Method for predicting adhesive force of ice layer on surface of icing insulator of power transmission line

Abstract

The invention provides a method for predicting the adhesive force of an ice layer on the surface of an icing insulator of a power transmission line. The method comprises the following steps: S1, obtaining an icing influence parameter of a sample insulator and the adhesive force of the ice layer of the icing of the sample insulator; s2, determining correlation coefficients of the icing influence parameters and the ice layer adhesive force, and selecting the icing influence parameters of which the correlation coefficients are greater than a set value as sample parameters; s3, constructing a random forest regression model, and inputting the sample parameters into the random forest regression model for training; s4, detecting icing influence parameters of a to-be-detected insulator of the power transmission line, selecting a parameter consistent with the sample parameter type from the icing influence parameters of the to-be-detected insulator as a prediction parameter, and inputting the prediction parameter into the trained random forest regression model to obtain the ice layer adhesive force of the to-be-detected insulator; accurate data support is provided for formulation of insulator surface deicing measures, and operation stability and operation safety of a power transmission line are ensured.

Description

Technical Field

[0001] The present invention relates to a method for predicting the icing state of a transmission line, and particularly to a method for predicting the ice adhesion force on the surface of an ice-covered insulator of a transmission line. Background Art

[0002] When the temperature near the ground is between 0°C and -15°C, the relative air humidity exceeds 80%, and the wind speed is greater than 1 m / s, supercooled water is likely to form ice when it contacts the surface of the insulator; the formation of ice will increase the weight of the insulator, change the mechanical stress of the transmission line, reduce the insulation performance, increase the risk of mechanical failures and ice flash accidents of the transmission line, and seriously endanger the safe and stable operation of the power system.

[0003] The ice adhesion force of the ice-covered insulator is related to the de-icing difficulty, the ice growth mode, and the line failure risk. In the prior art, the treatment of the ice-covered insulator is monitored based on images, the tension of the transmission line, etc., and then de-icing measures are taken. However, in this process, the factor of ice adhesion force is not considered, which often leads to incomplete de-icing of the insulator during the de-icing process, and the surface of the insulator will quickly form ice again after de-icing, which is not conducive to the stable operation of the transmission line.

[0004] Therefore, in order to solve the above technical problems, it is urgent to propose a new technical means. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a method for predicting the ice adhesion force on the surface of an ice-covered insulator of a transmission line, which can accurately predict the ice adhesion force of the insulator according to the changes of the environmental parameters, the ice shape and thickness of the insulator to be measured, so as to provide accurate data support for formulating the de-icing measures on the surface of the insulator, which is conducive to thoroughly de-icing the surface of the insulator and ensuring the operation stability and safety of the transmission line.

[0006] A method for predicting the ice adhesion force on the surface of an ice-covered insulator of a transmission line provided by the present invention includes the following steps:

[0007] S1. Obtain the icing influence parameters of the sample insulator and the ice adhesion force of the ice-covered sample insulator;

[0008] S2. Determine the correlation coefficient between the icing influence parameters and the ice adhesion force, and select the icing influence parameters with a correlation coefficient greater than the set value as the sample parameters;

[0009] S3. Construct a random forest regression model, and input the sample parameters into the random forest regression model for training;

[0010] S4. Detect the icing influence parameters of the insulators to be measured on the transmission line, select the parameters with the same type as the sample parameters from the icing influence parameters of the insulators to be measured as the prediction parameters, and input the prediction parameters into the trained random forest regression model to obtain the ice adhesion force of the insulators to be measured.

[0011] Furthermore, the icing influence parameters of the sample insulators include quantitative parameters and qualitative parameters; the quantitative parameters include environmental temperature, environmental humidity, wind speed, insulator voltage level, ice thickness, and ice temperature; the qualitative parameters include insulator material type and ice shape; among them, the quantitative parameters are used for the calculation of the correlation coefficient, and the qualitative parameters are used to determine the target insulators to be measured.

[0012] Furthermore, determining the correlation coefficient between the icing influence parameters and the ice adhesion force specifically includes:

[0013]

[0014] Among them: X represents the quantitative parameter of the icing influence parameter, Y represents the ice adhesion force under the current icing influence parameter; Cov(X,Y) represents the covariance, S X 、S Y respectively represent the standard deviations of X and Y, where:

[0015]

[0016] Among them: N represents the number of parameters, represents the average value of X, represents the average value of Y, x i represents the i-th qualitative parameter, y i represents the i-th parameter in Y.

[0017] Furthermore, constructing the random forest regression model specifically includes:

[0018] Call RandomForestRegressor based on the scikit-learn library of the Python platform to construct the random forest regression model;

[0019] Among them: The initial value of the decision tree of the random forest regression model is set to 100, and it is decreased by the set value each time until the decision tree curve of the random forest regression model shows overfitting at a certain point, and then the iteration stops, and the value of the decision tree before overfitting is used as the value of the random forest regression model.

[0020] Furthermore, the completion of the training of the random forest regression model is determined by the following method:

[0021] Divide the qualitative parameters into a test set and a training set, and input the test set and the training set into the random forest regression model;

[0022] Calculate the root mean square error, mean absolute error, and coefficient of determination of the random forest regression model:

[0023]

[0024] where: n is the amount of data in the test set; y i is the actual measured value, is the predicted value; MSE represents the root mean square error, MAE represents the mean absolute error, and R 2 represents the coefficient of determination;

[0025] When both the root mean square error and the mean absolute error are less than the set value or the coefficient of determination is greater than the set value, the training of the random forest regression model is completed.

[0026] Advantages of the present invention: Through the present invention, according to the changes in the environmental parameters of the insulator to be measured, as well as the shape and thickness of the ice coating, the adhesion force of the ice layer on the insulator can be accurately predicted, thereby providing accurate data support for the formulation of ice removal measures on the insulator surface, facilitating thorough ice removal on the insulator surface, and ensuring the operation stability and safety of the transmission line. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The present invention will be further described below in conjunction with the drawings and embodiments:

[0028] Figure 1 is the flow chart of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0029] The following further elaborates on the present invention in detail:

[0030] A method for predicting the adhesion force of the ice layer on the surface of an ice-covered insulator of a transmission line provided by the present invention includes the following steps:

[0031] S1. Obtain the ice-covered influence parameters of the sample insulator and the adhesion force of the ice layer on the sample insulator;

[0032] S2. Determine the correlation coefficient between the ice-covered influence parameters and the adhesion force of the ice layer, and select the ice-covered influence parameters with a correlation coefficient greater than the set value as the sample parameters;

[0033] S3. Construct a random forest regression model, and input the sample parameters into the random forest regression model for training;

[0034] S4. Detect the icing influence parameters of the insulator to be measured on the transmission line, select the parameters consistent with the sample parameter types from the icing influence parameters of the insulator to be measured as the prediction parameters, and input the prediction parameters into the trained random forest regression model to obtain the ice adhesion force of the insulator to be measured. Through the above method, according to the changes in the environmental parameters, icing shape and thickness of the insulator to be measured, the ice adhesion force of the insulator can be accurately predicted, thereby providing accurate data support for the de-icing measures on the insulator surface, providing accurate data support for the formulation of subsequent de-icing measures, facilitating thorough de-icing of the insulator surface, and ensuring the operation stability and safety of the transmission line.

[0035] In this embodiment, the icing influence parameters of the sample insulator include quantitative parameters and qualitative parameters; the quantitative parameters include environmental temperature, environmental humidity, wind speed, insulator voltage level, icing thickness and icing temperature; the qualitative parameters include insulator material type and icing shape; among them, the quantitative parameters are used for calculating the correlation coefficient, and the qualitative parameters are used to determine the target insulator to be measured. For example, if the materials of the sample insulators are porcelain, glass, silicone rubber, etc., then when conducting icing experiments on the sample insulators, icing experiments need to be carried out on different materials respectively, and then the adhesion force test is carried out. The random forest regression model is trained through the response data. Then, when the insulator to be measured is a porcelain insulator, the random forest regression model trained with the data of porcelain sample insulators is selected for prediction.

[0036] In this embodiment, determining the correlation coefficient between the icing influence parameter and the ice adhesion force specifically includes:

[0037]

[0038] Among them: X represents the quantitative parameter of the icing influence parameter, Y represents the ice adhesion force under the current icing influence parameter; Cov(X,Y) represents the covariance, S X 、S Y respectively represent the standard deviations of X and Y, where:

[0039]

[0040] Among them: N represents the number of parameters, represents the average value of X, represents the average value of Y, x i represents the i-th qualitative parameter, y i represents the i-th parameter in Y.

[0041] In this embodiment, constructing the random forest regression model specifically includes:

[0042] The RandomForestRegressor in the scikit-learn library based on the Python platform is called to construct a random forest regression model;

[0043] Among them: The initial value of the decision tree of the random forest regression model is set to 100, and it decreases by the set value each time until the decision tree curve of the random forest regression model shows overfitting at a certain point, then the iteration stops, and the value of the decision tree before the overfitting occurs is used as the value of the random forest regression model.

[0044] In this embodiment, the random forest regression model is determined to be trained through the following method:

[0045] The qualitative parameters are divided into a test set and a training set, and the test set and the training set are input into the random forest regression model; among them, when dividing, 70% of the sample data is used as the training set, 10% of the data is used as the test set, and then 20% of the sample data is used as the validation set;

[0046] Calculate the root mean square error, mean absolute error, and coefficient of determination of the random forest regression model:

[0047]

[0048] Among them: n is the amount of data in the test set; y i is the actual measured value, is the predicted value; MSE represents the root mean square error, MAE represents the mean absolute error, R 2 represents the coefficient of determination, where, R 2 The value range is from 0 to 1, and the closer it is to 1, the better the fitting degree of the random forest regression model to the data.

[0049] When both the root mean square error and the mean absolute error are less than the set value or the coefficient of determination is greater than the set value, the random forest regression model is trained.

[0050] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the purpose and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting the adhesion force of ice layer on the surface of icing insulators of transmission lines, characterized in that: Including the following steps: S1. Obtain the icing influence parameters of the sample insulator and the ice layer adhesion force of the sample insulator under icing; S2. Determine the correlation coefficient between the icing influence parameters and the ice layer adhesion force, and select the icing influence parameters with a correlation coefficient greater than the set value as sample parameters; S3. Construct a random forest regression model, and input the sample parameters into the random forest regression model for training; S4. Detect the icing influence parameters of the insulators to be measured on the transmission line, select the parameters with the same type as the sample parameters from the icing influence parameters of the insulators to be measured as prediction parameters, and input the prediction parameters into the trained random forest regression model to obtain the ice layer adhesion force of the insulators to be measured.

2. The method for predicting the adhesion force of ice layer on the surface of ice-covered insulators of transmission lines according to claim 1, wherein: The icing influence parameters of the sample insulator include quantitative parameters and qualitative parameters; the quantitative parameters include ambient temperature, ambient humidity, wind speed, insulator voltage level, icing thickness, and icing temperature; the qualitative parameters include insulator material type and icing shape; among them, the quantitative parameters are used for calculating the correlation coefficient, and the qualitative parameters are used to determine the target insulators to be measured.

3. The method for predicting the adhesion force of ice layer on the surface of an ice-covered insulator of a transmission line according to claim 2, wherein: Determining the correlation coefficient between the icing influence parameters and the ice layer adhesion force specifically includes: Where: X represents the quantitative parameter of the ice accretion influence parameter, Y represents the ice layer adhesion force under the current ice accretion influence parameter; Cov(X, Y) represents the covariance, S X and S Y respectively represent the standard deviations of X and Y, where: Where: N represents the number of parameters, represents the average value of X, represents the average value of Y, x i represents the i-th qualitative parameter, y i represents the i-th parameter in Y.

4. The method for predicting the adhesion force of ice layer on the surface of ice-covered insulators of transmission lines according to claim 1, wherein: Constructing the random forest regression model specifically includes: Calling RandomForestRegressor based on the scikit-learn library of the Python platform to construct a random forest regression model; Wherein: the initial value of the decision tree of the random forest regression model is set to 100, and it is decreased by the set value each time until the decision tree curve of the random forest regression model shows overfitting at a certain point, and then the iteration stops, and the value of the decision tree of the previous time before overfitting is used as the value of the random forest regression model.

5. The method for predicting the adhesion force of ice layer on the surface of icing insulators of transmission lines according to claim 1, wherein: The completion of the training of the random forest regression model is determined by the following method: Divide the qualitative parameters into a test set and a training set, and input the test set and the training set into the random forest regression model; Calculate the root mean square error, mean absolute error, and coefficient of determination of the random forest regression model: Where: n is the amount of data in the test set; y i is the actual measured value, is the predicted value; MSE represents the root mean square error, MAE represents the mean absolute error, R 2 represents the coefficient of determination; When both the root mean square error and the mean absolute error are less than the set value or the coefficient of determination is greater than the set value, the training of the random forest regression model is completed.

Images