CN-121982605-A - Multi-parameter real-time monitoring and abnormality diagnosis method and device for high-voltage direct-current transmission system

CN121982605ACN 121982605 ACN121982605 ACN 121982605ACN-121982605-A

Abstract

The invention relates to the technical field of automation and intelligent operation and maintenance of a power system, and discloses a multi-parameter real-time monitoring and abnormality diagnosis method and device of a high-voltage direct-current power transmission system, wherein the method comprises the steps of acquiring monitoring interface image frames according to a preset period, and generating semantic state vectors containing color zone bits and working condition mode codes by utilizing a semantic analysis module; the numerical value processing module combines the character confidence coefficient and the semantic state vector to determine a state self-adaptive filter coefficient to execute self-adaptive state gating filtering on the original numerical value, and the anomaly analysis module executes energy conservation calculation and bipolar gradient analysis based on the smooth monitoring value to generate an alarm signal. The invention effectively solves the contradiction between non-invasive acquisition steady-state jitter and transient delay by fusing semantic perception and self-adaptive filtering, and improves the diagnosis reliability of system logic conflict and dynamic asynchronous exception by utilizing physical mechanism constraint.

Inventors

ZHANG ZICONG
SHEN QIYAO
CUI MENG
LI LIKUN
LI SHAOSEN
ZHANG ZHE
HUANG JIANXIANG
LI HAO
LIU CHAO
ZHAO SHIWEI
WU XINWEN
HE ZHAONENG
WANG JIALEI
Fu tianyi

Assignees

中国南方电网有限责任公司超高压输电公司昆明局

Dates

Publication Date: 20260505
Application Date: 20260119

Claims (10)

1. The multi-parameter real-time monitoring and abnormality diagnosis method for the high-voltage direct-current transmission system is characterized by comprising the following steps of: triggering a screen image intercepting instruction according to a preset sampling period to acquire a monitoring interface image frame at the current moment; Transmitting the monitoring interface image frame to a semantic analysis module, and extracting multi-mode features of semantic state areas divided in advance in the monitoring interface image frame to generate semantic state vectors containing color state zone bits and discrete working condition mode codes; Transmitting the monitoring interface image frame to a numerical value processing module, and executing optical character recognition on numerical value data areas divided in advance in the monitoring interface image frame to acquire an original numerical value and a character confidence coefficient; the numerical value processing module determines a state self-adaptive filter coefficient according to the semantic state vector, and performs self-adaptive state gating filtering on the original numerical value by utilizing the state self-adaptive filter coefficient to output a smooth monitoring value; and the anomaly analysis module receives the smooth monitoring value, performs feature calculation and multi-strategy anomaly judgment based on a physical mechanism by combining the semantic state vector, and generates an anomaly alarm signal.
2. The method for monitoring and diagnosing abnormality of a hvdc transmission system according to claim 1, wherein the step of extracting multi-modal features comprises: Converting pixel data within a visual indication sub-region in the semantic state region from RGB color space to HSV color space; Calculating the pixel ratio of the number of pixels in the visual indication subarea, which fall in a preset target color tone interval, to the total number of pixels in the visual indication subarea; When the pixel duty ratio is larger than a preset color judgment threshold value, the color state mark position is set to be a first numerical value, otherwise, the color state mark position is set to be a second numerical value; Executing optical character recognition processing on the text state subareas in the semantic state area to acquire text character strings; and mapping the text character string into the discrete working condition mode code through a preset working condition keyword library.
3. The method for multi-parameter real-time monitoring and anomaly diagnosis of a hvdc transmission system according to claim 1, further comprising the step of calculating weighted observations prior to performing the adaptive state gating filtering: processing the character confidence coefficient by using a confidence coefficient weight function, setting the confidence coefficient weight to zero when the character confidence coefficient is lower than a preset effective confidence coefficient lower limit threshold value, otherwise setting the confidence coefficient weight to the character confidence coefficient; maintaining a short-time sampling buffer queue, and calculating the sum of products of original values of historical moments and corresponding confidence weights in the short-time sampling buffer queue as a molecule; calculating the sum of the corresponding confidence weights and the sum of the minimum positive numbers in the short-time sampling buffer queue as a denominator; And calculating the ratio of the numerator to the denominator to obtain a weighted observation value.
4. The method for multi-parameter real-time monitoring and anomaly diagnosis of a hvdc transmission system according to claim 3, wherein the step of adaptive state gating filtering comprises: calculating the product of the state adaptive filter coefficient and the weighted observation value to obtain a first partial value; Calculating a difference value of subtracting the state adaptive filter coefficient from the first value, and calculating a product of the difference value and a filter output value at the last moment to obtain a second partial value; Calculating the sum of the first partial value and the second partial value to be used as a current moment smooth monitoring value; the step of determining the state adaptive filter coefficients comprises: when the color state flag bit indicates an operation state and the discrete working condition mode code indicates a steady-state operation mode, selecting a locking filter coefficient with a numerical range between zero and five to zero as the state self-adaptive filter coefficient; When the discrete working condition mode code indicates a transient adjustment mode, selecting a tracking filter coefficient with a numerical range between zero point six and zero point nine as the state self-adaptive filter coefficient; And when the confidence coefficient weight is zero, selecting a holding filter coefficient with a value of zero as the state adaptive filter coefficient.
5. The method for monitoring and diagnosing abnormality of a hvdc transmission system according to claim 1, wherein the step of calculating characteristics based on physical mechanism comprises: calculating the total power of the direct current side by using the direct current voltage and the direct current in the smooth monitoring value; Calculating the absolute value of the difference between the input power of the alternating current side and the total power of the direct current side, and taking the absolute value as the absolute value of the power deviation; Calculating the maximum value of the input power of the alternating current side and the total power of the direct current side, and adding the maximum value and a non-zero bias constant to obtain a normalized base number; and calculating the ratio of the absolute value of the power deviation to the normalized base number to obtain a normalized difference ratio.
6. The method for multi-parameter real-time monitoring and abnormality diagnosis of a hvdc transmission system according to claim 5, wherein the step of calculating characteristics based on physical mechanisms further comprises: Calculating the difference value between the normalized difference ratio at the current moment and the normalized difference ratio at the previous moment, and dividing the difference value by the sampling period to obtain a monopole difference ratio gradient; When the first pole direct current and the second pole direct current of the HVDC transmission system are both larger than a preset minimum operation current threshold value, calculating the absolute value of the difference value between the first pole monopole difference ratio gradient and the second pole monopole difference ratio gradient, and taking the absolute value as the bipolar relative gradient deviation.
7. The method for monitoring and diagnosing abnormality of a hvdc transmission system according to claim 6, wherein the step of determining abnormality of the plurality of strategies comprises: Maintaining a historical data sliding window, calculating the mean value and standard deviation of the normalized difference ratio in the historical data sliding window, and generating a dynamic alarm threshold according to the sum of the mean value and the standard deviation; When the discrete working condition mode code indicates a steady-state operation mode, determining that the static drift of the measuring device is abnormal in response to the normalized difference ratio exceeding the dynamic alarm threshold or the bipolar relative gradient deviation exceeding a preset static micro-variation threshold; When the discrete working condition mode code indicates a transient adjustment mode, judging that bipolar dynamic response is asynchronous abnormal in response to the bipolar relative gradient deviation exceeding a preset dynamic tolerance threshold; And when the color state flag bit indicates a non-running state and the total power of the direct current side is larger than the preset minimum residual voltage power, judging that the logic conflict of the human-computer interface is abnormal.
8. The multi-parameter real-time monitoring and abnormality diagnosis device for the HVDC system is characterized by comprising: The image acquisition module is configured to be connected to the display output end of the high-voltage direct-current transmission system, and is used for intercepting real-time image frames of the monitoring interface according to a preset sampling period; the semantic analysis module is connected to the image acquisition module and is configured to perform multi-mode feature analysis on the semantic state area of the real-time image frame and output a semantic state vector; The numerical processing module is connected with the image acquisition module and the semantic analysis module and is configured to execute numerical extraction based on confidence coefficient on a numerical data area of the real-time image frame, dynamically adjust internal filter coefficients according to the semantic state vector and output a smooth monitoring value; and the anomaly analysis module is connected with the numerical processing module and is configured to execute energy conservation calculation and bipolar gradient analysis based on the smooth monitoring value and the semantic state vector and output an anomaly alarm signal.
9. The multi-parameter real-time monitoring and abnormality diagnosis device of the HVDC transmission system according to claim 8, wherein the device is integrated into an intelligent monitoring terminal for HVDC transmission, and the intelligent monitoring terminal comprises an industrial shielding shell (1), a heat radiation assembly (2), a communication interface group (3), a status indication assembly (4), a local monitoring screen group (5) and a video capturing interface group (6); The video capture interface group (6) is arranged on the surface of the industrial shielding shell (1) and is configured to form a physical input interface of the image acquisition module; The heat dissipation component (2) is arranged on the side surface of the industrial shielding shell (1); the local monitoring screen group (5) is arranged on the surface of the industrial shielding shell (1) and is configured to display the smooth monitoring value output by the numerical value processing module; The communication interface group (3) and the status indication component (4) are connected to the anomaly analysis module and configured to output the anomaly alarm signal.
10. The apparatus for monitoring and diagnosing a fault in a hvdc transmission system according to claim 8, wherein the numerical processing module stores therein a filter coefficient map and is configured to: When the color state flag bit in the semantic state vector indicates an operation state and the discrete working condition mode code indicates a steady-state operation mode, selecting a locking filter coefficient as an internal filter coefficient; When the discrete working condition mode codes in the semantic state vector indicate a transient adjustment mode, selecting a tracking filter coefficient as an internal filter coefficient; And selecting the retention filter coefficient as an internal filter coefficient when the confidence coefficient weight in the numerical value extraction process is zero.

Description

Multi-parameter real-time monitoring and abnormality diagnosis method and device for high-voltage direct-current transmission system Technical Field The invention relates to the technical field of automation and intelligent operation and maintenance of power systems, in particular to a multi-parameter real-time monitoring and abnormality diagnosis method and device for a high-voltage direct-current power transmission system. Background Non-invasive screen identification based on computer vision is a key means for acquiring data under physical isolation conditions of a high-voltage direct-current transmission system, but the prior art has a bottleneck. On the one hand, the recognition data accompanies random noise due to screen refresh asynchronism and moire interference. The traditional fixed parameter filtering can not achieve both steady-state precision and transient response, namely dynamic characteristic hysteresis is caused by too strong filtering, and false alarm is caused by too weak filtering. On the other hand, the existing scheme lacks fusion of an electrical physical rule and working condition semantics, ignores bipolar symmetry and energy conservation characteristics, is difficult to identify complex anomalies such as system logic deadlock, measurement drift or bipolar response asynchronism, and cannot meet high-reliability monitoring requirements. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a multi-parameter real-time monitoring and abnormality diagnosis method and device for a high-voltage direct-current transmission system, which solve the problems that steady-state numerical jitter and transient response delay are difficult to be compatible in the optical character recognition process and the system logic conflict and measurement drift abnormality recognition capability are insufficient due to the lack of physical mechanism constraint in single data statistics and discrimination. The first aspect of the invention provides a multi-parameter real-time monitoring and abnormality diagnosis method for a high-voltage direct-current transmission system. The method comprises the steps of firstly obtaining an image frame of a monitoring interface at the current moment according to a preset sampling period, and respectively transmitting the image frame to a semantic analysis module and a numerical value processing module. The semantic analysis module extracts multi-mode features of semantic state areas divided in advance in the image frames to generate semantic state vectors containing color state zone bits and discrete working condition mode codes, and the numerical processing module executes optical character recognition on the numerical data areas to acquire original numerical values and character confidence coefficients. The value processing module then determines state adaptive filter coefficients from the semantic state vector, and performs adaptive state gating filtering on the raw values using the coefficients to output smoothed monitored values. Finally, the anomaly analysis module combines the smooth monitoring value and the semantic state vector, performs feature calculation and multi-strategy anomaly judgment based on a physical mechanism, and generates an anomaly alarm signal. In the feature extraction stage, the multi-mode feature extraction involves converting pixel data in a visual indication subarea from RGB color space to HSV color space, calculating the pixel duty ratio in a target color tone interval, setting a color state zone bit according to the pixel duty ratio, identifying and acquiring a text character string from a text state subarea, and mapping the text character string into a discrete working condition mode code through a preset working condition keyword library. Before the filtering process, the method further comprises the step of calculating a weighted observation value, namely processing the confidence coefficient of the character by using a confidence coefficient weight function, maintaining a short-time sampling buffer queue, and obtaining the weighted observation value by calculating a weighted average value of the original numerical value at the historical moment and the corresponding confidence coefficient weight, wherein the weight of the low-confidence coefficient data is forcedly zeroed. The adaptive state gating filtering process is to calculate the product of the state adaptive filtering coefficient and the weighted observation value and the product of the difference value of the coefficient and the filtering output value at the previous moment, and sum the two to obtain the current moment smooth monitoring value. The selection logic of the filter coefficients is that a smaller locking filter coefficient is selected when the semantics indicate a steady-state operation mode, a larger tracking filter coefficient is selected when the semantics indicate a transient adjustment mode, and a zero keeping filter coefficient