of the invention, the comprehensive power supply testing method for integrated fault prediction further comprises the following steps: Matching a fault knowledge graph according to the initial test result, acquiring a re-inspection weight according to the corresponding parameter expression, testing the power supply equipment again, weighting the periodic fusion characteristic by using the re-inspection weight to obtain a re-inspected periodic fusion characteristic, carrying out fault prediction by a fault prediction model, and determining whether to output a correction result or a final result according to the number of times of circulation analysis exceeds a circulation limit value; if the output result is a correction result, matching the fault knowledge graph by using the correction result, acquiring a re-inspection weight, and testing again until a final result is acquired or the number of re-tests exceeds a set threshold; The re-verification weights are fused into personalized equipment test samples after weighting the cycle fusion characteristics each time, wherein the personalized equipment test samples are used for weigh