Gas station danger early warning method and system based on Internet of Things

Abstract

The invention relates to the technical field of data processing, in particular to a gas station danger early warning method and system based on the Internet of Things. The method comprises the following steps: acquiring gas-liquid ratio data of the oil gas recovery pipeline of the oil gun, wherein the gas-liquid ratio data comprises gas-liquid ratio numerical values of a plurality of samples arranged according to a time sequence; marking any sample as a target sample, and determining the mutation degree of the target sample; determining a fault factor of the target sample; determining the number of adaptive samples used for an isolated forest algorithm when the target sample is detected; and based on the adaptive sample number, performing anomaly detection on a target sample by using an isolated forest algorithm to realize gas station danger early warning based on the Internet of Things. Self-adaptive adjustment of the sample number is achieved by calculating the fault factors, normal gradual change and abnormal sudden change are effectively distinguished, the gas-liquid ratio data anomaly detection false alarm rate is reduced, and the accuracy and reliability of gas station danger early warning are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for early warning of dangers at gas stations based on the Internet of Things. Background Technology

[0002] Gas stations are key urban infrastructure, and their safe operation is of paramount importance. Therefore, they are key safety supervision units. During refueling operations, gas pumps are generally equipped with vapor recovery systems. The vapor recovery system aims to recover the vapors that evaporate during the refueling process in order to reduce environmental pollution and ensure operational safety. A core indicator for measuring the working status of the vapor recovery system is the vapor-liquid ratio, which is the ratio of the volume of recovered vapors to the volume of gasoline dispensed.

[0003] In monitoring gas stations, gas-liquid ratio data is typically collected and analyzed using anomaly detection algorithms. The Isolation Forest algorithm has become a mainstream technique in anomaly detection due to its high computational efficiency and the fact that it does not require pre-labeling of samples. However, directly applying the standard Isolation Forest algorithm to anomaly detection of gas-liquid ratio data has significant technical drawbacks. The standard Isolation Forest algorithm usually sets a fixed number of samples for all test samples when constructing the tree. The number of samples determines the ease with which a test sample is isolated, thus affecting the calculation of its anomaly score.

[0004] The fixed processing method makes it difficult for the standard Isolation Forest algorithm to effectively distinguish between two different situations that may both appear as high values: one is a dangerous spike signal caused by a sudden jump from a stable baseline due to faults such as accidental detachment of the refueling nozzle or momentary blockage of the pipeline; the other is a normal high value generated during a normal refueling process due to a smooth transition of the flow rate from low to high. Since normal high values ​​are relatively rare in the entire dataset, the standard Isolation Forest algorithm will frequently misjudge them as abnormal, generating false alarms, which not only reduces the reliability of hazard warnings but also increases the burden on maintenance personnel. Summary of the Invention

[0005] To address the problem that the standard isolated forest algorithm, due to its fixed sample size, has a high false alarm rate for detecting anomalies in the gas-liquid ratio data of gas stations, thus reducing the reliability of hazard warnings, this invention provides a gas station hazard warning method and system based on the Internet of Things.

[0006] In a first aspect, the present invention provides a gas station hazard warning method based on the Internet of Things, which adopts the following technical solution: A gas station hazard warning method based on the Internet of Things (IoT) includes: acquiring gas-liquid ratio data of the fuel nozzle vapor recovery pipeline, wherein the gas-liquid ratio data includes gas-liquid ratio values ​​of multiple samples arranged in chronological order; designating any sample as a target sample, and determining the degree of mutation of the target sample based on the gas-liquid ratio values ​​of each sample within the data segment to which the target sample belongs; determining a fault factor of the target sample based on the symmetry characteristics of the data segments on both sides of the target sample and the degree of mutation; determining an adaptive number of samples for the Isolation Forest algorithm when detecting the target sample based on the fault factor, wherein the adaptive number of samples is negatively correlated with the fault factor; and using the Isolation Forest algorithm to perform anomaly detection on the target sample based on the adaptive number of samples, thereby realizing a gas station hazard warning based on the IoT.

[0007] This invention constructs an evaluation index for the degree of abrupt change by analyzing the gas-liquid ratio values ​​of each sample within the data segment to which the target sample belongs, effectively identifying abnormal jumps in gas-liquid ratio data. By analyzing the symmetry characteristics of the data segments on both sides of the target sample, a comprehensive evaluation model for fault factors is constructed, effectively distinguishing between normal gradual changes and abnormal abrupt changes. Through the negative correlation between fault factors and adaptive sample quantity, intelligent adjustment of the parameters of the isolated forest algorithm is achieved. When the fault factor is high, it indicates an abnormal abrupt change, and the sample quantity is reduced to improve detection sensitivity; when the fault factor is low, it indicates a normal gradual change, and the sample quantity is increased to avoid false alarms. This effectively solves the problem that the standard isolated forest algorithm struggles to distinguish between normal high values ​​and dangerous spike signals, improving the accuracy and reliability of anomaly detection. Through adaptive sample quantity adjustment, differentiated processing of different types of gas-liquid ratio data is achieved, providing more accurate data support for gas station hazard warnings.

[0008] Furthermore, the gas-liquid ratio value is obtained by performing analog-to-digital conversion on the collected raw data.

[0009] This invention converts the acquired analog gas-liquid ratio signal into a digital value through analog-to-digital conversion, realizing the digital processing of data. This provides a standardized data foundation for subsequent intelligent analysis and algorithm processing, ensuring the accuracy and consistency of the data, and avoiding noise interference and signal attenuation problems during analog signal transmission.

[0010] Furthermore, the data segment to which the target sample belongs is a data segment composed of the target sample and a preset number of samples preceding the target sample.

[0011] Furthermore, the degree of mutation satisfies: In the formula, For the first The degree of mutation in each sample For the first The gas-liquid ratio values ​​for each sample. For the first The mean of all gas-liquid ratio values ​​within the data segment to which each sample belongs. For the first The range of all gas-liquid ratio values ​​within the data segment to which each sample belongs. For the first The data segment to which this sample belongs does not contain the first... The range of the residual gas-liquid ratio values ​​for each sample. For hyperparameters, To normalize the maximum and minimum values, It is the absolute value symbol.

[0012] This invention achieves a comprehensive assessment of the degree of mutation by constructing a composite function containing a deviation term and a range variation term. The deviation term reflects the degree of deviation of the target sample from the average value of the data segment, while the range variation term reflects the contribution of the target sample to the overall range of the data segment. The introduction of hyperparameters ensures the numerical stability of the calculation and avoids calculation anomalies caused by zero values. Normalization eliminates the influence of differences in magnitude between different data segments, making the degree of mutation highly comparable. Through the above methods, normal gradual changes and abnormal mutations are effectively distinguished: during normal gradual changes, although the deviation at a single point may be large, its contribution to the overall range is relatively small, while during abnormal mutations, not only is the deviation at a single point large, but its contribution to the range also increases significantly, improving the accuracy and robustness of mutation detection and providing reliable input parameters for subsequent fault factor calculation.

[0013] Furthermore, the data segments on both sides of the target sample are selected with the target sample as the center, and the same preset length data segments are selected on both sides of the target sample.

[0014] Furthermore, the failure factor satisfies: In the formula, For the first Fault factors for each sample For the first The degree of mutation in each sample In the first The number of samples within a data segment on either side of a given sample. and The first The sample and the first The gas-liquid ratio values ​​for each sample. To normalize the maximum and minimum values, It is a natural exponential function. It is the absolute value symbol.

[0015] This invention constructs a comprehensive evaluation model for fault factors by normalizing the product of the degree of mutation and the symmetry index. This model effectively integrates local mutation features and global symmetry features, enabling accurate identification of abnormal events. The symmetry index effectively measures the degree of symmetry between the two sides of the target sample: when the data on both sides are highly symmetrical, the difference sum approaches 0, and the exponent term approaches 1; when the data on both sides are asymmetrical, the difference sum increases, and the exponent term decreases. This helps to distinguish between normal gradual changes and abnormal mutations, thereby effectively suppressing false alarms under normal operating conditions while maintaining high sensitivity to real faults. Through the negative correlation between the fault factor and the adaptive sample size, the accuracy and reliability of the isolated forest algorithm in gas station hazard warning are improved.

[0016] Furthermore, the adaptive sample number satisfies: In the formula, For use in detecting the first The adaptive number of samples for the Isolation Forest algorithm when there are 1 sample. This is the preset initial number of samples in the Isolation Forest algorithm. For the first Fault factors for each sample It is a function with maximum value. It is a natural exponential function. This is for rounding operations.

[0017] This invention constructs an adaptive sample size calculation model by rounding the product of the initial sample size and the fault factor to the nearest integer and setting a minimum lower limit. This ensures a negative correlation between the sample size and the fault factor, enabling intelligent parameter adjustment. When the fault factor is high, it indicates an abnormal mutation, and the adaptive sample size is reduced accordingly, improving the sensitivity of the Isolation Forest algorithm to detect abnormal samples and avoiding false negatives. When the fault factor is low, it indicates a normal gradual change, and the adaptive sample size is increased, reducing the false alarm rate and improving detection accuracy. The minimum constraint ensures that the sample size is not too small, which would affect the statistical reliability of the algorithm. The rounding operation ensures that the sample size is an integer, meeting the algorithm's implementation requirements. Through the above mechanism, the contradiction between detection sensitivity and accuracy is effectively balanced, improving the applicability and reliability of the Isolation Forest algorithm in the analysis of gas-liquid ratio data at gas stations, and providing more accurate parameter adaptation capabilities for hazard warning systems.

[0018] Furthermore, the initial number of samples in the isolated forest algorithm is preset to 256.

[0019] Furthermore, the anomaly detection threshold in the preset parameters of the isolated forest algorithm is preset to 0.75, and the number of trees is preset to 50.

[0020] Secondly, the present invention provides a gas station hazard warning system based on the Internet of Things, which adopts the following technical solution: A gas station hazard warning system based on the Internet of Things (IoT) includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the aforementioned IoT-based gas station hazard warning method.

[0021] By adopting the above technical solution, a computer program for the above-mentioned Internet of Things-based gas station hazard warning method is generated and stored in a memory so that it can be loaded and executed by a processor. In this way, a terminal device can be made based on the memory and the processor for convenient use.

[0022] The present invention has the following technical effects: (1) Breaking through the limitation of the fixed number of samples in the standard isolated forest algorithm, the fault factor is constructed by characterizing the gas-liquid ratio value from the intensity of the jump in the baseline and the symmetry of the data segments on both sides. The adaptive number of samples in the isolated forest is dynamically adjusted. When the fault factor is high, the number of samples is reduced, making it easier for the isolated forest algorithm to identify it as an anomaly. When the fault factor is low, the number of samples is increased to avoid it being misjudged as an anomaly due to the scarcity of values. This solves the problem of the traditional algorithm mis-distinguishing between dangerous peaks and normal high values ​​and reduces the false alarm rate.

[0023] (2) The negative correlation mechanism between the adaptive sample quantity and the fault factor makes the isolation difficulty of the isolated forest algorithm accurately match the actual danger level of the sample. For real fault signals, such as instantaneous spikes caused by pipeline blockage, the small number of samples corresponding to high fault factors will amplify their abnormal features and ensure stable identification. For high-value fluctuations in normal operation, such as the smooth transition of the refueling flow rate increase, the large number of samples corresponding to low fault factors will weaken their rarity effect and avoid being incorrectly labeled, thereby improving the accuracy of anomaly detection, making the danger warning more reliable, and providing key support for gas stations to deal with real risks in a timely manner.

[0024] (3) The calculation of the fault factor integrates the degree of mutation and symmetry characteristics, which can adapt to the diverse operation scenarios and equipment status of gas stations. Compared with the standard isolated forest algorithm, which only relies on the judgment logic of numerical rarity, the present invention is more adaptable to complex working conditions. When the gas-liquid ratio data is affected by factors such as ambient temperature and oil type, it can still maintain stable early warning performance and ensure reliable application in various scenarios. Attached Figure Description

[0025] Figure 1 This is a flowchart of a gas station hazard warning method based on the Internet of Things according to an embodiment of the present invention.

[0026] Figure 2This is a line graph illustrating the gas-liquid ratio hazard warning results of a gas station in an IoT-based gas station hazard warning method according to an embodiment of the present invention. Detailed Implementation

[0027] 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, not all, of the embodiments of the present invention. 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.

[0028] This invention discloses a method for early warning of hazards at gas stations based on the Internet of Things (IoT), referring to... Figure 1 This includes steps S01-S05: S01: Obtain the gas-liquid ratio data of the fuel nozzle vapor recovery pipeline, wherein the gas-liquid ratio data includes the gas-liquid ratio values ​​of multiple samples arranged in chronological order.

[0029] In this embodiment, the gas-liquid ratio is collected in real time by a gas-liquid ratio sensor deployed on the gas pump vapor recovery pipeline in the gas station. For example, the collection frequency can be set to 10 times per second to ensure that instantaneous abnormal fluctuations can be captured. The analog signal collected by the sensor is converted into a digital gas-liquid ratio value by an analog-to-digital converter.

[0030] S02: Designate any sample as the target sample and determine the degree of mutation of the target sample.

[0031] It should be noted that in order to assess the degree of change of each sample compared to its surrounding times, this step will analyze the performance of the gas-liquid ratio. In the specific analysis, the more extreme the value of the gas-liquid ratio at each sample is in its data segment, the greater the degree of change, and the greater the possibility that the refueling nozzle malfunctioned at the time corresponding to that sample.

[0032] Any sample is designated as the target sample. The degree of mutation of the target sample is determined based on the gas-liquid ratio of each sample in the data segment to which the target sample belongs.

[0033] The implementers can set the length of the data segment according to the specific implementation situation, for example, 200. It should be noted that, since historical data needs to be analyzed, the first 199 samples collected will be used for the construction of the data segment and will not be used for calculations in subsequent steps.

[0034] Specifically, the data segment to which the target sample belongs is a data segment consisting of the target sample and a preset number of samples preceding the target sample.

[0035] Specifically, the degree of mutation satisfies: ; In the formula, For the first The degree of mutation in each sample For the first The gas-liquid ratio values ​​for each sample. For the first The mean of all gas-liquid ratio values ​​within the data segment to which each sample belongs. For the first The range of all gas-liquid ratio values ​​within the data segment to which each sample belongs. For the first The data segment to which this sample belongs does not contain the first... The range of the residual gas-liquid ratio values ​​for each sample. For hyperparameters, To normalize the maximum and minimum values, It is the absolute value symbol.

[0036] Implementers can set hyperparameters according to the specific implementation situation, such as 0.001. The existence of hyperparameters is to prevent... or This can lead to situations where the calculation results become meaningless.

[0037] in, The larger the value, the more likely it is to be the first. The more extreme the gas-liquid ratio value of a sample is within its data range, the greater the degree of abrupt change. The more samples there are, the greater the likelihood that the fuel nozzle will malfunction at a given time; conversely, the less samples there are, the less likely the fuel nozzle will malfunction at a given time. The larger it is, the more likely it is to be the first The more extreme the gas-liquid ratio value of a sample is within its corresponding data range, the greater its reliability, indicating a greater degree of abrupt change. The more samples there are, the greater the likelihood that the fuel nozzle will malfunction at a given time; conversely, the less samples there are, the less likely the fuel nozzle will malfunction at a given time.

[0038] S03: Determine the failure factors of the target sample.

[0039] It should be noted that the difference between instantaneous spike signals caused by equipment failure and high values ​​in normal flow rate changes lies in the following: When an instantaneous spike signal caused by equipment failure occurs, the values ​​of the data segments on both sides of the sampling time will show a more obvious symmetrical characteristic, while when a high value occurs in normal flow rate changes, the value changes of the data segments on both sides of the sampling time will show a unidirectional upward or downward characteristic. Therefore, in this step, the analysis of the numerical change characteristics of the gas-liquid ratio data, combined with the degree of change of each sample, yields the fault factor for each sample. The logic behind this is: the more obvious the symmetrical characteristic of the values ​​of the data segments on both sides of a sample, the greater the possibility that the cause of the sample's value protrusion is equipment failure rather than normal flow rate change, the greater the possibility that the refueling nozzle malfunctioned at the corresponding moment, and the larger the fault factor of that sample will be.

[0040] The failure factor of the target sample is determined based on the symmetry of the data segments on both sides of the target sample and the degree of mutation.

[0041] Specifically, the data segments on both sides of the target sample are selected with the target sample as the center, and the same preset length data segments are selected on both sides of the target sample.

[0042] Implementers can set the length of the data segment according to the specific implementation situation, for example, 50.

[0043] Specifically, the failure factor satisfies: ; In the formula, For the first Fault factors for each sample For the first The degree of mutation in each sample In the first The number of samples within a data segment on either side of a given sample. and The first The sample and the first The gas-liquid ratio values ​​for each sample. To normalize the maximum and minimum values, It is a natural exponential function. It is the absolute value symbol.

[0044] in, The larger the value, the better the performance analysis based on the gas-liquid ratio. The more prominent the value of a sample, the greater the possibility of an anomaly, and the greater the corresponding failure factor; conversely, the less prominent the value, the greater the possibility of an anomaly. The smaller the value, the better. The more pronounced the symmetry in the values ​​of the data segments on both sides of a sample, the better the symmetry. If a sample exhibits a significant numerical abnormality, the greater the likelihood that the cause is equipment malfunction rather than normal flow rate variation, indicating that the sample... The higher the probability of the fuel nozzle malfunctioning at the corresponding time point for each sample, the higher the probability of the fuel nozzle malfunctioning at the corresponding time point. The larger the failure factor of a sample, the greater the failure factor; conversely, the smaller the sample, the greater the failure factor.

[0045] S04: Determine the adaptive number of samples to use for the Isolation Forest algorithm when detecting target samples.

[0046] It should be noted that the logic used in this step to adaptively adjust the number of subsamples for each sample based on the fault factor of each sample is as follows: the larger the fault factor of each sample, the greater the probability that the sample will fail at the corresponding time. Therefore, in order to more accurately detect it as an abnormal sample, a smaller adaptive sample number needs to be set to highlight the abnormality.

[0047] Based on the fault factor, an adaptive number of samples is determined for the Isolation Forest algorithm when detecting target samples, wherein the adaptive number of samples is negatively correlated with the fault factor.

[0048] Specifically, the adaptive sample number satisfies: ; In the formula, For use in detecting the first The adaptive number of samples for the Isolation Forest algorithm when there are 1 sample. This is the preset initial number of samples in the Isolation Forest algorithm. For the first Fault factors for each sample It is a function with maximum value. It is a natural exponential function. This is for rounding operations.

[0049] Implementers can set the initial number of samples in the Isolation Forest algorithm according to the specific implementation situation, for example, 256.

[0050] S05: Based on the adaptive sample size, the isolated forest algorithm is used to detect anomalies in the target samples in order to realize IoT-based gas station hazard warning.

[0051] Implementers can set the anomaly detection threshold and the number of trees according to the specific implementation situation. For example, the anomaly detection threshold is 0.75 and the number of trees is 50.

[0052] After obtaining the adaptive number of samples to detect each sample and the preset parameters for the isolated forest algorithm, the isolated forest algorithm is used to detect anomalies in the gas-liquid ratio data. The subsequent operations of the isolated forest algorithm are existing well-known steps and will not be described in detail here.

[0053] like Figure 2 As shown, the vertical axis represents the gas-liquid ratio (A / L Ratio), with a value range of 1.0-4.0, and the horizontal axis represents the timestamp (covering the period from 01:00:00 to 01:00:10). Figure 2 The invention clearly distinguishes two types of gas-liquid ratio data characteristics: the first is the normal high-value area, in which the gas-liquid ratio value shows a smooth and gradual trend over time, without any instantaneous sudden rises or falls. This corresponds to the normal operating condition of the flow rate transitioning from low to high during refueling. Due to the low fault factor, this invention avoids being misjudged as abnormal by increasing the number of adaptive samples in the isolated forest algorithm. The second is the detected dangerous peak, which is manifested as an instantaneous jump in the gas-liquid ratio value at a specific time point. This type of signal has a high degree of abrupt change and obvious symmetry between the data segments on both sides, resulting in a high fault factor. The isolated forest algorithm amplifies its abnormal characteristics by reducing the number of adaptive samples, thereby achieving accurate identification, reducing the false alarm rate, and improving the reliability of hazard warnings.

[0054] This invention also discloses an Internet of Things (IoT)-based gas station hazard warning system, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement an IoT-based gas station hazard warning method according to the present invention.

[0055] 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.

[0056] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for early warning of hazards at gas stations based on the Internet of Things, characterized in that, include: Obtain the gas-liquid ratio data of the fuel nozzle vapor recovery pipeline, wherein the gas-liquid ratio data includes the gas-liquid ratio values ​​of multiple samples arranged in chronological order; Any sample is designated as the target sample. The degree of mutation of the target sample is determined based on the gas-liquid ratio of each sample in the data segment to which the target sample belongs. Based on the symmetry of the data segments on both sides of the target sample and the degree of mutation, the failure factor of the target sample is determined. Based on the failure factor, an adaptive number of samples is determined for the Isolation Forest algorithm when detecting target samples, wherein the adaptive number of samples is negatively correlated with the failure factor; Based on the adaptive sample size, the isolated forest algorithm is used to detect anomalies in the target samples, so as to realize the Internet of Things-based gas station hazard warning.

2. The method for early warning of hazards at gas stations based on the Internet of Things according to claim 1, characterized in that, The gas-liquid ratio value is obtained by performing analog-to-digital conversion on the collected raw data.

3. The method for early warning of hazards at gas stations based on the Internet of Things according to claim 1, characterized in that, The data segment to which the target sample belongs is a data segment consisting of the target sample and a predetermined number of samples preceding the target sample.

4. The method for early warning of hazards at gas stations based on the Internet of Things according to claim 1, characterized in that, The degree of mutation satisfies: ； In the formula, For the first The degree of mutation in each sample For the first The gas-liquid ratio values ​​for each sample. For the first The mean of all gas-liquid ratio values ​​within the data segment to which each sample belongs. For the first The range of all gas-liquid ratio values ​​within the data segment to which each sample belongs. For the first The data segment to which this sample belongs does not contain the first... The range of the residual gas-liquid ratio values ​​for each sample. For hyperparameters, To normalize the maximum and minimum values, It is the absolute value symbol.

5. The method for early warning of hazards at gas stations based on the Internet of Things according to claim 1, characterized in that, The data segments on both sides of the target sample are selected with the same preset length on both sides of the target sample as the center.

6. The method for early warning of hazards at gas stations based on the Internet of Things according to claim 1, characterized in that, The failure factor satisfies: ； In the formula, For the first Fault factors for each sample For the first The degree of mutation in each sample In the first The number of samples within a data segment on either side of a given sample. and The first The sample and the first The gas-liquid ratio values ​​for each sample. To normalize the maximum and minimum values, It is a natural exponential function. It is the absolute value symbol.

7. The method for early warning of hazards at gas stations based on the Internet of Things according to claim 1, characterized in that, The adaptive sample size satisfies: ； In the formula, For use in detecting the first The adaptive number of samples for the Isolation Forest algorithm when there are 1 sample. This is the preset initial number of samples in the Isolation Forest algorithm. For the first Fault factors for each sample It is a function with maximum value. It is a natural exponential function. This is for rounding operations.

8. A gas station hazard warning method based on the Internet of Things according to claim 7, characterized in that, The initial sample size in the isolated forest algorithm is preset to 256.

9. A gas station hazard warning method based on the Internet of Things according to claim 1, characterized in that, The preset parameters of the isolated forest algorithm include an anomaly detection threshold of 0.75 and a preset number of trees of 50.

10. A gas station hazard warning system based on the Internet of Things, characterized in that, include: A processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement an Internet of Things-based hazard warning method for gas stations according to any one of claims 1-9.

Images