CN-121980472-A - Early-stage abnormality detection method and medium for insulation degradation of power communication line

CN121980472ACN 121980472 ACN121980472 ACN 121980472ACN-121980472-A

Abstract

The invention provides an early-stage abnormality detection method and medium for insulation degradation of an electric power communication line, which relate to the field of abnormality detection, and provide EWED an abnormality detection model aiming at long-sequence-dependent electric power communication line data, wherein the abnormality detection model consists of a long-time dependent module, a spectrum coding module and an abnormality judgment module, the long-time dependent module enhances the identification capability of the model on early-stage abnormality trend, the spectrum coding module enhances the resolution of early-stage characteristics of insulation degradation, the abnormality judgment module enhances the accuracy and response sensitivity of abnormality judgment, the probability expression enhances the interpretability and the controllability of results, and a threshold mechanism enables a system to have the capability of flexibly adjusting and adapting to different running environments.

Inventors

ZHANG MINMIN
TAO WEIWEI
ZHANG YAN

Assignees

山东宏业发展集团有限公司

Dates

Publication Date: 20260505
Application Date: 20260403

Claims (10)

1. A method for detecting early-stage abnormality of insulation degradation of an electric power communication line, comprising the steps of: collecting relevant data of the power communication line, and constructing preprocessing division of a data set; Constructing a long-term dependence module, namely constructing a power communication line and a historical power communication line sequence by utilizing a sliding window, constructing square and linear magnitude parts for local variation, and calculating an inertial diffusion kernel; Calculating a global inertial memory field through an inertial diffusion kernel and a comprehensive characterization value, and obtaining a time field in a window by combining time position offset, an acting factor and a time gradient; the construction spectrum coding module is used for inputting a power communication line sequence, constructing an integer sequence through a discrete integer set and calculating an integer lattice point sequence; constructing a standardized power communication line sequence based on the window time position sequence, constructing an adjacency and degree matrix calculation graph Laplace matrix, and outputting a quantization factor; Calculating a module space mapping sequence and an integer lattice point sequence according to the time position offset to calculate an indication quantity, and calculating a resonance accumulation quantity; Calculating a local fractal rough index based on time position deviation and a power communication line sequence, processing a track increment and geometric consistency deviation index, and combining a basic attenuation constant to obtain a flexible attenuation factor; Calculating flexible matching factors by using the model space mapping, the integer lattice point sequence and the flexible attenuation factors, and accumulating and summing to obtain flexible resonance intensity data maximum value quantization transformation to obtain a pseudo-spectrum vector; and constructing an abnormality judging module, namely splicing the time field and the pseudo spectrum vector in the window to be combined characteristics, and judging an abnormal result of the power communication line through linear scoring and probability and combining a threshold value.
2. The method for detecting early-stage abnormality of insulation deterioration of a power communication line according to claim 1, wherein power communication line data is inputted, a power communication line sequence and a history power communication line sequence are constructed using a sliding window, and a local variation is obtained for each time position in the history power communication line sequence based on a difference between the position and a value of its neighboring history power communication line; the method comprises the steps of constructing a square magnitude part and a linear magnitude part based on local variation, respectively combining square weight and linear weight to obtain a comprehensive characterization value, constructing an exponential decay term based on time interval amounts between each time position in a historical power communication line sequence and the time position of the power communication line sequence, combining a linear expansion term, obtaining a core value under the current time position and the historical time position by combining the two terms, and arranging all time interval core value sets to obtain an inertial diffusion core.
3. The method for detecting early-stage insulation degradation abnormality of an electric power communication line according to claim 2, wherein the inertial diffusion kernel and the comprehensive characterization value are subjected to item-by-item multiplication, and accumulated according to a time sequence, a global inertial memory field is constructed, and a time gradient is obtained by comparing and calculating global inertial memory fields of adjacent time positions; Defining time position offset according to the sliding window, sequentially numbering each time position in the window relative to the window starting time position, distinguishing the relative position relation of each time position in the window according to the numbers, normalizing the window length to obtain an action factor, and combining the action factor corresponding to the time offset with a time gradient to obtain a time field in the window.
4. The method for detecting early-stage abnormality of insulation degradation of an electric power communication line according to claim 2, wherein the electric power communication line sequence is input, the integer sequence is constructed by using a discrete integer set as frequency indexes, the integer sequence is constructed according to each frequency index, the time position in a window is used as an independent variable, the time position and the frequency index are calculated according to integer transformation, and a rounding constraint is applied according to integer-mode space parameters, so that the integer lattice point sequence is obtained.
5. The method for detecting early-stage abnormality of insulation degradation of an electric power communication line according to claim 3, characterized in that a quantization factor is designed, translation and scale normalization processing are performed on the basis of an electric power communication line sequence at each time position in a window to obtain a standardized electric power communication line sequence, a corresponding adjacency relation matrix is constructed by performing similarity measurement on the standardized electric power communication line sequence between any two time positions in the window, a metric value of each time position is calculated on the basis of the matrix to obtain a degree matrix, and then a drawing Laplace matrix of the window is calculated through the adjacency relation matrix and the degree matrix; and selecting a second characteristic value and a third characteristic value in the graph Laplace matrix spectrum structure, extracting a difference ratio between the characteristic values to obtain an intermediate index, inputting the intermediate index into a nonlinear mapping unit for processing, performing range limiting processing according to upper and lower boundary parameters, and outputting a quantization factor corresponding to a current window.
6. The method for early abnormality detection of insulation degradation of a power communication line according to claim 4, wherein a power communication line sequence of a window start time position is obtained based on a time position offset, a linear scaling process is performed by a quantization factor to obtain an intermediate mapping quantity, and then a rounding constraint is performed by using an integer modulo space parameter, and a modulo space mapping sequence is output; and comparing the integer lattice point sequence with the modulo space mapping sequence at the same offset position one by one, obtaining an indication quantity according to a comparison result, sequentially generating the indication quantity by taking time position offset in a window as a range, and performing item-by-item accumulation operation on the indication quantity to obtain a resonance accumulation quantity for the frequency index k.
7. The method for detecting early-stage abnormal insulation degradation of an electric power communication line according to claim 6, wherein a flexible attenuation factor is provided, a plurality of adjacent value pairs with different intervals are selected from an electric power communication line sequence for each time position deviation in a window, differences among the adjacent value pairs are calculated to obtain a plurality of scale difference amounts, scale combination is carried out on the scale difference amounts and bias parameters, and the difference amounts with different scales are processed to obtain a local fractal roughness index; And carrying out differential calculation on track increment of adjacent integer lattice point sequences, carrying out normalization, differential comparison and offset processing on amplitude relations among the adjacent track increment, combining adjustment coefficients to adjust to obtain geometric consistency deviation indexes, and then combining basic attenuation constants to carry out multiplicative coupling processing on local roughness indexes and geometric consistency deviation indexes to obtain flexible attenuation factors.
8. The method for detecting early-stage insulation degradation anomalies of a power communication line according to claim 6, wherein the method is characterized in that offset difference calculation is carried out on a modular space mapping sequence and an integer lattice point sequence, amplitude scaling and indexing processing are carried out by combining a flexible attenuation factor, and a flexible matching factor is obtained; Taking the length of the sliding window as a traversal range, accumulating and summing flexible matching factors in the window, defining a discrete resonance relation by taking the resonance accumulated quantity as a discrete reference, and providing a baseline reference and consistency alignment structure for the flexible resonance intensity to obtain the flexible resonance intensity; And determining the maximum value of the flexible resonance intensity, taking the lattice index as a traversal range, and carrying out magnitude transformation operation on the maximum value of the flexible resonance intensity by the flexible resonance intensity to obtain a pseudo-spectrum vector.
9. The method for detecting early-stage abnormality of insulation degradation of an electric power communication line according to claim 1, wherein a time field and a pseudo-spectrum vector in an input window are spliced to obtain a joint characteristic, a linear scoring function is used to obtain an electric power communication line abnormality score, then an electric power communication line abnormality probability value is output through a probability mechanism, a threshold mechanism is introduced, and the electric power communication line abnormality probability value is judged according to a preset threshold value to obtain an electric power communication line abnormality result.
10. A computer-readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor realizes the steps of the method for early detection of insulation degradation of an electric power communication line according to any one of claims 1 to 9.

Description

Early-stage abnormality detection method and medium for insulation degradation of power communication line Technical Field The invention belongs to the field of anomaly detection, and particularly relates to a method and a medium for detecting early-stage anomaly of insulation degradation of a power communication line. Background The power communication line is used as a basic support for power system dispatching control, relay protection, state monitoring and information transmission, the operation reliability of the power system is directly related to the safety and stability of a power grid, the data amount and real-time requirements borne by the communication line are obviously improved along with the continuous expansion of the scale of the power system and the improvement of the intelligent level, however, the line is exposed in complex environments such as high temperature, humidity and the like for a long time, and the insulating material is easy to generate degradation phenomena such as microcracks, interface debonding and the like, so that early anomalies with strong concealment and uneven development speed are formed, and hidden dangers are buried for subsequent fault accumulation risks. At present, the insulation state detection of the power communication line mainly depends on manual inspection, periodic insulation test or a threshold judgment method based on single physical quantity, and the methods have the defects of long detection period, insufficient instantaneity, low sensitivity to weak degradation characteristics and the like, meanwhile, the operation environment of the communication line is complex, the insulation degradation process has multi-scale performance, concealment performance and progressive performance, early degradation signals often show weak disturbance, irregular fluctuation or fuzzy characteristics, the reliable identification is difficult to realize by the traditional statistical analysis method and the univariate monitoring method, and in addition, the characteristics of noise interference, strong non-stationarity, multi-physical factor coupling and the like exist in data acquisition, so that the early abnormal detection of the insulation degradation is more challenging. With the development of sensing technology, data acquisition equipment and intelligent algorithm, an intelligent early-stage abnormality detection method based on multidimensional operation characteristics gradually becomes a research focus in the field of power communication, a multiscale characteristic model can be established for normal behaviors of a communication line under different loads, climates and operation modes by introducing a data-driven abnormality modeling mechanism, early-stage weak degradation signals are enhanced and extracted, detection sensitivity and accuracy are improved, deep learning is realized by a method which can be used for fusing multiple physical quantity characteristics, has strong robustness, high instantaneity and strong self-adaptation capability, potential hidden hazards can be timely found in the background of early-stage abnormality detection of insulation degradation of the power communication line, equipment maintenance cost is reduced, and the method has important significance for improving the overall safety and reliability of a power communication system. Disclosure of Invention The invention provides a method and a medium for detecting early-stage abnormality of insulation degradation of an electric power communication line, which are used for providing EWED an abnormality detection model aiming at long-sequence-dependent electric power communication line data, and comprise a long-time dependent module, a frequency spectrum coding module and an abnormality judging module. The technical scheme adopted by the invention for realizing the purpose comprises the following steps: collecting relevant data of the power communication line, and constructing preprocessing division of a data set; Constructing a long-term dependence module, namely constructing a power communication line and a historical power communication line sequence by utilizing a sliding window, constructing square and linear magnitude parts for local variation, and calculating an inertial diffusion kernel; Calculating a global inertial memory field through an inertial diffusion kernel and a comprehensive characterization value, and obtaining a time field in a window by combining time position offset, an acting factor and a time gradient; the construction spectrum coding module is used for inputting a power communication line sequence, constructing an integer sequence through a discrete integer set and calculating an integer lattice point sequence; constructing a standardized power communication line sequence based on the window time position sequence, constructing an adjacency and degree matrix calculation graph Laplace matrix, and outputting a quantization factor; Calculating a module space mappi