Fire-fighting equipment material optimization scheduling method based on big data

Abstract

The invention relates to the technical field of fire fighting, and discloses a big data-based fire fighting equipment material optimization scheduling method, which comprises the following steps of: 1, acquiring temperature, voltage and accumulated working hour data of fire fighting equipment, and acquiring real-time wind speed, humidity and temperature data of a meteorological department as well as building density, road connectivity and historical fire point coordinates in geographic information; and step 2, predicting a future fire risk probability according to the acquired meteorological data, geographic information and historical fire point coordinates, and meanwhile, based on temperature, voltage and accumulated man-hour data of fire-fighting equipment. According to the method, a dynamic risk prediction and cross-regional collaborative pre-deployment technical scheme is adopted, the fire risk probability is predicted by fusing meteorological, geographic and historical disaster data in real time, and the pre-deployment scheme is generated based on the risk level and the equipment health state, so that the technical effects of improving the high-risk region resource coverage precision and optimizing the cross-regional collaborative efficiency are achieved.

Description

Technical Field

[0001] The present invention relates to the field of firefighting technology, and in particular to a method for optimizing the dispatch of firefighting equipment and materials based on big data. Background Art

[0002] Firefighting equipment and material dispatch is the process of scientifically allocating resources such as fire trucks, rescue equipment, and fire-fighting equipment through an intelligent system. With the help of the Internet of Things, big data, and artificial intelligence technologies, it collects equipment status, traffic conditions, and disaster data in real time, uses predictive models to evaluate needs, and automatically generates the best dispatch plan through optimization algorithms. It also supports dynamic adjustment, significantly improving response speed, resource utilization, and rescue success rate, realizing the transformation from traditional experience-based decision-making to data-driven intelligent decision-making, and providing efficient and reliable guarantees for urban emergency management.

[0003] For example, the existing Chinese patent publication number CN118569448B discloses a method for optimizing the scheduling of firefighting equipment and materials. It adopts a multi-objective optimization model, integrates genetic algorithms, and balances factors such as time, equipment utilization, and failure rate to achieve efficient scheduling. By integrating multiple real-time data sources, such as sensors, traffic, and fire alarm information, in-depth analysis is performed to ensure that the scheduling plan can dynamically adapt to emergency situations. Combining AI and machine learning, it intelligently analyzes historical data, identifies scheduling problems and provides suggestions, reduces human decision-making errors, realizes digital management of the entire life cycle of equipment, accurately predicts maintenance and replacement cycles, improves efficiency and safety, and introduces dynamic optimization and self-learning mechanisms to enable the system to continuously adjust and improve itself to adapt to changes in various scenarios and task requirements, ensuring that the accuracy and efficiency of scheduling decisions continue to improve.

[0004] For example, the existing Chinese patent with publication number CN109865231B discloses a movable modular intelligent fire-fighting duty support equipment method and related products, including: obtaining liquid level sensor data in the target fire truck through a liquid level sensor, the position acquisition module obtains the current first position of the target fire truck, and the fire cloud platform obtains the operating status of the target fire truck. When the liquid level sensor data meets the preset conditions, the dispatching strategy of the target fire truck is determined according to the operating status and the first position. In this way, when the liquid level sensor data meets the preset conditions, the corresponding dispatching strategy can be determined according to the position information and operating status of the fire truck, thereby realizing unified monitoring and dispatching of fire trucks in the fire station, and carrying out fire rescue and fire protection more intelligently and efficiently.

[0005] The shortcomings of the above patents are:

[0006] On the one hand, traditional firefighting material dispatch relies primarily on historical data and static rules, lacking the ability to dynamically predict impending fire zones. Furthermore, it is unable to promptly deploy equipment to the borders of cities and provinces adjacent to the fire zone. This results in a slow response to large-scale fires, which in turn reduces firefighting efficiency.

[0007] On the other hand, current firefighting equipment dispatch is primarily based on material type and inventory quantity, while ignoring key information such as the equipment's actual condition, including its age, maintenance intervals, remaining battery life, and frequency of use. This neglected information can lead to substandard and near-endurance equipment being mistakenly dispatched to the front lines, delaying the optimal fire response and even causing secondary fires.

[0008] To this end, the present invention proposes a firefighting equipment and material optimization scheduling method based on big data to solve the above-mentioned problems. Summary of the Invention

[0009] In view of the shortcomings of the existing technology, the present invention provides a method for optimizing the dispatch of firefighting equipment and materials based on big data to solve the problems raised in the above background technology.

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for optimizing the dispatch of firefighting equipment and materials based on big data, comprising:

[0011] Step 1: Collect the temperature, voltage, and accumulated working hours data of firefighting equipment, and obtain real-time wind speed, humidity, and temperature data from the meteorological department, as well as building density, road connectivity, and historical fire point coordinates from geographic information;

[0012] Step 2: Based on the collected meteorological data, geographic information, and historical fire point coordinates, the probability of future fire risks is predicted. Based on the temperature, voltage, and accumulated working hours data of the firefighting equipment, the equipment health score is calculated and classified as usable, warning, or disabled.

[0013] Step 3: Generate a cross-regional firefighting equipment pre-deployment plan based on the predicted fire risk probability and equipment health score;

[0014] Step 4: Generate dynamic rescue routes and equipment allocation instructions based on the generated pre-deployment plan, combined with real-time traffic conditions and dangerous area coordinates;

[0015] Step 5: Based on the actual equipment usage data and path execution results, update the fire risk prediction model and equipment health scoring model, and generate equipment maintenance and retirement instructions.

[0016] Preferably, in step 1, collecting firefighting equipment status data and disaster risk data further includes:

[0017] Sub-step 1.1, firefighting equipment sensor data collection and preprocessing:

[0018] Deploy temperature sensors, voltage sensors and working time counters in fire fighting equipment to collect temperature T, voltage V, and accumulated working time t use , calculation formula:

[0019]

[0020] Among them, T i is the i-th temperature sampling value, N is the number of sampling times per unit time, M is the number of sampling times per unit time, V j is the jth voltage sampling value, δ active (t) is the working status of the equipment at time t;

[0021] Sub-step 1.2, Disaster risk data integration and spatiotemporal alignment:

[0022] Obtain real-time wind speed v, humidity h, and ambient temperature T from the meteorological department interface env , and extract building density ρ, road connectivity c, historical fire point coordinates (x k ,y k ), calculation formula:

[0023]

[0024] (x k ,y k ) = GPS coordinates of the kth fire,

[0025] Data time alignment: Linear interpolation is used to compensate for time deviation of asynchronous data:

[0026]

[0027] Among them, v(t) is the real-time wind speed value at time t, t is the current timestamp, v(t i ) is the discrete time point t i Observed wind speed value, t i is the time point of data collection for the i-th time, v(t i+1 ) is the discrete time point t i+1 Observed wind speed value, t i+1 is the time point of the i+1th data collection;

[0028] Sub-step 1.3, multi-source data fusion and feature storage:

[0029] After aligning the equipment data and disaster data by timestamp, a spatiotemporal fusion data block is generated and stored in the spatiotemporal database. The index fields are time t and coordinates (x, y). The calculation formula is:

[0030] D fusion ={T,V,t use ,v,h,ρ,c,x k ,y k},

[0031] Among them, D fusion is the multi-source fusion data block after spatiotemporal alignment, x k 、y k are the latitude and longitude coordinates of the kth historical fire point.

[0032] Preferably, in step 2, predicting the fire risk probability and calculating the equipment health score further includes:

[0033] Sub-step 2.1: Construction of multi-source disaster risk prediction model:

[0034] Input the meteorological data collected in step 1 [wind speed v, humidity h, ambient temperature T env ] and geographic information [building density ρ, road connectivity c, historical fire point coordinates (x k ,y k )], construct spatiotemporal fusion features:

[0035] F risk =LSTM(v,h,T env )⊕GCN(ρ,c,x k ,y k ),

[0036] Among them, F risk LSTM is a long short-term memory network that extracts meteorological time series features, GCN is a graph convolutional network that models geographic spatial associations, and ⊕ is a feature concatenation operation.

[0037] Output regional risk probability:

[0038] P risk (x,y,t)=σ(W h ·F risk +b h ),

[0039] Among them, σ is the Sigmoid function, W h 、b h are trainable weights and biases, P risk is the fire risk probability of area (x, y) at time t;

[0040] Sub-step 2.2, calculation of fire equipment health score:

[0041] Input the equipment temperature T, voltage V, and accumulated working hours t collected in step 1 use , calculate the health score:

[0042]

[0043] Among them, S health is the health score, α and β are the weights of lifespan factor and state deviation factor, t total is the equipment design life, T normal 、V normal The temperature and voltage reference values when the equipment is working normally;

[0044] Health score categories:

[0045] Available: S health ≥80;

[0046] Warning: 60≤S health <80;

[0047] Disable: S health <60;

[0048] Sub-step 2.3, risk-health data fusion and classification output:

[0049] The risk probability P in sub-step 2.1 is risk The health score S in substep 2.2 health Generate a risk-health matrix by region association:

[0050] M(x,y)=[P risk (x,y,t),max(S health (x,y))],

[0051] Among them, M(x,y) is the correlation matrix between the risk probability of region (x,y) and the equipment health score, P risk (x,y,t) is the probability of fire occurring in area (x,y) at time t, S health (x,y) is the maximum health score of the equipment available in the area (x,y);

[0052] Output classification results:

[0053] High-risk areas: P risk ≥0.7 and max(S health )≥80;

[0054] Medium-risk area: 0.4≤P risk <0.7 or max(S health )<80;

[0055] Low-risk area: P risk <0.4.

[0056] Preferably, in step 3, generating a cross-regional firefighting equipment pre-deployment plan further includes:

[0057] Sub-step 3.1, regional risk prioritization:

[0058] Input the fire risk probability P predicted in step 2 risk (x,y,t), prioritize regions by threshold:

[0059]

[0060] Output target region set R = {(x,y)|P risk (x,y,t)≥0.4};

[0061] Sub-step 3.2: Health equipment screening and demand matching:

[0062] Enter the equipment health score S calculated in step 2 health , filter available equipment [S health ≥80], matching regional requirements by type:

[0063] ε j ={e k |S health (e k )≥80, type (e k ) = type (j)},

[0064] Among them, ε j is the set of available equipment that can be allocated to region j, e k For the kth firefighting equipment individual, type (e k ) is firefighting equipment k The functional category of , type (j) is the type of equipment required for area j;

[0065] Sub-step 3.3, solving the multi-objective resource optimization model:

[0066] Construct a mixed integer programming model with the objective function of minimizing the allocation cost and health penalty:

[0067] min∑ i∈W ∑ j∈R C ij x ij +λ∑ k∈ε (I k -μS health (k)) + ,

[0068] Among them, x ij is the quantity of equipment transferred from warehouse i to region j, C ij is the allocation cost, λ is the health penalty coefficient, μ is the health score conversion factor, Ik is the inventory status of equipment k,

[0069] W is the set of all firefighting equipment warehouses, and R is the set of target areas where firefighting equipment needs to be pre-deployed;

[0070] Constraints include:

[0071] Coverage requirements: ∑ i x ij ≥D j , D j Estimate demand for region j;

[0072] Warehouse capacity: ∑ j x ij ≤Q i , i is the maximum capacity of warehouse i.

[0073] Preferably, in step 4, generating a dynamic rescue path and equipment allocation instruction further includes:

[0074] Sub-step 4.1, real-time traffic and hazard zone data integration:

[0075] Input the target area coordinates in the pre-deployment plan generated in step 3 and obtain the traffic speed v in real time. road and congestion coefficient η, and the coordinates of the dangerous area (x d ,y d ), calculate the geofence radius and generate the dangerous area geofence:

[0076] r=v·t response ,

[0077] Where r is the radius of the geo-fence, v is the diffusion speed of the hazardous substance, and t response It is a preset security response time threshold;

[0078] Sub-step 4.2, multi-objective dynamic path optimization model construction:

[0079] Based on the reinforcement learning model, the state space is defined as S = {v road ,η,R,S health}, the action space is path node selection, and the reward function is:

[0080]

[0081] Among them, R is the reward function, t route is the estimated time of the path, t max is the maximum allowed time, w1, w2, w3 are time, health, and safety weights, and I(·) is the indicator function;

[0082] Sub-step 4.3, health equipment allocation instruction generation:

[0083] Enter the equipment health score S calculated in step 2 health , filter available equipment [S health ≥80], combined with the optimized path of sub-step 4.2, generate the transfer instruction, and instruct the additional path navigation coordinate sequence {(x1,y1),(x2,y2),...,(x n ,y n )}:

[0084] I j ={e k |e k ∈ε j ,S health (e k )≥80, type (e k ) = type (j)},

[0085] Among them, I j is the equipment list transferred to area j, ε j is the set of available equipment that can be allocated to region j, e k For the kth firefighting equipment individual, type (e k ) is firefighting equipment k The functional category of area j is type (j), and type (j) is the equipment type required for area j.

[0086] Preferably, in step 5, updating the model and generating maintenance and decommissioning instructions further include:

[0087] Sub-step 5.1, actual scheduling data collection and feature extraction:

[0088] Collect the usage data of the equipment actually transferred in step 4, including the cumulative working time increment Δt use , number of failures n fault , and the path execution time difference Δt route ;

[0089] Extract features: F update ={Δt use ,n fault ,Δt route},

[0090] Among them, F update is the feature set, Δt use is the cumulative working hours increment, n fault is the number of failures, Δt route The difference between the actual time and the planned time;

[0091] Sub-step 5.2, incremental update of risk prediction and health assessment model:

[0092] Updated fire risk prediction model:

[0093]

[0094] in, is the actual fire state, L risk is the mean square error loss function of the fire risk prediction model, is the actual fire status of the area (x, y);

[0095] Updated equipment health score model:

[0096]

[0097] in, According to the actual number of failures n fault The inferred health score, L health is the mean square error loss function of the equipment health scoring model, is the actual health score of equipment k;

[0098] Joint Optimization:

[0099] Among them, η is the learning rate, θ new is the new parameter updated by gradient descent, θ old are the old parameters before the model is updated, is the gradient of the loss function with respect to the model parameters θθ;

[0100] Sub-step 5.3, equipment retirement priority calculation and instruction generation:

[0101] Calculate retirement priority:

[0102] Among them, R retire (k) is the retirement priority score of equipment k, t use (k) is the cumulative working hours of equipment k, t total (k) is the design life of equipment k, S health (k) is the predicted health score of equipment k;

[0103] Generate instructions:

[0104] Retirement Order: R retire (k) ≥ 1.2;

[0105] Maintenance instruction: 0.8<R retire (k)≤1.2.

[0106] Preferably, in step 2, the calculation of the equipment health score includes a life factor and a state deviation factor, wherein the life factor is the ratio of the remaining life to the total life, and the state deviation factor is the normalized Euclidean distance between the sensor data and the reference value;

[0107] In step 3, the objective function of the mixed integer programming model is to minimize the sum of the allocation cost and the health penalty term, and the constraints include covering the requirements of high-risk areas and only calling equipment with a health score higher than a preset value.

[0108] Preferably, in step 4, the reward function of the multi-objective reinforcement learning model includes an exponential decay term for path time consumption, an equipment health score accumulation term, and a penalty term for entering a dangerous area;

[0109] In step 5, the calculation of equipment retirement priority includes a linear combination of the cumulative working hours ratio and the health score decay value, and automatic retirement is triggered when the priority exceeds the threshold.

[0110] A terminal device includes a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the method for optimizing the dispatch of firefighting equipment and materials based on big data is implemented.

[0111] A storage medium stores a computer program, which, when executed by a processor, implements a method for optimizing the dispatch of firefighting equipment and materials based on big data.

[0112] The present invention provides a method for optimizing the dispatch of firefighting equipment and materials based on big data. It has the following beneficial effects:

[0113] 1. The present invention adopts a dynamic risk prediction and cross-regional collaborative pre-deployment technical solution. It predicts the probability of fire risk by real-time integration of meteorological, geographical and historical disaster data, and generates a pre-deployment plan based on risk level and equipment health status, so as to achieve the technical effect of improving the resource coverage accuracy in high-risk areas and optimizing cross-regional collaborative efficiency. Compared with the technical solutions in the existing technology that rely on static plans or manual experience scheduling, it solves the problems of mismatch between resource allocation and dynamic disaster risks and delayed cross-regional collaborative response.

[0114] 2. The present invention adopts a technical solution for dynamic assessment of equipment health status and closed-loop management of the entire life cycle. It calculates health scores in real time through multi-source sensor data and classifies and manages them. It combines a retirement priority algorithm to trigger maintenance or retirement instructions, thereby achieving the technical effect of ensuring the availability of allocated equipment and reducing inefficient resource occupation. Compared with the technical solution of regular manual inspection or fixed life determination in the existing technology, it solves the shortcomings of delayed equipment status perception and increased rescue risks caused by the misallocation of faulty equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0115] Figure 1 Flowchart of the present invention. DETAILED DESCRIPTION

[0116] To help those skilled in the art understand the present invention, the following will provide a clear and complete description of the technical solutions in the embodiments of the present invention, in conjunction with the accompanying drawings. Obviously, the described embodiments are only partial embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0117] The present invention is described in detail below with reference to the accompanying drawings:

[0118] Example:

[0119] Please see the attached Figure 1 The embodiment of the present invention provides a method for optimizing the dispatch of firefighting equipment and materials based on big data, comprising:

[0120] Step 1: Collect the temperature, voltage, and accumulated working hours data of firefighting equipment, and obtain real-time wind speed, humidity, and temperature data from the meteorological department, as well as building density, road connectivity, and historical fire point coordinates from geographic information;

[0121] Sub-step 1.1, firefighting equipment sensor data collection and preprocessing:

[0122] Deploy temperature sensors, voltage sensors and working time counters in fire fighting equipment to collect temperature T, voltage V, and accumulated working time t use , calculation formula:

[0123]

[0124]

[0125] Among them, T i is the i-th temperature sampling value, N is the number of sampling times per unit time, M is the number of sampling times per unit time, V j is the jth voltage sampling value, δ active (t) is the working status of the equipment at time t;

[0126] Sub-step 1.2, Disaster risk data integration and spatiotemporal alignment:

[0127] Obtain real-time wind speed v, humidity h, and ambient temperature T from the meteorological department interface env , and extract building density ρ, road connectivity c, historical fire point coordinates (x k ,y k ), calculation formula:

[0128]

[0129] (x k ,y k ) = GPS coordinates of the kth fire,

[0130] Data time alignment: Linear interpolation is used to compensate for time deviation of asynchronous data:

[0131]

[0132] Among them, v(t) is the real-time wind speed value at time t, t is the current timestamp, v(t i ) is the discrete time point t i Observed wind speed value, t i is the time point of data collection for the i-th time, v(t i+1 ) is the discrete time point t i+1 Observed wind speed value, t i+1 is the time point of the i+1th data collection;

[0133] Sub-step 1.3, multi-source data fusion and feature storage:

[0134] After aligning the equipment data and disaster data by timestamp, a spatiotemporal fusion data block is generated and stored in the spatiotemporal database. The index fields are time t and coordinates (x, y). The calculation formula is:

[0135] D fusion ={T,V,t use ,v,h,ρ,c,x k ,y k},

[0136] Among them, D fusion is the multi-source fusion data block after spatiotemporal alignment, x k 、y k is the latitude and longitude coordinates of the kth historical fire point;

[0137] Step 2: Based on the collected meteorological data, geographic information, and historical fire point coordinates, the probability of future fire risks is predicted. Based on the temperature, voltage, and accumulated working hours data of the firefighting equipment, the equipment health score is calculated and classified as usable, warning, or disabled.

[0138] Sub-step 2.1: Construction of multi-source disaster risk prediction model:

[0139] Input the meteorological data collected in step 1 [wind speed v, humidity h, ambient temperature T env ] and geographic information [building density ρ, road connectivity c, historical fire point coordinates (x k ,y k)], construct spatiotemporal fusion features:

[0140] F risk =LSTM(v,h,T env )⊕GCN(ρ,c,x k ,y k ),

[0141] Among them, F risk LSTM is a long short-term memory network that extracts meteorological time series features, GCN is a graph convolutional network that models geographic spatial associations, and ⊕ is a feature concatenation operation.

[0142] Output regional risk probability:

[0143] P risk (x,y,t)=σ(W h ·F risk +b h ),

[0144] Among them, σ is the Sigmoid function, W h 、b h are trainable weights and biases, P risk is the fire risk probability of area (x, y) at time t;

[0145] Sub-step 2.2, calculation of fire equipment health score:

[0146] Input the equipment temperature T, voltage V, and accumulated working hours t collected in step 1 use , calculate the health score:

[0147]

[0148] Among them, S health is the health score, α and β are the weights of lifespan factor and state deviation factor, t total is the equipment design life, T normal 、V normal The temperature and voltage reference values when the equipment is working normally;

[0149] Health score categories:

[0150] Available: S health ≥80;

[0151] Warning: 60≤S health <80;

[0152] Disable: S health <60;

[0153] Sub-step 2.3, risk-health data fusion and classification output:

[0154] The risk probability P in sub-step 2.1 is risk The health score S in substep 2.2 health Generate a risk-health matrix by region association:

[0155] M(x,y)=[P risk (x,y,t),max(S health (x,y))],

[0156] Among them, M(x,y) is the correlation matrix between the risk probability of region (x,y) and the equipment health score, P risk (x,y,t) is the probability of fire occurring in area (x,y) at time t, S health (x,y) is the maximum health score of the equipment available in the area (x,y);

[0157] Output classification results:

[0158] High-risk areas: P risk ≥0.7 and max(S health )≥80;

[0159] Medium-risk area: 0.4≤P risk <0.7 or max(S health )<80;

[0160] Low-risk area: P risk <0.4;

[0161] In step 2, the calculation of the equipment health score includes the life factor and the state deviation factor. The life factor is the ratio of the remaining life to the total life, and the state deviation factor is the normalized Euclidean distance between the sensor data and the reference value.

[0162] Step 3: Generate a cross-regional firefighting equipment pre-deployment plan based on the predicted fire risk probability and equipment health score;

[0163] Sub-step 3.1, regional risk prioritization:

[0164] Input the fire risk probability P predicted in step 2 risk (x,y,t), prioritize regions by threshold:

[0165]

[0166] Output target region set R = {(x,y)|P risk (x,y,t)≥0.4};

[0167] Sub-step 3.2: Health equipment screening and demand matching:

[0168] Enter the equipment health score S calculated in step 2health , filter available equipment [S health ≥80], matching regional requirements by type:

[0169] ε j ={e k |S health (e k )≥80, type (e k ) = type (j)},

[0170] Among them, ε j is the set of available equipment that can be allocated to region j, e k For the kth firefighting equipment individual, type (e k ) is firefighting equipment k The functional category of , type (j) is the type of equipment required for area j;

[0171] Sub-step 3.3, solving the multi-objective resource optimization model:

[0172] Construct a mixed integer programming model with the objective function of minimizing the allocation cost and health penalty:

[0173] min∑ i∈W Σ j∈R C ij x ij +λ∑ k∈ε (I k -μS health (k)) + ,

[0174] Among them, x ij is the quantity of equipment transferred from warehouse i to region j, C ij is the allocation cost, λ is the health penalty coefficient, μ is the health score conversion factor, I k is the inventory status of equipment k,

[0175] W is the set of all firefighting equipment warehouses, and R is the set of target areas where firefighting equipment needs to be pre-deployed;

[0176] Constraints include:

[0177] Coverage requirements: Σ i x ij ≥D j , D j Estimate demand for region j;

[0178] Warehouse capacity: Σ j x ij ≤Q i , Q i is the maximum capacity of warehouse i;

[0179] In step 3, the objective function of the mixed integer programming model is to minimize the sum of the allocation cost and the health penalty term. The constraints include covering the high-risk area requirements and only calling equipment with a health score higher than the preset value.

[0180] Step 4: Generate dynamic rescue routes and equipment allocation instructions based on the generated pre-deployment plan, combined with real-time traffic conditions and dangerous area coordinates;

[0181] Sub-step 4.1, real-time traffic and hazard zone data integration:

[0182] Input the target area coordinates in the pre-deployment plan generated in step 3 and obtain the traffic speed v in real time. road and congestion coefficient η, and the coordinates of the dangerous area (x d ,y d ), calculate the geofence radius and generate the dangerous area geofence:

[0183] r=v·t response ,

[0184] Where r is the radius of the geo-fence, v is the diffusion speed of the hazardous substance, and t response It is a preset security response time threshold;

[0185] Sub-step 4.2, multi-objective dynamic path optimization model construction:

[0186] Based on the reinforcement learning model, the state space is defined as S = {v road ,η,R,S health}, the action space is path node selection, and the reward function is:

[0187]

[0188] Among them, R is the reward function, t route is the estimated time of the path, t max is the maximum allowed time, w1, w2, w3 are time, health, and safety weights, and I(·) is the indicator function;

[0189] Sub-step 4.3, health equipment allocation instruction generation:

[0190] Enter the equipment health score S calculated in step 2 health , filter available equipment [S health ≥80], combined with the optimized path of sub-step 4.2, generate the transfer instruction, and instruct the additional path navigation coordinate sequence {(x1,y1),(x2,y2),...,(x n ,y n )}:

[0191] I j ={ek |e k ∈ε j ,S health (e k )≥80, type (e k ) = type (j)},

[0192] Among them, I j is the equipment list transferred to area j, ε j is the set of available equipment that can be allocated to region j, e k For the kth firefighting equipment individual, type (e k ) is firefighting equipment k The functional category of , type (j) is the type of equipment required for area j;

[0193] In step 4, the reward function of the multi-objective reinforcement learning model includes an exponential decay term for path time, an accumulation term for equipment health score, and a penalty term for entering a dangerous area.

[0194] Step 5: Based on the actual equipment usage data and path execution results, the fire risk prediction model and equipment health score model are updated to generate equipment maintenance and decommissioning instructions;

[0195] Sub-step 5.1, actual scheduling data collection and feature extraction:

[0196] Collect the usage data of the equipment actually transferred in step 4, including the cumulative working time increment Δt use , number of failures n fault , and the path execution time difference Δt route ;

[0197] Extract features: F update ={Δt use ,n fault ,Δt route},

[0198] Among them, F update is the feature set, Δt use is the cumulative working hours increment, n fault is the number of failures, Δt route The difference between the actual time and the planned time;

[0199] Sub-step 5.2, incremental update of risk prediction and health assessment model:

[0200] Updated fire risk prediction model:

[0201]

[0202] in, is the actual fire state, L riskis the mean square error loss function of the fire risk prediction model, is the actual fire status of the area (x, y);

[0203] Updated equipment health score model:

[0204]

[0205] in, According to the actual number of failures n fault The inferred health score, L health is the mean square error loss function of the equipment health scoring model, is the actual health score of equipment k;

[0206] Joint Optimization:

[0207] Among them, η is the learning rate, θ new is the new parameter updated by gradient descent, θ old are the old parameters before the model is updated, is the gradient of the loss function with respect to the model parameters θθ;

[0208] Sub-step 5.3, equipment retirement priority calculation and instruction generation:

[0209] Calculate retirement priority:

[0210] Among them, R retire (k) is the retirement priority score of equipment k, t use (k) is the cumulative working hours of equipment k, t total (k) is the design life of equipment k, S health (k) is the predicted health score of equipment k;

[0211] Generate instructions:

[0212] Retirement Order: R retire (k) ≥ 1.2;

[0213] Maintenance instruction: 0.8<R retire (k)≤1.2;

[0214] In step 5, the calculation of equipment retirement priority includes a linear combination of the cumulative working hours ratio and the health score decay value. When the priority exceeds the threshold, automatic retirement is triggered.

[0215] The benefits of step 1 are that, through comprehensive multi-source data collection and preprocessing, this method establishes a high-quality spatiotemporal fusion data foundation in a timely manner. Synchronous alignment of sensor data with meteorological and geographic information eliminates temporal bias caused by asynchronous data collection and reduces data noise and redundancy. Once the data is stored in the spatiotemporal database, retrieval efficiency is significantly improved. Overall, step 1 provides complete and reliable data support for subsequent risk prediction and health assessment, ensuring the accuracy and real-time nature of the scheduling method.

[0216] The benefit of step 2 is that, by deeply integrating the two dimensions of disaster risk and equipment health, step 2 enables dual classification and assessment of both regions and equipment status. The risk prediction model quickly identifies high-risk areas, while the health scoring mechanism ensures that allocated equipment is in optimal condition. The combined risk-health matrix provides a basis for refined scheduling decisions, avoiding the waste of centralized resources and preventing faulty equipment from reaching the front lines of rescue.

[0217] The benefits of step 3 are that cross-regional pre-deployment, based on mixed integer programming, balances the dual objectives of minimizing costs and health penalties. The model fully considers warehouse capacity, regional demand, and equipment health to achieve optimal resource allocation. Furthermore, prioritization and type matching ensure that the necessary equipment is promptly delivered to high-risk areas, reducing allocation conflicts. Overall, step 3 effectively manages transportation costs and equipment lifespan while ensuring rescue efficiency.

[0218] Step 4 benefits from integrating real-time traffic congestion and hazardous area information to dynamically generate safe and efficient rescue routes. A multi-objective reinforcement learning model autonomously learns and optimizes after weighing route time, equipment health, and access risk, enhancing the intelligence of route decision-making. Generated dispatch instructions are accompanied by navigation coordinates, shortening response time and enhancing on-site operability for rescuers. This significantly improves rescue flexibility and on-site safety.

[0219] Step 5 benefits from a closed-loop feedback mechanism that enables continuous self-correction of the risk prediction and health assessment model. Based on incremental updates of actual allocation and execution data, the model's accuracy continues to improve. The equipment retirement priority algorithm automatically triggers maintenance and retirement, providing precise control over the entire equipment lifecycle. This step optimizes maintenance resource allocation, preventing inefficient or faulty equipment from occupying storage and transportation channels, and laying a solid foundation for the next round of scheduling.

[0220] A terminal device includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, a method for optimizing the dispatch of firefighting equipment and materials based on big data is implemented.

[0221] A storage medium stores a computer program, which, when executed by a processor, implements a method for optimizing the dispatch of firefighting equipment and materials based on big data.

[0222] The terminal device integrates a high-performance processor and a large-capacity memory, and is pre-installed with the big data scheduling algorithm of the present invention. It can quickly complete the whole process calculation of data collection, pre-processing, risk prediction, optimized scheduling and path planning locally. Compared with solutions that rely on cloud computing, localized processing greatly shortens the response delay, ensuring that scheduling decisions can be executed autonomously under network constraints or emergencies; at the same time, by caching models and historical data in the device, the system can achieve edge intelligence, thereby reducing communication bandwidth pressure and improving the robustness and security of the system. The terminal device supports modular expansion and remote upgrades, can flexibly adapt to the needs of various fire stations, and greatly enhances the convenience and maintainability of equipment deployment.

[0223] The storage medium completely encapsulates the firefighting equipment and material optimization and scheduling software of the present invention, facilitating rapid deployment and installation on various general-purpose computing platforms. Its plug-and-play nature allows users to run it directly on servers, workstations, or portable terminal devices without complex environment configuration. Furthermore, the storage medium supports partitioning and encryption management, ensuring the integrity and confidentiality of algorithms and data, and enabling data isolation based on different permission requirements, thereby enhancing the system's information security and protection capabilities. Through centralized storage and a unified update mechanism, maintenance personnel can easily push new versions and continuously iterate algorithms to ensure that scheduling methods are always in optimal condition.

[0224] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing the dispatch of firefighting equipment and materials based on big data, characterized in that: include: Step 1: Collect the temperature, voltage, and accumulated working hours data of firefighting equipment, and obtain real-time wind speed, humidity, and temperature data from the meteorological department, as well as building density, road connectivity, and historical fire point coordinates from geographic information; Step 2: Based on the collected meteorological data, geographic information, and historical fire point coordinates, the probability of future fire risks is predicted. Based on the temperature, voltage, and accumulated working hours data of the firefighting equipment, the equipment health score is calculated and classified as usable, warning, or disabled. Step 3: Generate a cross-regional firefighting equipment pre-deployment plan based on the predicted fire risk probability and equipment health score; Step 4: Generate dynamic rescue routes and equipment allocation instructions based on the generated pre-deployment plan, combined with real-time traffic conditions and dangerous area coordinates; Step 5: Based on the actual equipment usage data and path execution results, update the fire risk prediction model and equipment health scoring model, and generate equipment maintenance and retirement instructions.

2. The method for optimizing the dispatch of firefighting equipment and materials based on big data according to claim 1, characterized in that: In step 1, collecting firefighting equipment status data and disaster risk data further includes: Sub-step 1.1, firefighting equipment sensor data collection and preprocessing: Deploy temperature sensors, voltage sensors and working time counters in fire fighting equipment to collect temperature T, voltage V, and accumulated working time t use , calculation formula: Among them, T i is the i-th temperature sampling value, N is the number of sampling times per unit time, M is the number of sampling times per unit time, V j is the jth voltage sampling value, δ active (t) is the working status of the equipment at time t; Sub-step 1.2, Disaster risk data integration and spatiotemporal alignment: Obtain real-time wind speed v, humidity h, and ambient temperature T from the meteorological department interface env , and extract building density ρ, road connectivity c, historical fire point coordinates (x k ,y k ), calculation formula: (x k ,y k ) = GPS coordinates of the kth fire, Data time alignment: Linear interpolation is used to compensate for time deviation of asynchronous data: Among them, v(t) is the real-time wind speed value at time t, t is the current timestamp, v(t i ) is the discrete time point t i Observed wind speed value, t i is the time point of data collection for the i-th time, v(t i+1 ) is the discrete time point t i+1 Observed wind speed value, t i+1 is the time point of the i+1th data collection; Sub-step 1.3, multi-source data fusion and feature storage: After aligning the equipment data and disaster data by timestamp, a spatiotemporal fusion data block is generated and stored in the spatiotemporal database. The index fields are time t and coordinates (x, y). The calculation formula is: D fusion ={T,V,t use ,v,h,ρ,c,x k ,y k }, Among them, D fusion is the multi-source fusion data block after spatiotemporal alignment, x k 、y k are the latitude and longitude coordinates of the kth historical fire point.

3. The method for optimizing the dispatch of firefighting equipment and materials based on big data according to claim 1 is characterized in that: In step 2, predicting the fire risk probability and calculating the equipment health score further includes: Sub-step 2.1: Construction of multi-source disaster risk prediction model: Input the meteorological data collected in step 1 [wind speed v, humidity h, ambient temperature T env ] and geographic information [building density ρ, road connectivity c, historical fire point coordinates (x k ,y k )], construct spatiotemporal fusion features: Among them, F risk To fuse features, LSTM is a long short-term memory network that extracts meteorological time series features, and GCN is a graph convolutional network that models geographic spatial associations. It is the feature splicing operation; Output regional risk probability: P risk (x,y,t)=σ(W h ·F risk +b h ), Among them, σ is the Sigmoid function, W h 、b h are trainable weights and biases, P risk is the fire risk probability of area (x, y) at time t; Sub-step 2.2, calculation of fire equipment health score: Input the equipment temperature T, voltage V, and accumulated working hours t collected in step 1 use , calculate the health score: Among them, S health is the health score, α and β are the weights of lifespan factor and state deviation factor, t total is the equipment design life, T normal 、V normal The temperature and voltage reference values when the equipment is working normally; Health score categories: Available: S health ≥80; Warning: 60≤S health <80; Disable: S health <60; Sub-step 2.3, risk-health data fusion and classification output: The risk probability P in sub-step 2.1 is risk The health score S in substep 2.2 health Generate a risk-health matrix by region association: M(x,y)=[P risk (x,y,t),max(S health (x,y))], Among them, M(x,y) is the correlation matrix between the risk probability of region (x,y) and the equipment health score, P risk (x,y,t) is the probability of fire occurring in area (x,y) at time t, S health (x,y) is the maximum health score of the equipment available in the area (x,y); Output classification results: High-risk areas: P risk ≥0.7 and max(S health )≥80; Medium-risk area: 0.4≤P risk <0.7 or max(S health )<80; Low-risk area: P risk <0.

4.

4. The method for optimizing the dispatch of firefighting equipment and materials based on big data according to claim 1, characterized in that: In step 3, generating a cross-regional firefighting equipment pre-deployment plan further includes: Sub-step 3.1, regional risk prioritization: Input the fire risk probability P predicted in step 2 risk (x,y,t), prioritize regions by threshold: Output target region set R = {(x,y)|P risk (x,y,t)≥0.4}; Sub-step 3.2: Health equipment screening and demand matching: Enter the equipment health score S calculated in step 2 health , filter available equipment [S health ≥80], matching regional requirements by type: ε j ={e k |S health (e k )≥80, type (e k ) = type (j)}, Among them, ε j is the set of available equipment that can be allocated to region j, e k For the kth firefighting equipment individual, type (e k ) is firefighting equipment k The functional category of , type (j) is the type of equipment required for area j; Sub-step 3.3, solving the multi-objective resource optimization model: Construct a mixed integer programming model with the objective function of minimizing the allocation cost and health penalty: min∑ i∈W ∑ j∈R C ij x ij +λ∑ k∈ε (I k -μS health (k)) + , Among them, x ij is the quantity of equipment transferred from warehouse i to region j, C ij is the allocation cost, λ is the health penalty coefficient, μ is the health score conversion factor, I k is the inventory status of equipment k, W is the set of all firefighting equipment warehouses, and R is the set of target areas where firefighting equipment needs to be pre-deployed; Constraints include: Coverage requirements: ∑ i x ij ≥D j , D j Estimate demand for region j; Warehouse capacity: ∑ j x ij ≤Q i , Q i is the maximum capacity of warehouse i.

5. The method for optimizing the dispatch of firefighting equipment and materials based on big data according to claim 1 is characterized in that: In step 4, generating a dynamic rescue path and equipment allocation instruction further includes: Sub-step 4.1, real-time traffic and hazard zone data integration: Input the target area coordinates in the pre-deployment plan generated in step 3 and obtain the traffic speed v in real time. road and congestion coefficient η, and the coordinates of the dangerous area (x d ,y d ), calculate the geofence radius and generate the dangerous area geofence: r=v·t response , Where r is the radius of the geo-fence, v is the diffusion speed of the hazardous substance, and t response It is a preset security response time threshold; Sub-step 4.2, multi-objective dynamic path optimization model construction: Based on the reinforcement learning model, the state space is defined as S = {v road ,η,R,S health }, the action space is path node selection, and the reward function is: Among them, R is the reward function, t route is the estimated time of the path, t max is the maximum allowed time, w1, w2, w3 are time, health, and safety weights, and I(·) is the indicator function; Sub-step 4.3, health equipment allocation instruction generation: Enter the equipment health score S calculated in step 2 health , filter available equipment [S health ≥80], combined with the optimized path of sub-step 4.2, generate the transfer instruction, and instruct the additional path navigation coordinate sequence {(x1,y1),(x2,y2),...,(x n ,y n )}: I j ={e k |e k ∈ε j ,S health (e k )≥80, type (e k ) = type (j)}, Among them, I j is the equipment list transferred to area j, ε j is the set of available equipment that can be allocated to region j, e k For the kth firefighting equipment individual, type (e k ) is firefighting equipment k The functional category of area j is type (j), and type (j) is the equipment type required for area j.

6. The method for optimizing the dispatch of firefighting equipment and materials based on big data according to claim 1, characterized in that: In step 5, updating the model and generating maintenance and retirement instructions further include: Sub-step 5.1, actual scheduling data collection and feature extraction: Collect the usage data of the equipment actually transferred in step 4, including the cumulative working time increment Δt use , number of failures n fault , and the path execution time difference Δt route ; Extract features: F update ={Δt use ,n fault ,Δt route }, Among them, F update is the feature set, Δt use is the cumulative working hours increment, n fault is the number of failures, Δt route The difference between the actual time and the planned time; Sub-step 5.2, incremental update of risk prediction and health assessment model: Updated fire risk prediction model: in, is the actual fire state, L risk is the mean square error loss function of the fire risk prediction model, is the actual fire status of the area (x, y); Updated equipment health score model: in, According to the actual number of failures n fault The inferred health score, L health is the mean square error loss function of the equipment health scoring model, is the actual health score of equipment k; Joint Optimization: Among them, η is the learning rate, θ new is the new parameter updated by gradient descent, θ old are the old parameters before the model is updated, is the gradient of the loss function with respect to the model parameters θθ; Sub-step 5.3, equipment retirement priority calculation and instruction generation: Calculate retirement priority: Among them, R retire (k) is the retirement priority score of equipment k, t use (k) is the cumulative working hours of equipment k, t total (k) is the design life of equipment k, S health (k) is the predicted health score of equipment k; Generate instructions: Retirement Order: R retire (k) ≥ 1.2; Maintenance instruction: 0.8<R retire (k)≤1.

2.

7. The method for optimizing the dispatch of firefighting equipment and materials based on big data according to claim 1 is characterized in that: In step 2, the calculation of the equipment health score includes a life factor and a state deviation factor, wherein the life factor is the ratio of the remaining life to the total life, and the state deviation factor is the normalized Euclidean distance between the sensor data and the reference value; In step 3, the objective function of the mixed integer programming model is to minimize the sum of the allocation cost and the health penalty term, and the constraints include covering the requirements of high-risk areas and only calling equipment with a health score higher than a preset value.

8. The method for optimizing the dispatch of firefighting equipment and materials based on big data according to claim 1, characterized in that: In step 4, the reward function of the multi-objective reinforcement learning model includes an exponential decay term for path time, an accumulation term for equipment health score, and a penalty term for entering a dangerous area; In step 5, the calculation of equipment retirement priority includes a linear combination of the cumulative working hours ratio and the health score decay value, and automatic retirement is triggered when the priority exceeds the threshold.

9. A terminal device, characterized in that: The invention comprises a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the method for optimizing and dispatching fire-fighting equipment and materials based on big data as described in any one of claims 1 to 8 is implemented.

10. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, it implements the method for optimizing and dispatching firefighting equipment and materials based on big data as described in any one of claims 1 to 8.

Images