Radio interference prediction method based on meteorological information granularity reduction and least squares support vector machine

Abstract

A radio interference prediction method based on granular reduction of meteorological information and a least squares support vector machine (LSSVM) algorithm comprises the following steps: Step 1: obtaining radio interference values ​​and corresponding meteorological information from a measured city over the past three years as a total data set based on actual transmission line measurements; Step 2: determining and filtering out data sets within extreme weather intervals from the total data set based on temperature criteria, and dividing the total data set into a training set and a validation set with a ratio of 9:1. This invention provides a radio interference prediction method based on granular reduction of meteorological information and a least squares support vector machine algorithm, which can relatively accurately predict radio interference values ​​and has good adaptability to extreme weather conditions, providing insights for the research of radio interference value prediction methods in practical engineering projects.

Description

Technical Field

[0001] The present invention relates to the field of electromagnetic compatibility of high-voltage power transmission and transformation projects, and in particular to a radio interference prediction method based on granular restoration of meteorological information and least squares support vector machine. Background Art

[0002] Compared to plains, high altitudes experience lower atmospheric pressure and air density, leading to a longer electron mean free path. This increases the energy accumulated before electron collisions, making effective impact ionization more likely and corona discharge more intense. Consequently, radio interference from high-voltage transmission lines is more severe. Radio interference from transmission lines can affect the amplitude and phase of useful signal waveforms received by radio receiving equipment within a certain range of the line during normal operation. Therefore, accurately predicting and suppressing the effects of radio interference from high-voltage AC transmission lines has become a major technical issue that must be considered in the design, construction, and operation of high-voltage AC transmission lines in high-altitude areas.

[0003] A meteorological information granule is a collection of meteorological elements that are bound together due to indistinguishability, similarity, proximity, or a certain function. Three main models for information granulation exist: those based on fuzzy set theory, those based on rough set theory, and those based on quotient space theory. These three models are closely related but also distinct. Fuzzy set theory and rough set theory are highly complementary, and the optimization and integration of these two theories have demonstrated greater power in addressing uncertainty and incompleteness in knowledge. Both quotient space theory and rough set theory use equivalence classes to describe granulation, which in turn describes concepts. The domain of rough set theory is simply a point set of objects; topological relationships between elements are not considered. Quotient space theory, on the other hand, focuses on spatial relationships, assuming that topological relationships exist between domain elements—that is, the domain is a topological space. In meteorological information granule reduction, the goal is to simplify predicted radio interference data by focusing on the key characteristics of information granules and ignoring irrelevant factors, effectively improving forecast accuracy.

[0004] In terms of radio interference prediction accuracy, although the prediction methods in existing research can better extract the interference value data characteristics and fit the interference value change law, thereby more accurately predicting the radio interference value, they fail to consider the impact of large meteorological information fluctuations caused by extreme weather on the interference value. Therefore, there is a situation where the prediction is inaccurate under extreme weather conditions. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a radio interference prediction method based on granular restoration of meteorological information and least squares support vector machine. By combining meteorological granular restoration with least squares support vector machine, the radio interference of conductors is predicted, which can improve the adaptability of the model to areas prone to extreme weather, thereby obtaining higher prediction accuracy.

[0006] To solve the above technical problems, the technical solution adopted by the present invention is: a radio interference prediction method based on granular restoration of meteorological information and least squares support vector machine, comprising the following steps:

[0007] Step 1: Obtain the radio interference values ​​and corresponding meteorological information of the measurement city in the past three years as a total data set according to the actual measurement method of the transmission line;

[0008] Step 2: Based on the temperature standard, we identify and filter out the datasets in the extreme weather range from the total dataset, and divide the total dataset into a training set and a validation set with a ratio of 9:1.

[0009] Step 3: Use the meteorological information granular restoration method to restore the meteorological information of the dataset within the extreme weather interval obtained in step 2 to obtain the overall meteorological information that is not affected by extreme weather;

[0010] Step 4: Using the overall meteorological information restored in step 3, a least squares support vector machine trained with the training set is used to predict the radio interference value within the extreme weather range, and together with the original data set, a basic model that is not affected by extreme weather interference is constructed;

[0011] Step 5: Consider the cumulative effect of temperature on the basic radio interference value and conduct a predictive modeling of the change in radio interference during extreme weather;

[0012] Step 6: Accumulate the radio interference prediction models obtained in step 4 and step 5 to obtain the final radio interference prediction model for the city, and use the validation set for testing.

[0013] Preferably, the atmospheric particle reduction method in step 3 is as follows:

[0014] (1) For each extreme temperature range affected by extreme weather, a reference restoration baseline value should be selected. The reference value should be the data every half hour under good weather conditions;

[0015] (2) Substitute the reference value into the meteorological particle restoration function to restore the meteorological information within the meteorological particles in the extreme weather interval:

[0016]

[0017] Where dC is the half-hourly meteorological information benchmark value; m is the number of meteorological data in a month that meets the benchmark value selection requirements, j = 1, 2…, m; l is the number of extreme weather data in the month; Dj is the jth successfully restored meteorological information;

[0018] (3) Using the successfully restored meteorological information, a least squares support vector machine is trained and prediction is performed. The meteorological information does not include the extreme weather meteorological information defined in the present invention.

[0019] Preferably, in step 4, the training optimization objective of the least squares vector machine used is:

[0020]

[0021] Among them, γ is a regularization parameter, which represents a penalty parameter for controlling the error; ω is a weight vector; φ(x i ) is the kernel function; b is the offset; e i is the error variable; x i is the i-th input variable, y i is the i-th output variable, i = 1, 2…, n;

[0022] After the introduction of the Lagrange multiplier α i After that, the optimization objective becomes a dual Lagrange polynomial:

[0023]

[0024] Among them, α i is the Lagrange multiplier of the i-th variable;

[0025] According to the Karush-Kuhn-Tucker (KKT) constraints, the following relationship can be obtained:

[0026]

[0027] in, They are function J for ω and e respectively. i , b, α i Operation to find partial derivatives.

[0028] Write equation (4) as a linear equation system:

[0029]

[0030] Where γ is the regularization parameter, I is the unit element, and the negative unit matrix L = -(1,1,…,1) T , error variable matrix e=(e1,e2,…,e n), input variable matrix y=(y1,y2,…,y n ) T , support vector machine coefficient matrix α=(α1,α2,…,α n ) T , kernel function matrix Z=(φ(x1),…,φ(x n )) T ,

[0031] After eliminating ω and e in equation (5), we can get i The relevant system of equations:

[0032]

[0033] Among them, ZZ T is the n-dimensional matrix of the kernel function, let ZZ T =g(x i ,x j ), g(x i ,x j ) is the kernel function of the least squares support vector machine. Different kernel functions determine different least squares support vector machines.

[0034] Due to the variability of climate, the nonlinear characteristics of the radio interference prediction model are more prominent, so the RBF radial basis function is selected as the kernel function of the vector machine, and its formula is:

[0035]

[0036] Among them, σ is the width of the kernel, x is the center point of the input variable, and x i is the i-th input variable, ||xx i || 2 is the bi-norm of the distance from the center point of the variable to the i-th variable, and exp(·) is the exponential operation function;

[0037] Finally, by solving equation (6), we can get the support vector coefficient α i and offset b;

[0038] The regression model of the least squares support vector machine finally used in the present invention is:

[0039]

[0040] Among them, K(x,x i ) radial basis function kernel function, α i is the support vector coefficient, and b is the offset.

[0041] Preferably, in step 5, the method for solving the variation of radio interference during extreme weather is:

[0042] (1) Calculate the time point of temperature mutation, that is, the temperature difference between the time when extreme weather starts and the temperature half an hour before and the temperature one hour before:

[0043] △T1=T0-T -1 (9)

[0044] △T2=T0-T -2 (10)

[0045] Among them, T0 is the temperature at the beginning of extreme weather; T -1 The temperature half an hour before the extreme weather; T -2 The temperature one hour before the extreme weather event;

[0046] (2) Based on the basic interference value one hour ago, the trained least squares support vector machine is used to predict the basic interference value half an hour ago, one hour ago, and at the time of the onset of extreme weather, and the difference △Φ between the basic interference value and the corresponding actual radio interference value is calculated;

[0047] (3) Based on △T1, △T2 and △Φ, the temperature cumulative effect is used to calculate the corrected radio interference value using the idea of ​​minimizing the residual error. The binary linear regression equation of △Φ changing with meteorological information is obtained by the least squares method:

[0048] △Φ=f(△T1,△T2)=q1△T1+q2△T2+q3 (11)

[0049] Among them, △T1 and △T2 are the temperature differences between the time when extreme weather starts and the previous half hour and one hour, respectively, and q1, q2, and q3 are the coefficients of the binary linear regression equation.

[0050] Preferably, in step 6, the radio interference prediction model finally obtained is:

[0051] Φ(x)=f(x)+f(△T1,△T2) (12)

[0052] Among them, Φ(x) is the final predicted radio interference value, f(x) is the preliminary predicted radio interference value, and f(△T1,△T2) is the cumulative effect of temperature.

[0053] Preferably, in step 6, the indicators used to verify the prediction model include:

[0054] Mean Absolute Error (MAE):

[0055]

[0056] Mean Absolute Percentage Error (MAPE):

[0057]

[0058] And the root mean square error RMSE:

[0059]

[0060] Wherein, actual(t) is the t-th actual measurement value, forecast(t) is the t-th predicted value, and t=1, 2…, N.

[0061] The present invention provides a radio interference prediction method based on granular restoration of meteorological information and least squares support vector machine, which can accurately predict radio interference values ​​and has good adaptability to extreme weather conditions, providing ideas for the research of radio interference value prediction methods in actual engineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] The present invention will be further described below with reference to the accompanying drawings and examples:

[0063] Figure 1 Schematic diagram of the method of the present invention;

[0064] Figure 2 The results of using least squares support vector machine to predict radio interference in the period without extreme weather;

[0065] Figure 3 The results of using least squares support vector machine to predict radio interference in extreme weather intervals;

[0066] Figure 4 This is a graph showing the radio interference results in extreme weather zones predicted using the final model. DETAILED DESCRIPTION

[0067] like Figure 1 As shown, a radio interference prediction method based on meteorological information granularity reduction and least squares support vector machine includes the following steps:

[0068] Step 1: Obtain the radio interference values ​​and corresponding meteorological information of the measurement city in the past three years as a total data set according to the actual measurement method of the transmission line;

[0069] Step 2: Based on the temperature standard, we identify and filter out the datasets in the extreme weather range from the total dataset, and divide the total dataset into a training set and a validation set with a ratio of 9:1.

[0070] Step 3: Use the meteorological information granular restoration method to restore the meteorological information of the dataset within the extreme weather interval obtained in step 2 to obtain the overall meteorological information that is not affected by extreme weather;

[0071] Step 4: Using the overall meteorological information restored in step 3, a least squares support vector machine trained with the training set is used to predict the radio interference value within the extreme weather range, and together with the original data set, a basic model that is not affected by extreme weather interference is constructed;

[0072] Step 5: Consider the cumulative effect of temperature on the basic radio interference value and conduct a predictive modeling of the change in radio interference during extreme weather;

[0073] Step 6: Accumulate the radio interference prediction models obtained in step 4 and step 5 to obtain the final radio interference prediction model for the city, and use the validation set for testing.

[0074] Taking the city of Golmud in Qinghai, which is 2,260 meters above sea level on the Qinghai-Tibet Plateau, as an example, extreme weather refers to extreme rainstorms, heavy snow, droughts and other weather conditions that cause drastic changes in temperature.

[0075] The data acquisition method and data composition are as follows: The German SCHWARZBECK FCKL1528 electromagnetic interference measurement receiver and the US HOBO U30 small weather station were used. The weather station acquired data every 1 minute, and the radio interference measurement system acquired data every 4 seconds. The data for each minute was obtained by averaging 15 consecutive data points. Radio interference values ​​and corresponding meteorological values ​​such as temperature, humidity, and air pressure were measured in minute increments in the Golmud area of ​​Qinghai Province, and the average values ​​were taken in half-hour increments to form the final data set.

[0076] The standard for judging whether the selected data belongs to extreme weather is: the temperature data is less than or equal to -10℃, or greater than or equal to 20℃.

[0077] The ratio of the training set and the test set is 9:1.

[0078] A weather granule refers to a collection of weather information that has an impact on radio interference. Specifically, in the present invention, a weather granule only contains all weather data values ​​every half hour.

[0079] Preferably, the atmospheric particle reduction method in step 3 is as follows:

[0080] (1) For each extreme temperature range affected by extreme weather, a reference restoration baseline value should be selected. The reference value should be the data every half hour under good weather conditions;

[0081] (2) Substitute the reference value into the meteorological particle restoration function to restore the meteorological information within the meteorological particles in the extreme weather interval:

[0082]

[0083] Where dC is the half-hourly meteorological information benchmark value; m is the number of meteorological data in a month that meets the benchmark value selection requirements, j = 1, 2…, m; l is the number of extreme weather data in the month; Dj is the jth successfully restored meteorological information;

[0084] (3) Using the successfully restored meteorological information, a least squares support vector machine is trained and prediction is performed. The meteorological information does not include the extreme weather meteorological information defined in the present invention.

[0085] Preferably, in step 4, the training optimization objective of the least squares vector machine used is:

[0086]

[0087] Among them, γ is a regularization parameter, which represents a penalty parameter for controlling the error; ω is a weight vector; φ(x i ) is the kernel function; b is the offset; e i is the error variable; x i is the i-th input variable, y i is the i-th output variable, i = 1, 2…, n;

[0088] After the introduction of the Lagrange multiplier α i After that, the optimization objective becomes a dual Lagrange polynomial:

[0089]

[0090] Among them, α i is the Lagrange multiplier of the i-th variable;

[0091] According to the Karush-Kuhn-Tucker (KKT) constraints, the following relationship can be obtained:

[0092]

[0093] in, They are function J for ω and e respectively. i , b, α i Operation to find partial derivatives.

[0094] Write equation (4) as a linear equation system:

[0095]

[0096] Where γ is the regularization parameter, I is the unit element, and the negative unit matrix L = -(1,1,…,1) T , error variable matrix e=(e1,e2,…,e n), input variable matrix y=(y1,y2,…,y n ) T , support vector machine coefficient matrix α=(α1,α2,…,α n ) T , kernel function matrix Z=(φ(x1),…,φ(x n )) T ,

[0097] After eliminating ω and e in equation (5), we can get i The relevant system of equations:

[0098]

[0099] Among them, ZZ T is the n-dimensional matrix of the kernel function, let ZZ T =g(x i ,x j ), g(x i ,x j ) is the kernel function of the least squares support vector machine. Different kernel functions determine different least squares support vector machines.

[0100] Due to the variability of climate, the nonlinear characteristics of the radio interference prediction model are more prominent, so the RBF radial basis function is selected as the kernel function of the vector machine, and its formula is:

[0101]

[0102] Among them, σ is the width of the kernel, x is the center point of the input variable, and x i is the i-th input variable, ||xx i || 2 is the bi-norm of the distance from the center point of the variable to the i-th variable, and exp(·) is the exponential operation function;

[0103] Finally, by solving equation (6), we can get the support vector coefficient α i and offset b;

[0104] The regression model of the least squares support vector machine finally used in the present invention is:

[0105]

[0106] Among them, K(x,x i ) radial basis function kernel function, α i is the support vector coefficient, and b is the offset.

[0107] Preferably, in step 5, the method for solving the variation of radio interference during extreme weather is:

[0108] (1) Calculate the time point of temperature mutation, that is, the temperature difference between the time when extreme weather starts and the temperature half an hour before and the temperature one hour before:

[0109] △T1=T0-T -1 (9)

[0110] △T2=T0-T -2 (10)

[0111] Among them, T0 is the temperature at the beginning of extreme weather; T -1 The temperature half an hour before the extreme weather; T -2 The temperature one hour before the extreme weather event;

[0112] (2) Based on the basic interference value one hour ago, the trained least squares support vector machine is used to predict the basic interference value half an hour ago, one hour ago, and at the time of the onset of extreme weather, and the difference △Φ between the basic interference value and the corresponding actual radio interference value is calculated;

[0113] (3) Based on △T1, △T2 and △Φ, the temperature cumulative effect is used to calculate the corrected radio interference value using the idea of ​​minimizing the residual error. The binary linear regression equation of △Φ changing with meteorological information is obtained by the least squares method:

[0114] △Φ=f(△T1,△T2)=q1△T1+q2△T2+q3 (11)

[0115] Among them, △T1 and △T2 are the temperature differences between the time when extreme weather starts and the previous half hour and one hour, respectively, and q1, q2, and q3 are the coefficients of the binary linear regression equation.

[0116] Preferably, in step 6, the radio interference prediction model finally obtained is:

[0117] Φ(x)=f(x)+f(△T1,△T2) (12)

[0118] Among them, Φ(x) is the final predicted radio interference value, f(x) is the preliminary predicted radio interference value, and f(△T1,△T2) is the cumulative effect of temperature.

[0119] Preferably, in step 6, the indicators used to verify the prediction model include:

[0120] Mean Absolute Error (MAE):

[0121]

[0122] Mean Absolute Percentage Error (MAPE):

[0123]

[0124] And the root mean square error RMSE:

[0125]

[0126] Wherein, actual(t) is the t-th actual measurement value, forecast(t) is the t-th predicted value, and t=1, 2…, N.

[0127] To verify the effectiveness of the radio interference prediction method based on granular restoration of meteorological information and least squares support vector machine in predicting radio interference values ​​in extreme weather areas, this specific embodiment mainly sets up three groups of experiments to predict two groups of one-day interference data:

[0128] The first set of experiments used the data of one day without extreme weather in Golmud, Qinghai, Tibet, and used the least squares support vector machine to make predictions and obtain the results;

[0129] The first set of experiments used the interference data of one day with extreme weather in Golmud, Qinghai, Tibet, and used the least squares support vector machine to make predictions and obtain the results;

[0130] The third group of experiments used the interference data of one day with extreme weather in Golmud, Qinghai, Tibet, as the object, and used the final prediction model proposed in the present invention to make predictions and obtain the results.

[0131] like Figure 2 The following is the prediction result of Experiment 1. It can be seen from the figure that the prediction effect is good. Its various indicators are shown in Table 1:

[0132] Table 1 Experiment 1 prediction result evaluation table

[0133]

[0134] This experiment proves that after the meteorological information of the original data set is restored through meteorological particle restoration, the trained least squares support vector machine has a high prediction accuracy for the radio interference value in the non-extreme meteorological range and has good practicality.

[0135] like Figure 3 The following table shows the prediction results of Experiment 2. It can be seen that the trained least squares support vector machine is not very effective in predicting the interference value in the extreme weather period. This is because the climate change caused by extreme weather leads to a drastic change in the interference value. Its various indicators are shown in Table 2:

[0136] Table 2 Experiment 2 prediction results evaluation table

[0137]

[0138] like Figure 4The following are the prediction results of Experiment 3, and the various indicators are shown in Table 3:

[0139] Table 3 Experiment 3 prediction result evaluation table

[0140]

[0141] It can be seen that after taking into account the variation of interference values ​​caused by extreme weather, the final prediction model of the present invention greatly improves the accuracy of radio interference prediction within the extreme weather range in Tibetan cities.

[0142] The above embodiments are merely preferred technical solutions of the present invention and should not be construed as limiting the present invention. The scope of protection of the present invention shall be the technical solutions set forth in the claims, including equivalent alternatives to the technical features of the technical solutions set forth in the claims. In other words, equivalent alternatives and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A radio interference prediction method based on granular restoration of meteorological information and least squares support vector machine, characterized in that: The following steps are included: Step 1: Obtain the radio interference values ​​and corresponding meteorological information of the measurement city in the past three years as a total data set according to the actual measurement method of the transmission line; Step 2: Based on the temperature standard, we identify and filter out the datasets in the extreme weather range from the total dataset, and divide the total dataset into a training set and a validation set with a ratio of 9:

1. Step 3: Use the meteorological information granular restoration method to restore the meteorological information of the dataset within the extreme weather interval obtained in step 2 to obtain the overall meteorological information that is not affected by extreme weather; Step 4: Using the overall meteorological information restored in step 3, a least squares support vector machine trained with the training set is used to predict the radio interference value within the extreme weather range, and together with the original data set, a basic model that is not affected by extreme weather interference is constructed; Step 5: Consider the cumulative effect of temperature on the basic radio interference value and conduct a predictive modeling of the change in radio interference during extreme weather; Step 6: Accumulate the radio interference prediction models obtained in step 4 and step 5 to obtain the final radio interference prediction model for the city, and use the validation set for testing; The method for reducing the atmospheric particles in step 3 is as follows: (1) For each extreme temperature range affected by extreme weather, a reference restoration baseline value should be selected. The baseline value should be the data every half hour under good weather conditions; (2) Substitute the reference value into the meteorological particle restoration function to restore the meteorological information within the meteorological particles in the extreme weather interval: ；（1） Where dC is the half-hourly meteorological information benchmark value; m is the number of meteorological data in a month that meets the benchmark value selection requirements, j = 1, 2…, m; l is the number of extreme weather data in the month; Dj is the jth successfully restored meteorological information; (3) Using the successfully restored meteorological information, train the least squares support vector machine and make predictions. The meteorological information does not contain the extreme weather information defined in step 2. In step 5, the method for solving the change in radio interference during extreme weather is: (1) Calculate the time point of temperature mutation, that is, the temperature difference between the time when extreme weather starts and the temperature half an hour before and the temperature before one hour: ；（9） ；（10） in, The temperature at the time when the extreme weather started; The temperature half an hour before the extreme weather; The temperature one hour before the extreme weather event; (2) Based on the basic interference value one hour ago, the trained least squares support vector machine predicts the basic interference value half an hour ago, one hour ago, and the time when the extreme weather starts, and calculates the difference between the basic interference value and the corresponding real radio interference value. ; (3) Based on 、 as well as The idea of ​​minimizing the residual error is used to calculate the temperature cumulative effect to correct the radio interference value, and the least squares method is used to obtain The binary linear regression equation that changes with meteorological information: ；（11） in, 、 The temperature differences between the time when extreme weather starts and the previous half hour and the previous hour, respectively. 、 、 is the coefficient of the binary linear regression equation.

2. The radio interference prediction method based on granular restoration of meteorological information and least squares support vector machine according to claim 1 is characterized in that: In step 4, the training optimization objective of the least squares vector machine used is: ；（2） in, is a regularization parameter, which represents a penalty parameter for controlling the error; is the weight vector; is the kernel function; b is the offset; is the error variable; is the i-th input variable, is the i-th output variable, i=1, 2…, n; After the introduction of the Lagrange multiplier After that, the optimization objective becomes a dual Lagrange polynomial: ；（3） in, is the Lagrange multiplier of the i-th variable; According to the Karush-Kuhn-Tucker (KKT) constraints, the following relationship can be obtained: ；（4） in, 、 、 、 Function right 、 、 、 Operations to find partial derivatives; Write equation (4) as a system of linear equations: ；（5） in, is the regularization parameter, is the unit element, the negative unit matrix , error variable matrix , input variable matrix , support vector machine coefficient matrix , kernel function matrix , Eliminate the equation (5) and After that, we get The relevant system of equations: ； （6） in, is the n-dimensional square matrix of the kernel function, let , It is the kernel function of the least squares support vector machine. Different kernel functions determine different least squares support vector machines. Due to the variability of climate, the nonlinear characteristics of the radio interference prediction model are more prominent, so the RBF radial basis function is selected as the kernel function of the vector machine, and its formula is: ；（7） in, is the width of the kernel, is the center point of the input variable, is the i-th input variable, is the two-norm of the distance from the center point of the variable to the i-th variable, is the exponential operation function; Finally, by solving equation (6), we can get the support vector coefficient and offset b; Least squares support vector machine regression model: ；（8） in, Radial basis function kernel function, is the support vector coefficient, and b is the offset.

3. The radio interference prediction method based on granular restoration of meteorological information and least squares support vector machine according to claim 1 is characterized in that: In step 6, the radio interference prediction model finally obtained is: ；（12） in, is the final predicted radio interference value, is the preliminary predicted radio interference value, is the cumulative effect of temperature.

4. The radio interference prediction method based on granular restoration of meteorological information and least squares support vector machine according to claim 1 is characterized in that: In step 6, the indicators used to verify the prediction model include: Mean Absolute Error (MAE): ；（13） Mean Absolute Percentage Error (MAPE): ；（14） And the root mean square error RMSE: ；（15） in, is the tth actual measurement value, is the t-th predicted value, t=1, 2…, N.

Images