A method for predicting development image of long gap streamer discharge under lightning impulse

Abstract

The application discloses a long-gap streamer discharge development image prediction method under lightning impulse, utilizes a streamer discharge test observation system, collects streamer discharge images, discharge voltage and time data shot by an ICCD, converts the images into discharge light intensity matrices, determines the relationship among the light intensity matrices, the discharge voltage and the time, adopts an XGBoost algorithm to establish a multi-output regression model, selects a loss function to evaluate the optimal model and parameters by adopting an n-fold cross-validation method, inputs the time-varying instantaneous voltage and the corresponding time into the trained model according to the voltage discharge sequence, obtains a series of predicted light intensity matrices, restores the light intensity matrices into images, and thus the discharge images of the streamer discharge process are obtained; the method can improve the deficiency that complete and continuous long-gap streamer development images are difficult to be obtained by using the test method in the prior art, and realizes the prediction of the nanosecond-level change development process of the streamer discharge light intensity of a long air gap under lightning impulse.

Description

Technical Field

[0001] This invention belongs to the field of long air gap discharge observation technology, specifically relating to a method for predicting the development of long gap streamer discharge under lightning impact. Background Technology

[0002] As the most widely used insulating medium, the insulation properties of air gaps have gradually become one of the key issues in the insulation design of ultra-high voltage transmission projects. As the initial stage of long air gap discharge, streamer discharge is a transient process with a time scale on the order of nanoseconds and a local spatial scale on the order of micrometers. By acquiring discharge images during this process, the streamer discharge morphology, spatial ionization region, and the time-varying characteristics of streamer development speed can be clearly understood. However, streamer discharge in long gaps under lightning strikes is rapid and ionization is very weak, making it difficult to observe its complete and continuous transient process. Furthermore, due to the randomness and diversity of long air gap discharges, it is difficult to fully understand the physical development process of the discharge using only experimental methods and data, and the experimental workload is enormous. Machine learning algorithms, due to their self-learning, self-organizing, and adaptive characteristics, are widely used in linear or nonlinear prediction and inference. They have also achieved good results in research fields such as power systems and high-voltage discharge prediction, but their application research in the image features of air gap discharges is still lacking. Summary of the Invention

[0003] The purpose of this invention is to provide a method for predicting the development of long-gap jet discharge under lightning impact, which can overcome the shortcomings of existing technologies that use experimental methods to obtain complete and continuous images of long-gap jet development, and realize the prediction of the nanosecond-level change in the discharge intensity of long-gap jets under lightning impact, thereby solving the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for predicting the development of long-gap streamer discharge under lightning impulse, comprising the following steps:

[0005] S1. Using a streamer discharge test observation system, collect streamer discharge images, discharge voltage and time data with nanosecond-level exposure time ICCDs during lightning impulse discharge processes of different voltage levels in long air gaps.

[0006] S2. Divide the ICCD image into prediction regions according to pixel blocks and convert them into a discharge intensity matrix;

[0007] S3. Based on the synchronous observation system, determine the relative relationship between each light intensity matrix and the instantaneous discharge voltage and discharge time to form a modeling dataset;

[0008] S4. Each element / light intensity value in the light intensity matrix is ​​a target / output object. The dynamic features include the corresponding discharge time and instantaneous voltage value. The XGBoost algorithm is used to establish a multi-output regression model. The optimal model and parameters are evaluated by selecting the loss function through the n-fold cross-validation method.

[0009] S5. Input the instantaneous voltage and corresponding time that change with time into the trained model according to the voltage discharge sequence to obtain the prediction results;

[0010] S6. Restore the series of light intensity matrices obtained from the prediction to obtain the discharge image of the streamer discharge process.

[0011] Preferably, an observation system for lightning impulse streamer discharge experiments is established based on an ICCD camera with nanosecond-level exposure time. In repeated discharges under the same conditions, different ICCD shooting delays are set to obtain discharge light intensity images of the streamer at different times. Furthermore, by changing the applied discharge voltage level, a large number of corresponding discharge images at different voltage levels and times are obtained.

[0012] Preferably, in order to reduce noise in the captured image, the ICCD image is divided into pixel blocks and converted into a discharge light intensity matrix. a ij The average light intensity of each point in each pixel block recorded by the camera is used to determine the relative relationship between each light intensity matrix and the instantaneous discharge voltage and discharge time, thus forming a modeling dataset.

[0013] Preferably, each element / light intensity value in the light intensity matrix is ​​a target / output object, and the dynamic features include the corresponding discharge time and instantaneous voltage value. The XGBoost algorithm is used to establish a multi-output regression model.

[0014] The optimal model and parameters are evaluated by selecting the loss function using the n-fold cross-validation method; the instantaneous voltage and corresponding time that change with time are input into the trained model to obtain a series of light intensity matrices, which are then restored to images to obtain the discharge image of the streamer discharge process.

[0015] The technical effects and advantages of this invention are as follows: This method for predicting the development of long-gap streamer discharge under lightning impact utilizes a streamer discharge experimental observation system to collect streamer discharge images, discharge voltage, and time data captured by an ICCD. After converting the images into a discharge intensity matrix, the relationship between the intensity matrix, discharge voltage, and time is determined. A multi-output regression model is established using the XGBoost algorithm. The optimal model and parameters are evaluated by selecting the loss function using an n-fold cross-validation method. The instantaneous voltage and corresponding time that change with time are input into the trained model according to the voltage discharge sequence to obtain a series of predicted intensity matrices. After restoring the images, the discharge images of the streamer discharge process can be obtained.

[0016] The above-described method for predicting the development of long-gap streamer discharge under lightning impact can overcome the shortcomings of existing technologies that rely on experimental methods to obtain complete and continuous images of long-gap streamer development, and achieve nanosecond-level prediction of the development process of long air gap streamer discharge intensity under lightning impact. Attached Figure Description

[0017] Figure 1 A schematic diagram of pixel block division for an ICCD image;

[0018] Figure 2 This is a schematic diagram showing the relative relationship between the light intensity matrix, instantaneous discharge voltage, and discharge time.

[0019] Figure 3 A flowchart of an image prediction method for the development of long-gap streamer discharge under lightning impact. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] This invention provides, for example Figure 1-3 The method for predicting the development of long-gap streamer discharge under lightning impulse, as shown, includes the following steps:

[0022] S1. Establish a streamer discharge test observation system, and use a nanosecond-level exposure time ICCD camera to take multiple discharge images at different times during the long air gap streamer discharge process under the same conditions. The shooting time interval is on the nanosecond level, and the discharge voltage is recorded synchronously.

[0023] Under the condition that the electrode structure and external environment remain unchanged, change the level of the applied discharge voltage, repeat step 1, and obtain the corresponding discharge images at different voltage levels and different times.

[0024] S2. To reduce noise in the captured images, the ICCD image is divided into pixel blocks and converted into a discharge light intensity matrix. a ij It is the average value of the light intensity of each point in each pixel block recorded by the camera, and the light intensity is normalized. n*m is the number of pixel blocks in the discharge image.

[0025] S3. Based on the measurement results of the synchronous observation system, determine the relative relationship between each light intensity matrix and the instantaneous discharge voltage and discharge time, and form a modeling dataset;

[0026] S4. For the same electrode structure and external environment, the applied voltage is the main factor affecting the development process of streamer discharge, and the discharges are similar. Based on the above assumptions, a multi-output regression model is established using the XGBoost algorithm. Each element (light intensity value) in the light intensity matrix is ​​a target (output) object, and the dynamic characteristics include the corresponding discharge time and instantaneous voltage value.

[0027] S5. For each element, the specific formula for the corresponding model objective function is as follows:

[0028] Where Ω(f) t ) represents the regularization term, and C is a constant. To calculate the mean squared error (MSE), an n-fold cross-validation method was used to select the loss function and evaluate the optimal model and parameters.

[0029] S6. By inputting the instantaneous voltage and corresponding time that change over time into the trained model, a series of light intensity matrices can be obtained. After restoring these matrices to an image, the discharge image of the streamer discharge process can be obtained. This improves upon the shortcomings of existing technologies that rely on experimental methods to obtain complete and continuous images of long-gap streamer development, while also significantly reducing the workload of experiments.

[0030] This method for predicting the development of long-gap streamer discharge under lightning impact utilizes a streamer discharge experimental observation system to acquire streamer discharge images, discharge voltage, and time data captured by an ICCD. After converting the images into discharge intensity matrices, the relationship between the intensity matrix, discharge voltage, and time is determined. A multi-output regression model is established using the XGBoost algorithm, and the optimal model and parameters are evaluated by selecting the loss function through an n-fold cross-validation method. The instantaneous voltage and corresponding time, which change with time, are input into the trained model according to the voltage discharge sequence to obtain a series of predicted intensity matrices. After restoring the images, the discharge images of the streamer discharge process can be obtained.

[0031] The above-described method for predicting the development of long-gap streamer discharge under lightning impact can overcome the shortcomings of existing technologies that rely on experimental methods to obtain complete and continuous images of long-gap streamer development, and achieve nanosecond-level prediction of the development process of long air gap streamer discharge intensity under lightning impact.

[0032] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting the development of long-gap streamer discharge under lightning impulse, characterized in that, Includes the following steps: S1. Using a streamer discharge test observation system, collect streamer discharge images, discharge voltage and time data with nanosecond-level exposure time ICCDs during lightning impulse discharge processes of different voltage levels in long air gaps. S2. Divide the ICCD image into prediction regions according to pixel blocks and convert them into a discharge intensity matrix. a ij The average light intensity of each point in each pixel block recorded by the camera is used to determine the relative relationship between each light intensity matrix and the instantaneous discharge voltage and discharge time, thus forming a modeling dataset. S3. Based on the synchronous observation system, determine the relative relationship between each light intensity matrix and the instantaneous discharge voltage and discharge time to form a modeling dataset; S4. Each element / light intensity value in the light intensity matrix is ​​a target / output object. The dynamic features include the corresponding discharge time and instantaneous voltage value. The XGBoost algorithm is used to establish a multi-output regression model. The optimal model and parameters are evaluated by selecting the loss function through the n-fold cross-validation method. S5. Input the instantaneous voltage and corresponding time that change with time into the trained model according to the voltage discharge sequence to obtain the prediction results; S6. Restore the series of light intensity matrices obtained from the prediction to obtain the discharge image of the streamer discharge process.

2. The method for predicting the development of long-gap streamer discharge under lightning impulse according to claim 1, characterized in that: Based on an ICCD camera, a lightning impulse streamer discharge experimental observation system with nanosecond-level exposure time was established. In multiple repeated discharges under the same conditions, by setting different ICCD shooting delays, discharge light intensity images of the streamer at different times were obtained. Furthermore, by changing the applied discharge voltage level, a large number of corresponding discharge images at different voltage levels and times were obtained.

3. The method for predicting the development of long-gap streamer discharge under lightning impulse according to claim 1, characterized in that: Each element / intensity value in the light intensity matrix is ​​a target / output object. The dynamic features include the corresponding discharge time and instantaneous voltage value. A multi-output regression model is established using the XGBoost algorithm. The optimal model and parameters are evaluated by selecting the loss function using the n-fold cross-validation method. The instantaneous voltage and corresponding time that change with time are input into the trained model to obtain a series of light intensity matrices. After being restored to an image, the discharge image of the streamer discharge process can be obtained.

Images