A method and system for predicting icing on transmission lines based on multi-source satellite remote sensing

Abstract

The present invention discloses a method and system for predicting icing on transmission lines based on multi-source satellite remote sensing, belonging to the technical field of predicting icing on transmission lines. The method includes: obtaining a historical data set; normalizing the indicators of remote sensing data; obtaining the misclassification rate of the indicators based on the random forest method; screening the main indicators according to the misclassification rate; screening the historical data set according to the main indicators to obtain a training set; training using the training set based on the multi-core relevance vector machine method to obtain an icing prediction model for transmission lines; predicting the remote sensing data according to the icing prediction model to obtain the icing thickness. Based on the random forest-multi-core relevance vector machine method, a multi-source satellite remote sensing data and actual icing data are used for training to construct an icing prediction model, and the icing prediction model is used to predict the icing thickness according to the remote sensing data; multiple kernel functions are combined to construct the icing prediction model to enhance the prediction effect of the model.

Description

Technical Field

[0001] The present invention relates to the technical field of transmission line icing prediction, and particularly relates to a transmission line icing prediction method and system based on multi-source satellite remote sensing. Background Art

[0002] Icing on transmission lines often causes power accidents, seriously affecting the normal operation of the power system. Light icing on transmission lines can lead to flashovers and tripping of transmission and substation equipment, while heavy icing can cause overhead transmission line breakage, pole collapse, and even large-scale power grid paralysis.

[0003] In response to many problems caused by transmission line icing, since the 1950s, technical forces have been successively invested at home and abroad to observe and study transmission line icing, and explore the icing mechanism, formation conditions, and prediction methods of transmission lines. At present, the existing icing prediction models mainly fall into three categories: physical theory models, statistical models, and intelligent models. Some important parameters of physical theory models are difficult to measure in conventional meteorological observations, and there are many limitations in practical applications. Due to the complexity of the indicators affecting icing, it is also difficult to achieve through statistical prediction methods for statistical models. Intelligent models are currently the best-performing models in icing prediction work. However, in practical applications, due to insufficient data sources and measurement technology limitations, existing models have disadvantages such as low prediction accuracy. Summary of the Invention

[0004] In view of the above technical problems existing in the prior art, the present invention provides a transmission line icing prediction method and system based on multi-source satellite remote sensing. By using the random forest method to screen the main indicators and the multi-kernel relevance vector machine method to establish an icing prediction model, high-precision and high-efficiency transmission line icing prediction can be achieved.

[0005] The present invention discloses a transmission line icing prediction method based on multi-source satellite remote sensing. The method includes: obtaining a historical data set, where the historical data set includes satellite remote sensing data and the corresponding transmission line icing thickness; normalizing the indicators of the remote sensing data to obtain the normalized values of the indicators; obtaining the misclassification rate of the indicators based on the random forest method; screening the main indicators according to the misclassification rate; screening the historical data set according to the main indicators to obtain a training set; training using the training set based on the multi-kernel relevance vector machine method to obtain an icing prediction model for the transmission line; and predicting the remote sensing data according to the icing prediction model to obtain the icing thickness.

[0006] Preferably, the normalized value is expressed as:

[0007]

[0008] where a iis the value of the index, a min is the minimum value of the index, a max is the maximum value of the index, A i is the normalized value.

[0009] Preferably, the method for screening the main indicators includes:

[0010] Sort the indicators from small to large according to the misclassification rate;

[0011] Select the threshold C according to the indicators, and take the first C indicators as the main indicators.

[0012] Preferably, the training set D is expressed as:

[0013] D = {(x1,y1),…,(x i ,y i ),…,(x N ,y N )} (3)

[0014] where N represents the number of samples, x i represents the column vector composed of the main indicator eigenvalue at a certain moment, and y i represents the ice coating thickness under the condition of x i .

[0015] Preferably, the ice coating prediction model is expressed as:

[0016]

[0017] where y i represents the ice coating thickness under the condition of x i , K C (x i ) is the kernel function of the ice coating prediction model, x i represents the column vector composed of the indicator eigenvalue, ω i is the kernel function weight, ε i represents the additive noise, and ε i obeys the zero-mean Gaussian distribution.

[0018] Preferably, the kernel function is expressed as:

[0019] K C (x i ) = λ·K G (x i )+(1-λ)·K P (x i )

[0020]

[0021] K P (xi ) = (r1·x i + r2) r3 (41)

[0022] where λ is the weight parameter, and K G (x i ) represents the polynomial kernel function, and K P (x i ) is the Gaussian kernel function. r1, r2, r3, r g are the parameters of the kernel respectively.

[0023] Preferably, the method for calculating the ice accretion amount on the transmission line:

[0024] M i = πbk(d + bk)ρ × 10 -3

[0025] b = (d i - d) / 2 (5)

[0026] where M i is the ice accretion amount per unit length of the conductor, ρ is the density of ice, d is the conductor diameter, d i is the diameter of the conductor after ice accretion, k is the conductor diameter coefficient, and b is the ice accretion thickness.

[0027] Preferably, the method for quantitatively describing the transmission line ice disaster process:

[0028] According to the load per unit length of the conductor, preset the ice accretion amount interval of the conductor and the corresponding disaster level;

[0029] According to the calculated ice accretion amount on the transmission line, match the corresponding disaster level;

[0030] According to the disaster level, use the corresponding identifier or color to display on the map.

[0031] The present invention also provides a system for implementing the above-mentioned transmission line ice prediction method, including an acquisition module, a normalization module, an index screening module, a training module, and a prediction module.

[0032] The acquisition module is used to obtain the historical data set;

[0033] The normalization module is used to normalize the indexes of the remote sensing data to obtain the normalized values of the indexes;

[0034] The index screening module is used to obtain the misclassification rate of the indexes based on the random forest method, and screen the main indexes according to the misclassification rate;

[0035] The training module is used to screen the historical data set according to the main indicators to obtain a training set; based on the method of multi-kernel relevance vector machine, the training set is used for training to obtain an icing prediction model for transmission lines;

[0036] The prediction module is used to predict the remote sensing data according to the icing prediction model to obtain the corresponding icing thickness.

[0037] Preferably, the system further includes an ice amount prediction module and a disaster prediction module,

[0038] The ice amount prediction module is used to calculate the ice amount on the conductor according to the icing thickness;

[0039] The disaster prediction module is used to obtain the corresponding disaster level of the ice amount according to the preset ice amount interval and its disaster program.

[0040] Compared with the prior art, the beneficial effects of the present invention are as follows: Based on the method of random forest - multi-kernel relevance vector machine (RF-MKRVM), multi-source satellite remote sensing data and actual icing data are used for training to construct an icing prediction model. The icing prediction model is used to predict the icing thickness according to the remote sensing data; multiple kernel functions are combined to construct the icing prediction model to enhance the prediction effect of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flowchart of the method for predicting icing on transmission lines of the present invention;

[0042] Figure 2 is a system logic block diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0044] The following further describes the present invention in detail with reference to the accompanying drawings:

[0045] A method for predicting icing on transmission lines based on multi-source satellite remote sensing, as Figure 1 shown, the method includes:

[0046] Step 101: Obtain a historical data set, which includes satellite remote sensing data and the corresponding ice coating thickness of transmission lines. The satellite remote sensing data includes time series data and non-time series data. The time series data includes meteorological temperature sensing data, such as humidity, temperature, wind speed, wind direction, etc.; the non-time series data includes high-resolution optical temperature sensing data and other temperature sensing data, such as line location, underlying surface type, altitude, slope, aspect (determined by combining elevation, slope and aspect information), and line elevation. It should be noted that the preprocessing of historical data is prior art, such as removing abnormal data and standardizing the input format of historical data, which will not be elaborated in this application. In a specific embodiment, historical data of overhead transmission lines in the northern mountainous area of Hebei is collected.

[0047] Step 102: Normalize the indicators of the remote sensing data to obtain the normalized values of the indicators. The normalized value is expressed as:

[0048]

[0049] where a i is the value of the indicator, a min is the minimum value of the indicator, a max is the maximum value of the indicator, and A i is the normalized value.

[0050] Step 103: Based on the random forest (RF) method, obtain the out-of-bag (OOB) error rate of the indicators. The formula for the error rate is expressed as: OOB error rate = number of misclassified / total number of samples. The error rate is an unbiased estimate of the generalization error of the random forest, and its result is approximately equal to the k-fold cross-validation that requires a large amount of calculation. The calculation method of the error rate is prior art and will not be elaborated in this application.

[0051] Step 104: Screen the main indicators according to the error rate.

[0052] Among them, the method for screening the main indicators is: sort the indicators from small to large according to the error rate; select the threshold C according to the indicators, and take the first C items of indicators as the main indicators. C is a positive integer, and the threshold C can be judged and selected according to the actual effect. In a specific embodiment, the main indicators include: temperature, humidity, wind speed, elevation at the location of the transmission line, the included angle between the transmission line and the wind direction (the included angle range is [0°, 90°]), the aspect of the transmission line, and the closest distance between the transmission line and lakes, rivers and other various water bodies, but not limited to this. The above main indicators are the main factors for water vapor sublimation.

[0053] Step 105: Screen the historical data set according to the main indicators to obtain a training set. The training set D is expressed as:

[0054] D = {(x1, y1), …, (x i , yi ),…,(x N ,y N )} (3)

[0055] Among them, N represents the number of samples, and x i represents the column vector composed of the main index characteristic values at a certain moment, and y i represents the ice coating thickness under the condition of x i .

[0056] Step 106: Based on the method of multi-kernel relevance vector machine (MKRVM), use the training set for training to obtain the ice coating prediction model of the transmission line.

[0057] The Relevance Vector Machine (RVM) is a machine learning model based on the Markov property, Bayes theory and maximum likelihood. Compared with traditional machine learning algorithms, it has stronger robustness. The kernel function is the key of the RVM model. In the prediction process, a suitable kernel function can better map the feature vector to a high-dimensional space and improve the prediction accuracy of the model.

[0058] Step 107: According to the ice coating prediction model, predict the remote sensing data to obtain the corresponding ice coating thickness.

[0059] Based on the method of random forest - multi-kernel relevance vector machine (RF-MKRVM), use multi-source satellite remote sensing data and actual ice coating data for training to construct an ice coating prediction model. The ice coating prediction model is used to predict the ice coating thickness according to the remote sensing data; use the random forest method to select the influencing factors or indicators strongly related to the ice coating thickness, and then use the method of multi-kernel relevance vector machine to construct the ice coating thickness prediction model for the strongly related influencing indicators and the ice coating thickness, which has the characteristics of good accuracy; the kernel function is the key of the vector machine. In the prediction process, a suitable kernel function can better map the feature vector to a high-dimensional space and improve the prediction accuracy of the model; fully consider the influencing factors of the transmission line ice coating and combine the historical data of the ice coating thickness, which has comprehensiveness and timeliness; in the process of the transmission line ice coating, the influencing factors are relatively complex, and it is difficult for a single kernel function to match the characteristics of all ice coating data. Combine multiple kernel functions to construct an ice coating prediction model to enhance the prediction effect of the model.

[0060] In step 106, the ice coating prediction model can be expressed as:

[0061]

[0062] Among them, y i represents the ice coating thickness under the condition of x i , K C (x i) is the kernel function of the icing prediction model, x i is represented as a column vector composed of index feature values, ω i is the kernel function weight, ε i represents additive noise, ε i obeys a zero-mean Gaussian distribution. ω i can be written as the weight vector ω i =(ω0, ω1, ω2, …, ωN), that is, each kernel function has a corresponding weight.

[0063] The kernel function selects a weighted combination of the polynomial kernel function and the Gaussian kernel function, which is expressed as:

[0064] K C (x i ) = λ·K G (x i )+(1 - λ)·K P (x i )

[0065]

[0066] K P (x i )=(r1·x i +r2) r3 (41)

[0067] Among them, λ is the weight parameter, K G (x i ) represents the polynomial kernel function, K P (x i ) is the Gaussian kernel function, r1, r2, r3, r g are the parameters of the kernel respectively. The icing prediction model has five parameters to be optimized, λ, r1, r2, r3 and r g , and the selection of parameters has a crucial impact on the performance of the icing prediction model. ||x i || represents taking the modulus of the column vector.

[0068] Step 108: Calculate the icing amount of the transmission line according to the icing thickness:

[0069] M i =πbk(d + bk)ρ×10 -3

[0070] b=(d i -d) / 2 (5)

[0071] Among them, M i is the icing amount per unit length of the wire (kg / m), ρ is the density of ice, d is the wire diameter, d iis the diameter of the wire after icing, k is the wire diameter coefficient, and b is the predicted icing thickness (i.e., y i ).

[0072] In a specific embodiment, the relationship between k and the wire diameter d is shown in the following table:

[0073] Wire diameter (mm) 5 10 20 30 k 1.1 1.0 0.9 0.8

[0074] Step 109: The method for quantitatively describing the icing disaster procedure of the transmission line according to the icing amount includes:

[0075] Step 201: According to the load per unit length of the wire, preset the icing amount interval of the wire and the corresponding disaster level. In a specific embodiment, the model of the disaster procedure is as follows:

[0076]

[0077] Among them, M0 is the load per unit length of the line (kg / m).

[0078] Step 202: Predict the icing thickness according to the icing prediction model of Formula 4; calculate the icing amount according to Formula 5; match the corresponding disaster level according to the calculated icing amount of the transmission line and Formula 6.

[0079] Step 203: According to the disaster level, use corresponding markings or colors to display on the map.

[0080] Quantify the icing prediction result, and reflect the icing disaster level according to the predicted icing amount. The above quantification method quantitatively describes the icing result, making the icing prediction result more intuitively expressed and enhancing the readability of the result. The spatial display of the icing prediction result, combined with satellite maps, geographical coordinates and other map elements, will further improve the effectiveness of the icing prediction result.

[0081] The present invention also provides a system for implementing the above method, as Figure 2 shown, including a collection module 1, a normalization module 2, an index screening module 3, a training module 4 and a prediction module 5,

[0082] The collection module 1 is used to obtain the historical data set;

[0083] The normalization module 2 is used to normalize the indexes of the remote sensing data to obtain the normalized values of the indexes;

[0084] The index screening module 3 is used to obtain the misclassification rate of the indexes based on the random forest method, and screen the main indexes according to the misclassification rate;

[0085] The training module 4 is used to screen the historical data set according to the main indicators to obtain a training set; based on the method of multi-core relevance vector machine, use the training set for training to obtain an icing prediction model for the transmission line;

[0086] The prediction module 5 is used to predict the remote sensing data according to the icing prediction model to obtain the corresponding icing thickness.

[0087] The system further includes an ice amount prediction module 6 and a disaster prediction module 7,

[0088] The ice amount prediction module 6 is used to calculate the ice amount on the conductor according to the icing thickness;

[0089] The disaster prediction module 7 is used to obtain the corresponding disaster level of the ice amount according to the mapping of the preset ice amount interval and the disaster program, and is used to visually display the disaster level.

[0090] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for predicting icing on transmission lines based on multi-source satellite remote sensing, characterized in that, The method includes: Obtaining a historical data set, where the historical data set includes satellite remote sensing data and the corresponding ice coating thickness of transmission lines; Normalizing the indexes of the remote sensing data to obtain the normalized values of the indexes; Obtaining the misclassification rate of the indexes based on the random forest method; Screening the main indexes according to the misclassification rate; Screening the historical data set according to the main indexes to obtain a training set; Based on the multi-kernel relevance vector machine method, training with the training set to obtain an ice coating prediction model for transmission lines; Predicting the remote sensing data according to the ice coating prediction model to obtain the ice coating thickness; The ice coating prediction model is expressed as: Among them, y i represents the ice accretion thickness under the condition of x i , K C (x i ) is the kernel function of the ice accretion prediction model, and x i represents the column vector composed of index feature values. ω i is the kernel function weight, and ε i represents the additive noise. ε i obeys the zero-mean Gaussian distribution; The kernel function is expressed as: K C (x i ) = λ·K G (x i ) + (1 - λ)·K P (x i ) K P (x i )=(r1·x i +r2) r3 (41) where λ is the weight parameter, K G (x i ) represents the polynomial kernel function, K P (x i ) is the Gaussian kernel function, and r1, r2, r3, r g are the parameters of the kernel respectively; The method for calculating the ice coating amount of transmission lines: M i = πbk(d + bk)ρ × 10 -3 b = (d i - d) / 2 (5) Among them, M i is the ice accretion amount per unit length of the wire, ρ is the density of ice, d is the wire diameter, d i is the diameter of the wire after icing, k is the wire diameter coefficient, and b is the ice thickness.

2. The method for predicting icing on transmission lines according to claim 1, characterized in that, The normalized value is expressed as: Among them, a i is the value of the index, a min is the minimum value of the index, a max is the maximum value of the index, and A i is the normalized value.

3. The method for predicting icing on transmission lines according to claim 1, characterized in that, The method for screening the main indexes includes: Sorting the indexes from small to large according to the misclassification rate; Selecting a threshold value C according to the indexes, and taking the first C indexes as the main indexes.

4. The method for predicting icing on transmission lines according to claim 1, characterized in that, The training set D is expressed as: D = {(x1, y1),…,(x i , y i ),…,(x N , y N )} (3) Among them, N represents the number of samples, and x i represents the column vector composed of the main index feature values at a certain moment, and y i represents the ice coating thickness under the condition of x i condition.

5. The method for predicting icing on transmission lines according to claim 1, characterized in that, It also includes a method for quantitatively describing the ice coating disaster process of transmission lines: Presetting an ice coating amount interval of the conductor and the corresponding disaster level according to the load per unit length of the conductor; Matching the corresponding disaster level according to the calculated ice coating amount of the transmission line; Displaying on the map with corresponding marks or colors according to the disaster level.

6. A system for implementing the transmission line icing prediction method according to any one of claims 1-5, characterized in that, It includes a collection module, a normalization module, an index screening module, a training module and a prediction module. The collection module is used to obtain the historical data set; The normalization module is used to normalize the indexes of the remote sensing data to obtain the normalized values of the indexes; The index screening module is used to obtain the misclassification rate of the indexes based on the random forest method, and screen the main indexes according to the misclassification rate; The training module is used to screen the historical data set according to the main indexes to obtain a training set; Based on the multi-kernel relevance vector machine method, training with the training set to obtain an ice coating prediction model for transmission lines; The prediction module is used to predict the remote sensing data according to the ice coating prediction model to obtain the corresponding ice coating thickness.

7. The system according to claim 6, characterized in that, It also includes an ice coating amount prediction module and a disaster prediction module. The ice coating amount prediction module is used to calculate the ice coating amount of the conductor according to the ice coating thickness; The disaster prediction module is used to obtain the corresponding disaster level of the ice coating amount according to the preset ice coating amount interval and its disaster level.

Images