Distribution cable aging positioning method based on digital grid

Abstract

The invention provides a cable insulation local aging real-time positioning method based on grid search in order to realize effective positioning of various local aging types. According to the main thought, steady-state electric signals, measured at different local aging positions and aging degrees, at the two ends of a cable are utilized to form a training data set, a neural network model is trained, and the trained neural network model is utilized to carry out local aging positioning on new measurement data. The relation between electric signals at the head end and the tail end of a cable and local aging parameters has the characteristic of high nonlinearity, and in order to enable a trained neural network to effectively extract local aging position information, it is necessary to widen the used frequency band and increase the measurement frequency number. Therefore, the invention provides a method of injecting a monitoring reference signal at the head end of the cable, measuring derived voltage at the tail end, and performing local aging positioning by using a grid search model. In the off-line mode, the transfer function of the cable is easy to measure, and the method is easy to implement.

Description

Technical Field

[0001] This application belongs to the field of power distribution cable aging location method based on digital grid. Background Technology

[0002] As a crucial component of urban power transmission, cables are increasingly being replaced by underground cables and internal conduit cables as urban economic levels rise. This trend is particularly evident in first- and second-tier cities. Power cables traverse cities, and their safe, stable, and reliable operation is a vital factor in the urban energy system. Cable faults directly impact the work and lives of urban residents, making the safe operation and maintenance of urban cables paramount. Cables possess excellent mechanical and electrical properties, and the vast majority of faults originate from the aging of the insulation layer. Many factors influence aging, including the cable material itself, the external operating environment, and human or foreign object damage. Each or a combination of these factors contributes to varying degrees of cable aging. Based on the stress that promotes insulation aging, aging can be broadly categorized into thermal aging, electrical aging, water aging, chemical corrosion, and mechanical aging. However, the root cause is ultimately high-temperature heating, with sustained high temperatures being the primary inducing factor for thermal and insulation aging. In this field, researchers from various sectors have conducted in-depth studies and developed testing methods. Generally, testing methods are divided into online and offline methods. Online monitoring combines historical and current data to assess insulation status, while offline testing combines laboratory preventative tests with the test results for evaluation. Commonly used measurement methods include withstand voltage tests, elongation at break measurement, absorption ratio measurement, leakage current measurement, and dielectric loss angle measurement, all of which can simulate the relationship between cable faults and insulation aging to a certain extent. However, due to the long distances of distribution cable laying and limited underground space, maintenance personnel find it difficult to access the lower ends to inspect specific fault nodes. They can only rely on the transmission of fault information from nodes back to the backend, calculating the relative location through data transmission and attenuation models. However, many current fault location methods are inaccurate, making it difficult for maintenance personnel to determine the precise fault location, significantly impacting fault repair. Large errors also increase maintenance costs considerably. Therefore, fault location methods for distribution cables are worthy of further research.

[0003] Based on an assessment of the overall insulation condition of the cable, further locating the locally aged sections helps to enable targeted maintenance, thereby reducing system operation and maintenance costs. However, current research on real-time methods for locating locally aged sections of cable insulation is limited, and most existing studies rely on traveling wave injection, which is difficult to perform real-time location of locally aged sections of cables with smoothly changing insulation electrical parameters. To address this issue, this invention proposes a digitally driven real-time aging location method for distribution cables. Unlike traditional traveling wave-based methods, the proposed method primarily estimates local aging parameters by measuring the voltage and leakage current at both ends of the cable, improving the adaptability of the proposed method to various local aging modes and making it easier to implement online. Summary of the Invention

[0004] The purpose of this application is to overcome the problem of the lack of accuracy in online life assessment of power distribution cables.

[0005] To achieve the above objectives, this application proposes a method for aging location of power distribution cables based on digital grids.

[0006] To effectively locate various types of localized aging, this invention proposes a grid search-based real-time localization method for cable insulation localized aging. The main idea is to use steady-state electrical signals measured at both ends of the cable under different localized aging locations and degrees to form a training dataset and train a neural network model. The trained neural network model is then used to locate localized aging based on new measurement data. The relationship between the electrical signals at both ends of the cable and localized aging parameters is highly nonlinear. To enable the trained neural network to effectively extract localized aging location information, it is necessary to broaden the used bandwidth and increase the number of measurement frequencies. Therefore, this invention proposes injecting a monitoring reference signal at the cable's beginning and measuring the derived voltage at the end, using a grid search model for localized aging location. In offline mode, the cable's transfer function is easily measured, and the proposed method is easy to implement. Attached Figure Description

[0007] Figure 1 The cable end voltage, first end leakage current spectrum and field measured impedance waveform are provided for the unit monitoring voltage injection in this application.

[0008] Figure 2 The algorithm for locating localized aging of cable insulation based on grid search provided in this application;

[0009] Figure 3 Flowchart of the cable insulation local aging location algorithm based on grid search provided in this application;

[0010] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. Detailed Implementation

[0011] The technical solution of this application will be described in detail below with reference to the accompanying drawings.

[0012] This invention proposes a digital grid-based method for locating the aging of power distribution cables, which models localized aging of cable insulation. However, this method is not applicable to non-uniform or smooth aging conditions. The grid search-based method for locating localized aging of cable insulation utilizes a training dataset composed of steady-state electrical signals measured at both ends of the cable under different localized aging locations and degrees. This dataset is then used to train a neural network model. The trained neural network model is then used to locate localized aging based on new measurement data, achieving effective location of various localized aging types. The relationship between the electrical signals at both ends of the cable and localized aging parameters is highly nonlinear. To enable the trained neural network to effectively extract localized aging location information, it is necessary to broaden the used bandwidth and increase the number of measurement frequencies. Therefore, this invention proposes a method that injects a monitoring signal at the cable's beginning and measures the resulting voltage at the end, using a grid search model for localized aging location. In offline mode, the cable's transfer function is easily measured, making the proposed method easy to implement.

[0013] This invention provides online, real-time cable aging monitoring, estimating local aging parameters based on measurements of voltage and leakage current at both ends of the cable. It injects n common-mode monitoring signals of different frequencies, where the i-th monitoring frequency is denoted as f. i At each injected monitoring frequency, the voltage at both ends of the cable and the magnitude of the common-mode leakage current are measured. The voltage at the beginning of the cable corresponding to monitoring frequency fi is denoted as Ui1(fi), the voltage at the end as Ui2(fi), the leakage current at the beginning as Ii1(fi), and the leakage current at the end as Ii2(fi). The following analysis assumes that the common-mode monitoring signal is injected from the beginning of the cable. There are three local aging parameters to be estimated, namely the starting position L of the local aging. b The termination position Le of local aging and the degree of local aging γ. The degree of local aging is defined as the ratio of the capacitance of the locally aged section to the capacitance of the intact section, i.e., γ = Caged / C0. This algorithm is mainly based on the following idea: given the known amplitude of the monitoring voltage injected at the cable head end, the closer the estimated local aging parameters are to the actual parameters, the smaller the difference between the measured values ​​of the cable end voltage and the leakage current at the head end and the estimated values ​​obtained based on the model of equation (1):

[0014] U i2 =H t U inj

[0015] I i1 =Y0U inj (1)

[0016] Among them, the equivalent admittance Y0 at the beginning of the entire cable, and the monitoring signal U injected from the beginning of the cable into the cable. inj At that time, the voltage signal U measured at the end of the cable i2 and the current signal I measured at the first end i1 .

[0017] Therefore, the problem of locating local aging of cable insulation can be transformed into an optimization problem as shown in equation (2).

[0018]

[0019] In the formula U m2 (f k ) and I m1 (f k ) represent monitoring frequencies of f k The terminal voltage and the beginning leakage current are calculated using the model and estimated local aging parameters. L still represents the total cable length. λ is a weighting coefficient used to adjust the weight between voltage estimation error and current estimation error, and is set to 1 in subsequent sections. Furthermore, because the method for estimating local aging parameters proposed in this chapter only uses the amplitudes of the voltage and leakage current at both ends of the cable and does not require phase information, it eliminates the need for real-time communication between the two ends of the cable. This avoids local aging parameter estimation errors caused by phase angle measurement errors, making the proposed method easier to implement in industrial applications.

[0020] Although the local aging parameters can be estimated by solving the optimization problem shown in equation (2), directly using optimization algorithms to solve the local aging location problem presents the following difficulties. First, there is a highly nonlinear relationship between the measured values ​​and the local aging parameters, and the aging location and aging degree are highly coupled. This makes it easy for common optimization algorithms to get stuck in local extrema when solving this problem, resulting in low positioning accuracy. In addition, the changes in the measured values ​​at both ends caused by local aging of the cable insulation are very weak (e.g., Figure 1 As shown in the figure, this results in a smoother gradient across the entire solution space for the optimization problem, increasing the search time for the optimal solution. Localized aging of cable insulation causes changes in the measured end voltage and beginning leakage current spectra. However, the changes in electrical parameters caused by localized insulation aging are usually weak, making the changes in voltage and leakage current measured at both ends insignificant. Furthermore, the trends in voltage and leakage current measurements differ at different frequencies, making it difficult to directly extract information about localized aging from the spectrum. Therefore, it is necessary to design corresponding algorithms to estimate the location and degree of localized aging.

[0021] To overcome the above challenges, this paper proposes a grid search-based algorithm for locating localized aging of cable insulation, which mainly consists of two steps: grid search and iterative filtering. The overall idea of ​​the algorithm is as follows: Figure 2 As shown below, the two steps of the algorithm will be described in detail.

[0022] Step 1: Grid search.

[0023] First, divide the entire cable into p segments, resulting in p+1 endpoints, including the cable start and end points. Selecting any two distinct endpoints forms a combination of possible local aging locations. Let L be the set of all possible local aging location combinations, and denote the i-th element of the set as (L...). b,i ,L e,i For each combination of positions in set L, solve the optimization problem as shown in equation (3).

[0024]

[0025] In the formula γ i This represents the estimated aging degree corresponding to the i-th position combination. It can be seen that for each position combination, since the location of local aging is already determined, the number of parameters to be optimized is reduced from three to one, making it easier to solve. For the optimization problem shown in equation (3), this paper uses the Sequential Quadratic Programming (SQP) method to solve it.

[115] As can be seen, by using the grid search method, the optimization problem in the entire solution space is transformed into a suboptimal problem on multiple location combinations L, which greatly improves the accuracy and efficiency of the computation.

[0026] Step 2: Filtering and Iteration.

[0027] After obtaining the aging degree corresponding to each location combination, the estimated values ​​of the terminal voltage and the first-terminal leakage current are calculated based on all the currently obtained local aging parameters, and the estimation error of each location combination is evaluated. Wherein, the estimation error E for the i-th location combination... i It can be represented as shown in equation (4).

[0028]

[0029] The smaller the estimation error, the closer the location combination is to the actual local aging location. To ensure the robustness of the results, the t location combinations with the smallest estimation errors are selected to obtain the new L. b Variation range [L] b1 ,L b2 ] and the new L e Variation range [L] e1 ,L e2 ]. Among them, L b1 The minimum value of the local aging initiation position in the t position combinations, and L b2L is the maximum value of the local aging initiation position in the t position combinations. e1 and L e2 The definition method is similar. If the threshold condition L is met... b2 -L b1 <α and L e2 -L e1 If α < α, the algorithm terminates and outputs L. b The range of change [L] b1 ,L b2 ] and L e The range of change [L] e1 ,L e2 The estimated local aging location is calculated, and the average of the local aging degrees corresponding to the t location combinations is output as the estimated local aging degree. If the threshold condition is not met, the new variation range [L] is used. b1 ,L b2 ] and [L e1 ,L e2 The algorithm is then divided into p segments again, and the above solution steps are repeated. The algorithm terminates if the number of iterations exceeds the set maximum number of iterations N. The overall flow of the proposed local aging localization algorithm is as follows: Figure 3 As shown.

[0030] For each measurement data point, the proposed grid-search-based method for locating localized aging of cable insulation is used to estimate the local aging location and degree of aging. The algorithm error is defined as shown in Equation (6-10). The location error of the starting position of localized aging is denoted as E. Lb The positioning error of the local aging termination position is denoted as E. Le The estimation error of the degree of local aging is denoted as E. γ .

[0031]

[0032] In the formula N S This represents the total number of data points in the test dataset. This represents the local aging start position of the i-th data in the dataset. This represents the range of local aging initiation position changes estimated using the proposed algorithm for the i-th data. and The parameters representing the termination point of localized aging are defined similarly to those representing the start point of localized aging. γ (i) This represents the local aging degree of the i-th data point in the dataset. This represents the degree of local aging estimated using the proposed algorithm.

[0033] To improve the computational efficiency of each optimization step, the number of grid divisions, p, is set to 8. Because the optimization problem is highly nonlinear, the position combination with the smallest estimated error obtained in each selection and iteration may not coincide with the actual local aging position. Therefore, during the selection and iteration phase, the number of position combinations with the smallest error, t, is set to 4 to ensure the robustness of the algorithm. Simultaneously, the value of t increases with the number of grid divisions, p. Furthermore, the maximum number of iterations N for algorithm termination and the estimated local aging position threshold α need to be set. Considering industrial requirements, α is set to 0.05 km in this chapter. Simulations show that on the test dataset, the algorithm's average number of iterations is 3; therefore, N is set to 10 in this chapter to ensure the convergence of the localization algorithm.

Claims

1. A method for locating aging power distribution cables based on digital grids, characterized in that, Model-based methods for modeling local aging of cable insulation are difficult to apply to non-uniform and smooth aging conditions. The grid search method for locating local aging of cable insulation uses steady-state electrical signals measured at both ends of the cable under different local aging locations and aging degrees to form a training dataset and train a neural network model. The trained neural network model is then used to locate local aging on new measurement data, achieving effective location of various types of local aging.

2. A method for locating aging power distribution cables based on digital grids, characterized in that, Given the known amplitude of the monitoring voltage injected at the cable head end, the closer the estimated local aging parameters are to the actual parameters, the smaller the difference between the measured values ​​of the cable end voltage and the leakage current at the cable head end and the estimated values ​​obtained from the grid search, and the more accurate the positioning.

3. A method for locating aging power distribution cables based on digital grids, characterized in that, By using a grid search method, the optimization problem in the entire solution space is transformed into a suboptimal problem on multiple location combinations L, which greatly improves the accuracy and efficiency of the computation.

Images