CN-122017718-A - Online health monitoring method and system for low-voltage transformer based on terminal side AI

CN122017718ACN 122017718 ACN122017718 ACN 122017718ACN-122017718-A

Abstract

The invention discloses an online health monitoring method and system for a low-voltage transformer based on an end side AI, which are used for realizing real-time, continuous and online autonomous monitoring of the running state of the transformer without continuous intervention of an external upper computer by performing steps of multi-period electric signal acquisition, end side AI reasoning, sliding window anti-shake post-processing, fault output and the like on a secondary side embedded section side processing unit of the low-voltage transformer, so that the traditional low-voltage transformer has independent health state sensing and evaluating capability.

Inventors

HU LEI
LEI JIAN
ZHANG LIANG
SUN LIXIA
WANG QINGLEI

Assignees

湖南智焜能源科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260416

Claims (10)

1. An online health monitoring method of a low-voltage transformer based on an end-side AI is characterized by comprising the following steps: Based on a preset time period, periodically acquiring N power frequency periodic electric signals output by a secondary side of the transformer, and acquiring M sampling points in each period to form an original sampling matrix; Transmitting the original sampling matrix to a preset end side for processing, and determining a health index and a fault confidence; if the health index is smaller than a preset first health index threshold or the fault confidence is larger than a preset first fault confidence threshold, judging that health abnormality exists at the current moment, and marking the acquired data of the current period as preliminary abnormality; based on a preset sliding time window, extracting the number of preliminary abnormal marks in the corresponding sliding time window; If the number of the preliminary abnormal marks in the sliding time window exceeds a preset anti-shake threshold, generating a final abnormal mark and triggering health warning information.
2. The method for online health monitoring of a low-voltage transformer based on an end-side AI according to claim 1, wherein the step of sending the original sampling matrix to a preset end-side for processing to determine a health index and a fault confidence specifically comprises: respectively carrying out normalization processing and waveform alignment on each period of an original sampling matrix to obtain a normalized waveform matrix and extracting characteristic parameters of each period of the normalized waveform matrix, wherein the characteristic parameters comprise an effective value and a harmonic amplitude; extracting odd harmonic amplitude values in the harmonic amplitude values, and determining the maximum value of the odd harmonic amplitude ratio according to the odd harmonic amplitude values; Determining the waveform quality index according to the maximum value of the odd harmonic amplitude ratio and the maximum value of the waveform distortion rate The formula is as follows: wherein 、 The maximum value of the odd harmonic amplitude ratio and the maximum value of the waveform distortion rate are respectively represented; 、 Health reference values respectively representing the odd harmonic amplitude ratio and the waveform distortion rate; Representing the corresponding weight coefficient; extracting the harmonic wave of the current period and the harmonic wave of the adjacent previous period; comparing and analyzing the harmonic wave of the current period and the harmonic wave of the adjacent previous period, determining the waveform similarity value of the current period, and determining a waveform similarity value set after traversing the harmonic waves of all periods; Determining a period stability index of the current harmonic according to the effective value and the waveform similarity value set; Based on a preset algorithm, determining a health index and a fault confidence according to the period stability index and the waveform quality index of the current harmonic.
3. The method for online health monitoring of a low-voltage transformer based on an end-side AI according to claim 2, wherein the step of determining the maximum value of the waveform distortion rate/determining the maximum value of the odd harmonic amplitude ratio specifically comprises: Dividing the harmonic amplitude according to the period, and determining the harmonic amplitude of different periods; let the waveform distortion rate be THD, then the waveform distortion rate corresponding to the ith period be THD (i), the formula is Wherein For the amplitude of the fundamental wave, For the h harmonic amplitude in the i-th period, A harmonic fluctuation weight coefficient representing an i-th period; Extracting odd harmonic amplitude values in the harmonic amplitude values of the same period; setting the odd harmonic amplitude ratio The odd harmonic amplitude corresponding to the ith period is the ratio of The formula is ; After traversing all the periods, a waveform distortion rate set and an odd harmonic amplitude ratio set are obtained; And extracting the maximum values in the waveform distortion rate set and the odd harmonic amplitude ratio set, namely the maximum value of the corresponding waveform distortion rate and the maximum value of the odd harmonic amplitude ratio.
4. The low-voltage transformer on-line health monitoring method based on end-side AI of claim 3, wherein the step of obtaining the harmonic fluctuation weight coefficient specifically comprises the following steps: extracting an amplitude sequence of characteristic harmonic frequencies of an original sampling matrix in each period; Comparing and analyzing the amplitude values of the characteristic harmonic frequencies of the current period and the previous period to determine a cross-correlation vector; If the cross-correlation vector is larger than a preset first similarity threshold, judging that the harmonic fluctuation of the current period is derived from the background harmonic change of the system side, and setting a preset first numerical value as a corresponding harmonic fluctuation weight; if the cross-correlation vector is smaller than or equal to a preset first similarity threshold, judging that the harmonic fluctuation of the current period is derived from the characteristic change of the transformer, and setting a preset second value as a corresponding harmonic fluctuation weight; The preset first value is smaller than the preset second value.
5. The method for online health monitoring of a low-voltage transformer based on an end-side AI according to claim 3, wherein after the waveform distortion rate set and the odd harmonic amplitude ratio set are obtained, the method further comprises: in each power frequency period, extracting a direct current component from an original sampling signal to obtain a direct current component value; Extracting even harmonic amplitude values in harmonic amplitude values of the same period; According to the even harmonic amplitude in the harmonic amplitude of the same period, determining the amplitude ratio of the even harmonic in the corresponding period; if the absolute value of the direct current component value is larger than a preset direct current component threshold value, the amplitude ratio of even harmonic is larger than a preset even harmonic duty ratio threshold value, and the amplitude of second harmonic is larger than the amplitude of third harmonic, judging that the current period is a direct current magnetic biasing period; And deleting the waveform distortion rate and the odd harmonic amplitude ratio in the direct current magnetic bias period.
6. The low-voltage transformer on-line health monitoring method based on end-side AI of claim 2, further comprising: Extracting the phase relation before and after the zero crossing point of the current in each power frequency period; Based on a preset voltage phase reference signal, determining that the current direction of the period is a positive direction or a reverse direction; Respectively constructing a forward direction characteristic buffer zone and a reverse direction characteristic buffer zone based on the current direction, wherein the forward direction characteristic buffer zone and the reverse direction characteristic buffer zone respectively and independently store corresponding characteristic parameters; And calculating the period stability index and the waveform quality index based on the characteristic data of the buffer area corresponding to the direction of the current period.
7. The low-voltage transformer on-line health monitoring method based on end-side AI of claim 2, further comprising: In a preset stable period, if the health index of n continuous periods is greater than a preset second health index threshold and the fault confidence is smaller than a preset second fault confidence threshold, triggering the reference value to update in a buffering way, wherein the updating formula is as follows: wherein The reference value after the update is represented, A reference value before updating, the reference value including a healthy reference value of an odd harmonic amplitude ratio and a healthy reference value of a waveform distortion rate; in order to correspond to the adjustment coefficient, The characteristic parameter of the jth period in the current n periods is set; the preset second health index threshold is larger than the preset first health index threshold, and the preset second fault confidence threshold is smaller than the preset first fault confidence threshold.
8. The low-voltage transformer on-line health monitoring method based on end-side AI of claim 1, further comprising: monitoring the mean value and the variance of waveform quality indexes in a window period of a preset first number in real time, wherein the mean value and the variance are respectively set as a first mean value and a first variance; If the first average value exceeds a preset first threshold value and the first difference is smaller than a preset stability threshold value, judging that the current environment is in a continuous high-harmonic state, and generating the length information of the extended sliding time window; correcting the length of the sliding time window based on a preset delay value to obtain a corrected sliding time window; And adjusting the anti-shake threshold based on the sliding time window after correction The formula is as follows: wherein And (3) indicating the sliding time window length after correction, wherein L indicates the preset sliding time window length, and K indicates the preset anti-shake threshold.
9. The low-voltage transformer on-line health monitoring method based on end-side AI of claim 8, further comprising: After the sliding time window after correction is used, extracting the average value of waveform quality indexes in a preset second number of window periods, and setting the average value as a second average value; If the second average value is smaller than a preset second threshold value, the length of the next sliding time window is restored to the length of the preset sliding time window; the preset second number is larger than the preset first number, and the preset second threshold value is smaller than the preset first threshold value.
10. The low-voltage transformer on-line health monitoring system based on the terminal side AI is characterized by further comprising a low-voltage current/voltage transformer, a signal conditioning circuit, an ADC acquisition unit, an MCU main control unit, a storage unit and a communication unit; The MCU main control unit is respectively connected with the ADC acquisition unit, the storage unit and the communication unit, the output end of the signal conditioning circuit is connected with the ADC acquisition unit, the output end of the low-voltage transformer is connected with the signal conditioning circuit, and the MCU main control unit is used for executing the low-voltage transformer on-line health monitoring method based on the terminal side AI according to any one of claims 1-9.

Description

Online health monitoring method and system for low-voltage transformer based on terminal side AI Technical Field The invention relates to the technical field of power equipment monitoring, in particular to an online health monitoring method and system for a low-voltage transformer based on an end-side AI. Background The low-voltage transformer is core sensing equipment for metering, load monitoring and fault protection in a power distribution system, and is widely applied to key scenes such as metering boxes, intelligent measuring switches, low-voltage power distribution loops and the like. The traditional low-voltage transformer can only complete basic electric signal acquisition and transmission work, is equivalent to 'eyes' of a power distribution system but has no 'self-perception' capability, and cannot carry out self-diagnosis on the running state of the low-voltage transformer. In the long-term operation process, various faults such as coil turn-to-turn short circuit, wire breakage, iron core saturation, aging, loose wiring, poor contact, zero drift, linearity abnormality, harmonic distortion, metering misalignment and the like of the transformer are very easy to occur, and the faults can directly lead to distribution metering data distortion, misoperation of a protection device and failure of monitoring data, so that the stable and reliable operation of a low-voltage distribution system is seriously influenced. At present, the current industry monitors the state of the transformer mainly by means of manual regular inspection and offline verification, so that a great amount of manpower and material resources are consumed, the running state of the transformer cannot be diagnosed on line, in real time and at the end side intelligently, faults can be found after occurrence, and obvious monitoring hysteresis exists. Disclosure of Invention In order to solve at least one technical problem, the invention aims to provide an online health monitoring method and system for a low-voltage transformer based on an end side AI, which can remarkably improve the speed and accuracy of fault discovery and reduce the labor cost of operation and maintenance. The first aspect of the invention provides an online health monitoring method of a low-voltage transformer based on an end side AI, which comprises the following steps: Based on a preset time period, periodically acquiring N power frequency periodic electric signals output by a secondary side of the transformer, and acquiring M sampling points in each period to form an original sampling matrix; Transmitting the original sampling matrix to a preset end side for processing, and determining a health index and a fault confidence; if the health index is smaller than a preset first health index threshold or the fault confidence is larger than a preset first fault confidence threshold, judging that health abnormality exists at the current moment, and marking the acquired data of the current period as preliminary abnormality; based on a preset sliding time window, extracting the number of preliminary abnormal marks in the corresponding sliding time window; If the number of the preliminary abnormal marks in the sliding time window exceeds a preset anti-shake threshold, generating a final abnormal mark and triggering health warning information. In this scheme, the step of sending the original sampling matrix to the preset end side for processing, and determining the health index and the fault confidence level specifically includes: respectively carrying out normalization processing and waveform alignment on each period of an original sampling matrix to obtain a normalized waveform matrix and extracting characteristic parameters of each period of the normalized waveform matrix, wherein the characteristic parameters comprise an effective value and a harmonic amplitude; extracting odd harmonic amplitude values in the harmonic amplitude values, and determining the maximum value of the odd harmonic amplitude ratio according to the odd harmonic amplitude values; Determining the waveform quality index according to the maximum value of the odd harmonic amplitude ratio and the maximum value of the waveform distortion rate The formula is as follows: wherein 、The maximum value of the odd harmonic amplitude ratio and the maximum value of the waveform distortion rate are respectively represented;、 Health reference values respectively representing the odd harmonic amplitude ratio and the waveform distortion rate; Representing the corresponding weight coefficient; extracting the harmonic wave of the current period and the harmonic wave of the adjacent previous period; comparing and analyzing the harmonic wave of the current period and the harmonic wave of the adjacent previous period, determining the waveform similarity value of the current period, and determining a waveform similarity value set after traversing the harmonic waves of all periods; Determining a period stability index of the