A method and system for analyzing the performance of a lightning protection device in a flammable and explosive place

Abstract

The present application relates to the field of data processing, more particularly, the present application relates to a kind of flammable and explosive place lightning protection device performance analysis method and system, method includes: first, the resistance, temperature and humidity data in the preset days are collected, and preprocessing is carried out to construct the target resistance time sequence of each day;Then, based on the day-night temperature difference, these days are clustered, and multiple cycle windows are divided;Then, the target resistance time sequence of all days is decomposed, the periodicity level and abnormal degree of each layer are calculated, and the performance evaluation value is obtained after being weighted after conversion into weight;Finally, if the value is less than or equal to the preset abnormal threshold, it is determined that the device is abnormal.The periodicity level and abnormal degree of resistance change are comprehensively considered, the performance of lightning protection grounding device can be comprehensively and accurately evaluated, potential abnormalities can be found in time, and misjudgment or omission caused by single factor evaluation can be avoided.

Description

Technical Field

[0001] This invention relates to the field of data processing. More specifically, this invention relates to a method and system for analyzing the performance of lightning protection devices in flammable and explosive environments. Background Technology

[0002] In flammable and explosive workshop environments, the performance of lightning protection devices is crucial. To ensure the proper functioning of lightning protection devices, their grounding resistance needs to be monitored regularly. Grounding resistance is a direct indicator of the electrical conductivity of a grounding connection, and changes in grounding resistance can reflect changes in the performance of the lightning protection device.

[0003] Existing technologies, such as the patent application with publication number CN114217136A, disclose a statistical method for detecting lightning protection grounding resistance based on big data. This method constructs a histogram of resistance value distribution at each detection point using historical observation data, employs a Gaussian mixture model to quantify the degree of anomaly in a single measurement, calculates the growth trend by weighting the resistance change within a time window with the degree of anomaly, and optimizes the anomaly judgment parameters through an iterative correction mechanism until the growth rate of all data points meets the standard. Finally, the detection priority is determined based on the average growth rate of the resistance value at each detection point, and a differentiated probability sampling strategy is formulated accordingly to complete the grounding resistance monitoring.

[0004] However, when obtaining the degree of anomaly in grounding resistance observations, relying solely on the number of histogram maxima to construct a Gaussian mixture model to fit the parameters and calculate the degree of anomaly cannot accurately capture the complex changing trends and potential anomalies in the grounding resistance data. It is easily affected by noise and random fluctuations, which can lead to incorrect identification of anomalies. Summary of the Invention

[0005] To address the technical problem that the existing technology cannot effectively capture changes in grounding resistance, leading to incorrect identification of abnormal situations, the present invention provides solutions in the following aspects.

[0006] In the first aspect, a method for analyzing the performance of lightning protection devices in flammable and explosive locations includes:

[0007] Collect and preprocess the resistance, temperature and humidity parameters of the lightning protection grounding device within a preset number of days, and extract and construct the target resistance time series for each day from the resistance parameters collected each day.

[0008] Based on the diurnal temperature difference, all days within a preset number of days are clustered to obtain multiple clusters. The clustering results are used to divide the target resistance time series of all days into multiple periodic windows. Based on the optimal decomposition layer, the target resistance time series of all days is decomposed. The periodicity level and anomaly degree of all periodic windows in each layer are calculated. The periodicity level is converted into the weight of the corresponding layer. The anomaly degree of each layer is multiplied by the weight of the corresponding layer and weighted and summarized to obtain the performance evaluation value of the target resistance time series of all days.

[0009] When the performance evaluation value is less than or equal to the preset abnormal threshold, the lightning protection grounding device is determined to be abnormal.

[0010] Preferably, the process of constructing the target resistance timing sequence includes:

[0011] The continuous sampling data of a day is divided into multiple sequential sampling intervals at certain intervals. Then, the point with the lowest humidity is found in each sampling interval, and the corresponding grounding resistance is extracted. This operation is repeated for all sampling intervals. Finally, the extracted grounding resistances are arranged in the sampling order to obtain the target resistance time sequence for the day.

[0012] Preferably, the clustering further includes adjusting the cluster radius according to the variation of the diurnal temperature range, and the adjustment process includes:

[0013] Select any day as the target day, and calculate the average of the sum of the absolute values ​​of the differences between the daytime and nighttime temperature differences of the target day and the two days before and after it, as the range of temperature difference change of the target day;

[0014] The mean of the absolute difference between the diurnal and daytime temperature differences of two consecutive days within all days is calculated, and combined with the normalized temperature difference variation range, the cluster radius required for clustering the target day is obtained.

[0015] Preferably, the process of obtaining the optimal number of decomposition layers includes:

[0016] Set the range of decomposition layers, perform multi-layer decomposition on the target resistance time series for all days, calculate the mean, median and standard deviation of the autocorrelation values ​​of all periodic windows in each layer, and calculate the periodicity level of all periodic windows in each layer based on the mean, median and standard deviation of the autocorrelation values.

[0017] The decomposition effect evaluation value corresponding to any decomposition layer is calculated based on the periodic average value of each layer's periodic window, the maximum value of the periodic level, and the reconstruction error after decomposition.

[0018] Traverse the range of decomposition levels and select the level that maximizes the decomposition effect evaluation value as the optimal decomposition level.

[0019] Preferably, the process of obtaining the periodicity level includes:

[0020] Divide the standard deviation of the autocorrelation value corresponding to each layer by the mean of the corresponding autocorrelation value to obtain the coefficient of variation for each layer. Multiply the result of subtracting the coefficient of variation from 1 by the median of the autocorrelation value of the corresponding layer to obtain the periodicity level of all periodic windows in each layer.

[0021] Preferably, the process of obtaining the degree of abnormality includes:

[0022] After decomposing the target resistance time series of all days based on the optimal decomposition layer, the target resistance time series of each layer is segmented according to the clustering operation. The slope of all segments in each layer is obtained by fitting using the least squares method. The standard deviation of the slope of all segments in each layer is calculated, as well as the difference between the maximum and minimum slope values ​​of all segments in each layer is calculated. The product of the standard deviation of the slope of all segments in each layer and the difference between the maximum and minimum slope values ​​of all segments in each layer is normalized using norm to obtain the anomaly degree of all periodic windows in each layer.

[0023] Preferably, the process of obtaining the degree of abnormality includes:

[0024] After decomposing the target resistance time series of all days based on the optimal decomposition layer, the target resistance time series of each layer is segmented according to the clustering operation. The slope of all segments in each layer is obtained by fitting using the least squares method, and the standard deviation and mean of the slope of all segments in each layer are calculated. The ratio of the standard deviation to the mean is used as the degree of variation of the slope of all segments in each layer.

[0025] Sort all segments of each layer by slope from smallest to largest, obtain the first quartile and the third quartile, and calculate the interquartile range.

[0026] The abnormality of all periodic windows in each layer is obtained by normalizing the product of the variation degree of all segments in each layer and the interquartile range.

[0027] Secondly, a performance analysis system for lightning protection devices in flammable and explosive locations includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the performance analysis method for lightning protection devices in flammable and explosive locations described in any one of the claims is implemented.

[0028] The beneficial effects of this invention are:

[0029] First, considering the influence of humidity on resistance, the grounding resistance corresponding to the lowest humidity point in each sampling interval is selected to construct the target resistance time series. This can accurately reflect the resistance change of the lightning protection grounding device under a relatively stable humidity environment, providing a more reliable data basis for subsequent analysis.

[0030] Secondly, by clustering diurnal temperature differences, the number of days with similar temperature change patterns is classified, capturing the regularity of resistance changes under different climatic conditions (such as high temperature leading to increased resistance and low temperature leading to decreased resistance), making the clustering results more reasonable, accurately dividing the periodic window, and further improving the accuracy of subsequent analysis.

[0031] Then, by using the optimal number of decomposition layers to decompose the target resistor timing sequence, different components in the timing sequence can be better separated, the periodic characteristics of resistance changes can be highlighted, noise interference can be reduced, and strong support can be provided for subsequent accurate performance evaluation, avoiding information loss or noise mixing due to improper number of decomposition layers.

[0032] Finally, by comprehensively considering the periodicity and degree of abnormality of resistance changes, the performance of lightning protection grounding devices can be comprehensively and accurately evaluated, potential anomalies can be detected in a timely manner, and misjudgments or omissions caused by evaluation of a single factor can be avoided. Attached Figure Description

[0033] Figure 1 This is a flowchart of steps S1-S3 in a method for analyzing the performance of lightning protection devices in flammable and explosive locations according to an embodiment of the present invention. Detailed Implementation

[0034] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0035] Reference Figure 1 A method for analyzing the performance of lightning protection devices in flammable and explosive locations includes steps S1-S3, as detailed below:

[0036] S1: Collect and preprocess the resistance, temperature and humidity parameters of the lightning protection grounding device within a preset number of days, and extract and construct the target resistance time series for each day from the resistance parameters collected each day.

[0037] To monitor grounding resistance, three types of sensors need to be installed on the lightning protection grounding system in the workshop: a resistance sensor, a temperature sensor, and a humidity sensor. These sensors collect data at fixed sampling intervals (e.g., once per minute). For example, data from the past year is collected, generating a set of data every minute, each set including the three parameters: grounding resistance, temperature, and humidity.

[0038] In the data collected above, it is considered that the grounding resistance is affected by temperature and humidity. The effect of humidity on the grounding resistance is irregular and unstable, while the effect of temperature changes (such as diurnal temperature difference and seasonal changes) on the grounding resistance is periodic and relatively stable.

[0039] To better analyze the long-term trend and short-term fluctuations of grounding resistance, reduce redundant data, and minimize the impact of humidity on the analysis, the continuous sampling data of a day is divided into multiple sequential sampling intervals at certain intervals (in this embodiment of the invention, it is divided according to one hour). Then, within each sampling interval (each sampling interval has 60 sets of data), the point with the lowest humidity is identified, and the corresponding grounding resistance is extracted. This operation is repeated for all sampling intervals. Finally, the extracted grounding resistances are arranged in the sampling order to obtain the target resistance time series for each day.

[0040] S2: Based on the diurnal temperature difference, cluster all days within a preset number of days to obtain multiple clusters. Use the clustering results to divide the target resistance time series of all days into multiple periodic windows. Decompose the target resistance time series of all days based on the optimal decomposition level, calculate the periodicity level and anomaly degree of all periodic windows in each layer, convert the periodicity level into the weight of the corresponding layer, multiply the anomaly degree of each layer by the weight of the corresponding layer, and perform weighted summation to obtain the performance evaluation value of the target resistance time series of all days.

[0041] Since the influence of humidity changes has been eliminated in S1 above, the influence of temperature changes on grounding resistance also needs to be considered when further analyzing the grounding resistance.

[0042] Although temperature changes (diurnal and seasonal variations) have a periodic effect on grounding resistance, the specific period length (i.e. how long it takes for a temperature change to cause the grounding resistance to complete a full periodic change) is unclear, as diurnal and seasonal temperature changes each have different period lengths.

[0043] Since the period length is not clear, in order to more accurately analyze the impact of temperature changes on grounding resistance, it is necessary to first divide the period window using temperature parameters.

[0044] In one embodiment, the temperature difference variation amplitude for each sampled day (reflecting the local fluctuation of the temperature difference each day) is calculated. That is, any day is selected as the target day, and the average of the sum of the absolute values ​​of the differences between the target day and the two days before and after it is calculated. This is expressed by the following formula:

[0045]

[0046] In the formula, For the first The range of temperature changes over the day For the first The temperature difference between day and night, For the first The temperature difference between day and night, For the first The temperature difference between day and night.

[0047] Further, the above Mapped to the interval [0, 1], after normalization, The larger the value, the more drastic the temperature fluctuation. The closer to 1; The smaller the value, the more stable the temperature difference after normalization. The closer it is to 0.

[0048] In addition, it is necessary to calculate the mean of the absolute differences in diurnal temperature range between two consecutive days within all sampled days, thus reflecting the average scale of the overall temperature difference change, i.e., satisfying the following relationship:

[0049]

[0050] This represents the mean of the absolute differences in diurnal temperature range between two consecutive days within all sampled days. For the first The temperature difference between day and night, For the first The temperature difference between day and night.

[0051] Furthermore, multiplying this average difference by the normalized temperature difference variation range for the target day yields the cluster radius required for clustering the target day, satisfying the following relationship:

[0052]

[0053] In the formula, For the first The cluster radius required when performing clustering. This represents the mean of the absolute differences in diurnal temperature range between two consecutive days within all sampled days. For the first The range of temperature changes over the day This indicates normalization processing.

[0054] Among them, when When it is big, Small, Smaller clusters result in more subdivided classes with smaller radii, avoiding the erroneous merging of days with large differences; conversely, when... Hour, big, Larger clusters will merge more adjacent clusters, forming clusters with larger radii.

[0055] Furthermore, utilizing the above-mentioned... The process of obtaining the cluster radius required for clustering involves clustering the diurnal temperature difference of all days (using DBSCAN clustering as an example) to obtain multiple clusters. Using the clustering results, the target resistance time series corresponding to the number of days contained in the same cluster are grouped into the same periodic window. In this way, the target resistance time series of all days can be divided into multiple periodic windows, and the periodic change pattern of resistance data affected by temperature in the same window is roughly the same.

[0056] After this clustering is completed, the timing sequence of the target resistance within each period window can be analyzed and processed more accurately.

[0057] Grounding resistance typically contains multiple components: short-term fluctuations, caused by environmental factors such as temperature and humidity changes or random noise, which are characterized by high frequency and rapid changes; and long-term trends, caused by slow processes such as equipment aging and corrosion, which are characterized by low frequency and stable changes.

[0058] DTW (Discrete Wavelet Transform) divides data into different frequency bands through a step-by-step decomposition. These frequency bands are generated by a series of wavelet functions and scaling functions, which can capture the details and trend information in the data.

[0059] These components can be separated using DTW decomposition. This decomposition breaks down the resistance data into multiple levels of different frequencies, each corresponding to a specific characteristic of the signal. Specifically:

[0060] First, the target resistance timing sequence for all days is pre-decomposed into I layers, where the value of I is an integer in the range [3, 10]. This range can be adjusted according to actual conditions such as sequence length and noise level.

[0061] Then, DTW is used to perform multi-level decomposition on the target resistance time series for all days (using orthogonal wavelet bases such as db4) to obtain approximation coefficients and detail coefficients at different levels. The approximation coefficients reflect the long-term trend of the resistance data (such as corrosion, aging, etc.), while the detail coefficients reflect the short-term fluctuations of the resistance data (such as changes in environmental humidity, temperature effects, etc.). Each level of the target resistance time series has its corresponding periodicity level, which is determined by the median and coefficient of variation of the autocorrelation value of the corresponding periodic window.

[0062] For example, the target resistance time series of all days is clustered into 5 periodic windows, which are decomposed into three layers, each of which also has 5 periodic windows. Then, the autocorrelation value of all periodic windows in each layer is calculated (i.e., the autocorrelation value of each periodic window in each layer is calculated using the existing autocorrelation function, and the maximum value of the autocorrelation function output is selected as the autocorrelation value of each periodic window), and the mean, median and standard deviation of all autocorrelation values ​​in each layer are calculated.

[0063] The periodicity level of each layer is then calculated based on the mean, median, and standard deviation of the autocorrelation, satisfying the following relationship:

[0064]

[0065] In the formula, For the first The periodicity level of the layer periodic window, For the first The median of the layer autocorrelation value, For the first Standard deviation of layer autocorrelation values For the first The mean of the layer autocorrelation values. Among them, Used to reduce the impact of resistance data in individual abnormal windows. It is the first The coefficient of variation (COP) of the autocorrelation values ​​across all periodic windows measures the degree of fluctuation in periodicity intensity across resistance data within different periodic windows. A smaller COP indicates more stable periodicity; a larger COP indicates weaker periodicity.

[0066] Next, the decomposition effect evaluation value corresponding to the decomposition layer number I is calculated. This value comprehensively considers the average periodicity level of the target resistance timing of each layer, the maximum periodicity level of each layer, and the reconstruction error after decomposition. By traversing the range of values ​​for the decomposition layer number I, the value I' that maximizes the decomposition effect evaluation value is selected as the optimal solution.

[0067] The decomposition effect evaluation value corresponding to the above decomposition level I can be expressed by the following formula:

[0068]

[0069] In the formula, Number of decomposition layers The corresponding decomposition effect evaluation value, This represents the periodic water average value for each periodic window. This represents the maximum periodic level of each periodic window after decomposition. To decompose the target resistance timing of all days into Reconstruction error after layer This represents an exponential function with base e.

[0070] in, The higher the value, the more significant the periodicity of the layer, which is more beneficial for analyzing the periodic behavior of grounding resistance. As a normalization benchmark, it avoids distortion of evaluation values ​​caused by differences in the periodic intensity of the data itself; by... Incorporate evaluation to ensure that evaluation results are comparable across different datasets or decomposition layers; The value measures the accuracy of signal reconstruction after decomposition. The smaller the value, the more completely the original signal is preserved in the decomposition process, avoiding information loss caused by over-decomposition or algorithm error. At the same time, the reconstruction error is penalized by an exponential term (the larger the error, the smaller the exponential term), thus achieving a balance between periodic feature extraction and signal fidelity.

[0071] The number of decomposition layers (I) in DWT directly affects the analysis results. Too few layers may result in insufficient separation of periodic features, while too many layers may introduce redundancy or reconstruction errors.

[0072] By calculating different I values , choose to The maximum corresponding decomposition level is taken as the optimal decomposition level. At this level, the combined effect of periodic feature extraction and reconstruction accuracy is the best, avoiding the blindness of subjectively choosing the number of levels.

[0073] After obtaining the optimal number of decomposition layers, the target resistance time series for all days is decomposed into multiple layers again, and the periodicity level of each periodic window is calculated according to the above calculation process.

[0074] Furthermore, in order to effectively capture the abnormal patterns of the target resistance timing, after decomposition, the target resistance timing of each layer is segmented according to the above S2 clustering operation. The least squares method is used to fit a linear model to obtain the slope of all segments in that layer. The standard deviation of the slope of all segments in each layer is calculated, as well as the difference between the maximum and minimum slope values ​​of all segments in each layer is calculated. The product of the standard deviation of the slope of all segments in each layer and the difference between the maximum and minimum slope values ​​of all segments in each layer is normalized using norm to obtain the degree of abnormality of all period windows in each layer.

[0075] For each segment, a linear model is fitted, and the slope is extracted to quantify the rate of change of resistance within each segment. Anomalies in the slope (such as a sudden increase or decrease) may directly reflect equipment failure or external interference. The degree of anomaly is comprehensively evaluated by combining the standard deviation (dispersion) and the range (variation). For example, if the standard deviation of a certain layer is high and the range is large, the product will significantly amplify the abnormal signal, avoiding errors caused by a single indicator.

[0076] After obtaining the anomaly level of each periodic window, the periodicity level corresponding to each layer is converted into an exponential term using an exponential function as a weight. Then, the anomaly levels of each layer are weighted and summed. Finally, the summation result is divided by the normalized weights to obtain the performance evaluation value of the target resistance time series for all days, which satisfies the following relationship:

[0077]

[0078] In the formula, For the target resistance timing performance evaluation values ​​for all days, For the first The degree of anomaly in the layer periodic window, For the first The periodicity level of the layer periodic window, To determine the optimal number of decomposition layers, This represents an exponential function with base e.

[0079] The above method uses the periodicity of each periodic window to weight the degree of anomaly. Since the exponential function has a monotonically increasing property, when... When smaller, The value is relatively large, meaning that the degree of timing anomaly of the target resistor with poor stability contributes more to the overall evaluation result of the periodic window.

[0080] In another embodiment, an alternative method for calculating the degree of anomaly in the timing of each target resistor layer is provided, as follows:

[0081] After decomposing the target resistance time series of all days based on the optimal decomposition layer, the target resistance time series of each layer is segmented according to the clustering operation. The slope of all segments in each layer is obtained by fitting using the least squares method, and the standard deviation and mean of the slope of all segments in each layer are calculated. The ratio of the standard deviation to the mean is used as the degree of variation of the slope of all segments in each layer.

[0082] Sort all segments of each layer by slope from smallest to largest, obtain the first quartile and the third quartile, and calculate the interquartile range.

[0083] The abnormality of all windows in each layer is obtained by normalizing the product of the degree of variation of all segments in each layer and the interquartile range.

[0084] That is, the relation is satisfied as follows:

[0085]

[0086] In the formula, For the first The degree of anomaly of all windows in the layer. No. The degree of variation in the slope of all segments of the layer. For the first The interquartile range of the slopes of all segments of the layer. This indicates normalization. The greater the degree of variation, the greater the dispersion of the local slope of the segment relative to the mean, the more significant the difference in the resistance change trend, and the higher the probability of anomalies; the larger the interquartile range, the greater the difference in the local slope of the middle segment, and the more unstable the resistance change trend.

[0087] In summary, clustering and decomposition are two complementary steps. Clustering provides a more meaningful data partition for subsequent decomposition, making the decomposition operation more efficient and accurate. Decomposition further refines the data within each periodic window, separating components of different frequencies to better analyze the periodicity and stability of the resistance data. Through these two steps, a more comprehensive understanding of the changing patterns of grounding resistance data can be achieved, improving the accuracy and robustness of the analysis.

[0088] It should be noted that the approximation coefficients, detail coefficients, autocorrelation values, and reconstruction errors involved in the above calculation process are all existing technologies, and will not be elaborated further here.

[0089] S3: When the performance evaluation value is less than or equal to the preset abnormal threshold, it is determined that there is an abnormality in the lightning protection grounding device.

[0090] After obtaining the overall target resistance timing anomaly level in S2 above, collect the lightning protection device's testing data from the past 3-5 years, such as grounding resistance, and calculate the normal operating conditions. Value distribution range, set the abnormal threshold to the upper limit of the normal distribution, such as the 95th percentile, to ensure that the false alarm rate is less than 5%.

[0091] When the above performance evaluation value is less than or equal to the set abnormal threshold, it is determined that there is an abnormality in the lightning protection grounding device, and emergency handling is performed, such as visually inspecting the grounding wire connection, lightning strike counter reading, using a grounding resistance tester to locate high resistance points, and using an insulation resistance meter to test the insulation performance of the line, etc.

[0092] The system includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the method for performance analysis of lightning protection devices in flammable and explosive locations according to the first aspect of the present invention.

[0093] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0094] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for analyzing the performance of lightning protection devices in flammable and explosive environments, characterized in that, include: The resistance, temperature and humidity parameters of the lightning protection grounding device within a preset number of days are collected and preprocessed. The continuous sampling data of a day is divided into multiple sequential sampling intervals at certain intervals. Then, in each sampling interval, the point with the lowest humidity is found and the corresponding grounding resistance is extracted. This operation is repeated for all sampling intervals. Finally, the extracted grounding resistances are arranged in the sampling order to obtain the target resistance time sequence for the day. Based on the diurnal temperature difference, all days within a preset number of days are clustered to obtain multiple clusters. The clustering results are used to divide the target resistance time series of all days into multiple periodic windows. Based on the optimal decomposition layer, the target resistance time series of all days is decomposed. The periodicity level and anomaly degree of all periodic windows in each layer are calculated. The periodicity level is converted into the weight of the corresponding layer. The anomaly degree of each layer is multiplied by the weight of the corresponding layer and weighted and summarized to obtain the performance evaluation value of the target resistance time series of all days. When the performance evaluation value is less than or equal to the preset abnormal threshold, the lightning protection grounding device is determined to be abnormal. The standard deviation of the autocorrelation value corresponding to each layer is divided by the mean of the corresponding autocorrelation value to obtain the coefficient of variation for each layer. The result of subtracting the coefficient of variation from 1 is multiplied by the median of the autocorrelation value of the corresponding layer to obtain the periodicity level of all periodic windows in each layer. The autocorrelation value for each layer is obtained as follows: The autocorrelation function is used to calculate the autocorrelation value of each periodic window in each layer, and the maximum value output by the autocorrelation function is selected as the autocorrelation value of each periodic window.

2. The method for performance analysis of lightning protection devices in flammable and explosive locations according to claim 1, characterized in that, The clustering also includes adjusting the cluster radius according to the variation range of the diurnal temperature range. The adjustment process includes: selecting any day as the target day, calculating the average of the sum of the absolute values ​​of the differences between the target day and the two days before and after it, as the variation range of the temperature range of the target day. The mean of the absolute difference between the diurnal and daytime temperature differences of two consecutive days within all days is calculated, and combined with the normalized temperature difference variation range, the cluster radius required for clustering the target day is obtained.

3. The method for performance analysis of lightning protection devices in flammable and explosive locations according to claim 1, characterized in that, The process of obtaining the optimal number of decomposition layers includes: setting the range of decomposition layers, performing multi-layer decomposition on the target resistance time series for all days, calculating the mean, median and standard deviation of the autocorrelation values ​​of all periodic windows in each layer, and calculating the periodicity level of all periodic windows in each layer based on the mean, median and standard deviation of the autocorrelation values. The decomposition effect evaluation value corresponding to any decomposition layer is calculated based on the periodic average value of each layer's periodic window, the maximum value of the periodic level, and the reconstruction error after decomposition. Traverse the range of decomposition levels and select the level that maximizes the decomposition effect evaluation value as the optimal decomposition level.

4. The method for performance analysis of lightning protection devices in flammable and explosive locations according to claim 1, characterized in that, The process of obtaining the degree of anomaly includes: decomposing the target resistance time series of all days based on the optimal decomposition layer number, segmenting the target resistance time series of each layer according to the clustering operation, using the least squares method to fit the slope of all segments in each layer, calculating the standard deviation of the slope of all segments in each layer, and calculating the difference between the maximum and minimum slope values ​​of all segments in each layer. The product of the standard deviation of the slope of all segments in each layer and the difference between the maximum and minimum slope values ​​of all segments in each layer is normalized using norm to obtain the degree of anomaly of all periodic windows in each layer.

5. The method for performance analysis of lightning protection devices in flammable and explosive locations according to claim 1, characterized in that, The process of obtaining the degree of anomaly includes: decomposing the target resistance time series of all days based on the optimal decomposition layer number, segmenting the target resistance time series of each layer according to the clustering operation, using the least squares method to fit the slope of all segments of each layer, and calculating the standard deviation and mean of the slope of all segments of each layer, and using the ratio of the standard deviation to the mean as the degree of variation of the slope of all segments of each layer. Sort all segments of each layer by slope from smallest to largest, obtain the first quartile and the third quartile, and calculate the interquartile range. The abnormality of all periodic windows in each layer is obtained by normalizing the product of the variation degree of all segments in each layer and the interquartile range.

6. A performance analysis system for lightning protection devices in flammable and explosive environments, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the performance analysis method for lightning protection devices in flammable and explosive locations according to any one of claims 1-5.

Images