Lightning protection monitoring method for natural gas station

Abstract

The invention relates to the technical field of lightning protection, and discloses a natural gas station lightning protection monitoring method. The method comprises the following steps: acquiring electric field and equipment state data in real time by using a triangular networking electric field instrument and a multi-type sensor; performing time-frequency domain decomposition on the data, and constructing a feature library containing historical experience; then, fusing physical model calculation and historical mode matching, and generating a lightning current predicted value; an early warning threshold value is adjusted according to the health state of the equipment, and accurate early warning is achieved; according to the method, a physical rule formed by thunder and lightning is combined with a historical data driven mode, meanwhile, the influence of the equipment state on early warning is considered, the problems that a traditional method is low in prediction precision and poor in equipment linkage performance are solved, and the method is suitable for popularization and application. And a dynamic and accurate thunder and lightning protection scheme is provided for the natural gas station.

Description

Technical Field

[0001] The present invention relates to the field of lightning protection technology, and in particular to a lightning protection monitoring method for a natural gas station. Background Art

[0002] Current lightning protection monitoring at natural gas stations mainly uses prediction methods based on physical equations, such as using Gauss's theorem to calculate the charge of thunderclouds to derive lightning current, statistical models based on historical data, such as using pattern matching to predict lightning current amplitude, and independent equipment status monitoring, such as surge protector leakage current detection and ground resistance measurement. These technologies provide support for protection from the perspectives of lightning physical mechanisms, historical law statistics, and equipment health status, but have not yet formed an integrated system in engineering applications.

[0003] The above technologies have the following shortcomings: pure physical models are difficult to adapt to complex terrain and extreme weather conditions, and the prediction errors are large; statistical models lack the ability to recognize new lightning patterns and are prone to underreporting; equipment status monitoring and lightning prediction are independent of each other, and it is impossible to dynamically adjust the early warning strategy according to the real-time performance of the equipment; multi-source monitoring data lacks an efficient fusion mechanism, and the collaborative prediction capability is weak. There is an urgent need for a dynamic monitoring solution that integrates physical laws, historical data and equipment status to improve protection accuracy and response efficiency. Summary of the Invention

[0004] The technical problem to be solved by the present invention is that the existing technology has the disadvantage of inaccurate prediction. For this reason, we propose a lightning protection monitoring method for a natural gas station.

[0005] To achieve the above objectives, the present application adopts the following technical solution: a method for monitoring lightning protection at a natural gas station, comprising the following steps:

[0006] S1. Data Collection: Three electric field meters are installed in an open area near the natural gas station, forming a triangular network. This allows the movement of charged clouds to be located, and the electric field intensity E and rate of change dE / dt to be monitored in real time. The leakage current of the surge protector is also uploaded in real time via IoT sensors. , grounding system resistance , status data of air terminal corrosion degree;

[0007] S2. Feature Library Construction: Based on historical lightning current data around the station, a historical probability library is constructed by season, time period, and weather type, storing the mean, standard deviation, and frequency domain characteristics of lightning current under each mode. Real-time electric field data is decomposed in the time and frequency domains to extract time and frequency domain variation characteristics and construct feature vectors. Feature vectors of time and frequency domain characteristics are then constructed, and a probability model that includes charge separation characteristics is constructed in combination with meteorological parameters.

[0008] S3. Fusion prediction: Match the real-time feature vector with the historical probability library, calculate the theoretical range of lightning current using the physical equations of thundercloud electrification, and simultaneously match historical lightning current patterns using a dynamic time warping algorithm to generate a lightning current amplitude prediction value and its confidence interval. The half-width of the confidence interval is defined as the ratio of half the difference between the upper and lower limits of the confidence interval to the predicted value.

[0009] S4. Equipment status correction warning: Dynamically adjust the warning threshold based on the aging degree of the surge protector and the grounding grid performance parameters;

[0010] S5. Optimization: After each warning, the actual lightning current, warning lead time, and equipment response status are recorded. Feature weights are updated through SHAP value analysis. The historical probability library is iterated quarterly to optimize the prediction model parameters.

[0011] Preferably, the networking parameters of the electric field meter are: 3 electric field meters are used to form an equilateral triangle with a side length of 5-15km, and the clock synchronization error is achieved based on Beidou timing to achieve ≤1μs; the electric field measurement range is ±50kV / m, the resolution is 0.1kV / m, the sampling rate is 10Hz, and a built-in 50Hz notch filter is used to suppress industrial interference; the surge protector monitoring uses a magnetic balance sensor to collect leakage current in real time, and when the leakage current is greater than 10μA, it is judged as an aging warning.

[0012] Preferably, the time domain features and frequency domain features are:

[0013] Time domain features: Extract the rising rate of the electric field curve 5 minutes before the flash , and are the maximum and minimum values ​​of the electric field intensity during the observation period, respectively, and 300 s is the time window used to quantify the time scale of the electric field change;

[0014] Frequency domain features: Perform 5-layer Db4 wavelet decomposition on dE / dt and calculate the proportion of high-frequency energy between 100-1000Hz , f is the frequency, It represents the frequency domain component at frequency f in the frequency domain after wavelet decomposition of the electric field change rate signal dE / dt, or called the frequency domain coefficient. Specifically, after 5-layer Db4 wavelet decomposition of dE / dt, the signal will be decomposed into different frequency bands. That is, the amplitude or energy distribution of each frequency point within a certain frequency band;

[0015] Construct feature vectors from time domain features and frequency domain features .

[0016] Preferably, the calculation formula for the lightning current amplitude prediction value is: , where Q is the charge of the thundercloud, and the peak value of the electric field before the lightning is monitored by the triangular network electric field instrument And the cloud top height h retrieved by radar, the thundercloud charge is derived by Gauss's theorem , is the vacuum dielectric constant; combined with the discharge efficiency and discharge duration Calculate the basic amplitude of lightning current , S is the terrain shielding angle between the thundercloud and the station calculated by the digital elevation model Generate shielding factor , Determine the physical boundary of the predicted lightning current amplitude for the terrain shielding angle between the thundercloud and the station , To generate the correction coefficient, a dynamic time warping algorithm is used to match the current electric field change rate characteristics with the patterns in the historical lightning current database to generate the correction coefficient. , , is the dynamic time-warped distance between the current feature and the historical pattern, The historical average distance updated quarterly, the final lightning current amplitude forecast is .

[0017] Preferably, Sakoe-Chiba bandwidth constraint is used for similarity matching, and the bandwidth value is 10% of the window length. The matching distance calculation formula is: ,in , .

[0018] Preferably, the correction model of the warning threshold is , , hour, , , hour, , α=0.5, β=0.8, determined by fitting historical accident data, triggering a graded warning.

[0019] Preferably, the steps of updating feature weights by SHAP value analysis are:

[0020] S51. Generate 100 feature perturbation samples for the current prediction sample;

[0021] S52. Calculate the prediction deviation caused by each feature perturbation ;

[0022] S53. Feature importance weight update, the update formula is: , is the deviation after the perturbation of the i-th feature, is the feature importance weight.

[0023] Preferably, the installation of the electric field meter also meets the following requirements: the sensor head is 5-10m from the ground, and there are no obstacles higher than 3m within 10m around it; it is equipped with a lightning surge protector with a protection level of In≥20kA and a response time of ≤1ns; it has a built-in GPS timing module, and the constant temperature crystal oscillator is used to keep time when the lock is lost, with a 24-hour error of ≤10μs.

[0024] Preferably, the cloud top height h is obtained using an S-band radar, and the physical constraint calculation is started when the echo intensity is greater than 45dBZ and h is greater than 10km; the radar and electric field data are aligned by time stamps, and the asynchronous data is corrected using cubic spline interpolation, with a correction error of ≤5%.

[0025] Preferably, When the confidence interval half width is greater than 20%, the drone is triggered to carry out a spiral scan at the cloud base altitude of 1-3 km. The scan data is fused into the Kalman filter. , the update step is 1 minute.

[0026] Technical effects and advantages of the present invention:

[0027] In the present invention, lightning protection is achieved through the steps of monitoring, analysis, prediction, correction and optimization: first, a triangular network electric field meter and multiple types of sensors are used to collect electric field and equipment status data in real time; then the data is decomposed in the time-frequency domain to construct a feature library containing historical experience; then the physical model calculation and historical pattern matching are integrated to generate a lightning current prediction value; then the warning threshold is adjusted according to the health status of the equipment to achieve accurate warning; finally, the model parameters are optimized through actual data feedback to form an adaptive evolutionary system. The present invention combines the physical laws of lightning formation with pattern recognition driven by historical data, while considering the impact of equipment status on warning, solving the problems of low prediction accuracy and poor equipment linkage of traditional methods, and providing a dynamic and accurate lightning protection solution for natural gas stations. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The disclosure of the present invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. In the drawings, the same reference numerals are used to refer to the same components:

[0029] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0030] It is easy to understand that according to the technical solution of the present invention, without changing the essential spirit of the present invention, a person skilled in the art can propose a variety of interchangeable structural modes and implementation modes. Therefore, the following specific embodiments and drawings are only exemplary descriptions of the technical solution of the present invention and should not be regarded as the entire invention or as a limitation or restriction of the technical solution of the present invention.

[0031] Reference Figure 1 As shown, the present invention provides a technical solution: a natural gas station lightning protection monitoring method, comprising the following steps:

[0032] S1. Data Collection: Three electric field meters are installed in an open area near the natural gas station, forming a triangular network. This allows the movement of charged clouds to be located, and the electric field intensity E and rate of change dE / dt to be monitored in real time. The leakage current of the surge protector is also uploaded in real time via IoT sensors. , grounding system resistance , lightning receptor corrosion status data; the electric field meter monitors the atmospheric electric field strength and its rate of change in real time, while IoT sensors collect equipment status data such as surge protector leakage, grounding system resistance, and lightning receptor corrosion status. This triangular network design can accurately locate the movement trajectory of charged clouds, just like three networks working together to capture cloud dynamics, and multi-type sensor data collection provides comprehensive raw information for subsequent analysis;

[0033] S2. Feature library construction: Based on historical lightning current data around the station, a historical probability library is constructed by season, time period, and weather type, storing the mean, standard deviation, and frequency domain characteristics of lightning current under each mode. Real-time electric field data is decomposed in the time and frequency domains to extract time and frequency domain variation characteristics and form a feature vector. Historical lightning current data is classified and organized by season, time period, and weather type to form a historical probability library, storing the average value, fluctuation range, and high-frequency energy characteristics of lightning current under various weather conditions. Real-time electric field data is decomposed in the time and frequency dimensions to extract the rising speed of the electric field within 5 minutes before the flash and the energy proportion in the 100-1000 Hz frequency band. These features are integrated into a feature vector to establish a "fingerprint library" for lightning events. After feature extraction, real-time data can be compared with historical fingerprints to determine the possible intensity of the current lightning.

[0034] S3. Fusion prediction: Match the real-time feature vector with the historical probability library, calculate the theoretical range of lightning current using the physical equation of thundercloud electrification, and simultaneously match the historical lightning current pattern through the dynamic time warping algorithm to generate a lightning current amplitude prediction value and its confidence interval; the half-width of the confidence interval is defined as the ratio of half the difference between the upper and lower limits of the confidence interval to the predicted value; compare the real-time extracted feature vector with the pattern in the historical probability library, and based on the physical laws of thundercloud electrification, estimate the thundercloud charge by the electric field peak and cloud top height, and then calculate the theoretical range of lightning current in combination with terrain shielding. Use the dynamic time warping algorithm to match the current electric field change characteristics with the characteristics of historical lightning current events for similarity. For example, when the current electric field rises rapidly and the high-frequency energy is high, if similar characteristics historically correspond to strong lightning currents, the prediction value will be corrected to improve the prediction accuracy;

[0035] S4. Equipment Status Correction Warning: Dynamically adjusts warning thresholds based on surge protector aging and grounding grid performance parameters. Based on the surge protector's aging, such as leakage current exceeding 10 microamperes, and the grounding grid's conductivity (greater resistance means lower performance), the lightning current warning threshold is dynamically adjusted. For example, when the surge protector ages, the equipment's lightning resistance decreases. In this case, the warning threshold is lowered, issuing an early warning to ensure the equipment is optimally prepared for lightning strikes.

[0036] S5. Optimization: After each warning, the actual lightning current, warning lead time, and equipment response status are recorded, and the feature weights are updated through SHAP value analysis. The historical probability library is iterated quarterly to optimize the prediction model parameters. After each warning, the actual lightning current size, warning lead time, and equipment response are recorded. The SHAP value is used to analyze the weights of various features, such as the electric field change rate and the proportion of high-frequency energy, that affect the prediction results. The historical probability library and model parameters are updated quarterly.

[0037] The networking parameters of the electric field meter are as follows: three electric field meters are used to form an equilateral triangle with a side length of 5-15km, and the clock synchronization error is achieved based on Beidou timing with a clock synchronization error of ≤1μs; the electric field measurement range is ±50kV / m, the resolution is 0.1kV / m, the sampling rate is 10Hz, and a built-in 50Hz notch filter is used to suppress industrial interference; the surge protector monitoring uses a magnetic balance sensor to collect leakage current in real time, and when the leakage current is greater than 10μA, it is judged as an aging warning; the installation of the electric field meter also meets the following requirements: the sensor head is 5-10m from the ground, and there are no obstacles higher than 3m within 10m around it; it is equipped with a lightning surge protector with a protection level In≥20kA and a response time of ≤1ns; it has a built-in GPS timing module, and the constant temperature crystal oscillator is used to keep time when the lock is lost, with a 24-hour error of ≤10μs.

[0038] The time domain features and frequency domain features are:

[0039] Time domain features: Extract the rising rate of the electric field curve 5 minutes before the flash , and are the maximum and minimum values ​​of the electric field intensity during the observation period, respectively, and 300 s is the time window used to quantify the time scale of the electric field change;

[0040] Frequency domain features: Perform 5-layer Db4 wavelet decomposition on dE / dt and calculate the proportion of high-frequency energy between 100-1000Hz , f is the frequency, It represents the frequency domain component at frequency f in the frequency domain after wavelet decomposition of the electric field change rate signal dE / dt, or called the frequency domain coefficient. Specifically, after 5-layer Db4 wavelet decomposition of dE / dt, the signal will be decomposed into different frequency bands. That is, the amplitude or energy distribution of each frequency point within a certain frequency band;

[0041] Construct feature vectors from time domain features and frequency domain features .

[0042] The calculation formula for the predicted value of lightning current amplitude is: According to the peak value of the electric field before the flash monitored by the triangular network electric field instrument And the cloud top height h retrieved by radar, the thundercloud charge is derived by Gauss's theorem , is the vacuum dielectric constant; combined with the discharge efficiency and discharge duration Calculate the basic amplitude of lightning current , S is the terrain shielding angle between the thundercloud and the station calculated by the digital elevation model Generate shielding factor , Determine the physical boundary of the predicted lightning current amplitude for the terrain shielding angle between the thundercloud and the station , As the correction coefficient, the dynamic time warping algorithm is used to perform similarity matching between the current electric field change rate characteristics and the patterns in the historical lightning current database. The Sakoe-Chiba bandwidth constraint is used for similarity matching, and the bandwidth value is 10% of the window length. The matching distance calculation formula is: ,in , , generate the correction coefficient , , is the dynamic time-warped distance between the current feature and the historical pattern, The historical average distance updated quarterly, the final lightning current amplitude forecast is .

[0043] The revised model of the warning threshold is: , , hour, , , hour, , α=0.5, β=0.8, determined by fitting historical accident data, triggering a graded warning.

[0044] The steps for updating feature weights through SHAP value analysis are:

[0045] S51. Generate 100 feature perturbation samples for the current prediction sample;

[0046] S52. Calculate the prediction deviation caused by each feature perturbation ;

[0047] S53. Feature importance weight update, the update formula is: , is the deviation after the perturbation of the i-th feature, is the feature importance weight.

[0048] The cloud top height h is obtained using S-band radar. Physical constraint calculations are initiated when the echo intensity is greater than 45 dBZ and h is greater than 10 km. The radar and electric field data are aligned by time stamps, and the asynchronous data are corrected using cubic spline interpolation with a correction error of ≤5%.

[0049] When the confidence interval half width is greater than 20%, the drone is triggered to carry out a spiral scan at the cloud base altitude of 1-3 km. The scan data is fused into the Kalman filter. , the update step is 1 minute.

[0050] Working principle: This method achieves lightning protection through the steps of monitoring, analysis, prediction, correction and optimization: first, a triangular network electric field meter and multi-type sensors are used to collect electric field and equipment status data in real time; then the data is decomposed in the time and frequency domain to build a feature library containing historical experience; then the physical model calculation and historical pattern matching are integrated to generate lightning current prediction values; then the warning threshold is adjusted according to the health status of the equipment to achieve accurate warning; finally, the model parameters are optimized through actual data feedback to form an adaptive evolutionary system. Its core innovation lies in combining the physical laws of lightning formation with pattern recognition driven by historical data, while considering the impact of equipment status on warning, solving the problems of low prediction accuracy and poor equipment linkage of traditional methods, and providing natural gas stations with dynamic and accurate lightning protection solutions.

[0051] The technical scope of the present invention is not limited to the contents of the above description. Those skilled in the art can make various deformations and modifications to the above embodiments without departing from the technical idea of ​​the present invention, and these deformations and modifications should all fall within the protection scope of the present invention.

Claims

1. A lightning protection monitoring method for a natural gas station, characterized in that: The following steps are involved: S1. Data Collection: Three electric field meters are installed in an open area near the natural gas station, forming a triangular network. This allows the movement of charged clouds to be located, and the electric field intensity E and rate of change dE / dt to be monitored in real time. The leakage current of the surge protector is also uploaded in real time via IoT sensors. , grounding system resistance , status data of air terminal corrosion degree; S2. Feature library construction: Based on the historical lightning current data around the station, a historical probability library is constructed by season, time period, and weather type, storing the mean, standard deviation, and frequency domain characteristics of lightning current in each mode; Perform time-frequency decomposition on real-time electric field data, extract time-domain and frequency-domain variation features, and construct feature vectors of time-domain and frequency-domain features. Combined with meteorological parameters, a probabilistic model including charge separation features is constructed. S3. Fusion prediction: Match the real-time feature vector with the historical probability library, calculate the theoretical range of lightning current using the physical equations of thundercloud electrification, and simultaneously match historical lightning current patterns using a dynamic time warping algorithm to generate a lightning current amplitude prediction value and its confidence interval. The half-width of the confidence interval is defined as the ratio of half the difference between the upper and lower limits of the confidence interval to the predicted value. S4. Equipment status correction warning: Dynamically adjust the warning threshold based on the aging degree of the surge protector and the grounding grid performance parameters; S5. Optimization: After each warning, the actual lightning current, warning lead time, and equipment response status are recorded. Feature weights are updated through SHAP value analysis. The historical probability library is iterated quarterly to optimize the prediction model parameters.

2. The method for monitoring lightning protection at a natural gas station according to claim 1, characterized in that: The networking parameters of the electric field meter are as follows: three electric field meters are used to form an equilateral triangle with a side length of 5-15 km, and the clock synchronization error is achieved ≤1μs based on Beidou timing; the electric field measurement range is ±50kV / m, the resolution is 0.1kV / m, the sampling rate is 10Hz, and a built-in 50Hz notch filter is used to suppress industrial interference; the surge protector monitoring uses a magnetic balance sensor to collect leakage current in real time, and when the leakage current is greater than 10μA, it is judged as an aging warning.

3. The method for monitoring lightning protection at a natural gas station according to claim 1, characterized in that: The time domain features and frequency domain features are: Time domain features: Extract the rising rate of the electric field curve 5 minutes before the flash , and are the maximum and minimum values ​​of the electric field intensity during the observation period, respectively, and 300 s is the time window used to quantify the time scale of the electric field change; Frequency domain features: Perform 5-layer Db4 wavelet decomposition on dE / dt and calculate the proportion of high-frequency energy between 100-1000Hz , f is the frequency, It represents the frequency domain component at frequency f in the frequency domain after wavelet decomposition of the electric field change rate signal dE / dt, or called the frequency domain coefficient. Specifically, after 5-layer Db4 wavelet decomposition of dE / dt, the signal will be decomposed into different frequency bands. That is, the amplitude or energy distribution of each frequency point within a certain frequency band; Construct feature vectors from time domain features and frequency domain features .

4. The method for monitoring lightning protection at a natural gas station according to claim 3, characterized in that: The calculation formula for the lightning current amplitude prediction value is: , where Q is the charge of the thundercloud, and the peak value of the electric field before the lightning is monitored by the triangular network electric field instrument And the cloud top height h retrieved by radar, the thundercloud charge is derived by Gauss's theorem , is the vacuum dielectric constant; combined with the discharge efficiency and discharge duration Calculate the basic amplitude of lightning current , S is the terrain shielding angle between the thundercloud and the station calculated by the digital elevation model Generate shielding factor , Determine the physical boundary of the predicted lightning current amplitude for the terrain shielding angle between the thundercloud and the station , To generate the correction coefficient, a dynamic time warping algorithm is used to match the current electric field change rate characteristics with the patterns in the historical lightning current database to generate the correction coefficient. , , is the dynamic time-warped distance between the current feature and the historical pattern, The historical average distance updated quarterly, the final lightning current amplitude forecast is .

5. The method for monitoring lightning protection at a natural gas station according to claim 4, characterized in that: The similarity matching adopts the Sakoe-Chiba bandwidth constraint, the bandwidth value is 10% of the window length, and the matching distance calculation formula is: ,in , .

6. The method for monitoring lightning protection at a natural gas station according to claim 4, characterized in that: The correction model of the warning threshold is: , , hour, , , hour, , α=0.5, β=0.8, determined by fitting historical accident data, triggering a graded warning.

7. The method for monitoring lightning protection at a natural gas station according to claim 6, characterized in that: The steps of updating feature weights by SHAP value analysis are as follows: S51. Generate 100 feature perturbation samples for the current prediction sample; S52. Calculate the prediction deviation caused by each feature perturbation ; S53. Feature importance weight update, the update formula is: , is the deviation after the perturbation of the i-th feature, is the feature importance weight.

8. The method for monitoring lightning protection at a natural gas station according to claim 2, characterized in that: The installation of the electric field meter also meets the following requirements: the sensor head is 5-10m from the ground, and there are no obstacles higher than 3m within 10m around it; it is equipped with a lightning surge protector with a protection level of In≥20kA and a response time of ≤1ns; it has a built-in GPS timing module, and when the lock is lost, the constant temperature crystal oscillator is used to keep time, with a 24-hour error of ≤10μs.

9. The method for monitoring lightning protection at a natural gas station according to claim 4, characterized in that: The cloud top height h is obtained using an S-band radar. Physical constraint calculation is initiated when the echo intensity is greater than 45 dBZ and h is greater than 10 km. The radar and electric field data are aligned through time stamps, and the asynchronous data is corrected using cubic spline interpolation, with a correction error of ≤5%.

10. The method for monitoring lightning protection at a natural gas station according to claim 4, characterized in that: described When the confidence interval half width is greater than 20%, the drone is triggered to carry out a spiral scan at the cloud base altitude of 1-3 km. The scan data is fused into the Kalman filter. , the update step is 1 minute.

Images