A method and system for safety evaluation of a natural gas filling station

Abstract

The application discloses a kind of natural gas filling station safety evaluation method and system, it is related to filling station safety evaluation technical field.The natural gas filling station safety evaluation method, by obtaining the time sequence state data of natural gas filling station's gas filling equipment and gas filling environment, and pretreatment is carried out;Respectively based on pre-trained equipment feature extraction model and environment feature extraction model carries out feature analysis, obtains gas filling equipment and gas filling environment time sequence feature set;And this respectively analyzes gas filling equipment and gas filling environment safety index;And respectively with the preset several groups of gas filling safety interval set carries out judgment analysis, the application is constructed equipment feature extraction model and environment feature extraction model, the time sequence state data of gas filling equipment and gas filling environment is analyzed in depth, to extract multiple important features related to safety, compared with the practice of traditional only relying on artificial patrol or fixed threshold alarm mechanism, the method can identify more subtle, earlier abnormal signal in equipment or environment state.

Description

Technical Field

[0001] The present invention relates to the technical field of gas filling station safety evaluation, and in particular to a natural gas filling station safety evaluation method and system. Background Art

[0002] With the acceleration of urbanization and the growing popularity of clean energy, natural gas, as a key energy source, has gained widespread application in various sectors, including automotive, industrial, and household applications. Natural gas filling stations, as a crucial link in the natural gas supply chain, are responsible for transporting natural gas from the pipeline network to vehicles or other consumers. The safety of these stations is crucial, directly impacting personal safety, environmental protection, and the proper functioning of equipment.

[0003] Natural gas filling stations present numerous safety risks during their daily operations. Firstly, flammable gas leaks are one of the most common and dangerous risks. Gas leaks can not only easily cause fires or explosions, but because natural gas is colorless and odorless, leaks are often difficult to detect in a timely manner, posing a serious risk to the station's operations. Secondly, equipment failure is also a significant source of safety hazards. Abnormal conditions such as excessive temperatures, excessive current, and mechanical vibration in filling equipment can easily cause equipment failure, compromising the station's safe operation. Furthermore, environmental factors such as climate conditions, gas flow rates, and equipment stability can also impact the station's safety.

[0004] The limitations of existing technologies include at least the following problems. When conducting safety assessments on natural gas filling stations, existing technologies usually rely on static monitoring point data or manual inspection methods, and can only make rough judgments on the status of filling equipment or the environment at a specific time point. They lack the ability to systematically model and evaluate the status change trends during operation, making it difficult to achieve a comprehensive, continuous and quantitative safety evaluation of the operating status of the gas station. Especially in the face of complex situations such as the coordinated changes of multiple parameters of equipment and dynamic interference of environmental factors, traditional methods find it difficult to judge whether the gas filling system is in a safe operating range based on real-time or time-series data, which can easily lead to potential risks not being identified in a timely manner, posing a major safety hazard. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the present invention provides a natural gas filling station safety evaluation method and system, which solves the problem that the existing technology is difficult to achieve dynamic quantitative safety evaluation of the operating status of the gas filling station.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a natural gas filling station safety evaluation method, comprising the following steps: obtaining the time series status data of the gas filling equipment and gas filling environment of the natural gas filling station and preprocessing it; performing feature analysis on the preprocessed time series status data of the gas filling equipment and gas filling environment based on the pre-trained equipment feature extraction model and environment feature extraction model, respectively, to obtain the time series feature set of the gas filling equipment and gas filling environment of the natural gas filling station; based on the time series feature set of the gas filling equipment and gas filling environment, respectively analyzing the safety index of the gas filling equipment and gas filling environment of the natural gas filling station; respectively comparing the safety index of the gas filling equipment and gas filling environment of the natural gas filling station with several preset groups of gas filling safety interval sets for judgment analysis, each group of gas filling safety interval sets respectively including the gas filling equipment safety interval and the gas filling environment safety interval, and each group of gas filling safety interval sets respectively corresponding to a safety level.

[0007] Furthermore, the time series status data of the gas filling equipment includes the gas pressure value of the gas filling pipeline, the gas flow rate value of the gas filling pipeline, the surface temperature value of the gas filling equipment, the vibration amplitude of the gas filling equipment structure, the motor operating current value and the opening and closing status value of the gas filling valve at several time points.

[0008] Furthermore, the equipment feature extraction model includes an input layer, a convolutional layer, an LSTM layer, a feature extraction layer, and an output layer. The time series feature set of the gas filling equipment includes the maximum drop rate of gas pressure in the gas filling pipeline, the average flow rate of gas in the gas filling pipeline, the peak value of the surface temperature of the gas filling equipment, the rate of rise of the surface temperature of the gas filling equipment, the peak value of the structural vibration of the gas filling equipment, the standard deviation of the motor operating current, and the opening and closing frequency of the gas filling valve.

[0009] Furthermore, the specific steps for obtaining the time series feature set of the gas filling equipment of the natural gas filling station are as follows: in the input layer of the equipment feature extraction model, the time series status data of the gas filling equipment is received and processed to obtain the initial feature vector of the gas filling equipment; in the convolution layer of the equipment feature extraction model, the input initial feature vector of the gas filling equipment is convolved to extract local features to obtain the convolution feature map of the gas filling equipment; in the LSTM layer of the equipment feature extraction model, the convolution feature map of the gas filling equipment output by the convolution layer is time series modeled to obtain the long-term dependency features of the gas filling equipment; in the feature extraction layer of the equipment feature extraction model, the long-term dependency features of the gas filling equipment output by the LSTM layer are statistically analyzed to obtain the time series feature set of the gas filling equipment; in the output layer of the equipment feature extraction model, the time series feature set of the gas filling equipment is output.

[0010] Furthermore, the specific steps for analyzing the safety index of the gas filling equipment of the natural gas filling station are as follows: read the maximum drop rate of gas pressure in the gas pipeline of the natural gas filling station, the average flow rate of gas in the gas pipeline, the peak temperature of the surface temperature of the gas filling equipment, the rate of increase of the surface temperature of the gas filling equipment, the peak value of the vibration of the structure of the gas filling equipment, and the standard deviation of the motor operating current, and perform standardization processing; the maximum drop rate of gas pressure in the gas pipeline of the natural gas filling station, the average flow rate of gas in the gas pipeline, the peak temperature of the surface temperature of the gas filling equipment, the rate of increase of the surface temperature of the gas filling equipment, the peak value of the vibration of the structure of the gas filling equipment, and the standard deviation of the motor operating current are combined with the corresponding opening and closing frequency of the gas valve for comprehensive analysis to obtain the safety index of the gas filling equipment of the natural gas filling station.

[0011] Furthermore, the time series state data of the aeration environment includes combustible gas concentration values, ambient temperature values, foundation structure micro-vibration amplitude values, electromagnetic interference intensity values, thermal infrared radiation temperature values ​​and air flow velocity values ​​at several time points.

[0012] Furthermore, the environmental feature extraction model includes an input layer, a convolutional layer, an RNN layer, a pooling layer, a feature extraction layer, an output layer, and a gasification environment time series feature set including the combustible gas concentration peak, the duration of combustible gas concentration exceeding the standard, the average ambient temperature value, the foundation structure micro-vibration peak, the electromagnetic interference intensity peak, the thermal infrared radiation temperature peak, the thermal infrared radiation temperature rising slope, and the average air flow speed value.

[0013] Furthermore, the specific steps for obtaining the time series feature set of the gas filling environment of the natural gas filling station are as follows: in the input layer of the environmental feature extraction model, the time series state data of the gas filling environment is received and processed to obtain the initial feature vector of the gas filling environment; in the convolution layer of the environmental feature extraction model, the input initial feature vector of the gas filling environment is convolved to extract local features to obtain the convolution feature map of the gas filling environment; in the RNN layer of the environmental feature extraction model, the convolution feature map of the gas filling environment output by the convolution layer is time series modeled to obtain the long-term dependence features of the gas filling environment; in the pooling layer of the environmental feature extraction model, the long-term dependence features of the gas filling environment output by the RNN layer are pooled to obtain the pooled feature set of the gas filling environment; in the feature extraction layer of the environmental feature extraction model, the pooled feature set of the gas filling environment output by the pooling layer is statistically analyzed to obtain the time series feature set of the gas filling environment; in the output layer of the environmental feature extraction model, the time series feature set of the gas filling environment is output.

[0014] Furthermore, the specific steps for analyzing the gas filling environment safety index of the natural gas filling station are as follows: read the gas filling environment time series feature set of the natural gas filling station and perform standardization processing; conduct a comprehensive analysis on the standardized gas filling environment time series feature set to obtain the gas filling environment safety index of the natural gas filling station.

[0015] The natural gas filling station safety evaluation system comprises: a filling data acquisition unit, which is used for acquiring time sequence state data of filling equipment and filling environment of a natural gas filling station and performing preprocessing; a filling data extraction unit, which is used for performing feature analysis on the preprocessed time sequence state data of the filling equipment and the filling environment based on a pre-trained equipment feature extraction model and an environment feature extraction model respectively, and obtaining a time sequence feature set of the filling equipment and the filling environment of the natural gas filling station; a filling data analysis unit, which is used for analyzing filling equipment and filling environment safety indexes of the natural gas filling station based on the time sequence feature set of the filling equipment and the filling environment; and a filling safety evaluation unit, which is used for judging and analyzing the filling equipment and filling environment safety indexes of the natural gas filling station with a plurality of preset filling safety interval sets respectively, wherein each filling safety interval set comprises a filling equipment safety interval and a filling environment safety interval, and each filling safety interval set corresponds to a safety level.

[0016] The application has the following beneficial effects:

[0017] (1) The natural gas filling station safety evaluation method can automatically extract a plurality of important safety-related features, such as maximum pressure drop rate, temperature peak value, concentration overproof duration, vibration peak value, etc., by constructing the equipment feature extraction model and the environment feature extraction model to perform deep analysis on the time sequence state data of the filling equipment and the filling environment. Compared with the traditional method of relying on manual inspection or fixed threshold alarm mechanism, the method can identify more subtle and earlier abnormal signals in the equipment or environment state, and is especially suitable for working conditions with short-time mutation or persistent abnormal trend. Since the feature extraction process is completely based on historical data training, the method has good adaptability and generalizability, so that the key features can be stably extracted even in a complex operation background, thereby significantly improving the accuracy and timeliness of the overall safety judgment.

[0018] (2) After the equipment and environment features are extracted, the filling equipment safety index and the filling environment safety index are constructed, and are compared and analyzed with a plurality of preset safety interval sets, so as to realize automatic grading judgment of the current safety state. Each safety interval set corresponds to a different level of safety situation, so that clear risk warning signals and response suggestions can be provided for the on-site management personnel. Unlike the traditional method of relying on manual experience judgment or fuzzy scoring, the method can help to avoid subjective intervention and the emergence of management blind spots, greatly improving the efficiency and standardization of safety evaluation, and establishing a replicable and generalizable evaluation standard for the safety operation of the filling station.

[0019] (3) The natural gas filling station safety evaluation system divides the entire evaluation process into four functional units: data acquisition, feature extraction, safety index analysis, and level determination. Each unit corresponds to an independent functional goal and calculation task. The system structure is clear and the logical division of labor is clear. This modular design not only improves the maintainability of the system, but also facilitates the independent development, testing, and upgrading of each functional unit, reducing the difficulty of system deployment. Especially in multi-site, cross-regional gas filling networks, it can achieve rapid migration and replication deployment, adapting to the data sources and operating environments of different sites. At the same time, the system structure is also conducive to docking with the existing gas station automation control system or monitoring platform, and integration can be achieved without major modifications to the underlying architecture, thereby enhancing the engineering practicality and scalability of the solution.

[0020] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a flow chart of a natural gas filling station safety evaluation method according to the present invention.

[0022] Figure 2 This is a flowchart of the specific steps for obtaining a time series feature set of gas filling equipment at a natural gas filling station in a natural gas filling station safety evaluation method of the present invention.

[0023] Figure 3 This is a block diagram of a natural gas filling station safety evaluation system of the present invention. DETAILED DESCRIPTION

[0024] See also Figure 1 , an embodiment of the present invention provides a technical solution: a natural gas filling station safety evaluation method, comprising the following steps: obtaining and preprocessing the time series state data of the gas filling equipment and the gas filling environment of the natural gas filling station; performing feature analysis on the preprocessed time series state data of the gas filling equipment and the gas filling environment based on a pre-trained equipment feature extraction model and an environment feature extraction model, respectively, to obtain a time series feature set of the gas filling equipment and the gas filling environment of the natural gas filling station; analyzing the safety index of the gas filling equipment and the gas filling environment of the natural gas filling station based on the time series feature set of the gas filling equipment and the gas filling environment; and performing judgment analysis on the safety index of the gas filling equipment and the gas filling environment of the natural gas filling station and a plurality of preset groups of gas filling safety interval sets, each group of gas filling safety interval sets including a gas filling equipment safety interval and a gas filling environment safety interval, and each group of gas filling safety interval sets corresponding to a safety level, including but not limited to the following examples:

[0025] Gas filling safety zone set 1 (low risk, high stability):

[0026] Aeration equipment safety range: 0.000–2.500 (indicates that the equipment is operating well and parameter fluctuations are minimal);

[0027] Gas filling environment safety range: 0.000–2.500 (indicates that the environment is safe and there is no obvious external disturbance);

[0028] Gas filling safety zone set 2 (medium-low risk, medium stability):

[0029] Gas filling equipment safety range: 2.501–3.000 (indicates that the equipment has certain load fluctuations and monitoring is recommended);

[0030] Gas filling environment safety range: 2.501–3.000 (indicates that the environmental impact begins to increase and is in the middle warning range);

[0031] Gas filling safety zone set 3 (medium risk, low stability):

[0032] Gas filling equipment safety range: 3.001–3.500 (indicating interference or unstable fluctuations in equipment operation);

[0033] Aeration environment safety range: 3.001–3.500 (indicates an increase in the frequency of abnormal environmental parameters, affecting equipment safety);

[0034] Gas filling safety zone set 4 (high risk, high interference):

[0035] Gas filling equipment safety range: 3.501–4.000 (indicates that the equipment is obviously abnormal and some parameters have exceeded the standard);

[0036] Aeration environment safety range: 3.501–4.000 (indicates that the external environment interference is serious and may trigger a system chain reaction);

[0037] Gas filling safety zone set 5 (serious risk, extreme instability):

[0038] Gas filling equipment safety range: 4.001 and above (indicates that the equipment is in a serious unstable state and needs to be stopped immediately);

[0039] Gas filling environment safety range: 4.001 and above (indicates that the environmental impact is extremely strong and there is a possibility of triggering safety accidents).

[0040] When the safety index of the gas filling equipment and the gas filling environment is within the preset gas filling safety range set, the natural gas filling station is marked as safe and the relevant staff are notified.

[0041] Specifically, the time series status data of the gas filling equipment includes the gas pressure value of the gas filling pipeline, the gas flow rate value of the gas filling pipeline, the surface temperature value of the gas filling equipment, the vibration amplitude of the gas filling equipment structure, the motor operating current value and the opening and closing status value (0 / 1) of the gas filling valve at several time points.

[0042] The equipment feature extraction model includes an input layer, a convolutional layer, an LSTM layer, a feature extraction layer, and an output layer. The time series feature set of the gas filling equipment includes the maximum drop rate of the gas pressure in the gas filling pipeline, the average flow rate of the gas in the gas filling pipeline, the peak temperature of the surface of the gas filling equipment, the rate of increase of the surface temperature of the gas filling equipment, the peak vibration of the structure of the gas filling equipment, the standard deviation of the motor operating current, and the opening and closing frequency of the gas filling valve.

[0043] The pre-training steps for the device feature extraction model are as follows:

[0044] First, a training dataset is constructed. The training dataset is generated based on multiple historical gas filling equipment operation data. The historical equipment data includes the gas pressure value of the gas filling pipeline, the gas flow rate value of the gas filling pipeline, the surface temperature value of the gas filling equipment, the vibration amplitude of the gas filling equipment structure, the motor operating current value, and the opening and closing status value of the gas filling valve (0 / 1). The historical data and the target gas filling equipment are similar in terms of operating status, equipment type, working environment, etc. Specifically, for each historical gas filling equipment, the equipment status data at multiple time points are first obtained, and the actual operating characteristics of these data points are recorded. Afterwards, a time series feature set is constructed based on the above data. Combined with the anomaly annotations in the historical data (such as equipment failure, equipment overtemperature, etc.), each data point is marked whether there is an abnormal state, as well as the specific type of anomaly (such as temperature anomaly, pressure change anomaly, etc.), thereby forming a complete input and output training sample set.

[0045] The equipment feature extraction model adopted in this embodiment is a deep neural network structure, which mainly includes an input layer, a convolution layer, an LSTM layer, a feature extraction layer and an output layer. Among them, the input layer is used to receive the time series state feature vector of the gas filling equipment corresponding to each time point; the convolution layer is used to extract local features from the input feature vector, the LSTM layer is used to perform time series modeling on the local features output by the convolution layer, the feature extraction layer is used to extract key safety features from the output of the LSTM layer, and the output layer is used to generate the final gas filling equipment time series feature set. Supervised learning is adopted in the training process. The input is the time series feature vector of the gas filling equipment, and the output is the corresponding gas filling equipment time series feature (such as the maximum pressure drop rate, flow rate, temperature peak, etc.). The loss function is optimized by the mean square error (MSE) or other loss functions suitable for regression tasks.

[0046] During the model training process, various optimization strategies are introduced, including batch normalization to stabilize the distribution of intermediate layers, the dropout mechanism to prevent overfitting, and adaptive learning rate adjustment strategies (such as cosine annealing or step decay) to enhance the robustness of the training process. In addition, an early stopping mechanism is introduced to terminate the training process early when the accuracy of the validation set no longer improves in several consecutive iterations to avoid overfitting. After training is completed, the model performance is evaluated based on the validation set, and the set of network parameters with the best performance is selected as the final pre-training model parameters. The trained device feature extraction model can be directly used for time series data analysis of new gas filling equipment, and can output the device's time series feature set for each device data point.

[0047] like Figure 2 As shown in Figure 2, the specific steps for obtaining the time series feature set of the gas filling equipment at the natural gas filling station are as follows:

[0048] In the input layer of the equipment feature extraction model, the time series status data of the gas filling equipment is received and processed to obtain the initial feature vector of the gas filling equipment. Specifically, the time series status data of the gas filling equipment is received, including the gas pressure value of the gas filling pipeline, the gas flow rate value of the gas filling pipeline, the surface temperature value of the gas filling equipment, the vibration amplitude of the gas filling equipment structure, the motor operating current value, and the opening and closing status value (0 / 1) of the gas filling valve, and is converted into the initial feature vector as the input for subsequent convolutional layer and LSTM layer processing;

[0049] In the convolution layer of the equipment feature extraction model, the input initial feature vector of the gas filling equipment is convolved to extract local features and obtain the convolution feature map of the gas filling equipment. Specifically, the input initial feature vector of the gas filling equipment is processed by a 1D convolution operation. The convolution layer uses the convolution kernel to perform weighted summation on the input feature vector at each time step, thereby extracting local time series features, such as sudden changes in pressure, flow rate fluctuations, temperature changes, etc. The size of the convolution kernel (usually 3 or 5) determines the time step range involved in each convolution operation. Each convolution operation extracts a local feature, which can effectively capture the mutation pattern or short-term changes in the time series data. The output after convolution processing is the convolution feature map of the gas filling equipment, which contains the local pattern features in the input data.

[0050] In the LSTM layer of the equipment feature extraction model, the convolutional feature map of the gas filling equipment output by the convolutional layer is subjected to time series modeling to obtain the long-term dependency features of the gas filling equipment. Specifically, the convolutional feature map of the gas filling equipment output by the convolutional layer is used as input for time series modeling. The LSTM layer (long short-term memory network) uses its special gating mechanism (input gate, forget gate, output gate) to perform time series processing on the input data to capture the long-term dependencies in the data. This layer can learn long-term trends in equipment operation from time series data, such as gradual temperature increases or periodic changes in pressure fluctuations. LSTM uses the memory and forgetting mechanisms to preserve long-term dependency information between time steps and output the long-term dependency features of the gas filling equipment.

[0051] In the feature extraction layer of the equipment feature extraction model, the long-term dependency features of the gas filling equipment output by the LSTM layer are statistically analyzed to obtain the time series feature set of the gas filling equipment. Specifically, the long-term dependency features of the gas filling equipment output by the LSTM layer are further statistically analyzed to extract key information and generate the time series feature set of the gas filling equipment. Specifically, the following features are extracted from the time series data:

[0052] Maximum rate of decrease of gas pressure in the gas filling pipeline: The maximum rate of pressure change is obtained by calculating the difference of the pressure values.

[0053] Average flow rate of gas in the gas pipeline: average the flow rate data and calculate the average flow rate of the equipment during the time period.

[0054] Gas filling equipment surface temperature peak: extract the maximum value in the temperature data to reflect whether the equipment is overheating.

[0055] Aeration equipment surface temperature rise rate: Calculate the slope of temperature change and identify the rate of temperature rise.

[0056] Aeration equipment structural vibration peak: Extract the maximum value in the vibration data to determine whether the equipment has a mechanical failure.

[0057] Motor operating current standard deviation: Calculates the standard deviation of current data to identify fluctuations in motor load.

[0058] Gas valve opening and closing frequency: Count the frequency of valve opening and closing to reflect the operating stability of the system.

[0059] The extraction of these features ensures a comprehensive assessment of the device status and provides sufficient information for subsequent safety assessments;

[0060] In the output layer of the equipment feature extraction model, the time series feature set of the gas filling equipment is output. Specifically, the various features calculated by the feature extraction layer are summarized into a feature vector. This vector contains the time series feature set of the gas filling equipment, including the maximum drop rate of gas pressure in the gas filling pipeline, the average flow rate of gas in the gas filling pipeline, the peak value of the surface temperature of the gas filling equipment, the rate of rise of the surface temperature of the gas filling equipment, the peak value of the structural vibration of the gas filling equipment, the standard deviation of the motor operating current, and the opening and closing frequency of the gas filling valve.

[0061] In this implementation, efficient and comprehensive feature extraction can be performed on the multi-dimensional time series status data of gas filling equipment in natural gas filling stations. By constructing a neural network structure of input layer, convolution layer, LSTM layer, feature extraction layer and output layer, the model can accurately capture the short-term mutation characteristics and long-term dependency characteristics contained in the data, and is particularly suitable for identifying potential abnormal trends in equipment operation. Compared with traditional methods based on static thresholds or expert rules, this model makes full use of historical data training and generation, and has stronger robustness and adaptability. At the same time, the feature extraction model can be deployed as an independent module to other similar industrial scenarios for migration and use, which greatly improves its engineering practicality and promotion value.

[0062] Specifically, the specific steps for analyzing the safety index of the gas filling equipment of a natural gas filling station are as follows: read the maximum drop rate of gas pressure in the gas pipeline of the natural gas filling station, the average flow rate of gas in the gas pipeline, the peak temperature of the surface temperature of the gas filling equipment, the rate of increase of the surface temperature of the gas filling equipment, the peak value of the vibration of the structure of the gas filling equipment, and the standard deviation of the motor operating current, and perform standardization processing (i.e., remove the unit processing); combine the standardized maximum drop rate of gas pressure in the gas pipeline of the natural gas filling station, the average flow rate of gas in the gas pipeline, the peak temperature of the surface temperature of the gas filling equipment, the rate of increase of the surface temperature of the gas filling equipment, the peak value of the vibration of the structure of the gas filling equipment, and the standard deviation of the motor operating current with the corresponding gas valve opening and closing frequency for comprehensive analysis to obtain the safety index of the gas filling equipment of the natural gas filling station.

[0063] The specific formula for calculating the safety index of natural gas filling equipment at a natural gas filling station is as follows: ;in, 、 、 、 、 、 、 、 The following are the safety index of the natural gas filling station's filling equipment, the maximum rate of decrease of gas pressure in the filling pipeline, the average flow rate of gas in the filling pipeline, the peak value of the surface temperature of the filling equipment, the rate of increase of the surface temperature of the filling equipment, the peak value of the vibration of the filling equipment structure, the standard deviation of the motor operating current, and the opening and closing frequency of the filling valve. is the temperature adjustment coefficient stored in the database, and in this embodiment, the value is 1.5.

[0064] The specific implementation example for calculating the safety index of gas filling equipment at a natural gas filling station is as follows, with the following parameters:

[0065] The maximum rate of decrease of gas pressure in the gas pipeline of a natural gas filling station is approximately: 0.827.

[0066] The average gas flow rate in the gas pipeline is approximately: 1.103.

[0067] The peak surface temperature of the gas filling equipment is approximately: 0.792.

[0068] The surface temperature rise rate of the gas filling equipment is approximately: 0.964.

[0069] The peak value of the structural vibration of the aeration equipment is approximately: 0.881.

[0070] The standard deviation of the motor operating current is approximately: 0.913.

[0071] The opening and closing frequency of the gas filling valve is approximately: 4.6.

[0072] The temperature adjustment coefficient stored in the database is: 1.5.

[0073] Substituting the above contents into the specific formula for calculating the safety index of natural gas filling equipment at a natural gas filling station, we obtain:

[0074] The safety index of gas filling equipment at a natural gas filling station = ((0.827 / 1.103) + ((0.792 / 0.964)^1.5)) × (0.881 / 0.913) + ln (1+4.6) ≈ 3.165.

[0075] In this implementation plan, a quantitative evaluation of the operating status of natural gas filling station equipment is achieved by constructing a calculation method for the gas filling equipment safety index. This method not only clarifies the standardized processing flow of various key characteristic parameters, but also guides the reasonable balance and unification between indicators by setting a temperature adjustment coefficient with adaptability to working conditions, ensuring the comparability and integration of various inputs. The designed formula has a clear structure and can reflect various safety risk factors such as pressure changes, flow rate characteristics, temperature evolution, motor current fluctuations and equipment vibrations. At the same time, the influence of the opening and closing frequency of the gas valve on the system fluctuations is considered, making the calculation results more sensitive and indicative.

[0076] Specifically, the time series status data of the aeration environment include combustible gas concentration values, ambient temperature values, foundation structure micro-vibration amplitude values, electromagnetic interference intensity values, thermal infrared radiation temperature values ​​and air flow velocity values ​​at several time points.

[0077] The environmental feature extraction model includes input layer, convolution layer, RNN layer, pooling layer, feature extraction layer, output layer, and the aeration environment time series feature set includes combustible gas concentration peak, combustible gas concentration exceeding the standard duration, environmental average temperature value, foundation structure micro-vibration peak, electromagnetic interference intensity peak, thermal infrared radiation temperature peak, thermal infrared radiation temperature rising slope, and air flow average speed value.

[0078] Among them, the pre-training steps of the environmental feature extraction model are as follows:

[0079] First, a training dataset was constructed. This dataset was generated based on multiple historical gas filling environment data, including combustible gas concentrations, ambient temperatures, ground structure microvibration amplitudes, electromagnetic interference intensity, thermal infrared radiation temperature, and air velocity. By analyzing the similarities between the historical data and the target gas filling station environment in terms of equipment type and operating environment, a set of environmental time series features was constructed. Based on historical annotations (e.g., abnormal gas concentrations, excessive temperatures, etc.), each data point was labeled with its abnormal state and specific type, thus forming a complete training sample set.

[0080] The environmental feature extraction model used in this embodiment is a deep neural network structure, primarily comprising an input layer, a convolutional layer, an RNN layer, a pooling layer, a feature extraction layer, and an output layer. The input layer is used to receive the time-series state feature vector of the aerated environment corresponding to each time point; the convolutional layer is used to extract local features from the input feature vector; the RNN layer is used to perform time-series modeling on the local features output by the convolutional layer; the pooling layer is used to pool the long-term dependency features output by the RNN layer; the feature extraction layer is used to extract key safety features from the output of the pooling layer; and the output layer is used to generate the final set of time-series features for the aerated environment. Supervised learning is employed during training, with the input being the time-series feature vector of the aerated environment and the output being the corresponding time-series features of the aerated environment (such as peak concentration or peak temperature). The loss function is optimized using mean square error (MSE) or other loss functions suitable for regression tasks.

[0081] During model training, various optimization strategies are introduced, including a layer-by-layer learning rate adjustment mechanism that dynamically adjusts the learning rate based on the training progress of each layer, accelerating the training process. Batch normalization is used to stabilize the distribution of intermediate layers, and a random dropout mechanism is used to prevent overfitting. Furthermore, periodic learning rate adjustment strategies (such as cosine annealing) are introduced to improve the convergence of the training process. To further prevent overfitting, an early stopping mechanism is introduced during training to terminate the training process prematurely if the validation set accuracy no longer improves over several consecutive iterations. After training, model performance is evaluated based on the validation set, and the optimal set of network parameters is selected as the final pre-trained model parameters. The trained gas filling environment feature extraction model can be directly used to analyze time series data of new gas filling environments, outputting a set of environmental time series features for each environmental data point.

[0082] The specific steps to obtain the time series feature set of the natural gas filling environment of a natural gas filling station are as follows:

[0083] In the input layer of the environmental feature extraction model, the time series state data of the aeration environment is received and processed to obtain the initial feature vector of the aeration environment. Specifically, the time series state data of the aeration environment, including the combustible gas concentration value, the ambient temperature value, the amplitude of the foundation structure micro-vibration, the electromagnetic interference intensity value, the thermal infrared radiation temperature value, and the air flow velocity value, is received and converted into the initial feature vector as the input for the subsequent convolutional layer and RNN layer processing;

[0084] In the convolution layer of the environmental feature extraction model, the input initial feature vector of the aerated environment is convolved to extract local features and obtain a convolution feature map of the aerated environment. Specifically, the input initial feature vector of the aerated environment is processed by a 1D convolution operation. The convolution layer uses the convolution kernel to perform weighted summation on the input feature vector at each time step, thereby extracting local time series features, such as combustible gas concentration fluctuations, temperature changes, vibration changes, etc. The size of the convolution kernel (usually 3 or 5) determines the time step range involved in each convolution operation. The output after convolution is the convolution feature map of the aerated environment, which contains the local pattern features in the input data.

[0085] In the RNN layer of the environmental feature extraction model, the convolutional layer outputs the gasification environment convolution feature map, performing time series modeling to obtain the long-term dependency features of the gasification environment. Specifically, the convolutional layer outputs the gasification environment convolution feature map as input for time series modeling. The RNN layer (recursive neural network) uses its structure to perform time series processing on the input data, capturing the long-term dependencies in the data. The RNN layer can handle long-term trend changes in the gasification environment, such as gradually increasing gas concentrations and temperature fluctuations. Using its internal recurrent structure to maintain dependency information between time steps, it outputs the long-term dependency features of the gasification environment, providing a foundation for subsequent feature extraction.

[0086] In the pooling layer of the environmental feature extraction model, the long-term dependency features of the aeration environment output by the RNN layer are pooled to obtain a pooled feature set of the aeration environment. Specifically, the pooling layer performs a pooling operation (such as maximum pooling or average pooling) on ​​the long-term dependency features output by the RNN layer to reduce the feature dimension and retain the most important temporal features. This operation helps simplify the model calculation and reduce the risk of overfitting. At the same time, it retains the core temporal features of the aeration environment, such as the temperature fluctuation range and the extreme values ​​of gas concentration, thereby extracting the pooled feature set of the aeration environment.

[0087] In the feature extraction layer of the environmental feature extraction model, the pooled feature set of the aerated environment output by the pooling layer is statistically analyzed to obtain the time series feature set of the aerated environment. Specifically, the pooled feature set of the aerated environment output by the pooling layer is further statistically analyzed to extract key information and generate the time series feature set of the aerated environment. Specifically, the following features are extracted from the time series data:

[0088] Combustible gas concentration peak: reflects the extreme fluctuation of combustible gas concentration by extracting the maximum value of gas concentration.

[0089] Duration of combustible gas concentration exceeding the standard: Calculate the duration of combustible gas concentration exceeding the standard to reflect the persistence of the exceeding standard event.

[0090] Average ambient temperature: Calculate the average value of the ambient temperature to reflect the overall temperature conditions of the gas station.

[0091] Foundation and structure microvibration peak value: Extracts the maximum amplitude from foundation and structure microvibration data to help identify potential infrastructure problems.

[0092] Electromagnetic interference intensity peak: Extract the maximum value of electromagnetic interference intensity and monitor the fluctuation of electromagnetic environment.

[0093] Thermal infrared radiation temperature peak value: extract the maximum value in the thermal infrared radiation temperature data to reflect the abnormality of the heat source.

[0094] Thermal infrared radiation temperature rising slope: the rising slope of the thermal infrared radiation temperature change is calculated to help identify the trend of rapid temperature rise.

[0095] Air flow average speed value: the average value of the air flow speed is calculated to reflect the air circulation and ventilation situation;

[0096] In the output layer of the environmental feature extraction model, the output gas filling environment time series feature set is output, which is specifically: the features calculated by the feature extraction layer are summarized into a feature vector, and this vector contains the gas filling environment time series feature set, including the combustible gas concentration peak value, the combustible gas concentration exceeding standard duration, the environmental average temperature value, the ground structure micro-vibration peak value, the electromagnetic interference intensity peak value, the thermal infrared radiation temperature peak value, the thermal infrared radiation temperature rising slope, and the air flow average speed value.

[0097] In this embodiment, by constructing a deep neural network architecture including an input layer, a convolution layer, an RNN layer, a pooling layer, a feature extraction layer and an output layer, efficient modeling and multi-dimensional feature extraction of the gas filling station environment data are realized. Compared with the traditional method relying on a single index or fixed threshold judgment, the model can autonomously learn typical abnormal patterns from historical data, such as abnormal fluctuations in gas concentration, dramatic changes in environmental temperature, sudden enhancement of electromagnetic interference and micro-vibration, etc. By introducing the convolution and recurrent network joint structure, both short-term mutations and long-term trends can be captured. The pooling mechanism further compresses redundant information, improves the accuracy and compactness of feature expression, and ensures that the model has good generalization ability by using supervised learning and various optimization strategies such as grouping normalization and early stopping mechanism during model training. Finally, the extracted feature set has clear physical meaning and time sequence indication, effectively improving the response sensitivity and evaluation accuracy of complex environmental risk changes.

[0098] Specifically, the specific steps of analyzing the gas filling environment safety index of the natural gas filling station are as follows: reading the gas filling environment time series feature set of the natural gas filling station and performing standardization processing (i.e. unit removal processing); comprehensively analyzing the standardized gas filling environment time series feature set to obtain the gas filling environment safety index of the natural gas filling station.

[0099] Wherein, the specific formula for calculating the gas filling environment safety index of the natural gas filling station is as follows: ; wherein, , , , , , , , , The indicators are the natural gas filling station's refueling environment safety index, combustible gas concentration peak value, duration of combustible gas concentration exceeding the standard, average ambient temperature, foundation structure micro-vibration peak value, electromagnetic interference intensity peak value, thermal infrared radiation temperature peak value, thermal infrared radiation temperature rising slope, and average air flow velocity value. 、 They are the combustible gas concentration adjustment coefficient and the thermal infrared radiation temperature adjustment coefficient stored in the database, which are 1.5 and 2 respectively in this embodiment.

[0100] In this implementation plan, by extracting representative environmental characteristics and constructing a standardized processing flow, effective fusion of data under different physical dimensions is achieved. The adjustment coefficients of combustible gas concentration and thermal infrared temperature are introduced into the formula to enhance the response sensitivity to high-risk environmental parameters, especially in abnormal leakage or temperature rise scenarios. The overall index calculation method enables various risk factors to be effectively integrated under a unified indicator system. This method can quantify the safety status of the gas filling environment, facilitate the system to quickly determine whether the current operation is in a safe range, and improve the early warning efficiency of environmental anomalies.

[0101] See also Figure 3 , an embodiment of the present invention provides a technical solution: a natural gas filling station safety evaluation system, comprising: a gas filling data acquisition unit, used to obtain the time series status data of the gas filling equipment and the gas filling environment of the natural gas filling station and perform preprocessing; a gas filling data extraction unit, used to perform feature analysis on the preprocessed time series status data of the gas filling equipment and the gas filling environment based on a pre-trained equipment feature extraction model and an environment feature extraction model, respectively, to obtain a time series feature set of the gas filling equipment and the gas filling environment of the natural gas filling station; a gas filling data analysis unit, used to analyze the safety index of the gas filling equipment and the gas filling environment of the natural gas filling station based on the time series feature set of the gas filling equipment and the gas filling environment; a gas filling safety evaluation unit, used to judge and analyze the safety index of the gas filling equipment and the gas filling environment of the natural gas filling station with several preset groups of gas filling safety interval sets, each group of gas filling safety interval sets including a gas filling equipment safety interval and a gas filling environment safety interval, and each group of gas filling safety interval sets corresponding to a safety level.

[0102] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0103] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A natural gas filling station safety evaluation method, characterized in that: The following steps are involved: Obtain the time series status data of the natural gas filling equipment and filling environment of the natural gas filling station and perform preprocessing; Based on the pre-trained equipment feature extraction model and environment feature extraction model, the pre-processed gas filling equipment and gas filling environment time series state data are analyzed to obtain the gas filling equipment and gas filling environment time series feature sets of the natural gas filling station; Based on the time series feature sets of gas filling equipment and gas filling environment, the safety index of gas filling equipment and gas filling environment of natural gas filling stations are analyzed respectively; The safety index of the natural gas filling station's filling equipment and filling environment is compared with several preset filling safety interval sets for judgment and analysis. Each filling safety interval set includes a filling equipment safety interval and a filling environment safety interval, and each filling safety interval set corresponds to a safety level. The equipment feature extraction model includes an input layer, a convolutional layer, an LSTM layer, a feature extraction layer, and an output layer. The time series feature set of the gas filling equipment includes the maximum drop rate of gas pressure in the gas filling pipeline, the average gas flow rate in the gas filling pipeline, the peak temperature of the gas filling equipment surface, the rate of rise of the gas filling equipment surface temperature, the peak vibration of the gas filling equipment structure, the standard deviation of the motor operating current, and the opening and closing frequency of the gas filling valve. The environmental feature extraction model includes input layer, convolution layer, RNN layer, pooling layer, feature extraction layer, output layer, and the aeration environment time series feature set includes combustible gas concentration peak, combustible gas concentration exceeding the standard duration, environmental average temperature value, foundation structure micro-vibration peak, electromagnetic interference intensity peak, thermal infrared radiation temperature peak, thermal infrared radiation temperature rising slope, and air flow average speed value.

2. The natural gas filling station safety evaluation method according to claim 1, characterized in that: The time series status data of the gas filling equipment includes the gas pressure value of the gas filling pipeline, the gas flow rate value of the gas filling pipeline, the surface temperature value of the gas filling equipment, the vibration amplitude of the gas filling equipment structure, the motor operating current value and the opening and closing status value of the gas filling valve at several time points.

3. The natural gas filling station safety evaluation method according to claim 1, characterized in that: The specific steps to obtain the time series feature set of the gas filling equipment at the natural gas filling station are as follows: In the input layer of the equipment feature extraction model, the time series state data of the gas filling equipment is received and processed to obtain the initial feature vector of the gas filling equipment; In the convolution layer of the equipment feature extraction model, the input initial feature vector of the gas filling equipment is convolved to extract local features and obtain the convolution feature map of the gas filling equipment; In the LSTM layer of the equipment feature extraction model, the convolutional feature map of the gas filling equipment output by the convolutional layer is subjected to time series modeling to obtain the long-term dependency features of the gas filling equipment. In the feature extraction layer of the equipment feature extraction model, statistical analysis is performed on the long-term dependency features of the gas filling equipment output by the LSTM layer to obtain the time series feature set of the gas filling equipment; In the output layer of the equipment feature extraction model, the time series feature set of the gas filling equipment is output.

4. The natural gas filling station safety evaluation method according to claim 1, characterized in that: The specific steps for analyzing the safety index of natural gas filling equipment at a natural gas filling station are as follows: Read the maximum gas pressure drop rate of the gas pipeline of the natural gas filling station, the average gas flow rate of the gas pipeline, the peak temperature of the gas filling equipment surface, the rate of increase of the gas filling equipment surface temperature, the peak value of the gas filling equipment structure vibration, and the standard deviation of the motor operating current, and perform standard processing; The maximum gas pressure drop rate in the gas pipeline of the natural gas filling station, the average gas flow rate in the gas pipeline, the peak surface temperature of the gas filling equipment, the rate of rise in the surface temperature of the gas filling equipment, the peak structural vibration of the gas filling equipment, and the standard deviation of the motor operating current are comprehensively analyzed with the corresponding opening and closing frequency of the gas valves to obtain the gas filling equipment safety index of the natural gas filling station.

5. The natural gas filling station safety evaluation method according to claim 1, characterized in that: The time series status data of the aeration environment include the combustible gas concentration value, ambient temperature value, foundation structure micro-vibration amplitude, electromagnetic interference intensity value, thermal infrared radiation temperature value and air flow velocity value at several time points.

6. The natural gas filling station safety evaluation method according to claim 1, characterized in that: The specific steps to obtain the time series feature set of the natural gas filling environment of a natural gas filling station are as follows: In the input layer of the environmental feature extraction model, the time series state data of the aeration environment is received and processed to obtain the initial feature vector of the aeration environment; In the convolution layer of the environmental feature extraction model, the input initial feature vector of the aeration environment is convolved to extract local features and obtain the convolution feature map of the aeration environment; In the RNN layer of the environmental feature extraction model, the convolutional feature map of the aeration environment output by the convolutional layer is subjected to time series modeling to obtain the long-term dependency features of the aeration environment. In the pooling layer of the environmental feature extraction model, the long-term dependent features of the aeration environment output by the RNN layer are pooled to obtain the aeration environment pooling feature set; In the feature extraction layer of the environmental feature extraction model, statistical analysis is performed on the pooled feature set of the aeration environment output by the pooling layer to obtain the time series feature set of the aeration environment; In the output layer of the environmental feature extraction model, the time series feature set of the gas filling environment is output.

7. The natural gas filling station safety evaluation method according to claim 1, characterized in that: The specific steps for analyzing the refueling environment safety index of a natural gas filling station are as follows: Read the time series feature set of the natural gas filling environment at the natural gas filling station and perform standardization processing; A comprehensive analysis is conducted on the standardized time series feature set of the refueling environment to obtain the refueling environment safety index of the natural gas refueling station.

8. A natural gas filling station safety evaluation system, applying the natural gas filling station safety evaluation method according to any one of claims 1 to 7, characterized in that: include: The gas filling data acquisition unit is used to obtain the time series status data of the gas filling equipment and gas filling environment of the natural gas filling station and perform preprocessing; A gas filling data extraction unit is used to perform feature analysis on the pre-processed gas filling equipment and gas filling environment time series state data based on the pre-trained equipment feature extraction model and environment feature extraction model, respectively, to obtain a time series feature set of the gas filling equipment and gas filling environment of the natural gas filling station; A gas filling data analysis unit is used to analyze the gas filling equipment and gas filling environment safety index of a natural gas filling station based on the time series feature sets of the gas filling equipment and the gas filling environment; The gas filling safety evaluation unit is used to judge and analyze the gas filling equipment and gas filling environment safety index of the natural gas filling station with several preset gas filling safety interval sets. Each gas filling safety interval set includes the gas filling equipment safety interval and the gas filling environment safety interval, and each gas filling safety interval set corresponds to a safety level.

Images