Electric wire icing prediction method, processor, prediction device and storage medium

Abstract

The invention relates to a wire icing prediction method, a processor, a prediction device and a storage medium, belongs to the technical field of meteorology, and solves the technical problem that a current wire icing prediction model cannot timely discover wire icing in a complex terrain. The prediction method comprises the following steps of: a, processing original icing record data of the power transmission line into a meteorological element feature set; b, outputting the feature importance degree of wire icing thickness prediction according to the meteorological element feature set and the initial wire icing thickness value, and establishing a feature vector; step c, downscaling analysis is carried out on historical grid point hour-by-hour weather forecast data of the electric wire icing disaster weather process by using a mesoscale weather forecast mode WRF; d, correcting the meteorological element prediction vector and the actual condition value, and outputting a time sequence correction matrix of wire icing; according to the method, the microclimate of wire icing under the complex terrain can be described more accurately, and the wire icing thickness in the future n hours is accurately predicted and evaluated through the LSTM model.

Description

Technical Field

[0001] The present invention belongs to the field of meteorological technologies, and particularly relates to a method for predicting wire icing, a processor, a prediction device and a storage medium. Background Art

[0002] Currently, the studied wire icing prediction models mainly establish a certain correlation model based on historical forecast data and wire icing thickness, so as to predict the wire icing thickness through NWP numerical weather prediction. However, the previous NWP numerical weather prediction has deficiencies in considering the influence of terrain, time accuracy and spatial characterization. Most wire lines are laid in complex terrain areas such as ridges or mountainsides, and the micro-meteorological environment formed in these areas poses higher requirements for wire icing. Once wire icing occurs, potential hazards cannot be discovered and eliminated in time. Summary of the Invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art, and provide a method for predicting wire icing, a processor, a prediction device and a storage medium, so as to solve the technical problems that the current wire icing prediction model cannot timely discover wire icing in complex terrain.

[0004] To solve the above problems, the technical solution of the present invention is: a method for predicting wire icing, including the following steps:

[0005] Step a: Select a wire icing disaster weather process in the target area, select the original record data of the transmission line icing in this process, and process the original record data of the transmission line icing into a meteorological element feature set;

[0006] Step b: Output the feature importance of wire icing thickness prediction for the meteorological element feature set and the initial wire icing thickness value through the filtering method correlation analysis method, and establish a feature vector. Test the significant correlation of the meteorological element feature set through the F-test method. The meteorological element feature set and the feature vector are used as the input data of the LSTM model;

[0007] Step c: Use the mesoscale weather prediction model WRF to perform downscaling analysis on the historical grid hourly weather forecast data of the wire icing disaster weather process, simulate and evaluate the results of the downscaling analysis through different parameterization schemes, and output a set of optimal parameterization schemes, that is, the meteorological element prediction vector is used as the input data of the LSTM model to provide optimal forecast support for wire icing prediction;

[0008] Step d: Correct the meteorological element prediction vector and the actual value to output a time series correction matrix of wire icing;

[0009] Step e: According to the input data and the time series correction matrix of wire icing, the LSTM model uses the rolling prediction method to predict and evaluate the wire icing thickness in the next n hours.

[0010] Optionally, in step a, the original recorded data of transmission line icing is evaluated and screened, and the data with large icing thickness and large change range is selected.

[0011] Optionally, the original recorded data of transmission line icing includes "longitude", "latitude", "line point name", "wire icing thickness value" and "acquisition time".

[0012] Optionally, the original recorded data of transmission line icing is preprocessed into historical icing data and historical weather forecast data.

[0013] Optionally, the historical icing data and historical weather forecast data form a meteorological element feature set.

[0014] Optionally, the historical icing data is processed into a model training label set and used as the input data of the LSTM model.

[0015] The second object of the present invention is to provide a processor configured to execute the above-mentioned prediction method for wire icing.

[0016] The third object of the present invention is to provide a prediction device for wire icing, including:

[0017] An acquisition module for acquiring the original recorded data of transmission line icing during a wire icing disaster weather process in a target area;

[0018] A preprocessing module for processing the original recorded data of transmission line icing into a meteorological element feature set;

[0019] A first analysis module for outputting the feature importance degree of wire icing thickness prediction through the correlation analysis method of the filtering method for the meteorological element feature set and the initial wire icing thickness value, and establishing a feature vector, and using the meteorological element feature set and the feature vector as the input data of the LSTM model;

[0020] A verification module for checking the significant correlation of the meteorological element feature set through the F-test method;

[0021] A second analysis module for performing downscaling analysis on the historical grid hourly weather forecast data of the above-mentioned wire icing disaster weather process by using the mesoscale weather forecast model WRF, and performing simulation evaluation on the results of the downscaling analysis, and outputting a meteorological element prediction vector as the input data of the LSTM model;

[0022] A correction module for correcting the meteorological element prediction vector and the actual value, and outputting a time series correction matrix for wire icing;

[0023] An LSTM module with an LSTM model. The LSTM module predicts and evaluates the wire icing thickness in the next n hours by using the rolling prediction method according to the input data and the time series correction matrix.

[0024] The fourth object of the present invention is to provide a machine-readable storage medium, on which instructions are stored for causing a machine to execute a prediction method for wire icing according to the above.

[0025] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0026] The present invention adjusts the parameterization scheme of the mesoscale weather forecast model WRF through specific terrain, extracts eigenvectors from the output results and wire icing records, and can more accurately depict the micro-meteorology of wire icing under complex terrain through the nested scheme advantage of the mesoscale weather forecast model WRF, and accurately predicts and evaluates the wire icing thickness in the next n hours through the neural network time series model LSTM. Description of the Drawings

[0027] Figure 1 It is a flowchart of the wire icing prediction method in the embodiment.

[0028] Figure 2 It is a connection schematic diagram of the wire icing prediction device in the embodiment.

[0029] Reference numerals: 1, acquisition module; 2, preprocessing module; 3, first analysis module; 4, inspection module; 5, second analysis module; 6, correction module; 7, LSTM module. Detailed Embodiments

[0030] The present invention will be further described in detail below with reference to the drawings and embodiments.

[0031] Embodiment 1: As Figure 1 shown, the present embodiment provides a prediction method for wire icing, including the following steps:

[0032] Step a: Select a wire icing disaster weather process in the target area, select the original record data of the transmission line icing in this process, and process the original record data of the transmission line icing into a meteorological element feature set;

[0033] Step b: Use the filtering method correlation analysis method to output the feature importance of wire icing thickness prediction for the meteorological element feature set and the initial wire icing thickness value, and establish a feature vector. Use the F-test method to test the significant correlation of the meteorological element feature set. The meteorological element feature set and the feature vector are used as the input data of the LSTM model;

[0034] Step c: Use the mesoscale weather forecasting model WRF to perform downscaling analysis on the historical grid hourly weather forecasting data of the above-mentioned once wire icing disaster weather process, and use different parameterization schemes to simulate and evaluate the results of the downscaling analysis, and output an optimal parameterization scheme, that is, the meteorological element prediction vector is used as the input data of the LSTM model to provide optimal forecasting support for wire icing prediction;

[0035] Step d: Modify the meteorological element prediction vector and the actual value to output a time series correction matrix for wire icing;

[0036] Step e: According to the input data and the time series correction matrix of wire icing, the LSTM model uses the rolling prediction method to predict and evaluate the wire icing thickness in the next n hours.

[0037] Through the above settings, the advantages of the nesting scheme of the mesoscale weather forecasting model WRF can more accurately depict the micro-meteorology of wire icing under complex terrain, and use the neural network time series model LSTM to accurately predict and evaluate the wire icing thickness in the next n hours.

[0038] In the wire icing prediction method of this embodiment, in step a, the original recorded data of transmission line icing is evaluated and selected, and the data with large icing thickness and large change range is selected.

[0039] In the wire icing prediction method of this embodiment, the original recorded data of transmission line icing includes "longitude", "latitude", "line point name", "wire icing thickness value" and "acquisition time".

[0040] In the wire icing prediction method of this embodiment, the original recorded data of transmission line icing is preprocessed into historical icing data and historical meteorological forecasting data. The historical icing data and historical meteorological forecasting data form a meteorological element feature set.

[0041] In the wire icing prediction method of this embodiment, the historical icing data is processed into a model training label set and used as the input data of the LSTM model.

[0042] Embodiment 2: This embodiment provides a processor configured to execute the wire icing prediction method in Embodiment 1.

[0043] Embodiment 3: As Figure 1 and Figure 2 shown, this embodiment provides a prediction device for wire icing, including an acquisition module 1, a preprocessing module 2, a first analysis module 3, an inspection module 4, a second analysis module 5, a correction module 6, and an LSTM module 7; the acquisition module 1 is used to acquire the original record data of the transmission line icing during a wire icing disaster weather process in the target area; the preprocessing module 2 is used to process the original record data of the transmission line icing into a meteorological element feature set; the first analysis module 3 is used to output the feature importance of the wire icing thickness prediction through the correlation analysis method of the filtering method for the meteorological element feature set and the initial wire icing thickness value, and establish a feature vector, and the meteorological element feature set and the feature vector are used as the input data of the LSTM model; the inspection module 4 is used to check the significant correlation of the meteorological element feature set through the F-test method; the second analysis module 5 is used to perform downscaling analysis on the historical grid hourly weather forecast data of the wire icing disaster weather process, and perform simulation evaluation on the results of the downscaling analysis, and output the meteorological element prediction vector as the input data of the LSTM model; the correction module 6 is used to correct the meteorological element prediction vector and the actual value, and output a time series correction matrix for wire icing; the LSTM module 7 has an LSTM model, and the LSTM module 7 uses the rolling prediction method to predict and evaluate the wire icing thickness in the next n hours according to the input data and the time series correction matrix.

[0044] Through the above settings, the advantages of the nesting scheme of the mesoscale weather forecast model WRF can more accurately depict the micro-meteorology of wire icing under complex terrain, and accurately predict and evaluate the wire icing thickness in the next n hours through the neural network time series model LSTM. Optionally, the acquisition module 1 is connected to the preprocessing module 2, the inspection module 4 is connected to the preprocessing module 2, the preprocessing module 2 is connected to the first analysis module 3, the correction module 6 is connected to the second analysis module 5, and both the first analysis module 3 and the second analysis module 5 are connected to the LSTM module 7.

[0045] Embodiment 4: This embodiment provides a machine-readable storage medium, characterized in that instructions are stored on the machine-readable storage medium, and the instructions are used to cause the machine to execute a prediction method for wire icing in Embodiment 1.

Claims

1. A method for predicting electric wire icing, characterized in that: The following steps are involved: Step a: Select a power line icing disaster weather process in the target area, select the original record data of power line icing in the process, and process the original record data of power line icing into a meteorological element feature set; Step b: The feature importance of the prediction of the ice thickness of the wire is outputted by the filtering correlation analysis method for the feature set of meteorological elements and the initial ice thickness value of the wire, and a feature vector is established. The significant correlation of the feature set of meteorological elements is tested by the F test method. The feature set of meteorological elements and the feature vector are used as the input data of the LSTM model. Step c: using the mesoscale weather forecast model WRF to perform downscaling analysis on the historical hourly forecast data of the power line icing disaster weather process, simulating and evaluating the results of the downscaling analysis through different parameterization schemes, and outputting a set of optimal parameterization schemes, that is, the meteorological element prediction vector as the input data of the LSTM model, to provide the best forecast support for power line icing prediction; Step d: By correcting the meteorological element prediction vector and the actual value, a set of wireline icing time series correction matrix is ​​output; Step e: Based on the input data and the time series correction matrix of the wire ice coverage, the LSTM model uses a rolling forecast method to predict and evaluate the wire ice coverage thickness in the next n hours.

2. A method for predicting electric wire icing according to claim 1, characterized in that: In step a, the original recorded data of ice coating on the transmission lines are evaluated and selected, and data with large ice coating thickness and large variation range are selected.

3. The method for predicting electric wire icing according to claim 1, characterized in that: The original recorded data of ice coating on transmission lines include "longitude", "latitude", "line point name", "wire ice coating thickness value" and "collection time".

4. The method for predicting electric wire icing according to claim 1, characterized in that: The original recorded data of transmission line icing are preprocessed into historical icing data and historical weather forecast data.

5. A method for predicting electric wire icing according to claim 4, characterized in that: Historical ice cover data and historical weather forecast data constitute the meteorological element feature set.

6. A method for predicting electric wire icing according to claim 4, characterized in that: The historical ice cover data is processed into a model training label set and used as the input data of the LSTM model.

7. A processor, characterized in that: The method is configured to execute the method for predicting electric wire icing as claimed in any one of claims 1 to 6.

8. A prediction device for electric wire icing, characterized in that: include: An acquisition module (1) is used to acquire the original recorded data of power transmission line icing during a power line icing disaster weather process in a target area; A preprocessing module (2) is used to process the original recorded data of ice coating on the power transmission line into a meteorological element feature set; The first analysis module (3) is used to output the feature importance of the wire ice thickness prediction for the meteorological element feature set and the initial wire ice thickness value through a filtering method correlation analysis method, and to establish a feature vector. The meteorological element feature set and the feature vector are used as input data of the LSTM model; A test module (4) is used to check the significant correlation of the meteorological element feature set by means of an F test; The second analysis module (5) is used to use the mesoscale weather forecast model WRF to perform downscaling analysis on the hourly data of the historical forecast data of the power line icing disaster weather process, simulate and evaluate the results of the downscaling analysis, and output the meteorological element prediction vector as the input data of the LSTM model; A correction module (6) is used to correct the meteorological element prediction vector and the actual value, and output a set of wire ice coverage time series correction matrix; The LSTM module (7) has an LSTM model. The LSTM module (7) predicts and evaluates the ice thickness of the power lines in the next n hours using a rolling forecast method based on input data and a time series correction matrix.

9. A machine-readable storage medium, characterized in that: The machine-readable storage medium stores instructions, which are used to enable a machine to execute a method for predicting electric wire icing according to any one of claims 1 to 6.

Images