Petrochemical harbor district safety risk analysis method based on deep learning

Abstract

The invention discloses a petrochemical harbor area safety risk analysis method based on deep learning, and relates to the technical field of evaluation and analysis. The method comprises the following steps: firstly, collecting associated data of a petrochemical harbor area operation scene and a storage tank area, and inputting an accident prediction model after preprocessing; the model comprises a three-dimensional working environment model based on the digital twin technology and an accident risk classification model, and the accident risk classification model analyzes the influence of historical accident information on a current working area through an accident time decay function. And if the risk probability exceeds a threshold value, the system gives an early warning, performs color marking on an accident area in the working environment model, and sends an early warning signal to a worker device for warning. According to the invention, potential safety risks can be identified and responded in advance.

Description

Technical Field

[0001] The present invention relates to the field of evaluation and analysis technology, and in particular to a petrochemical port area safety risk analysis method based on deep learning. Background Art

[0002] Accidents causing casualties in petrochemical ports often occur, and the situation of personnel safety risk control is extremely severe. It is urgent to strengthen the safety risk control of port storage tank areas and ensure the safe operation of ports. The key to ensuring safe production in petrochemical ports lies in the safety implementation of various related events at all levels during the operation of petrochemical ports. During the operation of petrochemical ports, evaluating and supervising the safety of various personnel, various links of operations and environmental safety in the safe production of petrochemical ports is one of the effective means to prevent safety accidents. However, in the existing field of safety risk analysis of petrochemical ports, traditional risk assessment methods mainly rely on manual experience judgment, regular inspections and statistical analysis of historical accident data. Although these methods can identify potential safety hazards to a certain extent, they have many limitations and shortcomings. Summary of the invention

[0003] The purpose of this invention is to provide a petrochemical port area safety risk analysis method based on deep learning, which can identify and respond to potential safety risks in advance.

[0004] To achieve the above object, the present invention provides the following solutions: A petrochemical port area safety risk analysis method based on deep learning, including: Obtaining the associated data of the petrochemical port operation scene and the tank area; the associated data includes the operation environment data, equipment status, operation process and personnel distribution; The associated data is preprocessed, and the preprocessed data is input into the accident prediction model trained at the current moment, and the accident prediction model determines the accident risk event; the accident prediction model includes an operating environment model and an accident risk classification model; the operating environment model is a three-dimensional model constructed based on digital twin technology; the accident risk classification model includes an accident time decay function; the accident time decay function is the degree of influence of the accident information of the entire petrochemical port area on the current working area; the accident information is the accident information within a set time range; the set time range is [t1, t2]; t2 represents the current moment; t2-t1≥1; The risk occurrence probability is assessed based on the accident risk event. When the risk occurrence probability exceeds a set threshold, a warning signal is generated and the accident area in the working environment model is rendered in color. The warning signal is transmitted to the staff's equipment for warning, prompting the staff to check the accident area.

[0005] Optionally, the preprocessing process includes data cleaning, format conversion and normalization.

[0006] Optionally, the calculation formula of the accident time decay function is: , in, T (t) is the accident time decay function; T (t0) is the attenuation degree of the impact of the accident occurring at time t0 on the event occurring at time t; k is the set time attenuation coefficient.

[0007] Optionally, the accident prediction model includes a feature branch and a time branch, and the feature branch and the time branch are also connected to a classifier; the feature branch is used to extract data features; the time branch is used to extract time series features of events; the classifier is used to output accident risk events according to the data features, the time series features and the accident time attenuation function; wherein the feature branch includes a first EffAttn module and a first BiLSTM module connected in sequence; the time branch includes a second EffAttn module and a second BiLSTM module connected in sequence; the first BiLSTM module and the second BiLSTM module are also connected to a gated fusion module; the gated fusion module is also connected to the classifier.

[0008] Optionally, the gated fusion module adopts a Gate module.

[0009] Optionally, the EffAttn module is composed of a linear layer and a Softmax function, and the linear layer performs a simplified linear transformation to directly weight the input data.

[0010] Optionally, the training process of the accident prediction model includes: Obtaining basic data and training data within the safety control area of ​​the petrochemical port area; the basic data includes the operating environment data within the target area; the training data includes accident information of the entire petrochemical port area, accident risk events of the entire petrochemical port area, and accident correlation information of the entire petrochemical port area; Construct pre-trained models of digital twin models and accident risk classification models; Rendering a pre-trained model of a digital twin model according to the basic data, and inputting the training data into a pre-trained model of an accident risk classification model, training with the goal of minimizing the loss between the model output and the training label, to obtain a rendered three-dimensional model and a trained accident risk classification model; the training label is an accident risk event corresponding to the input data; The rendered three-dimensional model and the trained accident risk classification model are fused to obtain a trained accident prediction model.

[0011] Optionally, the staff member's equipment includes at least one of a personal computer, a laptop, a smart phone, a tablet computer and a portable wearable device.

[0012] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: The present invention discloses a petrochemical port area safety risk analysis method based on deep learning. The method first collects the associated data of the petrochemical port area operation scene and the tank area, and inputs the data into the accident prediction model after preprocessing. The model includes a three-dimensional operation environment model based on digital twin technology and an accident risk classification model, and the latter analyzes the impact of historical accident information on the current work area through the accident time decay function. If the risk probability exceeds the threshold, the system will issue an early warning, color-mark the accident area in the operation environment model, and send an early warning signal to the staff device for warning. The present invention helps to identify and respond to potential safety risks in advance. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0014] Figure 1 This is a flow chart of the petrochemical port area safety risk analysis method based on deep learning of the present invention. DETAILED DESCRIPTION

[0015] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0016] The purpose of this invention is to provide a petrochemical port area safety risk analysis method based on deep learning, which can identify and respond to potential safety risks in advance.

[0017] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] like Figure 1 As shown, the present invention provides a petrochemical port area safety risk analysis method based on deep learning, comprising: Step 100: Obtain the associated data of the petrochemical port operation scene and the tank area; the associated data includes the operation environment data, equipment status, operation process and personnel distribution.

[0019] Step 200: Preprocess the associated data, and input the preprocessed data into the accident prediction model trained at the current moment, and determine the accident risk event by the accident prediction model; the accident prediction model includes an operating environment model and an accident risk classification model; the operating environment model is a three-dimensional model built based on digital twin technology; the accident risk classification model includes an accident time decay function; the accident time decay function is the degree of influence of the accident information of the entire petrochemical port area on the current working area; the accident information is the accident information within the set time range; the set time range is [t1, t2]; t2 represents the current moment; t2-t1≥1. Among them, the preprocessing process includes data cleaning, format conversion and normalization processing.

[0020] Step 300: Evaluate the probability of risk occurrence based on the accident risk event. When the probability of risk occurrence exceeds a set threshold, generate a warning signal and render the accident area in the working environment model in color. Transmit the warning signal to the staff's equipment for warning, prompting the staff to check the accident area.

[0021] As a specific implementation, the calculation formula of the accident time decay function is: , in, T (t) is the accident time decay function; T (t0) is the attenuation degree of the impact of the accident occurring at time t0 on the event occurring at time t; k is the set time attenuation coefficient.

[0022] As a specific implementation, the accident prediction model includes a feature branch and a time branch, and the feature branch and the time branch are also connected to a classifier; the feature branch is used to extract data features; the time branch is used to extract the time series features of the event; the classifier is used to output the accident risk event according to the data features, the time series features and the accident time attenuation function; wherein the feature branch includes a first EffAttn module and a first BiLSTM module connected in sequence; the time branch includes a second EffAttn module and a second BiLSTM module connected in sequence; the first BiLSTM module and the second BiLSTM module are also connected to a gated fusion module; the gated fusion module is also connected to the classifier. wherein the gated fusion module adopts a Gate module, the EffAttn module is composed of a linear layer and a Softmax function, and the linear layer performs a simplified linear transformation to directly weight the input data.

[0023] As a specific implementation, the training process of the accident prediction model includes: Obtain basic data and training data within the safety control area of ​​the petrochemical port area; the basic data includes the operating environment data within the target area; the training data includes accident information of the entire petrochemical port area, accident risk events of the entire petrochemical port area, and accident correlation information of the entire petrochemical port area.

[0024] Build pre-trained models of digital twin models and accident risk classification models.

[0025] The pre-trained model of the digital twin model is rendered according to the basic data, and the training data is input into the pre-trained model of the accident risk classification model. Training is performed with the goal of minimizing the loss between the model output and the training label to obtain a rendered three-dimensional model and a trained accident risk classification model; the training label is the accident risk event corresponding to the input data.

[0026] The rendered three-dimensional model and the trained accident risk classification model are fused to obtain a trained accident prediction model.

[0027] As a specific implementation, the staff's equipment includes at least one of a personal computer, a laptop, a smart phone, a tablet computer, and a portable wearable device.

[0028] Based on the above technical solution, the following embodiments are provided to describe each step in detail.

[0029] Step 100: Obtain the associated data of the petrochemical port operation scene and the tank area.

[0030] In the petrochemical port area, various sensors and monitoring systems are deployed to collect relevant data of operation scenes and tank areas in real time. These data include but are not limited to: Working environment data: temperature, humidity, wind speed, wind direction, combustible gas concentration, etc.

[0031] Equipment status: tank pressure, liquid level, valve switch status, pump operating status, etc.

[0032] Job process: current job type, job progress, job permit status, etc.

[0033] Personnel distribution: personnel location, work team distribution, emergency evacuation routes, etc.

[0034] All of the above data are transmitted to the data center via the Internet of Things (IoT) technology, providing a basis for subsequent analysis.

[0035] Step 200: Data preprocessing and accident prediction model application.

[0036] Data preprocessing: Data cleaning: Remove invalid, erroneous or duplicate data records.

[0037] Format conversion: Ensure that all data conforms to model input requirements, such as converting timestamps to a unified format.

[0038] Normalization processing: Convert data of different dimensions to the same scale to facilitate model processing.

[0039] Accident prediction model application: Model structure: The accident prediction model consists of an operating environment model and an accident risk classification model. The operating environment model uses digital twin technology to build a three-dimensional virtual model of the petrochemical port area to reflect the on-site conditions in real time. The accident risk classification model contains feature branches and time branches, which process spatial features and temporal features respectively.

[0040] Feature extraction: The first EffAttn module of the feature branch weights the input data through a linear layer and a Softmax function to extract key features; then, the first BiLSTM module processes these features to capture dependencies between sequences. The second EffAttn module and the second BiLSTM module of the time branch perform similar operations, but focus on extracting the temporal features of events.

[0041] Gated fusion: Use the Gate module to fuse information from the feature branch and the time branch to ensure that the model can comprehensively consider spatial and temporal features.

[0042] Accident time decay function: This function calculates the impact of accidents in the entire petrochemical port area on the current working area within the set time range [t1, t2], enhancing the prediction ability of the model.

[0043] Model output: The classifier outputs possible accident risk events based on the fused features, time series features and accident time decay function.

[0044] Step 300: Risk probability assessment and early warning.

[0045] Risk assessment: Based on the output of the accident prediction model, the probability of occurrence of each risk event is evaluated.

[0046] Early warning signal: When the probability of a risk event exceeds the preset threshold, the system triggers the early warning mechanism.

[0047] Color rendering: In the work environment model, high-risk areas are rendered with specific colors (such as red) to visually display the risk location.

[0048] Warning transmission: Send warning signals to staff in real time via personal computers, laptops, smartphones, tablets or portable wearable devices to ensure rapid communication of information.

[0049] Emergency response: Based on the early warning information, the staff will immediately check the accident area and take necessary preventive measures or emergency response actions.

[0050] The model training process includes: Data preparation: Collect basic data and training data within the safety control area of ​​the petrochemical port area, including operating environment data, accident information, accident risk events and accident-related information.

[0051] Pre-trained model construction: Build pre-trained networks for the digital twin model and accident risk classification model respectively.

[0052] Model training: The digital twin model is rendered using basic data to accurately reflect the actual situation of the petrochemical port area. The training data is input into the pre-trained model of the accident risk classification model to optimize the model parameters by minimizing the loss between the model output and the training label.

[0053] Model fusion: Integrate the rendered 3D model and the trained accident risk classification model to form a complete accident prediction model.

[0054] Through the above steps, this method can achieve accurate prediction and timely warning of safety risks in petrochemical port areas, and effectively improve the safety management level of the port areas.

[0055] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0056] This article uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only used to help understand the core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A petrochemical port area safety risk analysis method based on deep learning, characterized in that: include: Obtaining the associated data of the petrochemical port operation scene and the tank area; the associated data includes the operation environment data, equipment status, operation process and personnel distribution; Preprocessing the associated data, and inputting the preprocessed data into the accident prediction model trained at the current moment, and determining the accident risk event by the accident prediction model; The accident prediction model includes an operating environment model and an accident risk classification model; the operating environment model is a three-dimensional model constructed based on digital twin technology; the accident risk classification model includes an accident time decay function; the accident time decay function is the degree of influence of accident information in the entire petrochemical port area on the current working area; the accident information is accident information within a set time range; the set time range is [t1, t2]; t2 represents the current time; t2-t1≥1; The risk occurrence probability is assessed based on the accident risk event. When the risk occurrence probability exceeds a set threshold, a warning signal is generated and the accident area in the working environment model is rendered in color. The warning signal is transmitted to the staff's equipment for warning, prompting the staff to check the accident area.

2. The petrochemical port area safety risk analysis method based on deep learning according to claim 1 is characterized in that: The preprocessing process includes data cleaning, format conversion and normalization.

3. The petrochemical port area safety risk analysis method based on deep learning according to claim 1 is characterized in that: The calculation formula of the accident time decay function is: ， in, T (t) is the accident time decay function; T (t0) is the attenuation degree of the impact of the accident occurring at time t0 on the event occurring at time t; k is the set time attenuation coefficient.

4. The petrochemical port area safety risk analysis method based on deep learning according to claim 1 is characterized in that: The accident prediction model includes a feature branch and a time branch, and the feature branch and the time branch are also connected to a classifier; the feature branch is used to extract data features; the time branch is used to extract the time series features of events; the classifier is used to output accident risk events according to the data features, the time series features and the accident time attenuation function; wherein the feature branch includes a first EffAttn module and a first BiLSTM module connected in sequence; the time branch includes a second EffAttn module and a second BiLSTM module connected in sequence; the first BiLSTM module and the second BiLSTM module are also connected to a gated fusion module; the gated fusion module is also connected to the classifier.

5. The petrochemical port area safety risk analysis method based on deep learning according to claim 4 is characterized in that: The gated fusion module adopts a Gate module.

6. The petrochemical port area safety risk analysis method based on deep learning according to claim 4 is characterized in that: The EffAttn module is composed of a linear layer and a Softmax function, and the linear layer performs a simplified linear transformation to directly weight the input data.

7. The petrochemical port area safety risk analysis method based on deep learning according to claim 1 is characterized in that: The training process of the accident prediction model includes: Obtaining basic data and training data within the safety control area of ​​the petrochemical port area; the basic data includes the operating environment data within the target area; the training data includes accident information of the entire petrochemical port area, accident risk events of the entire petrochemical port area, and accident correlation information of the entire petrochemical port area; Construct pre-trained models of digital twin models and accident risk classification models; Rendering a pre-trained model of a digital twin model according to the basic data, and inputting the training data into a pre-trained model of an accident risk classification model, training with the goal of minimizing the loss between the model output and the training label, to obtain a rendered three-dimensional model and a trained accident risk classification model; the training label is an accident risk event corresponding to the input data; The rendered three-dimensional model and the trained accident risk classification model are fused to obtain a trained accident prediction model.

8. The petrochemical port area safety risk analysis method based on deep learning according to claim 1 is characterized in that: The staff member's equipment includes at least one of a personal computer, a laptop, a smart phone, a tablet computer and a portable wearable device.

Images