Fire emergency management method and system

Abstract

The invention relates to the technical field of fire-fighting emergency management, and discloses a fire-fighting emergency management method and system, and the method comprises the steps: S1, collecting environment information; s2, the environment information is preprocessed, then the preprocessed environment information is input into a dynamic multi-level self-adaptive risk assessment model, and the dynamic multi-level self-adaptive risk assessment model assesses a risk state and outputs a risk level; and S3, starting a corresponding emergency management plan according to the risk level. According to the invention, the automation and intelligence level of emergency management is obviously improved, a corresponding emergency management plan can be started according to the risk level, the response is timely, and the reliability is high.

Description

Technical Field

[0001] The present invention belongs to the technical field of fire emergency management, and in particular relates to a fire emergency management method and system. Background Art

[0002] With the rapid development of urbanization and technological advancement, the demand for intelligent security management systems has become increasingly prominent, especially in the protection of high-risk public buildings and important infrastructure. Traditional security management systems often rely on manual monitoring and basic automation technologies, such as video surveillance systems and simple alarm devices, which mainly operate based on preset rules and lack flexibility and adaptability. In addition, the response delay and accuracy of these systems often cannot meet the needs in high-risk environments, especially when facing complex or hidden emergencies.

[0003] Although existing technologies can provide a certain degree of monitoring and early warning functions, most are still limited to traditional data processing methods and cannot perform in-depth data analysis or real-time decision support. For example, many systems use basic sensor networks to monitor specific environmental parameters, such as smoke or temperature, but these systems can usually only trigger alarms when preset thresholds are reached, and cannot understand or predict complex time patterns or environmental changes. They often lack effective data integration capabilities and cannot comprehensively consider data from different sources, which in turn affects the timeliness and adaptability of response measures.

[0004] In view of this, how to design a technology that can timely and accurately perform emergency management in the face of complex or hidden emergency situations is a technical problem to be solved by the present invention. Summary of the invention

[0005] The present invention provides a fire emergency management method and system, which can realize timely and accurate emergency management in the face of complex or hidden emergency situations.

[0006] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0007] In one aspect, the present invention provides a fire emergency management method and system, comprising the following steps:

[0008] S1, collect environmental information;

[0009] S2, preprocessing the environmental information, and then inputting the preprocessed environmental information into a dynamic multi-level adaptive risk assessment model, wherein the dynamic multi-level adaptive risk assessment model assesses the risk status and outputs a risk level;

[0010] S3: Initiate a corresponding emergency management plan based on the risk level.

[0011] In the fire emergency management method as described above, the environmental information includes temperature, humidity, smoke concentration and thermal energy.

[0012] In the fire emergency management method as described above, preprocessing the environmental information includes data cleaning, missing value processing and formatting.

[0013] In the fire emergency management method as described above, the dynamic multi-level adaptive risk assessment model includes:

[0014] An input layer, which receives the environment information;

[0015] A feature extraction layer, which extracts features from the environmental information received by the input layer;

[0016] A dynamic adaptation layer, which dynamically adjusts the weights of the features of different environmental information extracted by the feature extraction layer by analyzing the contribution of the features of different environmental information;

[0017] A fusion layer, which uses a risk assessment function to evaluate the risk status of the features adjusted by the dynamic adaptation layer;

[0018] The output layer outputs the risk level evaluated by the fusion layer.

[0019] In the fire emergency management method as described above, the feature extraction layer is a multi-layer structure, including:

[0020] A first feature extraction layer, which extracts low-level features, wherein the low-level features are basic change features directly collected;

[0021] A second feature extraction layer extracts high-level features, wherein the high-level features are time dependencies and long-term trends captured in the low-level features;

[0022] The feature fusion layer integrates the output of the second feature extraction layer to generate a unified feature representation.

[0023] For the fire emergency management method described above, the formula of the risk assessment function is as follows:

[0024]

[0025] Where R(t) represents the risk assessment value at time t; α i (t) represents the adaptive weight of the i-th sensor at time t; represents the risk characteristic function based on environmental information; represents the reading of the i-th sensor at time t; β(t) represents the importance weight of historical environmental information; g(h(t)) represents the weighted moving average function of historical environmental information; h(t) represents the vector of historical environmental information.

[0026] The fire emergency management method as described above, initiating a corresponding emergency management plan according to the risk level includes:

[0027] When R(t)≤R threshold1 , then the risk level is low risk, and the emergency management plan includes regular inspections of areas or equipment;

[0028] When R threshold1 <R(t)≤R threshold2 , then the risk level is medium risk, and the emergency management plan includes issuing an early warning notice to remind immediate inspection of the area or equipment to see if there are any abnormalities;

[0029] When R threshold2 <R(t), the risk level is high risk, and the emergency management plan includes issuing an alarm to notify evacuation of personnel and activating fire-fighting equipment;

[0030] Where R(t) represents the risk assessment value at time t; R threshold1 represents the first risk threshold; R threshold2 represents a second risk threshold, and the second risk threshold is greater than the first risk threshold.

[0031] The fire emergency management method as described above further includes:

[0032] S4, continuously updating the model parameters to optimize the dynamic multi-level adaptive risk assessment model in step S2.

[0033] In the fire emergency management method described above, the model parameters are continuously updated using an enhanced hybrid gradient descent algorithm, and the formula for updating the model parameters is as follows:

[0034] w(t+1)=w(t)-η(t)·(H(w(t)) -1 v(t)

[0035] Where w(t+1) is the model parameter vector after the t+1th iteration of the model; w(t) represents the model parameter vector after the tth iteration of the model; η(t) represents the adaptive learning rate, which is adjusted as the number of iterations increases; H(w(t)) -1 represents the inverse of the Hessian matrix, which is used to adjust the step size and reverse direction; v(t) represents the momentum term.

[0036] In another aspect, the present invention further provides a fire emergency management system, comprising:

[0037] A data acquisition module, which is used to collect environmental information in real time;

[0038] A data processing and analysis module receives the environmental information collected by the data collection module, processes the environmental information, and then evaluates the risk status to obtain a risk level;

[0039] A decision support module receives the risk level output by the data analysis unit, activates a corresponding emergency management plan according to the risk level, and generates a historical record;

[0040] The optimization and updating module receives the environmental information of the data acquisition module and the historical records sent by the decision support module to optimize and update the data analysis module.

[0041] Compared with the prior art, the advantages and positive effects of the present invention are: by collecting environmental information, the environmental status can be monitored in real time; by inputting environmental information into a dynamic multi-level adaptive risk assessment model, multi-source data can be integrated, and the dynamic multi-level adaptive risk assessment model can evaluate the risk status in the area and output the risk level, which can reduce underreporting or false alarms; according to different risk levels, the corresponding emergency management plan is initiated, which greatly improves the accuracy of early warning and the timeliness of response, thereby improving the efficiency and safety of emergency management. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0043] Figure 1 A flowchart of an embodiment of a fire emergency management method of the present invention;

[0044] Figure 2 This is a system block diagram of an embodiment of a fire emergency management system of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] It should be noted that in the description of the present invention, the terms "upper", "lower", "inner", "outer" and the like indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings, which are only for the convenience of description, and do not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0047] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0048] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may include that the first and second features are in direct contact, or may include that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, a first feature being "above", "above" and "above" a second feature includes that the first feature is directly above and obliquely above the second feature, or simply indicates that the first feature is higher in level than the second feature. A first feature being "below", "below" and "below" a second feature includes that the first feature is directly below and obliquely below the second feature, or simply indicates that the first feature is lower in level than the second feature.

[0049] The disclosure below provides many different embodiments or examples to realize different structures of the present invention. In order to simplify the disclosure of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present invention. In addition, the present invention can repeat reference numbers and / or reference letters in different examples, and this repetition is for the purpose of simplicity and clarity, which itself does not indicate the relationship between the various embodiments and / or settings discussed. In addition, the present invention provides various specific examples of processes and materials, but those of ordinary skill in the art can be aware of the application of other processes and / or the use of other materials.

[0050] As shown in Figure 1, the present application provides a fire emergency management method that can achieve efficient identification and response to emergency situations, which not only improves the reaction speed, but also improves the accuracy of handling emergency situations.

[0051] A fire emergency management method comprises the following steps:

[0052] S1, collect environmental information, specifically:

[0053] Environmental information includes temperature, humidity, smoke concentration and thermal energy, which are collected in real time by setting temperature sensors, humidity sensors, smoke concentration sensors and infrared sensors in the area.

[0054] S2, preprocessing the environmental information, and then inputting the preprocessed environmental information into a dynamic multi-level adaptive risk assessment model, which evaluates the risk status and outputs the risk level.

[0055] Preprocessing includes data cleaning, missing value processing and formatting. The preprocessed information is in vector form.

[0056] The preprocessed information is input into the dynamic multi-layer adaptive risk assessment model. The network structure of the dynamic multi-layer adaptive risk assessment model (DMARAM) is a multi-level, dynamically adaptable neural network architecture. The neural network architecture processes the input data in layers and combines a dynamic adjustment mechanism to ensure that risks can be assessed in real time and accurately under different environmental conditions. The following is the composition of the dynamic multi-layer adaptive risk assessment model:

[0057] Input layer,The input layer receives data from multiple sensors, such as temperature, humidity, smoke concentration, thermal energy value, etc. After preprocessing, these environmental information are input into the input layer in the form of vectors. Each input data point corresponds to an input node.

[0058] The feature extraction layer extracts features from the environmental information received by the input layer. The feature extraction layer has a multi-layer structure, including:

[0059] The first feature extraction layer extracts low-level features by setting up multiple convolutional neural networks (CNNs). Low-level features are basic change features directly collected, that is, simple statistics directly obtained by the sensor during data collection, such as changes and fluctuations in temperature, humidity, smoke concentration, and thermal energy. The formula for extracting low-level features through CNN is as follows:

[0060]

[0061] in, represents the output of the kth convolution kernel at position (i, j); x i+m，j+n Indicates an area for inputting environmental information data; represents the weight of the kth convolution kernel; b kis the bias parameter.

[0062] In the above formula, the size of the convolution kernel M×N and the number of convolution kernels can be adjusted according to the dimension of the environmental information: if the environmental information has multi-dimensional characteristics (such as temperature, humidity and other information collected simultaneously), the size of the convolution kernel can be appropriately increased to cover more input areas; the number of convolution kernels depends on the richness of the features, which is usually determined according to the number of channels of the input environmental information data (such as the number of different sensors) and the feature complexity to ensure that low-level features from different sensors can be captured.

[0063] After the convolution operation is completed, the ReLU (Rectified Linear Unit) function is applied to the output to retain positive values ​​and suppress negative values, thereby introducing nonlinear features. This step enables the network to learn the nonlinear relationship in the input data, so that the dynamic multi-level adaptive risk assessment model can better capture complex features. This nonlinear processing is reflected after each convolution operation, helping the model to cope with the diversity and complexity that may exist in environmental data. The formula for applying the ReLU function to the output is as follows:

[0064]

[0065] Among them, x t Represents the low-level features extracted by the first feature extraction layer; Represents the output of the kth convolution kernel at position (i, j).

[0066] The second feature extraction layer is used to extract high-level features. High-level features are processed by low-level features to capture the time dependency and long-term trend of low-level features. They are used to process the changing patterns of environmental data over time, such as temperature fluctuations, humidity changes, smoke concentration changes, or the periodicity of thermal energy changes. They can be extracted through a long short-term memory network (LSTM) or a gated recurrent unit (GRU). The following is an explanation of the two methods of extracting high-level features.

[0067] The formula for extracting high-level features using LSTM is as follows:

[0068] f t =σ(W f ·[h(t-1),x t ]+b f )

[0069] i t =σ(W i ·[h(t-1),x t ]+bi )

[0070]

[0071] o t =σ(W o ·[h(t-1),x t ]+b o )

[0072] h t =o t ⊙tanh(C t )

[0073] Among them, f t W represents the forget gate; f represents the weight matrix of the forget gate; h(t-1) represents the historical environment information data at the past time point t-1; b f Represents the bias term of the forget gate; i t represents the input gate; W i represents the weight matrix of the input gate; b i represents the bias term of the input gate; represents the candidate memory state; W c The weight matrix representing the candidate memory state; b c represents the bias term of the candidate memory state; C t Indicates the current memory status; C t-1 Indicates the memory state of the previous moment; o t represents the output gate; h t Indicates hidden state; W o represents the weight matrix of the output gate; b o represents the bias term of the output gate; σ represents the activation function, and the sigmoid function can be used here; ⊙ represents the product at the element level; x t Represents the low-level features extracted by the first feature extraction layer.

[0074] The formula for extracting high-level features using GRU is as follows:

[0075] z t =σ(W z ·[h(t-1),x t ])

[0076] r t =σ(W r ·[h(t-1),x t ])

[0077]

[0078] Among them, z trepresents the update gate; σ represents the activation function, and the sigmoid function is used here; W z represents what is used to calculate the weight matrix of the update gate; h(t-1) represents the historical environment information data at the past time point t-1; x t Represents the low-level features extracted by the first feature extraction layer; r t Represents the reset gate; W r represents the weight matrix used to calculate the reset gate; represents the candidate hidden state; W h represents the weight matrix used to calculate the candidate hidden state; ⊙ represents the element-level product; h t Indicates the hidden state.

[0079] Through the above-mentioned LSTM or GRU, any pair of low-level feature processing, the second feature extraction layer can capture the time series information of the environmental information, and further provide more refined features for the dynamic multi-level adaptive risk assessment model.

[0080] Feature fusion layer: This layer uses a multi-head attention mechanism to integrate the output of the second feature extraction layer, so that the environmental information collected by different sensors can generate a unified feature representation.

[0081] The formula for the multi-head attention mechanism to integrate the output of the second feature extraction layer is as follows:

[0082]

[0083] Among them, Attention(Q, K, V) represents the output of a single attention operation performed by the multi-head attention mechanism; Q represents the query matrix, which can be regarded as a partial feature vector in the high-level feature, used to query other related features; K represents the key matrix, which can be regarded as the key feature vector in the high-level feature, which represents the matching features related to the query; V represents the value matrix, which is the feature matrix corresponding to the input high-level feature, used to generate the output feature; K T represents the transposed matrix of the key matrix; softmax(·) represents the normalized exponential function; d k Represents the dimension of the key matrix, which is used to scale the inner product to prevent excessive values ​​from causing the softmax function to be too smooth.

[0084] The multi-head attention mechanism performs multiple attention operations, concatenates the results and linearly transforms them into the final output. The formula is as follows:

[0085] MultiHead(Q,K,V)=Concat(head1,head2,…,head h )·W O

[0086] Among them, MultiHead(Q, K, V) represents the final output after the multi-head attention mechanism performs multiple attention operations; Concat(head1, head2, ..., head h ) represents the concatenation of the outputs of multiple attention heads calculated separately, merging the output results of all attention heads into a whole; W o Represents the linear transformation matrix of the output; h represents the number of heads; head h represents the output of the h-th attention head; head h The formula is as follows:

[0087] head h =Attention(QW i Q , KW i K , VW i V )

[0088] Among them, Attention(QW i Q , KW i K , VW i V ) represents the calculation of the single-head attention mechanism; Q represents the query matrix; W i Q represents the linear transformation matrix of the query matrix; K represents the key matrix; W i K represents the linear transformation matrix of the key matrix; V represents the value matrix; W i V Represents the linear transformation matrix of the representation value matrix.

[0089] The dynamic adaptation layer analyzes the contribution of the features of different environmental information extracted by the feature extraction layer through an adaptive mechanism, and dynamically adjusts the weights of the features of different environmental information in the risk assessment function, thereby ensuring that under different environmental conditions, the dynamic multi-level adaptive risk assessment model can automatically adjust the weights to adapt to different situations.

[0090] The dynamic adaptation layer needs to assign an adaptive weight to each feature. Used to adjust the contribution of this feature in the risk assessment function. The calculation formula is as follows:

[0091]

[0092] Among them, α i(t) represents the adaptive weight of the i-th sensor at time t; k i represents the adjustment parameter used to balance the contribution of each sensor to the risk assessment result; It represents the feature function based on environmental information, and its value represents the influence of the feature under the current environmental state; represents the reading of the i-th sensor at time t; exp represents the indicator function, which is used to ensure the non-negativity of the weight; n represents the total number of sensors.

[0093] The fusion layer uses the risk assessment function to evaluate the risk status of the features output by each layer after adjustment by the dynamic adaptation layer. The formula of the risk assessment function is as follows:

[0094]

[0095] Where R(t) represents the risk assessment value at time t; α i (t) represents the adaptive weight of the i-th sensor at time t; represents the risk characteristic function based on environmental information; represents the reading of the i-th sensor at time t, which represents specific environmental information, such as temperature, humidity, smoke concentration, and thermal energy; β(t) represents the importance weight of historical environmental information, and historical environmental information refers to the environmental information collected before the current time point; g(h(t)) represents the weighted moving average function of historical environmental information; h(t) represents the vector of historical environmental information.

[0096] α i (t) represents the adaptive weight of the i-th sensor at time t, which can represent the influence of the sensor on the overall risk assessment in the current environment. Its calculation formula is as follows;

[0097]

[0098] Among them, k i represents the adjustment parameter used to balance the contribution of each sensor to the risk assessment result; Represents the absolute value of the risk characteristic function corresponding to the environmental information data of the i-th sensor; It represents the characteristic function based on environmental information, which captures the complex risk level through nonlinear combination function. Its calculation formula is as follows:

[0099]

[0100] Among them, γ iIt represents parameter 1 that is continuously adjusted through the system adaptive learning process. Parameter 1 is used to adjust the linear relationship between the risk characteristic function and the sensor reading. Parameter 1 reflects the linear weight of the sensor data on the overall risk assessment. The larger the value, the greater the impact of the sensor data on the risk assessment. i represents parameter 2 that is continuously adjusted through the system adaptive learning process. Parameter 2 is used to adjust the nonlinear relationship in the risk characteristic function. It reflects the nonlinear part of the risk characteristic function by affecting the tanh term. The adjustment of parameter 2 can capture the complex characteristics of the sensor readings, such as extreme environmental changes or abnormal fluctuations, and plays a key role in the nonlinear analysis ability of the model; i It represents parameter 3 which is continuously adjusted through the system adaptive learning process. Parameter 3 is used to adjust the scaling degree of sensor readings in the nonlinear function tanh. The size of parameter 3 affects the sensitivity of the tanh function to the sensor readings, which further affects the response speed and accuracy of the model to the input data. Represents the reading of the i-th sensor at time t.

[0101] β(t) represents the importance weight of historical environmental information, reflecting the importance of historical data in the current risk assessment function. Its calculation formula is as follows:

[0102]

[0103] Among them, θ represents the model parameters; h(tj) represents the historical environmental information data at the past time point tj, which is a time series data used to reflect the state of the environment at a certain moment in the past. For example, it can contain a variety of information such as past temperature, humidity, smoke concentration, thermal energy, etc. By analyzing these historical data, the model can make a more accurate assessment of the current risk status; T represents the length of the time window used to calculate the weighted moving average; it determines how much past historical data will be referenced when calculating β(t). The larger T is, the model will consider historical data over a longer time range; the smaller T is, the model will only focus on recent environmental changes.

[0104] g(h(t)) represents the weighted moving average function of historical environmental information, which is used to capture long-term trend changes. Its calculation formula is as follows:

[0105]

[0106] Where T represents the length of the time window used to calculate the weighted moving average; w jrepresents the dynamic weight, which indicates the degree of influence of historical environmental information data at different time points on the current moment. Generally speaking, the closer the time point is, the greater the weight; h(tj) represents the historical environmental information data at the past time point tj; Z represents the normalization factor, which is used to ensure that the sum of the weighted moving average is 1. The calculation formula of Z is as follows:

[0107]

[0108] Where T represents the length of the time window used to calculate the weighted moving average; w j Represents dynamic weight.

[0109] The output layer outputs the risk level evaluated by the fusion layer. The output layer can use the Softmax function (normalized exponential function) to map the risk assessment value R(t) into a standardized range for further decision support.

[0110] The output layer uses the Softmax function to map the risk assessment value to a standardized range. The formula is as follows:

[0111]

[0112] Among them, Softmax(R(t) i ) indicates that the Softmax function is used to map the risk assessment value to an output within a standardized range; R(t) i represents the risk assessment value corresponding to the i-th sensor at time t; j represents the index of all sensors, Indicates the sum of the risk assessment values ​​of all sensors.

[0113] S3: Initiate corresponding emergency management plans according to risk levels, which include low risk, medium risk and high risk. Specifically:

[0114] When R(t)≤R threshold1 , then the risk level is low risk, and the emergency management plan includes regular inspections of areas or equipment;

[0115] When R threshold1 <R(t)≤R threshold2 , then the risk level is medium risk, and the emergency management plan includes issuing early warning notices to remind immediate inspections of areas or equipment for abnormalities;

[0116] When R threshold2 <R(t), the risk level is high, and the emergency management plan includes issuing an alarm to notify evacuation of personnel and activating fire-fighting equipment;

[0117] Where R(t) represents the risk assessment value at time t; Rthreshold1 represents the first risk threshold; R threshold2 represents a second risk threshold, and the second risk threshold is greater than the first risk threshold.

[0118] S4, continuously update the model parameters to optimize the dynamic multi-level adaptive risk assessment model in step S2. This application uses an enhanced hybrid gradient descent algorithm (Enhanced Hybrid Gradient Descent, EHGDA) to optimize the model parameters. The algorithm optimizes the model parameters by combining the information of the first-order derivative and the second-order derivative and the momentum method, so that the model can converge to the optimal solution faster and more accurately when facing variable data. The following is a detailed description of the steps of the EHGDA algorithm:

[0119] S41, initialize model parameters.

[0120] First, the parameter vector of the model parameter θ is w(0). The model parameter θ refers to the variables that need to be optimized during the training process, such as weights and biases. The parameter vector w(0) is the set of model parameters θ. w(0) represents the value of all model parameters θ at the time of initialization, and w(t) represents the parameter value of the model parameter after the tth iteration of the model. The change process from w(0) to w(t) is the process of model optimization.

[0121] Initialize w(0) and set the initial learning rate η0, momentum parameter ρ and other necessary algorithm hyperparameters, such as the number of iterations.

[0122] Setting hyperparameters such as the initial efficiency η0 and momentum parameter ρ can control how the model updates weights during the optimization process, thereby ensuring the stability and convergence of model training. Initializing the model parameters can avoid falling into the local optimal solution and improve the training efficiency of the model.

[0123] S42, calculate the loss function and its gradient.

[0124] In each iteration, the loss function L(w(t)) corresponding to the current parameter w(t) is calculated, and then the gradient of the loss function with respect to the parameter is calculated gradient Provides the fastest path to reduce loss in the direction of the current parameters.

[0125] First, calculate the error between the current model prediction value and the true value, then sum and average the squares of the errors of all samples to get the loss function value corresponding to the current parameter w(t). The calculation formula of the loss function L(w(t)) is as follows:

[0126]

[0127] Where N represents the number of samples; y i represents the true value; f(x i ; w(t)) represents the model prediction value; w(t) represents the parameter value of the model parameters after the tth iteration of the model.

[0128] gradient The calculation formula is as follows:

[0129]

[0130] Where N represents the number of samples; y i represents the true value; f(x i ; w(t)) represents the model prediction value; w(t) represents the parameter value of the model parameters after the tth iteration of the model.

[0131] Through the enhanced hybrid gradient descent algorithm, the model can adjust parameters along the negative direction of the gradient to achieve the purpose of reducing the loss function.

[0132] S43, calculate the momentum term.

[0133] The momentum term v(t) is used to smooth the path of gradient descent and avoid falling into the local minimum in the area with small gradient. The update formula of the momentum term is as follows:

[0134]

[0135] Among them, ρ represents the momentum parameter, which is generally set between 0 and 1 and is used to control the influence of the gradient of the previous iteration on the current iteration; v(t-1) represents the momentum term at the previous moment; represents the parameter gradient.

[0136] S44, introduce second-order derivative information.

[0137] To improve convergence speed and accuracy, the EHGDA algorithm introduces the Hessian matrix H(w(t)) of the loss function, which is a second-order derivative information. Through the Hessian matrix, the algorithm can better adjust the step size to avoid slow convergence in complex terrain. Complex terrain is often used to describe the behavior of optimization algorithms in multi-parameter space.

[0138] The calculation formula of H(w(t)) is as follows:

[0139]

[0140] Where L(w(t)) represents the loss function; w i represents the i-th parameter of the model; w j Denotes the jth parameter of the model.

[0141] In this step, by combining the second-order derivative information and momentum term of the Hessian matrix, the EHGDA algorithm can better cope with complex terrain and improve the efficiency and accuracy of model training.

[0142] S45, update model parameters.

[0143] Using the momentum term and second-order derivative information, the formula for updating the model parameter vector is as follows:

[0144] w(t+1)=w(t)-η(t)·(H(w(t)) -1 v(t)

[0145] Where w(t+1) is the model parameter vector after the t+1th iteration of the model; w(t) represents the model parameter vector after the tth iteration of the model; η(t) represents the adaptive learning rate, which is adjusted as the number of iterations increases; H(w(t)) -1 represents the inverse of the Hessian matrix, which is used to adjust the step size and reverse direction; v(t) represents the momentum term, where the calculation formula of the adaptive learning rate η(t) is as follows:

[0146]

[0147] Among them, η0 represents the initial learning rate; Represents the adjustment factor, which is used to control the decay rate of the learning rate; t represents the number of iterations.

[0148] S46, check convergence conditions.

[0149] After each iteration, check whether the convergence conditions are met, such as whether the modulus of the gradient is small enough or whether the change of the loss function is lower than the set threshold. If the conditions are met, the algorithm terminates and the current parameters are the optimal solution; otherwise, return to step S42 to continue iteration.

[0150] S47, the algorithm terminates.

[0151] When the algorithm reaches the convergence condition or the preset maximum number of iterations, the iteration stops and the final model parameter w is output. * .

[0152] The EHGDA algorithm combines first-order and second-order derivative information, avoids falling into local minima through momentum method, and controls the convergence speed through adaptive learning rate. This algorithm can not only speed up the model training process, but also improve the generalization ability of the model, especially when dealing with complex emergency management data.

[0153] As shown in 2, the present application also provides a fire emergency management system, including a data acquisition module, a data processing and analysis module, a decision support module, and an optimization and update module:

[0154] The data acquisition module includes various sensors for collecting environmental information, including temperature sensors, humidity sensors, smoke sensors, infrared sensors, etc. It is used to collect environmental information in real time and can monitor the environmental status in real time. The sensor data is transmitted to the data processing and analysis module in real time via the wireless network.

[0155] The data processing and analysis module receives the environmental information collected by the data acquisition module and pre-processes the environmental information. The data processing and analysis module has a built-in dynamic multi-level adaptive risk assessment model (DMARAM). The pre-processed environmental information is input into the model, and the model can evaluate the risk status to obtain the risk level.

[0156] The decision support module receives the risk level output by the data analysis unit, initiates the corresponding emergency management plan according to the risk level, and generates historical records. Managers can prepare risk assessment reports based on the historical records and adjust the emergency management plan through historical records. The module also has a data visualization function and can set up a display screen. The display screen displays environmental information, risk levels and other content in real time, which can help managers quickly understand the current environmental status and risk status.

[0157] The optimization and update module receives the environmental information from the data acquisition module and the historical records sent by the decision support module. The optimization and update module has a built-in enhanced hybrid gradient descent algorithm (EHGDA). The EHGDA algorithm continuously optimizes and updates the DMARAM model parameters in the data analysis module, thereby improving the system's vacancy capacity and adaptability.

[0158] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0159] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for a person skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to replace some of the technical features therein by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions claimed to be protected by the present invention.

Claims

1. A fire emergency management method, characterized in that: The following steps are involved: S1, collect environmental information; S2, preprocessing the environmental information, and then inputting the preprocessed environmental information into a dynamic multi-level adaptive risk assessment model, wherein the dynamic multi-level adaptive risk assessment model assesses the risk status and outputs a risk level; S3: Initiate a corresponding emergency management plan based on the risk level.

2. The fire emergency management method according to claim 1, characterized in that: The environmental information includes temperature, humidity, smoke concentration and thermal energy.

3. The fire emergency management method according to claim 1, characterized in that: Preprocessing the environmental information includes data cleaning, missing value processing and formatting.

4. The fire emergency management method according to claim 1, characterized in that: The dynamic multi-level adaptive risk assessment model includes: An input layer, which receives the environment information; A feature extraction layer, which extracts features from the environmental information received by the input layer; A dynamic adaptation layer, which dynamically adjusts the weights of the features of different environmental information extracted by the feature extraction layer by analyzing the contribution of the features of different environmental information; A fusion layer, which uses a risk assessment function to evaluate the risk status of the features adjusted by the dynamic adaptation layer; The output layer outputs the risk level evaluated by the fusion layer.

5. The fire emergency management method according to claim 4, characterized in that: The feature extraction layer is a multi-layer structure, comprising: A first feature extraction layer, which extracts low-level features, wherein the low-level features are basic change features directly collected; A second feature extraction layer extracts high-level features, wherein the high-level features are time dependencies and long-term trends captured in the low-level features; The feature fusion layer integrates the output of the second feature extraction layer to generate a unified feature representation.

6. The fire emergency management method according to claim 4, characterized in that: The formula for the risk assessment function is as follows: Where R(t) represents the risk assessment value at time t; α i (t) represents the adaptive weight of the i-th sensor at time t; represents the risk characteristic function based on environmental information; represents the reading of the i-th sensor at time t; β(t) represents the importance weight of historical environmental information; g(h(t)) represents the weighted moving average function of historical environmental information; h(t) represents the vector of historical environmental information.

7. The fire emergency management method according to claim 6, characterized in that: Initiate corresponding emergency management plans according to the risk level, including: When R(t)≤R threshold1 , then the risk level is low risk, and the emergency management plan includes regular inspections of areas or equipment; When R threshold1 <R(t)≤R threshold2 , then the risk level is medium risk, and the emergency management plan includes issuing an early warning notice to remind immediate inspection of the area or equipment to see if there are any abnormalities; When R threshold2 <R(t), the risk level is high risk, and the emergency management plan includes issuing an alarm to notify evacuation of personnel and activating fire-fighting equipment; Where R(t) represents the risk assessment value at time t; R threshold1 represents the first risk threshold; R threshold2 represents a second risk threshold, and the second risk threshold is greater than the first risk threshold.

8. The fire emergency management method according to any one of claims 1 to 7, characterized in that: Also includes: S4, continuously updating the model parameters to optimize the dynamic multi-level adaptive risk assessment model in step S2.

9. The fire emergency management method according to claim 8, characterized in that: The model parameters are continuously updated using an enhanced hybrid gradient descent algorithm, and the formula for updating the model parameters is as follows: w(t+1)=w(t)-η(t)·(H(w(t)) -1 ·v(t)) Where w(t+1) is the model parameter vector after the t+1th iteration of the model; w(t) represents the model parameter vector after the tth iteration of the model; η(t) represents the adaptive learning rate, which is adjusted as the number of iterations increases; H(w(t)) -1 represents the inverse of the Hessian matrix, which is used to adjust the step size and reverse direction; v(t) represents the momentum term.

10. A fire emergency management system, characterized in that: include: A data acquisition module, which is used to collect environmental information in real time; A data processing and analysis module receives the environmental information collected by the data collection module, processes the environmental information, and then evaluates the risk status to obtain a risk level; A decision support module receives the risk level output by the data analysis unit, activates a corresponding emergency management plan according to the risk level, and generates a historical record; The optimization and updating module receives the environmental information of the data acquisition module and the historical records sent by the decision support module to optimize and update the data analysis module.

Images