Overhead transmission line super-short-term icing prediction method and system

Abstract

The application discloses an overhead transmission line super-short-term icing prediction method, which comprises the following steps: collecting historical data information of a target transmission line; establishing a mapping relationship between the obtained data information and the actual icing thickness of the target transmission line; constructing a training data set; constructing an initial model for super-short-term icing prediction of the overhead transmission line and training the initial model by using the training data set to obtain a super-short-term icing prediction model of the overhead transmission line; and using the super-short-term icing prediction model of the overhead transmission line obtained in the step to perform super-short-term icing prediction on the target transmission line. The application discloses a system for realizing the super-short-term icing prediction method of the overhead transmission line. The application realizes the super-short-term icing prediction of the overhead transmission line by virtue of an innovative neural network model design and overhead line historical data processing, and has higher reliability and better accuracy.

Description

Technical Field

[0001] This invention belongs to the field of electrical automation, specifically relating to a method and system for predicting ultra-short-term icing of overhead transmission lines. Background Technology

[0002] With economic and technological development and the improvement of people's living standards, electricity has become an indispensable secondary energy source in people's production and daily life, bringing endless convenience. Therefore, ensuring a stable and reliable supply of electricity has become one of the most important tasks of the power system.

[0003] Transmission lines are the power transmission lines of a power system and are of great significance to the system. Severe icing on overhead transmission lines increases the load on the lines. When the load stress reaches a certain level, it can lead to pole collapse and line breakage. Under special operating conditions, it can cause faults such as ice shedding, ice flashover, and insulator bridging, resulting in tripping. This seriously affects the safe and stable operation of transmission lines and the power system, and has a significant impact on the reliability of the power system. Icing areas on overhead transmission lines are mostly concentrated in mountainous or hilly areas with complex terrain. Conductors in some areas with "micro-topography and micro-climate" are also prone to icing. The severity of icing on transmission lines varies greatly depending on the terrain, altitude, and climate.

[0004] Current research on icing of transmission lines mainly focuses on changes in the transmission line itself or a simulated line to calculate the equivalent ice thickness; or it uses machine learning to train predictive models based on meteorological and geographical data to predict icing of transmission lines. However, existing methods still suffer from low reliability and poor accuracy. Summary of the Invention

[0005] One of the objectives of this invention is to provide a highly reliable and accurate method for predicting ultra-short-term icing of overhead transmission lines.

[0006] The second objective of this invention is to provide a system for implementing the ultra-short-term icing prediction method for overhead transmission lines.

[0007] The method for predicting ultra-short-term icing of overhead transmission lines provided by this invention includes the following steps:

[0008] S1. Collect historical data information of the target transmission line;

[0009] S2. Establish a mapping relationship between the data information obtained in step S1 and the actual icing thickness of the target transmission line;

[0010] S3. Based on the mapping relationship obtained in step S2, construct a training dataset using one-hot encoding;

[0011] S4. Based on the ARIMA model and the FCNN neural network model, construct an initial model for predicting ultra-short-term icing of overhead transmission lines;

[0012] S5. Using the training dataset constructed in step S3, train the initial model for predicting ultra-short-term icing of overhead transmission lines constructed in step S4 to obtain the ultra-short-term icing prediction model for overhead transmission lines.

[0013] S6. Using the ultra-short-term icing prediction model for overhead transmission lines obtained in step S5, perform ultra-short-term icing prediction on the target transmission line.

[0014] Step S2, which establishes the mapping relationship between the data information obtained in step S1 and the actual icing thickness of the target transmission line, specifically includes the following steps:

[0015] Based on the data obtained in step S1, establish the mapping relationships between the tower topography data of the target transmission line and the actual ice thickness of the conductor, the mapping relationships between meteorological data and the actual ice thickness of the conductor, the mapping relationships between time data and the actual ice thickness of the conductor, and the mapping relationships between conductor tension data and the actual ice thickness of the conductor.

[0016] Step S3, which involves constructing a training dataset based on the mapping relationship obtained in step S2 and using one-hot encoding, specifically includes the following steps:

[0017] Based on the mapping relationship between tower topography data and actual ice thickness on conductors:

[0018] The terrain includes routes outside mountain peaks, horizontal routes outside mountain ridges, routes across valleys, routes through valleys, uphill routes over mountain peaks, uphill routes over mountain ridges, mountain peak routes, and ridge top routes. The eight types of terrain are sequentially encoded with one-hot codes and the elevation values ​​of the tower locations are integrated.

[0019] Based on the mapping relationship between meteorological data and the actual icing thickness of the conductor:

[0020] The meteorological data includes real-time wind speed, real-time temperature, real-time humidity, average wind speed, average temperature, and average humidity data collected at set time points, and the wind direction is one-hot encoded according to a 16-direction wind rose diagram.

[0021] Based on the mapping relationship between conductor tension data and actual icing thickness:

[0022] The conductor tension data includes the difference between the real-time conductor tension value collected at set time points and the conductor tension sensor value, which is processed using the maximum and minimum difference feature vector normalization method.

[0023] The processed data is then subjected to principal component analysis to extract principal vectors, and the obtained principal vectors are used to construct the training dataset.

[0024] Step S4, which involves constructing an initial model for predicting ultra-short-term icing of overhead transmission lines based on the ARIMA model and the FCNN neural network model, specifically includes the following steps:

[0025] Step S3 generates two different types of data: a continuous variable dataset x. mn and discrete variable dataset y ab Among them, the continuous variable dataset x mn This includes real-time wind speed, real-time temperature, real-time humidity, average wind speed, average temperature, average humidity data, and conductor tension data; the initial data sampling interval for real-time variable data is set to Δt, and the average variable... The average value of the first three sampling intervals is expressed as:

[0026]

[0027] average value Located at x mn The dataset is in the i-th row and j-th column, and the corresponding real-time data is in the k-th column. The average value is calculated using the (i-1), (i-2), and (i-3)-th sampled values. A continuous variable dataset x with dimension M×N is constructed. mn Where m = 0, 1, ..., M-1, n = 1, 2, ..., N-1;

[0028]

[0029] For a continuous variable dataset x mn The maximum-minimum difference eigenvector normalization method is used to reduce the impact of differences in data volume on subsequent model training and prediction.

[0030] Discrete variable dataset y ab Construct a discrete variable dataset y of type A×B, including one-hot discrete variables for terrain and wind direction. ab :

[0031]

[0032] Take the continuous variable dataset x mn and discrete variable dataset y ab Perform concatenation to construct dataset z st Where s = m = a, t = n + b; the processed data is affected by one-hot encoding, and the data matrix set Z stThe dataset is relatively sparse, so principal component analysis is used to extract principal vectors. These principal vectors are then used to construct the training dataset, which is then combined with the resulting sample set c. k =[c1 … c i …c s ];

[0033] The dataset z was analyzed using both the ARIMA model and the FCNN neural network model. st c k Training and prediction are performed to form prediction sequences p. k =[p1 … p i … p s ] and q k =[q1 … q i … q s ];

[0034] Because the impact of "micro- and micro-climate" weather on transmission line icing is significant and exhibits considerable variability and unique characteristics, a weighted approach is used to fuse the transmission line icing prediction data by adjusting the difference coefficients u, v, and w to correct for existing biases.

[0035] r i =logu Δt (w Δt p i +(1-w Δt )q i )+v Δt

[0036] In the formula, r i To achieve the predicted value of line icing, the data sampling interval Δt was optimized based on the prediction results, with the initial value set to 1 hour.

[0037] Step S5 involves using the training dataset constructed in step S3 to train the initial model for predicting ultra-short-term icing of overhead transmission lines constructed in step S4, resulting in the ultra-short-term icing prediction model for overhead transmission lines. This process specifically includes the following steps:

[0038] The training dataset constructed in step S3 is divided into a training set and a test set;

[0039] The initial model for predicting ultra-short-term icing of overhead transmission lines constructed in step S4 is trained using a training set, and the trained model is tested using a test set until the model reaches its optimal performance, thus obtaining the ultra-short-term icing prediction model for overhead transmission lines.

[0040] This invention also provides a system for implementing the ultra-short-term icing prediction method for overhead transmission lines, comprising a data acquisition module, a mapping relationship establishment module, a training set construction module, a model construction module, a model training module, and an icing prediction module; the data acquisition module, mapping relationship establishment module, training set construction module, model construction module, model training module, and icing prediction module are connected in series; the data acquisition module is used to collect historical data information of the target transmission line and upload the data to the mapping relationship establishment module; the mapping relationship establishment module is used to establish a mapping relationship between the acquired data information and the actual icing thickness of the target transmission line based on the received data information, and upload the data to the training set construction module; the training set construction module is used to establish a mapping relationship between the acquired data information and the actual icing thickness of the target transmission line based on the received data information; the training set construction module is used to establish a mapping relationship between the acquired data information and the actual icing thickness of the target transmission line based on the received data information. The system receives information, constructs a training dataset based on one-hot encoding, and uploads the data to the model building module. The model building module, based on the received data, constructs an initial model for predicting ultra-short-term icing of overhead transmission lines using the ARIMA and FCNN neural network models, and uploads the data to the model training module. The model training module, using the received data and the constructed training dataset, trains the initial model for predicting ultra-short-term icing of overhead transmission lines to obtain an ultra-short-term icing prediction model, and uploads the data to the icing prediction module. The icing prediction module, using the obtained ultra-short-term icing prediction model, performs ultra-short-term icing prediction on the target transmission line based on the received data.

[0041] The method and system for predicting ultra-short-term icing of overhead transmission lines provided by this invention, through innovative neural network model design and processing of historical data of overhead lines, not only realizes ultra-short-term icing prediction of overhead transmission lines, but also has higher reliability and better accuracy. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0043] Figure 2 This is a topographical diagram illustrating the method of the present invention.

[0044] Figure 3 This is a schematic diagram of a 16-direction wind rose diagram for the method of the present invention.

[0045] Figure 4 This is a schematic diagram illustrating the effect of an embodiment of the method of the present invention.

[0046] Figure 5 This is a schematic diagram of the functional modules of the system of the present invention. Detailed Implementation

[0047] like Figure 1The diagram shown is a flowchart of the method of the present invention: The method for predicting ultra-short-term icing of overhead transmission lines provided by the present invention includes the following steps:

[0048] S1. Collect historical data information of the target transmission line;

[0049] S2. Establish the mapping relationship between the data information obtained in step S1 and the actual icing thickness of the target transmission line; specifically including the following steps:

[0050] Based on the data obtained in step S1, establish the mapping relationships between the tower topography data of the target transmission line and the actual ice thickness of the conductor, the mapping relationship between meteorological data and the actual ice thickness of the conductor, the mapping relationship between time data and the actual ice thickness of the conductor, and the mapping relationship between conductor tension data and the actual ice thickness of the conductor.

[0051] S3. Based on the mapping relationship obtained in step S2, construct the training dataset using one-hot encoding; specifically including the following steps:

[0052] Based on the mapping relationship between tower topography data and actual ice thickness on conductors:

[0053] The terrain described includes routes along the outer edges of mountain peaks, horizontal routes along the outer edges of ridges, routes crossing valleys, routes passing through valley floors, uphill routes crossing mountain peaks, uphill routes crossing ridges, mountain peak routes, and ridge top routes (corresponding to, in order, routes on the outer edges of mountain peaks and ridges). Figure 2 In (a) to (h), one-hot codes are constructed for the eight terrain types in sequence, and the elevation values ​​of the tower locations are integrated;

[0054] The one-hot encoding of the eight terrain types is shown in Table 1:

[0055] Table 1. Schematic diagram of one-hot encoding for 8 types of terrain construction

[0056]

[0057] Based on the mapping relationship between meteorological data and the actual icing thickness of the conductor:

[0058] The meteorological data includes real-time wind speed, real-time temperature, real-time humidity, average wind speed, average temperature, and average humidity data collected at set time points. The data is processed using the maximum-minimum difference eigenvector normalization method, and the wind direction is represented by a 16-direction wind rose diagram (e.g., Figure 3 (As shown) Perform one-hot encoding;

[0059] Wind direction, according to the one-hot encoding of the 16-direction wind rose diagram, is shown in Table 2:

[0060] Table 2. Wind direction according to the one-hot encoding of the 16-direction wind rose diagram.

[0061] 16-Directional Wind Rose Diagram due east East Northeast northeast North Northeast One-Hot encoding 00001 00010 00011 00100 16-Directional Wind Rose Diagram North Northwest northwest West Northwest One-Hot encoding 00101 00110 00111 01000 16-Directional Wind Rose Diagram due west West Southwest southwest Southwest One-Hot encoding 01001 01010 01011 01110 16-Directional Wind Rose Diagram due south South Southeast southeast East Southeast One-Hot encoding 01111 10000 10001 10010

[0062] Based on the mapping relationship between conductor tension data and actual icing thickness:

[0063] The conductor tension data includes the difference between the real-time conductor tension value collected at set time points and the conductor tension sensor value, which is processed using the maximum and minimum difference feature vector normalization method.

[0064] S4. Based on the ARIMA model and the FCNN neural network model, construct an initial model for predicting ultra-short-term icing of overhead transmission lines; specifically including the following steps:

[0065] Step S3 generates two different types of data: a continuous variable dataset x. mn and discrete variable dataset y ab Among them, the continuous variable dataset x mn This includes real-time wind speed, real-time temperature, real-time humidity, average wind speed, average temperature, average humidity data, and conductor tension data; the initial data sampling interval for real-time variable data is set to Δt, and the average variable... The average value of the first three sampling intervals is expressed as:

[0066]

[0067] average value Located at x mn The dataset is in the i-th row and j-th column, and the corresponding real-time data is in the k-th column. The average value is calculated using the (i-1), (i-2), and (i-3)-th sampled values. A continuous variable dataset x with dimension M×N is constructed. mn Where m = 0, 1, ..., M-1, n = 1, 2, ..., N-1;

[0068]

[0069] For a continuous variable dataset x mn The maximum-minimum difference eigenvector normalization method is used to reduce the impact of differences in data volume on subsequent model training and prediction.

[0070] Discrete variable dataset y ab Construct a discrete variable dataset y of type A×B, including one-hot discrete variables for terrain and wind direction. ab :

[0071]

[0072] Take the continuous variable dataset xmn and discrete variable dataset y ab Perform concatenation to construct dataset z st Where s = m = a, t = n + b; the processed data is affected by one-hot encoding, and the data matrix set Z st The dataset is relatively sparse, so principal component analysis is used to extract principal vectors. These principal vectors are then used to construct the training dataset, which is then combined with the resulting sample set c. k =[c1 … c i …c s ];

[0073] The dataset z was analyzed using both the ARIMA model and the FCNN neural network model. st c k Training and prediction are performed to form prediction sequences p. k =[p1 … p i … p s ] and q k =[q1 … q i … q s ];

[0074] Because the impact of "micro- and micro-climate" weather on transmission line icing is significant and exhibits considerable variability and unique characteristics, a weighted approach is used to fuse the transmission line icing prediction data by adjusting the difference coefficients u, v, and w to correct for existing biases.

[0075] r i =logu Δt (w Δt p i +(1-w Δt )q i )+v Δt

[0076] In the formula, r i To achieve the predicted value of line icing, the data sampling interval Δt was optimized based on the prediction results, with the initial value set to 1 hour;

[0077] S5. Using the training dataset constructed in step S3, train the initial model for predicting ultra-short-term icing of overhead transmission lines constructed in step S4 to obtain the ultra-short-term icing prediction model for overhead transmission lines; specifically including the following steps:

[0078] The training dataset constructed in step S3 is divided into a training set and a test set;

[0079] The initial model for predicting ultra-short-term icing of overhead transmission lines constructed in step S4 is trained using the training set, and the trained model is tested using the test set until the model reaches its optimal performance, thus obtaining the ultra-short-term icing prediction model for overhead transmission lines.

[0080] S6. Using the ultra-short-term icing prediction model for overhead transmission lines obtained in step S5, perform ultra-short-term icing prediction on the target transmission line.

[0081] The effects of the method of the present invention will be illustrated below with reference to an embodiment:

[0082] Through on-site verification, at 19:28 on February 9, the ice thickness of a 110 kV line was measured to be 13 mm using a simulated line. After applying for de-icing, the ice thickness was predicted to be 16.8 mm at 19:30. At 23:48, the de-icing current was supplied, and the predicted ice thickness was 19.3 mm. After de-icing, the ice sample was picked up, weighed, and measured to be 17.34 mm thick.

[0083] At 17:28 on February 10, the ice layer on a 110 kV line reached 9 mm. After applying for de-icing, it was predicted that the ice layer thickness on the line would be 8.7 mm at 17:30. At 23:02, the de-icing current was supplied, and the predicted ice layer thickness was 13.1 mm. After the ice was melted, the ice sample was picked up, weighed, and the ice thickness was measured to be 15.85 mm.

[0084] By comparing the actual conductor icing data from two trials with the predicted line icing data from this invention, it was found that the trends of the actual and predicted icing data were basically consistent, with the growth curves nearly parallel. Data obtained from ice samples after melting verified that the difference between the predicted icing value and the dominant line data was stable. Therefore, the prediction model can be improved by adjusting the difference coefficient. Comparative data are shown below. Figure 4 As shown in the figure, curve 1 is the equivalent icing thickness curve, and curve 2 is the comprehensive suspension load curve.

[0085] like Figure 5The diagram shows the functional modules of the system of the present invention: The system for implementing the ultra-short-term icing prediction method for overhead transmission lines disclosed in this invention includes a data acquisition module, a mapping relationship establishment module, a training set construction module, a model construction module, a model training module, and an icing prediction module; these modules are connected in series. The data acquisition module collects historical data information of the target transmission line and uploads the data to the mapping relationship establishment module. The mapping relationship establishment module establishes a mapping relationship between the acquired data information and the actual icing thickness of the target transmission line based on the received data information and uploads the data to the training set construction module. The training set construction module is used to... Based on the received data, a training dataset is constructed using one-hot encoding and uploaded to the model building module. The model building module, based on the received data and using the ARIMA and FCNN neural network models, constructs an initial model for predicting ultra-short-term icing of overhead transmission lines and uploads the data to the model training module. The model training module, using the received data and the constructed training dataset, trains the initial model for predicting ultra-short-term icing of overhead transmission lines to obtain an ultra-short-term icing prediction model, and uploads the data to the icing prediction module. The icing prediction module, using the obtained ultra-short-term icing prediction model, performs ultra-short-term icing prediction on the target transmission line based on the received data.

Claims

1. A method for predicting ultra-short-term icing on overhead transmission lines, comprising the following steps: S1. Collect historical data information of the target transmission line; S2. Establish a mapping relationship between the data obtained in step S1 and the actual icing thickness of the target transmission line; S3. Based on the mapping relationship obtained in step S2, construct the training dataset using one-hot encoding; specifically, this includes the following steps: Based on the mapping relationship between tower topography data and actual ice thickness on conductors: The terrain includes routes outside mountain peaks, horizontal routes outside mountain ridges, routes across valleys, routes through valleys, uphill routes over mountain peaks, uphill routes over mountain ridges, mountain peak routes, and ridge top routes. The eight types of terrain are sequentially encoded with one-hot codes and the elevation values ​​of the tower locations are integrated. Based on the mapping relationship between meteorological data and the actual icing thickness of the conductor: The meteorological data includes real-time wind speed, real-time temperature, real-time humidity, average wind speed, average temperature, and average humidity data collected at set time points, and the wind direction is one-hot encoded according to a 16-direction wind rose diagram. Based on the mapping relationship between conductor tension data and actual icing thickness: The conductor tension data includes the difference between the real-time conductor tension value collected at set time points and the conductor tension sensor value, which is processed using the maximum and minimum difference feature vector normalization method. The processed data is used to extract principal vectors using principal component analysis, and the obtained principal vectors are used to construct the training dataset. S4. Based on the ARIMA model and the FCNN neural network model, construct an initial model for predicting ultra-short-term icing of overhead transmission lines; specifically including the following steps: Step S3 generates two different types of data: a continuous variable dataset. and discrete variable datasets Among them, continuous variable datasets This includes real-time wind speed, real-time temperature, real-time humidity, average wind speed, average temperature, average humidity data, and conductor tension data; the initial data sampling interval for real-time variable data is set to [value missing]. Average variable The average value of the first three sampling intervals is expressed as: average value lie in The dataset's first Line 1 The column corresponds to the real-time data in the [number]th column. Column; through the first , , Calculate the average of each sample value; construct the dimension as follows: Continuous variable dataset ,in , ; For continuous variable datasets The maximum and minimum difference eigenvector normalization method is adopted for processing; Discrete variable dataset Constructing one-hot discrete variables including topography and wind direction discrete variable dataset : Data sets of continuous variables and discrete variable datasets Perform concatenation to construct a dataset. ,in , Principal component analysis (PCA) is used to extract principal vectors, which are then used to construct a training dataset. This dataset is then combined with the result sample set. ; The dataset was analyzed using both the ARIMA model and the FCNN neural network model. , Training and prediction are performed to generate prediction sequences. and ; By adjusting the difference coefficient , and To correct existing biases, the line icing prediction data are fused using a weighted method: In the formula, To achieve the predicted value of line icing; S5. Using the training dataset constructed in step S3, train the initial model for predicting ultra-short-term icing of overhead transmission lines constructed in step S4 to obtain the ultra-short-term icing prediction model for overhead transmission lines. S6. Using the ultra-short-term icing prediction model for overhead transmission lines obtained in step S5, perform ultra-short-term icing prediction on the target transmission line.

2. The method for predicting ultra-short-term icing of overhead transmission lines according to claim 1, characterized in that... Step S2, which establishes the mapping relationship between the data information obtained in step S1 and the actual icing thickness of the target transmission line, specifically includes the following steps: Based on the data obtained in step S1, establish the mapping relationships between the tower topography data of the target transmission line and the actual ice thickness of the conductor, the mapping relationships between meteorological data and the actual ice thickness of the conductor, the mapping relationships between time data and the actual ice thickness of the conductor, and the mapping relationships between conductor tension data and the actual ice thickness of the conductor.

3. The method for predicting ultra-short-term icing of overhead transmission lines according to claim 2, characterized in that... Step S5 involves using the training dataset constructed in step S3 to train the initial model for predicting ultra-short-term icing of overhead transmission lines constructed in step S4, resulting in the ultra-short-term icing prediction model for overhead transmission lines. This process specifically includes the following steps: The training dataset constructed in step S3 is divided into a training set and a test set; The initial model for predicting ultra-short-term icing of overhead transmission lines constructed in step S4 is trained using a training set, and the trained model is tested using a test set until the model reaches its optimal performance, thus obtaining the ultra-short-term icing prediction model for overhead transmission lines.

4. A system for implementing the ultra-short-term icing prediction method for overhead transmission lines as described in any one of claims 1 to 3, characterized in that... It includes a data acquisition module, a mapping relationship establishment module, a training set construction module, a model construction module, a model training module, and an icing prediction module; the data acquisition module, mapping relationship establishment module, training set construction module, model construction module, model training module, and icing prediction module are connected in series; the data acquisition module is used to collect historical data information of the target transmission line and upload the data to the mapping relationship establishment module; The mapping relationship establishment module is used to establish a mapping relationship between the acquired data information and the actual ice thickness of the target transmission line based on the received data information, and upload the data to the training set construction module; the training set construction module is used to construct a training dataset based on one-hot encoding based on the received data information, and upload the data to the model construction module. The model building module is used to construct an initial model for predicting ultra-short-term icing of overhead transmission lines based on the received data information, using the ARIMA model and the FCNN neural network model, and upload the data to the model training module. The model training module is used to train the initial model for ultra-short-term icing prediction of overhead transmission lines using the constructed training dataset based on the received data information, thereby obtaining the ultra-short-term icing prediction model for overhead transmission lines, and uploading the data to the icing prediction module; the icing prediction module is used to perform ultra-short-term icing prediction on the target transmission line using the obtained ultra-short-term icing prediction model based on the received data information.

Images