An all-dielectric self-supporting optical cable wire sag calculation optimization method based on support vector regression

Abstract

This invention discloses an optimization method for calculating conductor sag in all-dielectric self-supporting optical cables based on support vector regression. Addressing the inaccuracies in conductor sag estimation using existing parabolic and catenary methods, this method trains a support vector regression model using extensive historical inspection data, resulting in a precise support vector machine model relating surrounding environmental parameters to sag corrections. In practical applications, relevant surrounding environmental parameters can be obtained from sensors to calculate the sag correction, improving conductor sag estimation accuracy and ensuring optimal line operation.

Description

Technical Field

[0001] This invention relates to the field of optical cable installation, and in particular to a method for calculating and optimizing conductor sag of all-dielectric self-supporting (ADSS) optical cables based on support vector regression. Background Technology

[0002] Conductor sag refers to the vertical distance from any point on the conductor's suspension curve to the lines connecting the two suspension points. As we know from physics and mechanics, the magnitude of sag is inversely proportional to the stress on the conductor and inversely proportional to the square of the span. Conductor stress is related to the conductor's own weight and external loads. Estimating conductor sag can determine the safe crossing distance of the line, thus providing some prior knowledge for estimating tower height.

[0003] Currently, there are two main algorithms for calculating conductor sag: the parabolic method and the catenary method. The parabolic method and the catenary method analyze the force on the conductor sag from a mechanical analysis perspective, and can provide sag calculation results with a certain degree of accuracy, but the accuracy is limited.

[0004] In actual engineering sites, sag is closely related to many factors besides stress and span, including wind speed in the vertical environment of the conductor, ambient temperature of the conductor, solar radiation intensity, kinematic viscosity of the air layer on the conductor surface, heat transfer coefficient of the air layer on the conductor surface, average temperature rise of the conductor surface, radiative heat dissipation power of the conductor, convective heat dissipation power of the conductor, and solar heat absorption power of the conductor.

[0005] Improving the accuracy of sag calculation results is an urgent technical problem that needs to be solved. Summary of the Invention

[0006] This invention addresses the problems existing in the background technology by proposing a method for calculating and optimizing the sag of conductors in all-dielectric self-supporting optical cables based on support vector regression, which includes the following steps:

[0007] S1. Data Acquisition: Obtain raw data. , Indicates the first Data collected by the group;

[0008] S2. Data cleaning, including outlier detection and filtering;

[0009] S3. Data preprocessing to obtain the sample dataset. , Indicates the first preprocessed step Group data;

[0010] S4. Model Design: The support vector regression machine model with slack variables is as follows:

[0011]

[0012] Represented as ,in, and For model parameters, As slack variables, Here is the regularization constant. For the Lagrange multiplier parameters, For kernel function, Represents the kernel function mapping function; Insensitive loss function; This represents the number of data sample groups;

[0013] Sag correction parameters , This represents the sag calculation result obtained from the catenary method. Indicates by the first Group data The corresponding sag data;

[0014] S5. Model training, using the sample dataset and each set of sample data Corresponding sag data The model parameters are obtained by training the support vector regression machine model designed by S4. and ;

[0015] S6. Use the model for sag optimization:

[0016] Obtaining the sag calculation results using the catenary method ;

[0017] Raw data is acquired through the set sensors, and the raw data is input into the parameter model obtained by S5 to obtain the sag correction parameters. ;

[0018] Sag Optimization Results .

[0019] Specifically, in S1 and S6, the types of raw data collected include: the location of the conductor. Ambient wind speed, location of the conductor Ambient temperature, location of the conductor Sunlight intensity, conductor surface Kinematic viscosity of air layer, surface of conductor Average temperature rise, conductor Radiant heat dissipation power, wires Convection heat dissipation power, wires Solar heat absorption power, conductor surface The heat transfer coefficient of the air layer.

[0020] Preferably, in S2, outlier detection is performed based on a t-test, and the steps are as follows:

[0021] Treating all original sample data as a whole that follows a normal distribution, calculate the expected value of the total original sample data. and standard deviation suspicious samples If a population with a total sample size of 1 has no significant difference from the rest of the populations, then it belongs to the same population as the original collected data. The t-test statistic is:

[0022]

[0023] If the statistic If the value is greater than the critical value of the t-test at the corresponding confidence level, It was determined to be an outlier.

[0024] Preferably, in S2, after a value is identified as an outlier, it is removed to correct the outlier.

[0025] Preferably, in S2, after a value is identified as an outlier, the outlier data is corrected using a linear interpolation algorithm.

[0026] Preferably, in S3, data preprocessing includes: data normalization and principal component analysis.

[0027] Specifically, data normalization is performed using Min-Max standardization. The Min-Max standardization transformation function is as follows:

[0028]

[0029] in This represents the maximum value of each corresponding element in the sample data. This represents the minimum value of the sample data.

[0030] Beneficial effects of the present invention

[0031] Accurate estimation of ADSS optical cable conductor sag is crucial for ensuring good cable operation and effectively guaranteeing safe distances between the cable and crossings. Based on the solution proposed in this invention, relevant environmental parameters obtained from sensing components can be used to calculate sag corrections, improving conductor sag estimation accuracy and ensuring good line operation. Attached Figure Description

[0032] Figure 1 This is a flowchart of the method of the present invention.

[0033] Figure 2 This is a schematic diagram of a scenario for the present invention. Detailed Implementation

[0034] The present invention will be further described below with reference to embodiments, but the scope of protection of the present invention is not limited thereto:

[0035] This patent proposes a method for calculating and optimizing optical cable sag based on support vector regression. Combined with... Figure 1 The proposed method and steps are as follows:

[0036] 1. Data Collection

[0037] The data acquisition phase is completed manually by inspecting the conductor. A drone carrying the corresponding measuring components is used to measure a variety of data, including the ambient wind speed, ambient temperature, solar radiation intensity, kinematic viscosity of the air layer on the conductor surface, average temperature rise of the conductor surface, radiative heat dissipation power, convective heat dissipation power, solar heat absorption power, and heat transfer coefficient of the air layer on the conductor surface, as well as the actual sag of the conductor. These raw data are used to prepare for the next step of data preprocessing.

[0038] Let the original collected data be ,in Indicates the first The data collected by the group, each data point Corresponding to a real conductor sag data This can be represented as...

[0039]

[0040] in

[0041] - Location of the conductor ambient wind speed

[0042] - Location of the conductor ambient temperature

[0043] - Location of the conductor Sunlight intensity

[0044] - Wire surface Kinematic viscosity of air layer

[0045] - Wire surface Average temperature rise

[0046] -wire Radiative heat dissipation power

[0047] -wire Convection heat dissipation power

[0048] -wire Sunlight absorption power

[0049] - Wire surface heat transfer coefficient of the air layer

[0050] 2. Data Cleaning

[0051] During the data acquisition process, factors such as equipment failure, signal interference, and human error in verification may lead to data distortion or loss. It is necessary to detect and screen outliers to ensure data quality.

[0052] To determine whether the original collected data has been distorted or missing, the following outlier detection method based on t-test is used.

[0053] Treating all original sample data as a whole that follows a normal distribution, calculate the expected value of the total original sample data. and standard deviation suspicious samples If a population with a total sample size of 1 has no significant difference from the rest of the populations, then it belongs to the same population as the original collected data, and the t-test statistic is:

[0054]

[0055] If the statistic If the value is greater than the critical value of the t-test at the corresponding confidence level, If a value is identified as an outlier, it is removed, or the outlier data is corrected using a linear interpolation algorithm.

[0056] 3. Data Preprocessing

[0057] After data cleaning, outliers and missing values ​​were properly corrected and the format was standardized. The data can now reflect the characteristics of conductor sag correction. However, before machine learning, data preprocessing is still required to make the data more suitable for use by the machine learning model. This includes data normalization and principal component analysis.

[0058] Min-Max standardization is used for data normalization. Also known as deviation standardization, this method performs a linear transformation on the original data, mapping it to a fixed interval [0, 1] while preserving the original data structure relatively well. The transformation function for Min-Max standardization is as follows:

[0059]

[0060] in This represents the maximum value of each corresponding element in the sample data, such as... The maximum ambient wind speed at the location of all included conductors. Min-Max standardization represents the minimum value of the sample data and can normalize data when the units of measurement are uniform.

[0061] Principal Component Analysis (PCA) is a linear dimensionality reduction method. Its principle is to project data into a lower-dimensional space, maximizing the variance of the data in the projected dimension. Since the original collected data has too many dimensions, PCA is used to analyze it, preserving as much information as possible from the original space after dimensionality reduction.

[0062] set up dimensional vector Let the coordinate axes be the directions of the spatial coordinates to be transformed. To maximize the variance after data mapping, let...

[0063]

[0064] in The number of data samples. For the sample The vector representation of , This is the expectation for all data samples.

[0065] 4. Model Design

[0066] Let the preprocessed sample dataset be... Each sample corresponds to the sag data as follows: Let the sag calculation result obtained by the catenary method be... The sag correction amount corresponding to the sample data can be obtained. .

[0067] The following steps use a support vector regression machine with slack variables to obtain the sample data. With sag correction amount The relationship between them.

[0068] The support vector regression model with slack variables can be represented as follows:

[0069]

[0070] Introducing a kernel parameter representation method, It can be represented as

[0071]

[0072] in and For model parameters, As slack variables, Here is the regularization constant. For the Lagrange multiplier parameters, This is the kernel function.

[0073] 5. Model Training

[0074] The sample dataset for data preprocessing is as follows: And the sag data corresponding to each sample is Input the support vector regression machine model for training and obtain the model parameters. and In actual training, a kernel function with better accuracy should be selected based on the regression accuracy returned by the test set.

[0075] 6. Practical Application

[0076] In practical applications, the following data is obtained by measuring the sensors installed on the tower:

[0077] The system measures a wide range of data, including the ambient wind speed at the conductor's location, the ambient temperature at the conductor's location, the solar radiation intensity at the conductor's location, the kinematic viscosity of the air layer on the conductor's surface, the average temperature rise of the conductor's surface, the conductor's radiative heat dissipation power, the conductor's convective heat dissipation power, the conductor's solar heat absorption power, and the heat transfer coefficient of the air layer on the conductor's surface.

[0078] Input the obtained data into the parameter model obtained in step 5, and the trained support vector regression machine will return the sag correction parameters under the current parameters. And the preliminary estimation results of sag were obtained by the catenary method. Therefore, the optimized calculation results of the corrected sag can be obtained.

[0079]

[0080] Combination Figure 2 A schematic diagram of the scenario of the present invention is given. Compared with the estimated value of the catenary method, the estimated value obtained by the present method is closer to the true value of the conductor sag, thus improving the estimation accuracy.

[0081] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A support vector regression-based all-dielectric self-supporting optical cable conductor sag calculation optimization method, characterized by It includes the following steps: S1. Data Acquisition: Obtain raw data. , Indicates the first Data collected by the group; S2. Data cleaning, including outlier detection and filtering; S3. Data preprocessing to obtain the sample dataset. , Indicates the first preprocessed step Group data; S4. Model Design: The support vector regression machine model with slack variables is as follows: Represented as ,in, and For model parameters, As slack variables, Here is the regularization constant. For the Lagrange multiplier parameters, For kernel function, Represents the kernel function mapping function; Insensitive loss function; This represents the number of data sample groups; Sag correction parameters , This represents the sag calculation result obtained from the catenary method. Indicates by the first Group data Corresponding actual measurement data of conductor sag; S5. Model training, using the sample dataset and each set of sample data Corresponding sag data The model parameters are obtained by training the support vector regression machine model designed by S4. and ; S6. Use the model for sag optimization: Obtaining the sag calculation results using the catenary method ; Raw data is acquired through the set sensors, and the raw data is input into the support vector regression model obtained by S5 to obtain the sag correction parameters. ; Sag Optimization Results .

2. The method according to claim 1, characterized in that... In S1 and S6, the types of raw data collected include: the location of the conductor. Ambient wind speed, location of the conductor Ambient temperature, location of the conductor Sunlight intensity, conductor surface Kinematic viscosity of air layer, surface of conductor Average temperature rise, conductor Radiant heat dissipation power, wires Convection heat dissipation power, wires Solar heat absorption power, conductor surface The heat transfer coefficient of the air layer.

3. The method according to claim 1, characterized in that... In S2, outlier detection is performed based on the t-test, and the steps are as follows: Treating all original sample data as a whole that follows a normal distribution, calculate the expected value of the total original sample data. and standard deviation suspicious samples If a population with a total sample size of 1 has no significant difference from the rest of the populations, then it belongs to the same population as the original collected data. The t-test statistic is: If the statistic If the value is greater than the critical value of the t-test at the corresponding confidence level, It was determined to be an outlier.

4. The method according to claim 1, characterized in that... In S2, after a value is identified as an outlier, it is removed to correct the outlier.

5. The method according to claim 1, characterized in that... In S2, after an outlier is identified, the outlier data is corrected using a linear interpolation algorithm.

6. The method according to claim 1, characterized in that... In S3, data preprocessing includes data normalization and principal component analysis.

7. The method according to claim 6, characterized in that... Data normalization is performed using Min-Max standardization. The Min-Max standardization transformation function is as follows: in This represents the maximum value of each corresponding element in the sample data. This represents the minimum value of the sample data.

Images