Energy combustion and explosion physical risk coping-public emotion regulation and control cooperation method and platform

Abstract

The invention provides an energy combustion and explosion physical risk response-public emotion regulation and control cooperation method and platform, and the method comprises the steps: carrying out the preprocessing of the real-time data of an edge layer through a risk analysis and early warning prevention and control unit, and building a gas combustion and explosion risk database; a standardized cloud database is constructed based on multi-source and multi-modal data by utilizing an online incremental learning unit, batch automatic preprocessing, storage and calling of the data are realized, and the purposes of improving the calculation efficiency and reducing the manual operation amount are achieved; the emotion learning and analysis unit is used for crawling related accident news and public opinions; based on emotion prediction and intervention, digital twins are combined with an emotion analysis process, and public emotion prediction, staged psychological support and resource scheduling are realized. The technical problem that a decision blind area exists in emergency response due to the fact that physical risk early warning and emotional evolution analysis cannot be dynamically correlated is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of energy explosion risk prevention and control, in particular to a public emotion regulation and physical risk response coordination method and platform for energy explosion. BACKGROUND

[0002] With the large-scale construction of city gas stations, natural gas pipelines and cable channels, the potential risk of oil and gas leakage coupled with natural gas leakage leading to explosion continues to rise. In order to achieve controllable risk, the existing safety risk analysis technology has focused on the physical environment level, but it has ignored the social risk dimension derived from the explosion accident. In the context of highly developed social networks, accident information spreads rapidly, and public panic emotions are easily spread and escalated, even inducing secondary crises such as group events. In the current social environment, by integrating physical environment risk dynamic early warning and public emotion evolution analysis, breaking through the limitations of traditional systems that only focus on physical risk, and realizing the coordinated prevention and control of physical and social risks in explosion accidents, it is beneficial to ensure the safe operation of urban lifelines and the positive of social public opinion, and it is crucial for public life, property and even psychological safety.

[0003] The existing patent application document with publication number CN119918929A, "A risk assessment method, system and terminal for a gas pipeline", the existing method includes: obtaining sample monitoring data of a plurality of gas monitoring devices attached to the gas pipeline, establishing a risk base model of the gas pipeline according to the sample monitoring data; using weight analysis method to distribute index weight of the risk base model, obtaining target risk base model, training the target risk base model according to the sample monitoring data, obtaining risk assessment model; obtaining current sample monitoring data, inputting the current sample monitoring data into the risk assessment model, outputting the gas risk assessment result of the gas pipeline. The present application establishes a risk model through monitoring data, performs weight analysis and training, and finally evaluates the risk of the gas pipeline. By establishing a risk assessment model, potential risks of the gas pipeline can be effectively identified and warned, safety management level can be improved, and accident probability can be reduced.

[0004] The risk analysis technology in the foregoing prior art often only focuses on the physical environment risk level, which leads to the passive situation of "emphasizing physical prevention and control and ignoring psychological counseling" in emergency disposal, which cannot control social public opinion in advance and in time, curb the deterioration of emotions, and make social risks continue to intensify.

[0005] The existing invention patent application document CN116976485A, entitled "A Method for Early Warning of Online Public Opinion Evolution in Emergencies," includes the following steps: Step 1, collecting online public opinion data and preprocessing the collected data; Step 2, constructing a sentiment classification model for public opinion content; Step 3, establishing a public sentiment evolution analysis model; Step 4, continuously tracking public opinion events; and Step 5, analyzing and issuing early warnings about the evolution of online public opinion in emergencies. The online public opinion evolution early warning method provided by this invention explores the impact of different emotions on information dissemination during emergencies through thematic analysis and sentiment mining of hot events on online social platforms, analyzes the evolutionary patterns of public sentiment and the evolutionary trends of online public opinion, thereby predicting and regulating emergencies, promptly identifying potential risks affecting social stability, and preventing the occurrence of emotional polarization.

[0006] While existing emergency public opinion analysis technologies categorize emotions and construct emotion evolution models to achieve emotion analysis, they do not consider the impact of the intensity of public emotional reactions. Accidents of different severity often elicit different reaction intensities, and changes in reaction intensity can influence public sentiment. Furthermore, current technologies are mostly limited to the prediction and early warning stage, neglecting psychological intervention and resource allocation during the emotion evolution process. Timely and appropriate psychological intervention and resource allocation can effectively reduce negative emotions.

[0007] With the continuous expansion of urban gas stations, natural gas pipelines, and cable tunnels, the potential risk of combustion and explosion due to the coupling of oil and gas leaks with sparks and static electricity is also increasing. To ensure the controllability of this potential risk, the safety monitoring of gas and oil is receiving increasing attention. However, current analyses of combustion and explosion risks and accidents only focus on the physical environment, neglecting the negative impact of public sentiment on society after an accident and the timely intervention of negative emotions. In today's world, with the rapid spread of information through social networks, public sentiment is easily influenced by the emotions of others, leading to escalating public panic and even secondary social crises. Therefore, there is a lack of granular, quantifiable, and phased predictions and proactive interventions for the social risks such as public panic and accountability issues arising from accidents.

[0008] Currently, explosion risk prevention and control technologies are in a state of separation between physical disposal and public opinion control, and cannot comprehensively consider dynamic early warning of physical risks, the amplification effect of public sentiment, and collaborative prevention and control decisions.

[0009] In summary, existing technologies suffer from a technical problem: the inability to dynamically link physical risk warnings with emotional evolution analysis leads to decision-making blind spots in emergency response. Summary of the Invention

[0010] The technical problem to be solved by this invention is: how to solve the technical problem that the existing technology cannot dynamically link physical risk warning and emotional evolution analysis, resulting in decision-making blind spots in emergency response.

[0011] This invention solves the above-mentioned technical problems by employing the following technical solution: a collaborative method for addressing physical risks of energy combustion and explosion and regulating public sentiment, comprising: S1. Using the analysis and early warning prevention and control unit, acquire and preprocess real-time data of the edge layer, establish a gas combustion and explosion risk database, construct a dynamic risk identification model based on the Transformer-CNN deep learning algorithm, and conduct dynamic risk assessment; acquire and use the concentration data of monitoring points to model the gas diffusion behavior area, obtain the gas diffusion behavior model, apply the gas diffusion behavior model to the dynamic risk identification model, and conduct dynamic early warning of regional risks; generate prevention and control suggestions. S2. Utilize online incremental learning units to perform database standardization operations; automatically update the gas combustion and explosion risk database and weights to obtain a cloud database; construct a genetic optimization algorithm based on the cloud database to detect layout optimization; S3. Utilize the emotion learning and analysis unit to crawl and preprocess relevant news and public opinion data on natural gas and oil and gas leaks and explosions in real time; based on the relevant news and public opinion data, use a large language model to conduct fine-grained emotion analysis, matching and associating the severity of the accident with the intensity of emotions to obtain matching data; conduct emotion evolution analysis, establish a vertical emotion evolution model for accidents, analyze the distribution of emotion categories at different evolution stages, quantify the duration and characteristics of each emotion evolution stage, and analyze the trend of emotion deterioration; based on the matching data and the trend of emotion deterioration, conduct emotion prediction and intervention.

[0012] This invention proposes a collaborative platform for physical risk response and public sentiment regulation based on oil and gas explosions. It breaks through the limitations of traditional safety monitoring systems that only focus on physical risks. By integrating dynamic early warning of the physical environment with the evolution analysis of public sentiment, it achieves collaborative prevention and control of "physical-social" dual-dimensional risks in natural gas / oil and gas explosions.

[0013] In a more specific technical solution, S3 introduces the DeepSeek Large Language Model (LLM) for fine-grained emotion analysis. Specifically, it establishes rules for quantifying emotion intensity, extracts emotion tags and intensity from public comments, quantifies the severity of related incidents, correlates incident severity with emotion intensity, obtains matching data, and processes it to obtain the intensity of public emotional reactions.

[0014] In a more specific technical solution, S3 calculates the intensity of public emotional response to the current incident. PER ; Leveraging DeepSeek's zero-shot or few-shot learning capabilities, sentiment categories are batch-processed for comment texts to be classified and sentiment tags are extracted. In accident news texts, objective data about the accident is extracted, and the severity of the explosion accident is quantified according to preset classification rules to obtain the severity of the accident. The severity of the accident and the intensity of public emotional reaction PER Correlate the incidents to conduct a comprehensive assessment of their severity. Based on the accident level, emotional risk index, and emotional evolution stage, an emotional twin based on Unity is established to realize the dynamic inference of emotional evolution and form an emotional evolution prediction path. By combining the Large Language Model (LLM) and based on the predicted path of emotion evolution, we can carry out phased psychological intervention and resource allocation, generate targeted psychological intervention plans for each stage of emotion evolution, and soothe public emotions through online social media channels.

[0015] This invention establishes a physical-social risk control platform for oil and gas combustion and explosion through cloud-based data intelligent learning. It enables coordinated control of physical environment combustion and explosion risks ("analysis-early warning-prevention") and public sentiment ("evolution analysis-psychological intervention-resource allocation"), providing technical guidance for efficient early warning and timely response to potential combustion and explosion risks in both physical environment and public opinion dimensions.

[0016] In more specific technical solutions, emotion scores are assigned based on each emotion category. S Using the following logic, calculate the intensity of public emotional reaction to the current incident. PER :

[0017] In the formula, S i For the first i The sentiment scores assigned to each sample of comments. N This represents the total number of comments in the sample.

[0018] In a more specific technical solution, during the S3 emotion evolution analysis process, public comment text data is collected, the dataset is divided according to a preset time unit, and public emotion categories are extracted respectively. Establish a vertical emotional evolution model for accidents, analyze the distribution of the proportion of emotional categories in different evolution stages, and quantify the duration and emotional category characteristics of each emotional evolution stage. And during the emotional phase changes, obtain and analyze the intensity of public emotional responses. PER By analyzing the changes in the proportion of different emotion categories, we can determine the trend of worsening emotions and the reasons for such worsening.

[0019] In a more specific technical solution, S3 involves emotion prediction and intervention; based on matching data and the trend of emotion deterioration, indicators of accident severity and emotional response are formed, and a post-accident public emotion database is constructed in conjunction with basic accident information; based on the post-accident public emotion database, an emotional digital twin is constructed to simulate the post-accident emotion evolution path and predict the emotion evolution.

[0020] In a more specific technical solution, S3 utilizes a digital twin of emotions for real-time prediction; it uses the Unity engine to build a visual geographic information platform, and connects to physical world data streams and post-accident public sentiment databases in real time via Apache Kafka to establish... t Momentary Emotional Risk Index E t Taking into account both historical changes in emotional intensity and trends of emotional deterioration, the degree of change in emotional risk is quantified.

[0021] To address the shortcomings of existing explosion and fire prevention technologies, which separate physical response from public opinion control, this invention constructs a cloud-based intelligent learning platform to achieve coordinated regulation of "physical-social" risks. Traditional systems cannot dynamically link physical risk warnings with emotional evolution analysis, leading to decision-making blind spots in emergency response. This invention, through the collaborative computation of deep learning algorithms and large language models, enables the system to simultaneously possess the ability to extrapolate physical risk chains and predict social emotional stages. It can achieve risk analysis and early warning prevention, and also generate psychological intervention plans in real time through an emotional digital twin, ultimately achieving a second-level coordinated response of physical prevention and social intervention. This transforms the traditional passive response model into a proactive "risk-emotion" dual-dimensional prevention and control system.

[0022] In a more specific technical solution, the emotional risk index at the current moment is calculated using the following logic:

[0023] In the formula, PER t This represents the intensity of the public's emotional response at that moment; E t-1 This represents the emotional risk index at the previous moment. R t The mood deterioration index can be generated in real time by LLM analysis; α , β and gamma These are calibration coefficients for the evolution stage.

[0024] In a more specific technical solution, the resource demand index is determined based on data on the intensity of emotional reactions and the trend of emotional deterioration, using the following logic. RE :

[0025] In the formula, RE This is a resource demand index. Adjustment coefficients for different stages of emotional evolution. For population density, r The radius that the resource can cover.

[0026] In a more specific technical solution, the collaborative platform for addressing physical risks of energy combustion and explosion and regulating public sentiment includes: The analysis and early warning prevention and control unit is used to acquire and preprocess real-time data from the edge layer, establish a gas combustion and explosion risk database, construct a dynamic risk identification model based on the Transformer-CNN deep learning algorithm, and conduct dynamic risk assessment; acquire and utilize the concentration data of monitoring points to model the gas diffusion behavior area, obtain the gas diffusion behavior model, apply the gas diffusion behavior model to the dynamic risk identification model, conduct dynamic early warning of regional risks, and generate prevention and control suggestions. An online incremental learning unit is used to perform database standardization operations; the gas combustion and explosion risk database and weights are automatically updated to obtain a cloud database; a genetic optimization algorithm is built based on the cloud database to detect layout optimization; the online incremental learning unit is connected to the analysis and early warning prevention unit. The Emotion Learning and Analysis Unit is used to crawl and preprocess news and public opinion data related to natural gas and oil and gas leaks and explosions in real time. Based on the relevant news and public opinion data, it uses a large language model to conduct fine-grained emotion analysis, matching and associating the severity of the accident with the intensity of emotions to obtain matching data. It then conducts emotion evolution analysis, establishes a vertical emotion evolution model for the accident, analyzes the distribution of emotion categories at different evolution stages, quantifies the duration and characteristics of each emotion evolution stage, and analyzes the trend of emotion deterioration. Based on the matching data and the trend of emotion deterioration, it performs emotion prediction and intervention. The Emotion Learning and Analysis Unit is connected to the online incremental learning unit.

[0027] The present invention has the following advantages over the prior art: This invention proposes a collaborative platform for physical risk response and public sentiment regulation based on oil and gas explosions. It breaks through the limitations of traditional safety monitoring systems that only focus on physical risks. By integrating dynamic early warning of the physical environment with the evolution analysis of public sentiment, it achieves collaborative prevention and control of "physical-social" dual-dimensional risks in natural gas / oil and gas explosions.

[0028] This invention establishes a physical-social risk control platform for oil and gas combustion and explosion through cloud-based data intelligent learning. It enables coordinated control of physical environment combustion and explosion risks ("analysis-early warning-prevention") and public sentiment ("evolution analysis-psychological intervention-resource allocation"), providing technical guidance for efficient early warning and timely response to potential combustion and explosion risks in both physical environment and public opinion dimensions.

[0029] To address the shortcomings of existing explosion and fire prevention technologies, which separate physical response from public opinion control, this invention constructs a cloud-based intelligent learning platform to achieve coordinated regulation of "physical-social" risks. Traditional systems cannot dynamically link physical risk warnings with emotional evolution analysis, leading to decision-making blind spots in emergency response. This invention, through the collaborative computation of deep learning algorithms and large language models, enables the system to simultaneously possess the ability to extrapolate physical risk chains and predict social emotional stages. It can achieve risk analysis and early warning prevention, and also generate psychological intervention plans in real time through an emotional digital twin, ultimately achieving a second-level coordinated response of physical prevention and social intervention. This transforms the traditional passive response model into a proactive "risk-emotion" dual-dimensional prevention and control system.

[0030] This invention solves the technical problem in the prior art that the inability to dynamically link physical risk warnings and emotional evolution analysis leads to decision-making blind spots in emergency response. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the basic steps of the collaborative method for responding to physical risks of energy combustion and explosion and regulating public sentiment in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of data flow processing for the collaborative method of energy combustion and explosion physical risk response and public sentiment regulation in Embodiment 1 of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. 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.

[0033] Example 1 like Figure 1 and Figure 2 As shown, the collaborative method for responding to physical risks of energy combustion and explosion and regulating public sentiment provided by this invention includes the following basic steps: S1. The risk analysis and early warning control unit preprocesses real-time data from the edge layer to establish a gas combustion and explosion risk database. A dynamic risk identification model based on the Transformer-CNN deep learning algorithm is constructed to achieve dynamic risk assessment. A gas concentration topology algorithm is built based on the TCNN inverted convolutional neural network. Gas diffusion behavior regional modeling is achieved using monitoring point concentration data, forming a gas diffusion behavior model, which is then applied to the dynamic risk identification model to achieve dynamic early warning of regional risks. Taking into account factors such as risk areas, environmental conditions, surrounding layout, and population flow, a digital twin model is constructed by introducing dynamic risk identification models, gas diffusion behavior models, and dynamic regional risk early warning systems. Through the evolutionary reasoning of the accident chain within the digital twin, the potential hazard level of the area is evaluated and high-risk sources are identified. Corresponding response suggestions are provided based on the type and location of the risk sources. Furthermore, an optimization algorithm pushes the optimal path, automatically matching the nearest fire-fighting resources to high-risk areas.

[0034] S2. Utilizing online incremental learning units and constructing a standardized cloud database based on multi-source, multi-modal data, automated batch preprocessing, storage, and retrieval of data are achieved, thereby improving computational efficiency and reducing manual intervention. When a new leak or explosion event occurs, it is promptly entered into the cloud database, and the model is retrained, automatically updating the weights of each model. A genetic optimization algorithm is constructed based on the cloud database to assist in optimizing sensor layout, maximizing monitoring coverage and efficiency.

[0035] S3. Utilizing the emotion learning and analysis unit, this system automatically crawls news and public opinion related to natural gas and oil / gas leaks and explosions using Python. It performs fine-grained analysis of public sentiment based on a large language model, quantifying emotion intensity and analyzing the correlation between accident severity and emotion. Furthermore, it analyzes the changes in emotion evolution stages, identifying potential trends and causes of deteriorating public sentiment based on the proportion of different emotion intensities within each stage. Finally, it combines digital twins with the emotion analysis process to achieve public sentiment prediction, phased psychological support, and resource allocation based on emotion prediction and intervention.

[0036] The specific implementation operations in the emotion learning and analysis unit include, but are not limited to: data collection and preprocessing, fine-grained emotion analysis, emotion evolution analysis, and emotion prediction and intervention.

[0037] Specifically, during data collection and preprocessing, Python was used to crawl online news and public opinion related to natural gas and oil and gas leaks and explosions in real time. The crawling process was divided into news report crawling and social media crawling. News report crawling was used to obtain news texts related to the accidents from relevant news websites, while social media crawling could obtain public comments on the accidents from social media websites based on accident keywords. Anonymous questionnaires could also be published on social media websites to further understand public sentiment towards the accidents. The obtained text data was preprocessed to remove advertisements, irrelevant content, and duplicate information to complete data cleaning, and then stored to provide data support for subsequent sentiment analysis.

[0038] In the process of fine-grained emotion analysis, the DeepSeek Large Language Model (LLM) is introduced to establish rules for quantifying emotion intensity. Emotional labels and intensity are extracted from the comments published by the public. At the same time, the severity of related accidents is quantified, thereby achieving a matching correlation between the severity of accidents and the intensity of emotions.

[0039] In the process of fine-grained emotion analysis, rules for quantifying emotion intensity were established. Specifically, rules for quantifying the emotion intensity of explosion accidents were constructed by combining the LIWC (Linguistic Inquiry and Word Count) emotion lexicon with the characteristics of explosion accidents, as shown in Table 1 below: Table 1. Quantification Rules for Emotion Intensity and Category

[0040] Emotional scores assigned based on each emotion category. S The intensity of the public's emotional reaction to the incident can be calculated using the following formula. PER :

[0041] in S i For the first i The sentiment scores assigned to each sample of comments. N This represents the total number of comments in the sample.

[0042] In the fine-grained emotion analysis process, emotion labels are extracted from the text. Specifically, DeepSeek's zero-shot or few-shot learning capabilities are used to perform batch processing of emotion category classification on the text. For example, the text: "The explosion was terrible, all my windows are shattered, and my hands are still shaking." is classified as "fear".

[0043] In the process of fine-grained sentiment analysis, the severity of the explosion accident is quantified. Specifically, relevant information such as the number of casualties and property damage is extracted from the accident news text, and the severity of the explosion accident is quantified according to the accident level classification method of the National Regulations on Reporting and Handling Production Safety Accidents.

[0044] In the process of fine-grained emotion analysis, the accident level is comprehensively evaluated; specifically, the severity of the accident and the intensity of public emotional response are considered. PER The system integrates objective data with public opinion to achieve a comprehensive assessment of accident levels. The comprehensive accident level assessment is shown in Table 2 below.

[0045] Table 2 Comprehensive Evaluation Table of Accident Levels

[0046] In the emotional evolution analysis of this embodiment, public comment text data from the time the accident occurred to its complete conclusion was collected. The dataset was divided into days, and public emotional categories were extracted for each day. A vertical emotional evolution model for the accident was established to analyze the distribution of emotional categories at different evolution stages, quantifying the duration and characteristics of each emotional evolution stage. Furthermore, during the emotional stage changes, through... PER The analysis of changes in the distribution of emotion categories reveals whether there is a trend of deteriorating emotions. When deterioration occurs, the comments at the current stage are statistically analyzed to determine the reasons for the deterioration.

[0047] A vertical model of emotional evolution in accidents is constructed. Specifically, explosion accidents are sudden, occur rapidly, and have a short occurrence period, often leading to a rapid outpouring of public opinion. Furthermore, the causes of gas and oil explosions are often related to human factors, and the emergency response measures of relevant government departments often guide public opinion. Therefore, the emotional evolution stage often features a period dominated by negative and angry emotions, known as the accountability phase. Based on the characteristics of explosion accidents, the emotional evolution stage is divided into four phases: shock, accountability, transition, and stability. Analysis of the emotional proportion distribution in each phase using a large language model allows for the construction of a method to quantify the characteristics of the emotional evolution stages, as shown in Table 3 below. Table 3. Quantification Methods for the Characteristics of Emotional Evolution Stages

[0048] The process involves statistically analyzing the time frame of emotional evolution stages. Specifically, it involves collecting public comment text data from the time the accident occurred until its complete resolution, dividing the dataset into daily segments, and extracting the distribution characteristics of public sentiment categories for each segment. Based on the aforementioned method for quantifying the characteristics of emotional evolution stages, changes in sentiment categories can be statistically analyzed, enabling the quantification of the duration characteristics of each emotional evolution stage in explosion accidents of different accident levels.

[0049] An analysis of the trend of worsening emotions was conducted. Specifically, the dominant emotion categories differed across different emotional stages. If, at a certain stage, the proportion of the dominant emotion category decreased while the proportion increased in the previous stage, then a trend of worsening emotions was observed. Furthermore, the intensity of public emotional reactions varied within the same stage. PER An increase in such phenomena indicates a worsening trend in emotional intensity. When two trends of emotional deterioration exist, feature extraction is performed on public comments from the current and previous days to extract high-frequency words and keywords. Keywords from the previous day are then removed from the current day's keywords to identify the keywords causing the emotional deterioration. Identifying the causes of emotional deterioration can provide guidance for intervention plans by relevant departments.

[0050] In the emotion prediction and intervention process of this embodiment, through data collection, fine-grained emotion analysis, and emotion evolution analysis, indicators of accident severity and emotional response are formed, and a post-accident public emotion database is constructed by combining basic accident information. Based on this data, an emotion digital twin is constructed to simulate the post-accident emotion evolution path and achieve emotion evolution prediction. Furthermore, based on the emotion evolution path, psychological intervention and resource allocation measures adapted to the current stage of emotion evolution are proposed.

[0051] A public sentiment database for the aftermath of an accident will be constructed. Specifically, based on the above three steps, basic accident information (accident name, time of occurrence), accident severity indicators (accident level, number of casualties, property damage) and emotional reaction indicators (intensity of public emotional reaction, distribution of emotional categories, stages of emotional evolution and their duration, and potential causes of emotional deterioration) will be linked to form a public sentiment database for the aftermath of an accident, providing data support for subsequent emotion prediction and intervention.

[0052] Real-time prediction of emotional digital twins; specifically, a visual geographic information platform is built using the Unity engine, and real-time access to physical world data streams and post-accident public sentiment databases is achieved through Apache Kafka, establishing... t Momentary Emotional Risk Index E t Taking into account both historical changes in emotional intensity and potential trends of emotional deterioration, the degree of change in emotional risk is quantified using the following formula:

[0053] in, PER t This represents the intensity of the public's emotional response at that moment; E t-1 This represents the emotional risk index at the previous moment. R t The mood deterioration index can be generated in real time by LLM analysis; α , β and gamma The calibration coefficients for the evolution stages are as follows: 0.6, 0.2, and 0.2 for the shock period; 0.3, 0.3, and 0.4 for the accountability period; and 0.3, 0.5, and 0.2 for the transition and stable periods.

[0054] Using accident level, emotional risk index, and emotional evolution stage as input accident parameters, an emotion twin based on Unity is established to dynamically extrapolate emotional evolution and form an emotional evolution prediction path. The extrapolation outputs the duration and intensity of emotional reaction, as well as the potential trend of emotional deterioration for each emotional evolution stage.

[0055] Phased psychological intervention and resource allocation will be implemented. Specifically, combining LLM (Life Principles) and based on the predicted path of emotional evolution, targeted psychological intervention plans can be generated for each stage of emotional evolution, and public emotions can be soothed through online social media. The evolution of online public sentiment lags behind the actual emotional evolution at the accident site; it can be considered that the emotional evolution at the accident site precedes the public sentiment by 3-5 hours. Based on this characteristic, a predicted path for the emotional evolution at the accident site can be constructed, generating proactive psychological intervention and resource allocation measures to provide timely and effective psychological support and resource supply to the accident site. Resource Demand Index RE Resource requirements were comprehensively considered in light of both the intensity of emotional reactions and the trend of emotional deterioration. The formula for quantifying resource requirements is as follows:

[0056] in, RE This is a resource demand index. Adjustment coefficients for different stages of emotional evolution. For population density, r The radius that the resource can cover.

[0057] Example 2 This invention provides a collaborative platform for responding to physical risks of energy combustion and explosion and regulating public sentiment, comprising an edge layer, a cloud platform, and an application layer. The edge layer performs real-time data capture and preliminary response, consisting of gas concentration sensors, temperature and humidity sensors, a multi-parameter weather station, and acoustic sensors. It collects real-time data on natural gas and oil / gas concentrations, environmental parameters (temperature, humidity, air pressure, wind speed, and wind force, etc.), and acoustic anomaly data, transmitting the data to the cloud platform. The cloud platform receives the data and performs preprocessing to achieve three functions: risk analysis and early warning prevention, online incremental learning, and sentiment learning and analysis. The application layer enables human-computer interaction and command and dispatch, providing early warning visualization, pushing early warning information, and emergency firefighting coordination.

[0058] The cloud platform described herein combines deep learning algorithms, digital twins, and large language models to achieve current risk analysis and early warning prevention, online incremental learning, and emotion learning and analysis. It serves as the intelligent analysis and decision-making hub of this patent. Its core functions consist of three units: risk analysis and early warning prevention, online incremental learning, and emotion learning and analysis.

[0059] In summary, this invention proposes a collaborative platform for physical risk response and public sentiment regulation based on oil and gas explosions. This platform breaks through the limitations of traditional safety monitoring systems that only focus on physical risks. By integrating dynamic early warning of the physical environment with the evolution analysis of public sentiment, it achieves collaborative prevention and control of "physical-social" dual-dimensional risks in natural gas / oil and gas explosions.

[0060] This invention establishes a physical-social risk control platform for oil and gas combustion and explosion through cloud-based data intelligent learning. It enables coordinated control of physical environment combustion and explosion risks ("analysis-early warning-prevention") and public sentiment ("evolution analysis-psychological intervention-resource allocation"), providing technical guidance for efficient early warning and timely response to potential combustion and explosion risks in both physical environment and public opinion dimensions.

[0061] To address the shortcomings of existing explosion and fire prevention technologies, which separate physical response from public opinion control, this invention constructs a cloud-based intelligent learning platform to achieve coordinated regulation of "physical-social" risks. Traditional systems cannot dynamically link physical risk warnings with emotional evolution analysis, leading to decision-making blind spots in emergency response. This invention, through the collaborative computation of deep learning algorithms and large language models, enables the system to simultaneously possess the ability to extrapolate physical risk chains and predict social emotional stages. It can achieve risk analysis and early warning prevention, and also generate psychological intervention plans in real time through an emotional digital twin, ultimately achieving a second-level coordinated response of physical prevention and social intervention. This transforms the traditional passive response model into a proactive "risk-emotion" dual-dimensional prevention and control system.

[0062] This invention solves the technical problem in the prior art that the inability to dynamically link physical risk warnings and emotional evolution analysis leads to decision-making blind spots in emergency response.

[0063] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. Energy fuel explosion physical risk response - public emotion regulation synergy method, characterized by, The method comprises: S1, using an analysis and early warning prevention unit, acquiring and preprocessing edge layer real-time data, establishing a gas explosion risk database, constructing a dynamic risk identification model based on a Transformer-CNN deep learning algorithm, and performing dynamic risk assessment; obtaining and using monitoring point concentration data to model a gas diffusion behavior region to obtain a gas diffusion behavior model, applying the gas diffusion behavior model to the dynamic risk identification model, and performing regional risk dynamic early warning; and generating a prevention and control proposal; S2, using an online incremental learning unit, performing database standardization operations; automatically updating the gas explosion risk database and weights to obtain a cloud database; constructing a genetic optimization algorithm based on the cloud database to detect layout optimization; S3, using an emotion learning and analysis unit, real-time crawling and preprocessing related news and public opinion data of natural gas, oil and gas leakage and explosion; using a large language model to perform fine-grained emotion analysis based on the related news and public opinion data, match the severity of the accident and the intensity of the emotion, obtain matching data, perform emotion evolution analysis, establish an accident vertical emotion evolution model, analyze the proportion distribution of emotion categories at different evolution stages, quantify the duration and emotion category characteristics of each emotion evolution stage, and analyze the emotion deterioration trend; and performing emotion prediction and intervention based on the matching data and the emotion deterioration trend.

2. The energy pyrophysical risk response-public sentiment regulation synergic method according to claim 1, characterized in that, In S3, a DeepSeek large language model LLM is introduced to perform fine-grained emotion analysis; wherein emotion intensity quantification rules are established to extract emotion labels and intensity from public comment texts, and the severity of related accidents is quantified, the severity of the accident is associated with the intensity of the emotion, the matching data is obtained, and the public emotion response intensity is processed.

3. The energy fuel explosion physical risk response-public emotion regulation synergic method according to claim 1, characterized in that, In the S3, the public emotional reaction intensity of the current incident is calculated PER ; Through the zero-shot or few-shot learning ability of DeepSeek, the comment texts are batch-processed for emotion category classification, and text emotion labels are extracted; In the accident news text, the objective data of the accident is extracted, the severity of the explosion accident is quantified according to the preset division rule, and the severity of the accident is obtained; correlating the accident severity with the public emotional reaction intensity PER to perform an accident level comprehensive evaluation; According to the accident level, the emotion risk index, and the emotion evolution stage, an emotion twin based on Unity is established to realize dynamic deduction of emotion evolution and form an emotion evolution prediction path; Combined with the large language model LLM and based on the emotion evolution prediction path, stage-based psychological intervention and resource scheduling are performed to generate a targeted psychological intervention scheme for each emotion evolution stage, and the public emotion is soothed through network social media.

4. The energy fuel explosion physical risk response-public emotion regulation synergic method according to claim 3, characterized in that, According to the emotional scores of each emotional category S , the public emotional reaction intensity of the current incident is calculated using the following logic PER : In the formula, S i For the first i The emotional score of the sample of comments, N The total number of samples of comments.

5. The energy pyrophysical risk response-public sentiment regulation synergic method according to claim 1, characterized in that, In the emotion evolution analysis process of S3, public comment text data is collected, data sets are divided according to preset time units, and public emotion categories are extracted respectively; An accident vertical emotion evolution model is established to analyze the proportion distribution of emotion categories at different evolution stages, quantify the duration and emotion category characteristics of each emotion evolution stage, and analyze the emotion deterioration trend. And in the process of emotional stage change, obtain and according to the public emotional response intensity PER And the change information of the proportion distribution of emotion categories, analyze the emotion deterioration trend and the emotion deterioration reason.

6. The energy fuel explosion physical risk response-public emotion regulation synergic method according to claim 1, characterized in that, In the S3, mood prediction and intervention are performed; according to the matching data and the mood deterioration trend, an accident severity index and a mood reaction index are formed, and an accident post-public mood database is constructed in combination with accident basic information; based on the accident post-public mood database, a mood digital twin is constructed, an accident post-mood evolution path is simulated and obtained, and mood evolution prediction is performed.

7. The energy fuel explosion physical risk response-public emotion regulation synergic method according to claim 1, characterized in that, In the S3, real-time prediction is performed using the emotional digital twin; a visual geographic information platform is built using the Unity engine, physical world data streams and post-accident public emotion databases are accessed in real time through Apache Kafka, and a real-time emotional risk index is established t Real-time emotional risk index E t The emotional risk change degree is quantified by comprehensively considering the historical emotional intensity change and the emotional deterioration trend.

8. The energy fuel explosion physical risk response-public emotion regulation synergic method according to claim 6, characterized in that, The mood risk index at the current time is calculated by using the following logic: In the formula, PER t is the public emotional response intensity at this moment; E t-1 is the emotional risk index at the previous moment; R t For the Emotion Deterioration Index, the LLM Real-Time Analysis can generate; α 、 β and γ are the evolution phase calibration coefficients.

9. The energy fuel explosion physical risk response-public emotion regulation synergic method according to claim 6, characterized in that, According to the emotional response intensity, the emotional deterioration trend change data, using the following logic, determine the resource demand index RE : In the formula, RE is a resource demand index, is a mood evolution stage correction coefficient, is a population density, r is a resource coverable radius.

10. An energy combustion physical risk response - public sentiment regulation collaborative platform, characterized in that, The platform comprises: An analysis and early warning prevention and control unit, which is configured to acquire and pre-process edge layer real-time data, establish a gas explosion risk database, construct a dynamic risk identification model based on a Transformer-CNN deep learning algorithm, and perform dynamic risk assessment; acquire and utilize monitoring point concentration data to model a gas diffusion behavior region, obtain a gas diffusion behavior model, apply the gas diffusion behavior model to the dynamic risk identification model, and perform regional risk dynamic early warning; and generate a prevention and control suggestion scheme; An online incremental learning unit, which is configured to perform database standardization operations; automatically update the gas explosion risk database and weights to obtain a cloud database; construct a genetic optimization algorithm based on the cloud database to detect layout optimization; and the online incremental learning unit is connected to the analysis and early warning prevention and control unit; An emotion learning and analysis unit, which is configured to real-time crawl and pre-process related news and public opinion data of natural gas, oil and gas leakage and explosion; according to the related news and public opinion data, utilize a large language model to perform fine-grained emotion analysis, match and associate accident severity and emotion intensity to obtain matching data; perform emotion evolution analysis, establish an accident vertical emotion evolution model, analyze emotion category proportion distribution at different evolution stages, quantify the duration of each emotion evolution stage and emotion category characteristics, and analyze to obtain an emotion deterioration trend; and according to the matching data and the emotion deterioration trend, perform mood prediction and intervention; and the emotion learning and analysis unit is connected to the online incremental learning unit.

Images