Fire risk assessment method based on Markov chain model

Abstract

The invention relates to the field of fire risk assessment, and discloses a Markov chain model-based fire risk assessment method, which comprises the following steps of: defining a fire state set S, constructing a fire state transition matrix P according to historical fire data, and calculating steady-state distribution # imgabs0 # of each fire state by using the fire state transition matrix P, according to the distribution # imgabs1 # of the current fire state, the fire state distribution # imgabs2 # of the next moment is predicted, the fire state changes of multiple moments in the future are obtained after multiple times of repeated prediction, the expected value # imgabs4 # of the fire risk is defined for each fire state # imgabs3 #, the total expected value R of the fire risk is calculated, the future fire occurrence situation is simulated for multiple times through the Monte Carlo method, and the fire state changes of multiple moments in the future are obtained. Evaluating a future risk scene; and obtaining probability distribution of states at different moments in the future according to a fire occurrence condition simulation result, and providing a basis for a fireproof plan. Real-time assessment of fire risks is realized through dynamic modeling, accuracy and scientificity of assessment are improved, and fire prevention and emergency management decisions are effectively supported.

Description

Technical Field

[0001] The invention relates to the technical field of fire risk assessment, in particular to a fire risk assessment method based on a Markov chain model. Background Art

[0002] With the acceleration of urbanization and the improvement of people's living standards, the frequency and destructiveness of fires have gradually increased, bringing huge property losses and personal safety risks to society. According to relevant statistics, the economic losses caused by fire accidents each year are as high as billions of RMB, and countless lives have also been lost. This situation has prompted researchers and decision makers to find more scientific and effective fire risk assessment methods.

[0003] Traditional fire risk assessment methods mainly rely on static models and empirical judgments. These methods are often based on historical data summaries and lack real-time response to the dynamic changes of current fires, which leads to assessment bias in complex fire situations and limits the timeliness and accuracy of emergency response and resource scheduling. For example, static data analysis usually cannot consider the complex relationship between multiple factors, and empirical judgments may be affected by individual subjective factors, leading to misjudgments.

[0004] In recent years, with the development of computer science and data analysis technology, the use of mathematical models for dynamic assessment of fire risk has gradually become a research hotspot. Through probabilistic models and dynamic models, the randomness and uncertainty in the fire process can be effectively captured. In this regard, the Markov chain model has received widespread attention due to its "memorylessness" and dynamic modeling capabilities. The Markov chain can predict the future state through the current state without considering past historical records, which makes it very suitable for the time-dynamic assessment of fire risk.

[0005] However, current research on Markov chains in fire risk assessment is still relatively limited, lacking a systematic risk assessment method, which makes it difficult to achieve accurate and real-time scheduling of fire management and resource allocation. Summary of the invention

[0006] In view of the shortcomings of the existing methods, the present invention provides a fire risk assessment method based on the Markov chain model. By introducing the Markov chain model to establish a transition probability matrix of the fire state, the dynamic prediction of the fire risk can be achieved, providing a more scientific decision-making basis for fire management.

[0007] To achieve the above object, the technical solution adopted by the present invention is as follows: A fire risk assessment method based on a Markov chain model comprises the following steps: Step 1: Define the fire state set S = {S0, S1, S2, S3}: where S0 is the no-fire state, S1 is the initial fire state, S2 is the fire spread state, and S3 is the fire extinguishing state; Step 2: Construct the fire state transfer matrix P based on historical fire data, and use the fire state transfer matrix P to calculate the steady-state distribution of each fire state , according to the distribution of current fire status Predict the fire status distribution at the next moment , introduce dynamic environmental variables and building material characteristics, further adjust the transfer probability, and obtain a more accurate state prediction model. After repeated predictions, the changes in fire states at multiple moments in the future can be obtained; Step 3: For each fire state , defining the expected value of fire risk , calculate the total fire risk expectation value R, simulate future fire occurrences multiple times through the Monte Carlo method, and optimize the transition probability by combining deep reinforcement learning, so that the model can automatically learn the optimal fire state prediction through historical data and real-time data, and further optimize the risk assessment results to evaluate future risk scenarios; Step 4: Based on the simulation results of the fire situation, combined with the use function of the building and the risk situation of the surrounding environment, dynamically adjust the resource allocation strategy according to different fire scenarios and the evolution prediction of the fire status.

[0008] Step 2 specifically includes the following steps: Step 201: Construct a fire state transfer matrix P based on historical fire data, structural characteristics of the building and environmental factors: ; Set the basic state transition probability By adjusting the calculation method of the state transfer matrix, the dynamic factors are integrated into the transfer matrix P, combined with the dynamic environment impact function , make adjustments; The expression is: ,in It is the comprehensive influence function of dynamic factors, and the function form adopts nonlinear combination: , a new P matrix is ​​obtained through continuous iterative updates; Among them is , , , All are correction factors, T is temperature, W is wind speed, H is humidity, and Z is material parameter; Each element in the matrix Indicates from the state Transfer to state The probability of, and the sum of the elements in each row of the matrix is ​​equal to 1, that is, the sum of the probabilities of each state is equal to 1; Step 202: By solving the equation Get the steady-state distribution , through the constraints ,in ,satisfy , get the long-term stability probability of each fire state; Step 203: Using the known current fire status distribution And the state transfer matrix P is used to predict the state at the next moment. Through matrix operation, the state distribution at a certain moment in the future is obtained. By repeating this process multiple times, the changes in fire risk at multiple future moments can be predicted. The formula is: .

[0009] Step 3 specifically includes the following steps: Step 301: For each fire state , defining the expected value of fire risk : ; in Yes Status potential losses or impacts caused; Step 302: Expected value of fire risk for each state Perform weighted calculation to obtain the weighted sum R of the total fire risk expectation value. ; Its weight is the steady-state distribution probability of each fire state, and an index reflecting the overall fire risk level is obtained; Step 303: Using the Monte Carlo method, multiple simulations are performed on future fire situations using the current fire state distribution and the state transition matrix P as random sampling parameters, and the state sequence generated by each simulation and the probability distribution of each fire state at each moment are output to obtain the probability of different fire states occurring at different moments in the future. Step 304: By introducing a correction factor for post-disaster recovery capability, the recovery process is modeled in a Monte Carlo simulation to evaluate the impact of the fire on the recovery process and further improve the risk assessment results.

[0010] In step 4, specifically: Step 401: Based on the fire status dynamic assessment result obtained in step 3, combined with the use function of the building and the risk situation of the surrounding environment, a resource allocation matrix M for formulating a resource allocation strategy is established, and the matrix includes the quantity or type of resources required under different fire states; Step 402: Optimize resource allocation according to different fire occurrence scenarios, and dynamically adjust resource allocation strategies according to the evolution prediction of the fire status, thereby improving the efficiency of fire fighting and post-disaster recovery.

[0011] The present invention has the following beneficial effects and advantages: The fire risk assessment method based on the Markov chain model of the present invention utilizes the "memoryless" feature of the Markov chain model to divide the fire state into multiple stages, estimates the transition probability between each state through historical fire data, and finally constructs a state transition matrix. The matrix is ​​used to predict the long-term steady-state distribution of each fire state, thereby providing a scientific basis for the dynamic assessment of fire risk; real-time assessment of fire risk is achieved through dynamic modeling, the accuracy and scientificity of the assessment are improved, and fire prevention and emergency management decisions are effectively supported. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 The present invention is a flow chart of a fire risk assessment method based on a Markov chain model; Figure 2 The state transfer matrix and its steady-state distribution solution process of the present invention; Figure 3 It is a schematic diagram of the fire state transition process of the present invention, illustrating the transition paths between states such as "no fire", "early fire", "fire spread", and "fire extinguished". DETAILED DESCRIPTION

[0013] The present invention is described in detail below with reference to the accompanying drawings and specific examples.

[0014] The present invention provides a fire risk assessment method based on a Markov chain model. Figure 1 As shown, the following steps are included: Step 1: Define the fire state set S = {S0, S1, S2, S3}: where S0 is the no-fire state, S1 is the initial fire state, S2 is the fire spread state, and S3 is the fire extinguishing state; Step 2: Construct the fire state transfer matrix P based on historical fire data, and use the fire state transfer matrix P to calculate the steady-state distribution of each fire state , according to the distribution of current fire status Predict the fire status distribution at the next moment , dynamic environmental variables and building material properties are introduced to further adjust the transfer probability to obtain a more accurate state prediction model. After repeated predictions, the changes in fire states at multiple moments in the future are obtained.

[0015] Further, step 201: construct a fire state transfer matrix P according to historical fire data: ; ; In mathematical modeling, by adjusting the calculation method of the state transfer matrix, the dynamic factors (temperature T, wind speed W, humidity H, material properties W) are integrated into the transfer matrix P, and the basic state transfer probability is set. :

[0016] Combined with dynamic environmental impact function , make adjustments, specifically as follows: ,in It is the comprehensive influence function of dynamic factors, and the function form adopts nonlinear combination: , a new P matrix is ​​obtained through continuous iterative updates; Among them is , , , All of them are correction factors, T is temperature, W is wind speed, H is humidity, and Z is material parameters; material parameters include material burning rate, fire resistance level, and building material area. The above parameters are optimized through historical data fitting and machine learning algorithms.

[0017] Each element in the matrix Indicates from the state Transfer to state The probability of, and the sum of the elements in each row of the matrix is ​​equal to 1, that is, the sum of the probabilities of each state is equal to 1; The state allocation of this embodiment is as follows: In this matrix, the transition probability of each state is specifically defined. For example, in the "no fire" state, there is a 20% probability of entering the "early fire" state and an 80% probability of maintaining the "no fire" state; in the "early fire" state, there is a 70% probability of maintaining the status quo and a 30% probability of entering the "fire spread" state. This matrix is ​​obtained by analyzing historical fire data statistics and reflects the dynamic changes between various fire states.

[0018] ; ; Step 202: By solving the equation Get the steady-state distribution , through the constraints ,in , satisfy , and obtain the long-term stable probability of each fire state; by introducing environmental influencing parameters such as temperature, wind speed, and humidity, the transfer matrix P is dynamically adjusted to simulate the spread and extinguishing process of fire under different environmental conditions; The system of equations can be expanded as: ; This equation indicates that the state probability distribution of the system does not change after long-term operation.

[0019] Step 203: Using the known current fire status distribution And the state transfer matrix P is used to predict the state at the next moment. Through matrix operation, the state distribution at a certain moment in the future is obtained. By repeating this process multiple times, the changes in fire risk at multiple future moments can be predicted. The formula is: .

[0020] Step 3: For each fire state , defining the expected value of fire risk , calculate the total fire risk expectation R, simulate future fire situations multiple times through the Monte Carlo method, and evaluate future risk scenarios; based on the Markov chain state transfer, combine deep reinforcement learning to optimize the transfer probability, so that the model can automatically learn the optimal fire state prediction through historical data and real-time data. Algorithm selection: A reinforcement learning model based on the Actor-Critic framework, which optimizes environmental feedback (loss after fire state transfer) as a reward function. Deep reinforcement learning can dynamically adjust and optimize the state transfer matrix in real time, , where is the influence function of sensor data. Compared with models that rely solely on historical data, this real-time data embedding method can significantly improve the timeliness and prediction accuracy of the model, providing a more scientific basis for rapid fire response and optimal resource allocation.

[0021] Further, step 301: for each fire state , defining the expected value of fire risk : ; in Yes Status Potential loss caused; risk expectation Reflects the state The following losses may result.

[0022] By multiplying the state transition probability and the potential loss of each state, the fire risk in each state can be quantified. This assessment enables decision makers to identify the potential impacts under different fire states and develop targeted preventive measures.

[0023] Step 302: Expected value of fire risk for each state Perform weighted calculation to obtain the weighted sum R of the total fire risk expectation value. ; Its weight is the steady-state distribution probability of each fire state, and an indicator reflecting the overall fire risk level is obtained. Through this comprehensive evaluation, an indicator reflecting the overall fire risk level can be obtained to help decision makers fully understand the fire risk.

[0024] Step 303: Using the Monte Carlo method, using the current fire state distribution The state transfer matrix P is used as a random sampling parameter to simulate the occurrence of future fires multiple times, and the state sequence generated by each simulation and the probability distribution of each fire state at each moment are output to obtain the probability of different fire states occurring at different moments in the future. This method can effectively capture the randomness of fire risks and provide more comprehensive analysis results.

[0025] Step 304: By introducing the correction factor of post-disaster recovery capacity, the recovery process is modeled in the Monte Carlo simulation. In the fire risk assessment model, embedding the real-time data collected by the Internet of Things (IoT) sensors into the model can significantly improve the real-time and accuracy of the prediction. By deploying edge computing devices, key parameters (temperature T, wind speed W, humidity H, material parameter Z) are collected in real time. After preprocessing, the state transition matrix P can be dynamically updated. Specifically, the time series prediction model (Transformer) is used to process these real-time data and dynamically adjust the state transition probability , expressed as As a dynamic adjustment function, this method can dynamically reflect the occurrence and development process of fire, make up for the shortcomings of static historical data models, and provide more accurate and efficient support for fire prediction, rapid response and optimal resource allocation.

[0026] Step 4: The dynamic fire status evaluation results obtained in step 3 are used to establish a resource allocation matrix M for formulating resource allocation strategies. M includes the number or type of resources required under different fire conditions to optimize resource allocation when a fire occurs.

[0027] Step 401: Based on the fire status dynamic assessment results obtained in step 3, combined with the use function of the building and the risk situation of the surrounding environment, the building repair time and economic loss recovery time after the fire are considered. Factors are taken into consideration to establish a dynamic evaluation model for post-disaster recovery. Recovery capacity can be quantified as the recovery rate. , ; ; In fire state economic losses or resource consumption, is the recovery time. In the Monte Carlo simulation process, the recovery rate is embedded in the dynamic model, so that the evaluation results not only reflect the destructive process of the fire, but also predict the time and resources required for recovery, and establish a resource allocation matrix M for formulating resource allocation strategies, which includes the number or type of resources required under different fire conditions; Step 402: Optimize resource allocation according to different fire occurrence scenarios, and dynamically adjust resource allocation strategies according to the evolution prediction of the fire status, thereby improving the efficiency of fire fighting and post-disaster recovery.

[0028] The solution of steady-state distribution can provide a scientific basis for the long-term assessment of fire risk. For example, in practical applications, by calculating the steady-state probability of each state, the long-term risk level of a certain area under different fire conditions can be determined, so that resource allocation and emergency preparation can be carried out in a targeted manner.

[0029] Dynamic assessment of fire risk, based on the obtained steady-state distribution and state transfer matrix, can be further carried out. Markov chain model is used to predict the possibility of future fire state and formulate prevention and control strategies accordingly.

[0030] The specific steps are as follows: Based on historical data and current status, the state transition matrix P is used to predict the fire state that may occur at the next moment; the process is repeated to predict the state changes at multiple moments in the future; based on the distribution of fire status at each moment, high-risk time periods and states are identified to formulate prevention and control strategies in advance.

[0031] Taking a certain urban area as an example, historical fire data shows that: In the "no fire" state, there is a 20% probability of entering the "early stage of fire" state; in the "early stage of fire" state, there is a 70% probability of maintaining the current state and a 30% probability of entering the "fire spreading" state.

[0032] Based on the above data, the state transfer matrix P was constructed, and Monte Carlo simulation was used to perform 1,000 simulations to evaluate future fire risk scenarios. Through multiple simulations, the probability distribution of each state at different times in the future was obtained, thereby predicting the distribution of future fire risk states and providing a basis for fire prevention plans. By analyzing the simulation results, managers can identify high-risk time periods and states, and formulate targeted fire prevention measures and emergency plans accordingly. For example, if the simulation shows that the probability of the "fire spread" state increases significantly during a certain period of time, relevant departments can deploy more resources in advance to deal with potential fires.

[0033] By calculating the long-term steady-state distribution, the long-term risk of each state is analyzed in the absence of effective control measures. For example, if the steady-state distribution π=[0.5, 0.3, 0.1, 0.1], the long-term probability of the "fire spread" state is 10%, suggesting that measures need to be taken to reduce the probability of this state.

[0034] Combined with the dynamic assessment results of fire risk, the allocation of fire protection resources can be optimized. Set the resource allocation matrix M: .

[0035] By analyzing the long-term steady-state distribution, it is found that the long-term probability of the "fire spread" state will increase significantly without effective prevention and control measures. Based on this result, the existing fire resource allocation and emergency strategy can be optimized to reduce the long-term fire risk.

[0036] The present invention proposes a dynamic fire risk assessment method based on Markov chain, which can effectively reflect the dynamic changes of fire risk and improve the accuracy and real-time performance of fire assessment. This method can not only provide a scientific basis for fire prevention and control strategies and emergency management, but also optimize the resource allocation and emergency plans of fire management in a complex and changing environment through data-driven methods. By combining with modern data analysis technologies, such as big data analysis and machine learning, the intelligent level of fire risk assessment can be further improved, and it has broad application prospects and practical value.

[0037] This method is expected to be widely used in urban fire management, public safety emergency response and other fields. With the development of data collection and analysis technology, the fire risk assessment model based on Markov chain will continue to improve and provide more comprehensive and accurate support for urban safety management.

Claims

1. A fire risk assessment method based on a Markov chain model, characterized in that: The following steps are involved: Step 1: Define the fire state set S = {S0, S1, S2, S3}: where S0 is the no-fire state, S1 is the initial fire state, S2 is the fire spread state, and S3 is the fire extinguishing state; Step 2: Construct the fire state transfer matrix P based on historical fire data, and use the fire state transfer matrix P to calculate the steady-state distribution of each fire state , according to the distribution of current fire status Predict the fire status distribution at the next moment , introduce dynamic environmental variables and building material characteristics, further adjust the transfer probability, and obtain a more accurate state prediction model. After repeated predictions, the changes in fire states at multiple moments in the future can be obtained; Step 3: For each fire state , defining the expected value of fire risk , calculate the total fire risk expectation value R, simulate future fire occurrences multiple times through the Monte Carlo method, and optimize the transition probability by combining deep reinforcement learning, so that the model can automatically learn the optimal fire state prediction through historical data and real-time data, and further optimize the risk assessment results to evaluate future risk scenarios; Step 4: Based on the simulation results of the fire situation, combined with the use function of the building and the risk situation of the surrounding environment, dynamically adjust the resource allocation strategy according to different fire scenarios and the evolution prediction of the fire status.

2. A fire risk assessment method based on a Markov chain model according to claim 1, characterized in that Step 2 specifically includes the following steps: Step 201: Construct a fire state transfer matrix P based on historical fire data, structural characteristics of the building and environmental factors: ; Set the basic state transition probability By adjusting the calculation method of the state transfer matrix, the dynamic factors are integrated into the transfer matrix P, combined with the dynamic environment impact function , make adjustments; The expression is: ,in It is the comprehensive influence function of dynamic factors, and the function form adopts nonlinear combination: , a new P matrix is ​​obtained through continuous iterative updates; Among them is , , , All are correction factors, T is temperature, W is wind speed, H is humidity, and Z is material parameter; Each element in the matrix Indicates from the state Transfer to state The probability of, and the sum of the elements in each row of the matrix is ​​equal to 1, that is, the sum of the probabilities of each state is equal to 1; Step 202: By solving the equation Get the steady-state distribution , through the constraints ,in ,satisfy , get the long-term stability probability of each fire state; Step 203: Using the known current fire status distribution And the state transfer matrix P is used to predict the state at the next moment. Through matrix operation, the state distribution at a certain moment in the future is obtained. By repeating this process multiple times, the changes in fire risk at multiple future moments can be predicted. The formula is: .

3. A fire risk assessment method based on a Markov chain model according to claim 1, characterized in that Step 3 specifically includes the following steps: Step 301: For each fire state , defining the expected value of fire risk : ； in Yes Status potential losses or impacts caused; Step 302: Expected value of fire risk for each state Perform weighted calculation to obtain the weighted sum R of the total fire risk expectation value. ; Its weight is the steady-state distribution probability of each fire state, and an index reflecting the overall fire risk level is obtained; Step 303: Using the Monte Carlo method, multiple simulations are performed on future fire situations using the current fire state distribution and the state transition matrix P as random sampling parameters, and the state sequence generated by each simulation and the probability distribution of each fire state at each moment are output to obtain the probability of different fire states occurring at different moments in the future. Step 304: By introducing a correction factor for post-disaster recovery capability, the recovery process is modeled in a Monte Carlo simulation to evaluate the impact of the fire on the recovery process and further improve the risk assessment results.

4. A fire risk assessment method based on a Markov chain model according to claim 1, characterized in that In step 4, specifically: Step 401: Based on the fire status dynamic assessment result obtained in step 3, combined with the use function of the building and the risk situation of the surrounding environment, a resource allocation matrix M for formulating a resource allocation strategy is established, and the matrix includes the quantity or type of resources required under different fire states; Step 402: Optimize resource allocation according to different fire occurrence scenarios, and dynamically adjust resource allocation strategies according to the evolution prediction of the fire status, thereby improving the efficiency of fire fighting and post-disaster recovery.

Images