CN-121978439-A - Power transmission line charged state judging method and system based on sliding window regularization inversion and multi-window fusion

CN121978439ACN 121978439 ACN121978439 ACN 121978439ACN-121978439-A

Abstract

The invention relates to a method and a system for judging the electrified state of a power transmission line based on sliding window regularization inversion and multi-window fusion, and belongs to the technical field of power transmission line online monitoring. The method and the system collect space electric field signals by using a non-contact three-dimensional electric field sensor, and identify the electrified state of the transmission line through a sliding window regularization inversion algorithm and a multi-window probability fusion model. The invention adopts Tikhonov regularization and multipoint augmentation strategy, so that under the working condition of containing 5% random noise, average inversion error is reduced by more than 50% compared with the traditional method, the jump problem under the measurement of safety distance is solved, abnormal interference in a track can be automatically identified through a sliding window weight fusion mechanism, accurate charged state information can be output when local data is damaged, meanwhile, a probability judgment model introduces judgment confidence, frequent jump error of a hard threshold at a boundary point is solved, and the acceptability of a diagnosis result is improved.

Inventors

WANG JINGANG
MAN WEI
YU CHUANXIANG
HUANG RUJIN
WANG XIAOTIAN
ZHAO PENGCHENG

Assignees

重庆大学

Dates

Publication Date: 20260505
Application Date: 20260126

Claims (7)

1. The method for judging the electrified state of the power transmission line based on sliding window regularization inversion and multi-window fusion is characterized by comprising the following steps of: s1, establishing a field source coupling model based on a sliding window, namely acquiring three-dimensional electric field data of a plurality of measuring points on a safety track of a power transmission line, constructing a linear equation set of an electric field observation vector and a source side voltage vector based on an electric field superposition principle, grouping the measuring points by adopting a sliding window strategy, and constructing a local inversion unit; S2, solving the inversion of the multi-window fusion Tikhonov regularization, namely solving a voltage estimated value for electric field observation data in each sliding window by adopting a Tikhonov regularized least square method, calculating window weights based on the window condition number and the residual ratio, and carrying out weighted fusion on inversion results of a plurality of windows to obtain a final voltage estimated value; S3, state probability determination based on softmax mapping, namely, the fused voltage estimated values are subjected to per unit, the deviation degree of each state interval is calculated through an interval distance function, the state probability of each window is obtained by combining the softmax function, weighting fusion is carried out by utilizing window weights, and the final electrified state and the confidence coefficient thereof are output.
2. The method for judging the electrified state of the power transmission line based on sliding window regularization inversion and multi-window fusion according to claim 1 is characterized in that in the step S1, the power transmission line is composed of N wires, each phase wire is represented by a straight line segment in space, the straight line segment is uniquely determined through two space coordinate points of a starting point and an end point, a single sensor is adopted to carry out moving sampling along a preset safe track, M groups of three-dimensional electric field vector data are collected, the space redundancy information provided by continuous sampling points is utilized to avoid transient interference, and the total electric field vector response at a measuring point is formed by linearly superposing electric field contributions respectively formed by all conductors under unit voltage excitation.
3. The method for determining the electrification state of the power transmission line based on sliding window regularization inversion and multi-window fusion according to claim 2, wherein the step S1 specifically includes: S11, setting the response vector of the electric field generated at the measuring point under the unit voltage excitation of the jth conducting wire at the ith measuring point P i =(x i ,y i ,z i ) as follows: (1) According to the electric field integral principle, when the voltage of the ith wire is the unit voltage V i =1V and the voltages of other wires are 0, the linear charge density is recorded as Obtaining decoupling matrix elements in all directions: the x-direction decoupling matrix elements are: (2) The y-direction decoupling matrix elements are: (3) The z-direction decoupling matrix elements are: (4) Wherein, the Is the spatial coordinate vector of the ith measuring point, The coordinate vector of the source point with the parameter t on the jth section of wire; S12, electric field intensity at measuring point With source side voltage The following relationship is satisfied: (5) wherein the coefficient is The three-dimensional electric field data of M measuring points are summarized only according to the geometric structure, the spatial position and the coordinates of the measuring points of the lead, and the following equation set is established: (6) The matrix form is: (7) constructing least square objective function based on L2 norm (8) In order to make the objective function Minimizing, relating its gradient to The gradient of (2) is zero: (9) The finishing method can obtain: (10) In the formula, I.e. matrix Moore-Penrose generalized inverse matrix; S13, for each discrete point, a single-loop line is a 3x3 square matrix, a double-loop 3x6 underdetermined matrix needs to adopt a continuous multi-point structure to amplify the matrix, the underdetermined system is converted into an adaptive system, a sliding window with the length of L is defined, a kth window covers a measuring point set { k, k+1, & gt, k+L-1}, electric field observation and decoupling matrix in the window are stacked in rows, and a window electric field quantity and amplification decoupling matrix is obtained: (11) Wherein n u =3 or n u =6, so that the window system can be changed from underdetermined to proper/overdetermined only by meeting the requirement that 3L is larger than or equal to n u , wherein L is larger than or equal to 1 for a single-circuit line, and at least L is larger than or equal to 2 for a double-circuit line.
4. The method for determining the electrification state of the power transmission line based on sliding window regularization inversion and multi-window fusion according to claim 3, wherein in step S2, specifically comprising: S21, solving a voltage estimated value by adopting Tikhonov regularized least square in each window : (12) In the above-mentioned method, the step of, >0 Is the regularization parameter, To make trade-off between accuracy of solution and stability of solution, adopting L curve method to self-adaptively define The optimal value, the optimal balance of solving precision and stability is represented at the inflection point of the L curve; The closed unwrapping of formula (12) is: (13) s22, outputting a group of local inversion results which change along the path by a sliding window Calculating a window augmentation matrix Condition number of (2) The window inversion method is used for measuring the sensitivity degree of the window inversion to noise; defining a window residual ratio: (14) For reflecting the interpretation degree of the current window model to the observation, and constructing window weights according to the interpretation degree: (15) Wherein the method comprises the steps of , >0 For adjusting parameters for balancing the pathological error with the fitting error penalty, as the window slides along the path, And (3) with Will be recalculated, weight And adaptively updating, wherein the process is used as a weight updating mechanism when the window meets the requirement Oversized or too large When the threshold is exceeded, the weight is attenuated, so that misleading of a low-quality window to final state judgment is avoided; S23, after inversion results and weights of all windows are obtained, weighting fusion is carried out to obtain final voltage estimation: (16) Wherein K is the total number of windows, and the fusion output outputs the final voltage estimation on one hand On the other hand, the window weight distribution is reserved and can be used as a confidence coefficient source for judging the subsequent state.
5. The method for determining the electrification state of the power transmission line based on sliding window regularization inversion and multi-window fusion according to claim 4, wherein in step S3, specifically comprising: s31, mapping the fused voltage estimated value to a standard value space, and setting the inversion voltage vector of the kth sliding window as To eliminate the influence of the voltage level difference, the corresponding per unit voltage is defined as Concretely, the method is as follows (17) Wherein the method comprises the steps of For the number of phases corresponding to the line structure, Rated voltage value of corresponding phase; s32, dividing the circuit into live states of voltage loss, voltage shortage, normal and overvoltage, and recording the live states as a set Each state is represented as an interval in per unit field In order to avoid jump error at the boundary of interval, interval distance function is introduced to describe the deviation degree between per unit voltage and each state interval And state interval Definition: (18) When (when) When the component j falls into a certain state interval, the distance is zero, the distance value is larger as the deviation interval is farther, so that a continuous and differentiable measurement basis is provided for state judgment, and a judgment score is defined for the component j in the kth window under the state c m on the basis of the continuous and differentiable measurement basis, wherein the judgment score is as follows: (19) Wherein the method comprises the steps of Constant greater than 0, for adjusting the smoothness of the probability distribution, mapping the score to a window-level state probability by a softmax function: (20) Thereby obtaining the state probability vector of the k window corresponding component j (21) S33, utilizing the constructed window weight Weighting and fusing window level probabilities, and defining the fused state probability as the component j (22) Finally, obtaining the state corresponding to the maximum fusion probability according to all windows 。
6. A transmission line charging state determination system based on sliding window regularization inversion and multi-window fusion, wherein the system adopts the method as claimed in any one of claims 1 to 5.
7. An electronic device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method according to any one of claims 1 to 5 when the computer program is executed.

Description

Power transmission line charged state judging method and system based on sliding window regularization inversion and multi-window fusion Technical Field The invention belongs to the technical field of online monitoring of power transmission lines, and relates to a power transmission line electrification state judgment method and system based on sliding window regularization inversion and multi-window fusion. Background In the intelligent inspection scene of the electric power system, non-contact monitoring in a safe distance by utilizing an electric field sensor is a core means for judging the electrified state of a line. However, due to the tight electrically insulating safety spacing between the transmission line and the sensor, the spatial electric field signals attenuate significantly and overlap in phase during long distance propagation, resulting in extremely high spatial coupling of the excitation contributions formed by the phase conductors at the measurement points. At the mathematical level, this appears to be extremely high Condition Number (Condition Number) of the decoupling matrix, with significant numerical morbidity. Under the inversion structure of the discomfort (ill-posed), the traditional pseudo-inverse solving algorithm can amplify the tiny power frequency noise in the environment, the sensor attitude disturbance and the systematic deviation (such as wire sag and split wire effect) caused by model simplification in an order of magnitude. This directly results in severe random jumps or spurious fluctuations in the source side voltage of the inversion, severely affecting the stability and usability of the inversion results, and making it difficult to support reliable state diagnostics at the industrial level. On the other hand, the identification of the electrified state of the power transmission line is essentially a classification decision problem, and various dynamic working conditions such as voltage loss, undervoltage, steady-state operation, overvoltage and the like are covered. Conventional decision logic typically relies on inverting the absolute magnitude of the voltage and employs a hard threshold of fixed proportion as the switching criterion. However, under the actual complex background electromagnetic field and non-ideal measurement environment, the hard threshold judgment which simply depends on a single inversion value is extremely easy to cause erroneous judgment in the threshold neighborhood, namely, the recognition result frequently jumps between different states due to small amplitude drift. In addition, due to the lack of quantitative evaluation of uncertainty factors in the measurement process, the existing decision system cannot provide reliability support for the results of operation and maintenance personnel, so that a higher misjudgment or false alarm risk exists when boundary conditions (such as slight undervoltage caused by high-resistance grounding) are processed. Therefore, how to inhibit the noise amplification effect in the inversion process through algorithm innovation on the premise of ensuring the physical isolation safety, and construct a set of probabilistic judgment system which can be compatible with measurement uncertainty and has fault tolerance decision making capability is a key bottleneck to be overcome in the current power transmission line online monitoring technology. Aiming at the on-line monitoring of a high-voltage transmission line, the prior art mainly relies on ground or short-distance electromagnetic field sampling and numerical analysis. However, under the constraint of ensuring the electrical safety distance, these traditional schemes often expose common problems such as single observation dimension, poor algorithm immunity, stiff decision strategy and the like, and it is difficult to maintain stable identification precision under a complex field electromagnetic background, and the main disadvantages include: 1) The static single-point measurement limitation is that a low-dimensional analysis matrix is built only by means of electric field information of a single space measuring point, the space redundancy information support is lacked, and inversion results are extremely easy to be subjected to local electromagnetic interference or strong traction of a single abnormal sampling point, so that judgment stability is seriously insufficient. 2) Pseudo-inverse algorithm destabilization the traditional Moore-Penrose pseudo-inverse method lacks suppression capability on observed noise when processing high condition number (pathological) decoupling matrices. Under the condition of long-distance measurement, weak measurement errors can be amplified by an algorithm, so that the estimated value of the output voltage deviates from a physical true value, and even an unexplained negative value or an overrun value occurs. 3) The hard threshold decision logic is stiff, and the prior art lacks statistical modeling of measurement noise and model errors