A big data-based city fire hazard early warning analysis method and system

Abstract

The application discloses a kind of city fire hazard early warning analysis method and system based on big data, it is related to disaster warning technical field, including based on feature dimension initialization KDE's Gaussian kernel function and bandwidth parameter, construct quenching function and coupling parameter as constraint in combination with GIS information, calculate simulation defect density, in combination with time series calculation actual critical index and standard deviation, construct heat diffusion equation and solve by two-stage normal adjacent method, generate space-time risk field, calculate risk propagation speed field.The application improves the risk identification accuracy of high-dimensional heterogeneous data through the multi-index clustering method combining entropy weight and kernel density, constructs the odd-symmetry anomaly point identification mechanism, improves the accuracy and stability of anomaly point identification, constructs the heat diffusion equation, generates the space-time risk field by two-stage normal adjacent method, realizes the dynamic evolution display of city fire risk, and improves the early warning response efficiency.

Description

Technical Field

[0001] This invention relates to the field of disaster early warning technology, and in particular to a method and system for early warning and analysis of urban fire hazards based on big data. Background Technology

[0002] With the deepening of urbanization and the rapid development of big data technology, urban fire hazard early warning and analysis technology has gradually evolved from the traditional passive response mode based on single sensor triggers to an active prediction system that integrates multimodal data. Early fire early warning mainly relied on temperature and smoke sensors, triggering alarms through fixed thresholds, which had limited accuracy. With the advancement of the Internet of Things, geographic information systems, and big data processing technologies, the system has begun to integrate multi-source data such as urban building information, meteorological data, historical fire records, and population density. Fuzzy C-means clustering, entropy weighting, and kernel density estimation are used to mine the spatiotemporal characteristics of high-dimensional data. The application of heat diffusion equations and fluid visualization technology has further promoted the simulation of risk propagation dynamics. Quantum computing-related methods, such as Trotter–Suzuki decomposition and spin glass models, have begun to be explored for modeling complex nonlinear systems, significantly improving the spatiotemporal resolution and predictive ability of fire hazard early warning, and providing data-driven decision support for urban fire management.

[0003] However, existing technologies still have shortcomings. In the process of clustering and kernel density estimation, existing technologies do not fully consider spatial coupling and dynamic evolution mechanisms, ignore the impact of risk transmission between neighboring regions and temporal trends on early warning accuracy, and cannot respond to environmental changes in a timely manner. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method and system for early warning analysis of urban fire hazards based on big data, which solves the problem that existing technologies do not fully consider spatial coupling and dynamic evolution mechanisms in the process of clustering and kernel density estimation, ignore the impact of risk transmission between neighboring areas and temporal trends on early warning accuracy, and cannot respond to environmental changes in a timely manner.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for early warning and analysis of urban fire hazards based on big data, which includes,

[0008] Multimodal data is collected and preprocessed to obtain feature vectors. Region clusters are generated through FCM clustering, and feature weights are obtained through entropy weighting. The Gaussian kernel function and bandwidth parameters of KDE are initialized based on the feature dimensions. The quenching function and coupling parameters are constructed as constraints in combination with GIS information. The simulated defect density is calculated. The actual critical index and standardization deviation are calculated in combination with time series. The scaling variable, dynamic threshold and comprehensive deviation index are calculated. Points greater than the threshold are screened as potential anomalies. The correlation function is calculated and the actual scaling index is extracted to identify potential odd symmetric anomalies.

[0009] The Trotter–Suzuki method is used to map the feature vectors to a quantum spin glass model to obtain the Hamiltonian and the initial transverse field strength. The transverse field strength is adjusted through iterative optimization. Potential odd symmetry anomaly subsets are screened by combining the band gap and energy fluctuations. The kernel density is calculated based on the feature weights. The kernel density statistics of the subset are used to set the comparison threshold to identify the final odd symmetry anomalies. An optimized feature vector is constructed. Based on the updated bandwidth parameter, the optimized bandwidth and optimized odd symmetry anomalies are obtained by optimizing the objective function.

[0010] The city is divided into regular grids according to the WGS84 coordinate system. The wind direction intensity field is initialized by combining optimized feature vectors, outliers and weights and the initial risk intensity is calculated. The heat diffusion equation is constructed and solved by the two-stage constant nearest neighbor method to generate a spatiotemporal risk field. The risk propagation velocity field is calculated and a three-dimensional streamline map is generated by combining fluid visualization technology. The dynamic risk threshold is calculated, the area exceeding the threshold is identified to trigger the early warning, the response priority is calculated and pushed to the fire command system via API.

[0011] As a preferred embodiment of the big data-based urban fire hazard early warning analysis method of the present invention, the calculation of simulated defect density, combined with time series calculation of actual critical index and standardized deviation, includes:

[0012] Based on GIS geographic information, a quenching function is defined, coupling parameters are calculated, and set as constraints to calculate the simulated defect density of the feature vector.

[0013] Based on the time series of eigenvectors, the actual critical index is calculated by combining simulated defect density, and the defect density deviation is calculated and standardized to obtain the standardized deviation.

[0014] As a preferred embodiment of the big data-based urban fire hazard early warning analysis method of the present invention, the step of calculating the correlation function and extracting the actual scaling index to identify potential odd-symmetric anomalies includes:

[0015] Based on the feature vectors, their correlation functions are calculated, the actual scaling exponent is calculated, and feature vectors whose difference between the actual scaling exponent and the theoretical scaling exponent is greater than the deviation threshold are marked as potential odd symmetric outliers.

[0016] As a preferred embodiment of the big data-based urban fire hazard early warning analysis method of the present invention, wherein: the step of obtaining the optimized bandwidth and optimized odd-symmetric outliers by optimizing the objective function based on the updated bandwidth parameter includes:

[0017] Based on the eigenvectors, the Trotter–Suzuki method is used to map them to a quantum spin glass model, and the Hamiltonian and initial transverse field intensity are obtained. They are adjusted through iterative optimization, and the energy fluctuations of the band gap and potential odd symmetry anomalies during the iterative optimization process are calculated. The final subset of odd symmetry anomalies and the optimized eigenvectors are obtained by screening.

[0018] Based on the optimized feature vector and the final odd symmetric outliers, the bandwidth parameter is updated to obtain the updated bandwidth. An optimization objective function is defined, and the updated bandwidth is optimized through grid search to obtain the optimized bandwidth. The kernel density is recalculated to obtain the optimized odd symmetric outliers.

[0019] As a preferred embodiment of the big data-based urban fire hazard early warning analysis method of the present invention, the step of constructing a heat diffusion equation and solving it using the two-stage constant nearest neighbor method to generate a spatiotemporal risk field includes:

[0020] The city is divided into grids based on the WGS84 coordinate system using regular grids. The initial wind direction intensity field is initialized based on optimized feature vectors, optimized odd symmetric anomalies, and feature weights, and the initial risk intensity is calculated.

[0021] Based on the initial risk intensity, a heat diffusion equation is constructed and solved using a two-stage constant neighbor method to obtain the spatiotemporal risk field and calculate the risk propagation velocity field.

[0022] As a preferred embodiment of the big data-based urban fire hazard early warning analysis method of the present invention, the step of identifying areas exceeding the threshold to trigger an early warning, calculating the response priority, and pushing the response to the fire command system via API includes:

[0023] Construct a 3D streamline diagram, calculate the risk threshold, filter out risks with intensity greater than the risk threshold, trigger an early warning, and calculate the response priority;

[0024] The warning area and its corresponding risk level, combined with the response priority, are pushed to the fire command system via API interface.

[0025] As a preferred embodiment of the big data-based urban fire hazard early warning analysis method of the present invention, the step of collecting multimodal data and preprocessing it to obtain feature vectors includes:

[0026] Multimodal sensors are deployed to collect multimodal data, and urban building information (materials), meteorological data and GIS geographic information are obtained using API interfaces to generate structured datasets. Flammability coefficients, CO concentration fluctuation rates, electrical load fluctuations and environmental risk factors are extracted and horizontally stacked to obtain feature vectors.

[0027] Secondly, this invention provides a big data-based urban fire hazard early warning and analysis system, comprising,

[0028] The data acquisition and processing module is used to collect multimodal data and preprocess it to generate feature vectors;

[0029] The preliminary identification module is used to calculate the simulated defect density, actual critical index, and standardized deviation to preliminarily identify potential risk points.

[0030] The fine screening and optimization module is used to optimize odd symmetric outliers by optimizing the objective function.

[0031] The spatiotemporal risk field generation module is used to construct the heat diffusion equation based on the optimized feature vector and outliers, generate the spatiotemporal risk field, and calculate the risk propagation velocity field.

[0032] The early warning and response module is used to generate a three-dimensional streamline diagram based on the spatiotemporal risk field, calculate the dynamic risk threshold, trigger an early warning and calculate the response priority, and push it to the fire command system via API.

[0033] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the big data-based urban fire hazard early warning analysis method described in the first aspect of the present invention.

[0034] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the big data-based urban fire hazard early warning analysis method as described in the first aspect of the present invention.

[0035] The beneficial effects of this invention are as follows: This invention initializes the Gaussian kernel function and bandwidth parameters of KDE based on feature dimensions, constructs a quenching function and coupling parameters as constraints using GIS information, calculates and simulates defect density, calculates the actual critical exponent and standardized deviation using time series, constructs a heat diffusion equation and solves it using the two-stage constant nearest neighbor method, generates a spatiotemporal risk field, and calculates the risk propagation velocity field; it improves the risk identification accuracy of high-dimensional heterogeneous data, enhances the accuracy and stability of anomaly point identification, realizes the dynamic evolution display of urban fire risk, and improves the efficiency of early warning response. Attached Figure Description

[0036] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart of the urban fire hazard early warning analysis method based on big data in Example 1.

[0038] Figure 2 This is a schematic diagram of the urban fire hazard early warning and analysis system based on big data in Example 1. Detailed Implementation

[0039] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0040] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0041] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0042] Example 1, referring to Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides a method for early warning and analysis of urban fire hazards based on big data, including the following steps:

[0043] S1. Collect multimodal data and preprocess it to obtain feature vectors. Generate region clusters through FCM clustering and obtain feature weights through entropy weight method. Initialize the Gaussian kernel function and bandwidth parameters of KDE based on feature dimensions. Combine GIS information to construct quenching function and coupling parameters as constraints. Calculate simulated defect density. Combine time series to calculate actual critical index and standardized deviation. Calculate scaling variable, dynamic threshold and comprehensive deviation index. Screen points greater than the threshold as potential anomalies. Calculate correlation function and extract actual scaling index to identify potential odd symmetric anomalies.

[0044] Specifically, multimodal data is collected and preprocessed to obtain feature vectors, including:

[0045] Deploy multimodal sensors to collect multimodal data and use API interfaces to obtain urban building information (materials), meteorological data, and GIS geographic information;

[0046] The multimodal sensor includes a temperature sensor, a smoke sensor, an infrared sensor, and a current sensor;

[0047] The multimodal data includes temperature (internal), smoke concentration, CO concentration, and electrical load data;

[0048] The meteorological data includes temperature and humidity (external) and wind speed data;

[0049] The materials were mapped to flammability coefficients according to national fire protection standards, and combined with other collected data to obtain a dataset. Linear interpolation was used to fill in the gaps, and non-coded data was standardized to generate a structured dataset.

[0050] Based on the structured dataset, flammability coefficient, CO concentration fluctuation rate (the ratio of the difference between the current time and the previous time to the time interval), electrical load fluctuation, and environmental risk factors are extracted and stacked horizontally to obtain a feature vector;

[0051] The environmental risk factor refers to the weighted fusion of temperature, humidity, temperature, and wind speed to obtain the environmental risk factor.

[0052] The electrical load fluctuation is expressed by the following formula:

[0053] ,

[0054] in, For time Electrical fluctuation load, For time The load change, in time and time The difference in electrical load, The largest load change in history was collected by a current sensor. For sliding windows The standard deviation of the load, For sliding windows The average load change.

[0055] By deploying multimodal sensors and combining them with API interfaces to acquire urban building information, meteorological data, and GIS geographic information, comprehensive collection of multi-source heterogeneous data was achieved. This enables the comprehensive capture of the dynamic characteristics of the environment, building materials, and electrical loads, providing high-dimensional and high-precision input data for fire risk assessment. Through the fusion of multimodal data, the breadth and depth of data coverage were significantly improved, reducing information bias caused by a single data source. Building materials were mapped to flammability coefficients according to national fire protection standards, and a structured dataset was generated by combining linear interpolation and standardization, facilitating subsequent algorithm processing. The application of linear interpolation effectively filled data gaps and improved the completeness of the dataset, while standardization eliminated interference from different data units, ensuring the comparability of features. Flammability coefficients, CO concentration fluctuation rates, electrical load fluctuations, and environmental risk factors were extracted and horizontally stacked into feature vectors, significantly enhancing the representational ability of the feature vectors and providing a solid foundation for subsequent fire risk zoning and anomaly detection.

[0056] Furthermore, the simulated defect density is calculated, and the actual critical exponent and standardized deviation are calculated by combining the time series data, including:

[0057] Based on feature vectors, randomly selected feature vectors are defined as initial cluster centers. The number of clusters is set based on city size. An objective function is defined, CED distance is calculated, and membership degrees are calculated. The initial cluster centers are updated until the objective function value of the current iteration and the previous iteration is less than a threshold, thus obtaining regional clusters. For example, high-density residential areas and industrial areas with similar fire trends are represented by the following formula:

[0058] ,

[0059] in, Features and characteristics The combined distance, denoted as CED, Features and characteristics The absolute distance is the absolute value of the difference between features. Features and characteristics The incremental distance is the absolute value of the difference between the previous and current times of one feature and the difference between the previous and current times of another feature. Features and characteristics The growth rate distance is the difference between the ratio of the difference between the previous and current times of one feature to the ratio of the current time of the previous feature, and the difference between the previous and current times of another feature to the ratio of the current time of the previous feature. , as well as The weights are the absolute distance, the incremental distance, and the growth rate distance, respectively. The fuzzy coefficient of FCM The objective function value, Features Belongs to cluster The membership degree, where N is the number of features. For the number of clusters, For the first 1 eigenvector For the first Cluster center, Cluster center and characteristics The overall distance;

[0060] Based on the feature vectors, feature weights are obtained using the entropy weighting method.

[0061] Based on the feature dimension of the feature vector, the kernel function (Gaussian kernel) and bandwidth parameter of KDE are initialized using the following formula:

[0062] ,

[0063] in, For Gaussian kernel function, For the first One characteristic, This refers to the dimension of the feature vector. For example, if the feature vector contains three types of data, then the dimension is 3. For dimension The One characteristic, For dimension bandwidth, For dimension The characteristic standard deviation is 4, which is a fixed constant of Silverman's rule of thumb. It is a mathematical constant;

[0064] Based on GIS geographic information, a quenching function is defined, and coupling parameters are calculated using the following formula:

[0065] The quenching function and coupling parameters are formulated as follows:

[0066] ,

[0067] in, For time The quenching function, This is the quenching time constant. It is an exponentially decaying function. Features and characteristics The coupling strength, This serves as the baseline value for coupling strength. For coupling parameters, This represents the initial value of the quenching function;

[0068] By setting the quenching function and coupling parameters as constraints, the simulated defect density of the eigenvectors is calculated using the following formula:

[0069] ,

[0070] in, Features Simulated defect density, For the first The features are denoted as N, where N is the number of features.

[0071] Based on the time series of feature vectors (obtained by horizontal stacking of timestamps), the actual critical index is calculated by combining simulated defect density, and the defect density deviation is calculated and standardized to obtain the standardized deviation, as shown in the formula:

[0072] ,

[0073] in, It is a time series. This is the actual critical index. This is the proportionality coefficient. Features The deviation of the simulated defect density.

[0074] Based on feature vectors, a fuzzy C-means clustering algorithm is used to calculate regional clusters. Combined with entropy weighting and kernel density estimation, simulated defect density is generated, enabling refined zoning and dynamic assessment of regional fire risk. The FCM algorithm, through CED distance and membership degree calculation, can adaptively identify regions with similar fire trends, thereby improving zoning accuracy. The introduction of entropy weighting ensures the objectivity of feature weights and avoids subjective biases imposed by humans. The Gaussian kernel function of KDE further optimizes the smoothness of feature distribution and improves the accuracy of simulated defect density. Based on GIS geographic information, a quenching function is defined and coupling parameters are calculated, which can effectively capture the spatial interaction effects between regions, such as the fire spread risk of neighboring areas. By using the quenching function and coupling parameters as constraints to calculate simulated defect density, and combining time series to generate actual critical index and standardized deviation, the dynamics and reliability of fire risk prediction are significantly improved. The actual critical index reflects the evolution trend of risk, while the standardized deviation quantifies the degree of abnormal fluctuations, providing a key basis for fire early warning.

[0075] Furthermore, the correlation function is calculated and the actual scaling exponent is extracted to identify potential odd-symmetric outliers, including:

[0076] Based on the scaling variable calculated using region clusters, a dynamic threshold and a comprehensive deviation index are calculated. Points with a comprehensive deviation index exceeding the dynamic threshold are identified as potential outliers. The formula is as follows:

[0077] ,

[0078] in, For scaling variables, The average Euclidean distance between features. As the reference distance, Dynamic critical index The spatial scale of the lower region cluster is calculated based on GIS latitude and longitude. Features The comprehensive deviation index, To simulate defect density The weight, Features Simulated defect density Standardization bias, For scaling variables The weight, Features scaling variable The standardized deviation is the absolute value of the difference between the current scaling variable and the historical scaling variable. After standardization, AQ is the dynamic threshold. and These are the simulated defect densities. and scaling variables Standard deviation;

[0079] Based on the feature vectors, the correlation function is calculated using the following formula:

[0080] ,

[0081] in, Time interval The correlation function, For and the Features Time interval Features Time interval The number of features within;

[0082] Based on the correlation function, the actual scaling index is calculated, and the feature vectors corresponding to the actual scaling index and the theoretical scaling index that are greater than the deviation threshold are marked as potential odd symmetric outliers.

[0083] By calculating scaling variables and dynamic thresholds to screen comprehensive deviation indicators, and extracting actual scaling indices based on correlation functions to identify potential odd-symmetric anomalies, accurate anomaly detection of fire risk is achieved. Scaling variables combined with GIS latitude and longitude information can capture dynamic changes in spatial scale, while the introduction of dynamic thresholds avoids the limitations of fixed thresholds and improves the sensitivity and adaptability of anomaly detection. The calculation of correlation functions reveals the correlation of feature vectors in the time dimension, and the deviation analysis between actual and theoretical scaling indices further enhances the accuracy of anomaly identification.

[0084] S2. The Trotter–Suzuki method is used to map the feature vectors to a quantum spin glass model to obtain the Hamiltonian and the initial transverse field strength. The transverse field strength is adjusted through iterative optimization. The potential odd symmetry anomaly subset is screened by combining the band gap and energy fluctuation. The kernel density is calculated based on the feature weights. The comparison threshold is set using the subset kernel density statistics to identify the final odd symmetry anomaly. An optimized feature vector is constructed. Based on the updated bandwidth parameter, the optimized bandwidth and optimized odd symmetry anomaly are obtained by optimizing the objective function.

[0085] Specifically, based on the updated bandwidth parameter, the optimized bandwidth and optimized odd-symmetric outliers are obtained by optimizing the objective function, including:

[0086] Based on the eigenvectors, the Trotter–Suzuki method is used to map them to a quantum spin glass model, yielding the Hamiltonian and initial transverse field strength, as shown in the following formula:

[0087] ,

[0088] in, For Hamiltonian, Features In direction spin operator, Features In direction spin operator, The transverse field strength is initially set to the coupling strength. Features In direction Spin operators;

[0089] Based on the initial transverse field intensity, adjustments are made through iterative optimization, and the energy fluctuations of the bandgap and potential odd-symmetric anomalies during the iterative optimization process are calculated, as shown in the following formula:

[0090] ,

[0091] in, For the band gap, For the first Features The covariance matrix, Features Energy fluctuations, For Hamiltonian, To extract the minimum eigenvalue from the covariance matrix, a covariance matrix is ​​constructed using the sample covariance matrix. To optimize the transverse field strength, For transverse field strength;

[0092] Select subsets of potential odd-symmetric anomalies whose energy fluctuations are greater than v times the standard deviation of the average energy fluctuations within the region cluster and whose band gaps are superalgebraically closed (exponentially decaying) during the iteration process.

[0093] Based on the eigenvectors and feature weights, the kernel density is calculated, and the mean and standard deviation of the kernel density are calculated using the following formula:

[0094] ,

[0095] in, For kernel density estimation, Features Feature weights;

[0096] Based on a subset of potential odd symmetric outliers, the difference between the mean and standard deviation of the kernel density is calculated and set as the comparison threshold. Feature vectors with kernel densities less than the comparison threshold and belonging to the subset of potential odd symmetric outliers are selected and marked as final odd symmetric outliers. Feature vectors of non-final odd symmetric outliers are selected and concatenated into optimized feature vectors.

[0097] Based on the optimized eigenvectors and the final odd symmetric outliers, the bandwidth parameters are updated to obtain the updated bandwidth. An optimization objective function is defined, and the updated bandwidth is optimized through grid search. The optimized bandwidth is obtained by setting the early stopping method. The kernel density is recalculated to obtain the optimized odd symmetric outliers. The formula is as follows:

[0098] ,

[0099] in, Features Update bandwidth, Features standard deviation This represents the final number of odd-symmetric outliers. To optimize the objective function, Let KL divergence be the KL divergence. The kernel density distribution of the final odd-symmetric anomalies. To optimize the kernel density distribution of eigenvectors, This is to minimize the update bandwidth.

[0100] By mapping complex feature vectors to a quantum spin glass model using the Trotter–Suzuki method, computational complexity is significantly reduced while preserving the core characteristics of high-dimensional data. The Hamiltonian formula provides a stable quantum state description through precise definitions of spin operators and transverse field strength, ensuring the model's robustness in high-noise environments. Iterative optimization of the transverse field strength, combined with bandgap and energy fluctuation analysis, effectively identifies odd-symmetric outliers in the data, exhibiting higher sensitivity and accuracy. By using kernel density estimation and KL divergence optimization as objective functions, the bandwidth parameter is dynamically adjusted, enabling outlier detection to adapt to different data distribution characteristics, reducing false alarm and false negative rates, and enhancing the model's adaptability in dynamic environments.

[0101] S3. Divide the city into regular grids according to the WGS84 coordinate system, combine optimized feature vectors, outliers and weights to initialize the wind direction intensity field and calculate the initial risk intensity, construct the heat diffusion equation and solve it through the two-stage constant nearest neighbor method, generate the spatiotemporal risk field, calculate the risk propagation velocity field, and combine fluid visualization technology to generate a three-dimensional streamline map, calculate the dynamic risk threshold, identify areas exceeding the threshold to trigger early warning, calculate the response priority, and push it to the fire command system through API.

[0102] Specifically, the heat diffusion equation is constructed and solved using the two-stage constant nearest neighbor method to generate a spatiotemporal risk field, including:

[0103] The city is divided into grids based on the WGS84 coordinate system using a regular grid. An initial wind direction intensity field is initialized based on optimized eigenvectors, optimized odd-symmetric anomalies, and feature weights, and the initial risk intensity is calculated using the following formula:

[0104] ,

[0105] in, For grid points At the initial time The initial risk intensity, To optimize odd symmetric outliers Feature weights, To optimize odd symmetric outliers eigenvalues, Initial spatial decay parameters, To optimize the number of odd-symmetric outliers, For the exponent term, To optimize odd symmetric outliers x-coordinate To optimize odd symmetric outliers The ordinate;

[0106] Based on the initial risk intensity, a thermal diffusion equation is constructed and solved using a two-stage constant neighbor method to obtain the spatiotemporal risk field, as shown in the formula:

[0107] ,

[0108] in, For the spatial Laplace operator of risk intensity, The symbol for partial derivative is , indicating a partial derivative. The dynamic diffusion coefficient is... For the source term weight, For the intensity of the risk source, Let Dirac be the function used to optimize odd symmetric outliers. The coordinate is 1 if the coordinate is 1, otherwise it is 0. As the reference diffusion coefficient, To normalize terrain factors, such as slope and building height, data is collected via GIS. For time Grid points Normalized wind speed, This is the terrain attenuation coefficient. This is the wind speed enhancement coefficient;

[0109] Based on the spatiotemporal risk field, the risk propagation velocity field is calculated using the following formula:

[0110] ,

[0111] in, For the risk propagation speed field, The risk gradient is obtained based on the central difference algorithm. For wind speed vectors, This represents the wind speed influence coefficient.

[0112] Based on the WGS84 coordinate system, regular grid partitioning, combined with optimized feature vectors and outliers, can accurately capture the spatial distribution characteristics of risks in urban environments. The initial risk intensity formula, through the introduction of feature weights and spatial attenuation parameters, significantly improves the model's adaptability to complex terrain and environmental factors. The heat diffusion equation, combined with the two-stage constant nearest neighbor method, can quickly solve high-dimensional spatiotemporal risk fields, improving computational efficiency while maintaining high accuracy, making it suitable for real-time application scenarios. By calculating the risk propagation velocity field and combining wind speed and terrain factors, the model can dynamically capture the propagation path and speed of risks, providing key information for rapid response.

[0113] Furthermore, the system identifies areas exceeding thresholds that trigger early warnings, calculates response priorities, and pushes these warnings to the fire command system via API, including:

[0114] Based on the risk propagation velocity field, a three-dimensional streamline map is generated through fluid visualization technology. This includes defining optimized odd-symmetric anomaly points as streamline starting points, encoding risk intensity as streamline color through segmented color mapping (e.g., if the risk intensity is greater than the high risk intensity threshold, it is red; if the risk intensity is less than or equal to the high risk intensity threshold and less than the low risk intensity threshold, it is yellow; otherwise, it is green), and defining the risk propagation velocity field as streamline density (e.g., high density for rapid propagation).

[0115] Based on the 3D streamline diagram, the risk threshold is calculated using the following formula:

[0116] ,

[0117] in, For time The risk threshold, For time The mean of risk intensity, For time Standard deviation of risk intensity This is the adjustment coefficient;

[0118] If the risk intensity exceeds the risk threshold, an alert is triggered, and the response priority is calculated using the following formula:

[0119] ,

[0120] in, For grid Priority For grid The intensity of the risk, For grid and neighboring areas The overall distance between For the neighborhood The environmental disturbance index, For grid The neighborhood set;

[0121] The neighborhood refers to a grid. Centered on a specific point in the scene, a radius is defined based on the actual scenario; all grids within this radius are considered grids. The field;

[0122] The warning area and its corresponding risk level, combined with the response priority, are pushed to the fire command system via API interface.

[0123] The 3D streamline diagram transforms complex risk data into intuitive visualizations through color coding and density mapping, enabling decision-makers to quickly understand risk distribution and propagation trends, thereby improving decision-making efficiency. The calculation of dynamic risk thresholds, which combines the mean and standard deviation over time, can adaptively adjust thresholds based on real-time environmental changes, reducing the possibility of false alarms and missed alarms. The response priority formula comprehensively considers risk intensity, distance, and environmental interference, enabling the scientific allocation of rescue resources and ensuring that high-risk areas receive priority responses, thus optimizing resource utilization efficiency.

[0124] This embodiment also provides a big data-based urban fire hazard early warning and analysis system, including:

[0125] The data acquisition and processing module is used to collect multimodal data and preprocess it to generate feature vectors;

[0126] The preliminary identification module is used to calculate the simulated defect density, actual critical index, and standardized deviation to preliminarily identify potential risk points.

[0127] The fine screening and optimization module is used to optimize odd symmetric outliers by optimizing the objective function.

[0128] The spatiotemporal risk field generation module is used to construct the heat diffusion equation based on the optimized feature vector and outliers, generate the spatiotemporal risk field, and calculate the risk propagation velocity field.

[0129] The early warning and response module is used to generate a three-dimensional streamline diagram based on the spatiotemporal risk field, calculate the dynamic risk threshold, trigger an early warning and calculate the response priority, and push it to the fire command system via API.

[0130] This embodiment also provides a computer device applicable to the urban fire hazard early warning and analysis method based on big data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the urban fire hazard early warning and analysis method based on big data proposed in the above embodiment.

[0131] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0132] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the big data-based urban fire hazard early warning analysis method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0133] In summary, this invention initializes the Gaussian kernel function and bandwidth parameters of KDE based on feature dimensions, constructs a quenching function and coupling parameters as constraints using GIS information, calculates and simulates defect density, calculates the actual critical exponent and standardized deviation using time series, constructs a heat diffusion equation and solves it using the two-stage constant nearest neighbor method, generates a spatiotemporal risk field, and calculates the risk propagation velocity field; thus improving the risk identification accuracy of high-dimensional heterogeneous data, enhancing the accuracy and stability of anomaly point identification, realizing the dynamic evolution display of urban fire risk, and improving early warning response efficiency.

[0134] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for early warning and analysis of urban fire hazards based on big data, characterized in that: include, Multimodal data is collected and preprocessed to obtain feature vectors. Region clusters are generated through FCM clustering, and feature weights are obtained through entropy weighting. The Gaussian kernel function and bandwidth parameters of KDE are initialized based on the feature dimensions. The quenching function and coupling parameters are constructed as constraints in combination with GIS information. The simulated defect density is calculated. The actual critical index and standardized deviation are calculated in combination with time series. The scaling variable, dynamic threshold and comprehensive deviation index are calculated. The comprehensive deviation index is screened and marked as potential anomalies. The correlation function is calculated and the actual scaling index is extracted to identify potential odd symmetric anomalies. The quenching function and coupling parameters are formulated as follows: ， in, For time The quenching function, This is the quenching time constant. It is an exponentially decaying function. Features and characteristics The coupling strength, This serves as the baseline value for coupling strength. For coupling parameters, This represents the initial value of the quenching function; The simulated defect density is given by the following formula: ， in, Features Simulated defect density, For the first The features are denoted as N, where N is the number of features. The calculation of the correlation function and extraction of the actual scaling index to identify potential odd symmetric outliers includes calculating the correlation function based on the feature vector, as shown in the formula: ， in, Time interval The correlation function, For and the Features Time interval Features Time interval The number of features within; Based on the correlation function, the actual scaling index is calculated, and the feature vectors corresponding to the actual scaling index and the theoretical scaling index that are greater than the deviation threshold are marked as potential odd symmetric outliers. The Trotter–Suzuki method is used to map the feature vectors to a quantum spin glass model to obtain the Hamiltonian and the initial transverse field strength. The transverse field strength is adjusted through iterative optimization. A subset of potential odd symmetry anomalies is screened by combining the band gap and energy fluctuation. The kernel density is calculated based on the feature weights. The difference between the mean and standard deviation of the kernel density of the subset of potential odd symmetry anomalies is set as the comparison threshold to identify the final odd symmetry anomalies. An optimized feature vector is constructed. Based on the updated bandwidth parameter, the optimized bandwidth and optimized odd symmetry anomalies are obtained by optimizing the objective function. The process of selecting a subset of potential odd-symmetric anomalies by combining band gap and energy fluctuation includes iterative optimization based on the initial transverse field intensity, and calculating the energy fluctuations of the band gap and potential odd-symmetric anomalies during the iterative optimization process, as shown in the formula: , in, For the band gap, For the first Features The covariance matrix, Features Energy fluctuations, For Hamiltonian, To extract the minimum eigenvalue from the covariance matrix, a covariance matrix is ​​constructed using the sample covariance matrix. To optimize the transverse field strength, For transverse field strength; Select subsets of potential odd-symmetric outliers whose energy fluctuations are greater than v times the standard deviation of the average energy fluctuations within the region cluster and whose band gaps are superalgebraically closed during the iteration process. The process of identifying final odd symmetry outliers and constructing optimized feature vectors includes selecting feature vectors whose kernel density is less than the comparison threshold and whose feature vectors belong to a subset of potential odd symmetry outliers, marking them as final odd symmetry outliers, and selecting feature vectors of non-final odd symmetry outliers and concatenating them into optimized feature vectors. The city is divided into regular grids according to the WGS84 coordinate system. The wind direction intensity field is initialized by combining optimized feature vectors, optimized odd symmetric anomalies and feature weights, and the initial risk intensity is calculated. The heat diffusion equation is constructed and solved by the two-stage constant nearest neighbor method to generate a spatiotemporal risk field. The risk propagation velocity field is calculated. A three-dimensional streamline map is generated by combining fluid visualization technology. The risk threshold is calculated. Areas with risk intensity greater than the risk threshold are identified to trigger early warning. The response priority is calculated and pushed to the fire command system via API. The formula for initializing the wind direction intensity field and calculating the initial risk intensity by combining optimized feature vectors, optimized odd-symmetric outliers, and feature weights is as follows: , in, For grid points At the initial time The initial risk intensity, To optimize odd symmetric outliers Feature weights, To optimize odd symmetric outliers eigenvalues, Initial spatial decay parameters, To optimize the number of odd-symmetric outliers, For the exponent term, To optimize odd symmetric outliers x-coordinate To optimize odd symmetric outliers The ordinate.

2. The urban fire hazard early warning and analysis method based on big data as described in claim 1, characterized in that: The calculation of the simulated defect density, combined with the time series calculation of the actual critical exponent and standardized deviation, includes: Based on GIS geographic information, a quenching function is defined, coupling parameters are calculated, and set as constraints to calculate the simulated defect density of the feature vector. Based on the time series of eigenvectors, the actual critical index is calculated by combining simulated defect density, and the defect density deviation is calculated and standardized to obtain the standardized deviation.

3. The urban fire hazard early warning and analysis method based on big data as described in claim 2, characterized in that: The process of obtaining optimized bandwidth and optimized odd-symmetric outliers based on updated bandwidth parameters and by optimizing the objective function includes: Based on the eigenvectors, the Trotter–Suzuki method is used to map them to a quantum spin glass model, and the Hamiltonian and initial transverse field intensity are obtained. They are adjusted through iterative optimization, and the energy fluctuations of the band gap and potential odd symmetry anomalies during the iterative optimization process are calculated. The final subset of odd symmetry anomalies and the optimized eigenvectors are obtained by screening. Based on the optimized feature vector and the final odd symmetric outliers, the bandwidth parameter is updated to obtain the updated bandwidth. An optimization objective function is defined, and the updated bandwidth is optimized through grid search to obtain the optimized bandwidth. The kernel density is recalculated to obtain the optimized odd symmetric outliers.

4. The urban fire hazard early warning and analysis method based on big data as described in claim 3, characterized in that: The process of constructing the thermal diffusion equation and solving it using the two-stage constant nearest neighbor method to generate a spatiotemporal risk field includes: The city is divided into grids based on the WGS84 coordinate system using regular grids. The initial wind direction intensity field is initialized based on optimized feature vectors, optimized odd symmetric anomalies, and feature weights, and the initial risk intensity is calculated. Based on the initial risk intensity, a heat diffusion equation is constructed and solved using a two-stage constant neighbor method to obtain the spatiotemporal risk field and calculate the risk propagation velocity field.

5. The urban fire hazard early warning and analysis method based on big data as described in claim 4, characterized in that: The process of identifying areas with risk intensity greater than a risk threshold to trigger an early warning, calculating response priority, and pushing the warning to the fire command system via API includes: Construct a 3D streamline diagram, calculate the risk threshold, filter out risks with intensity greater than the risk threshold, trigger an early warning, and calculate the response priority; The warning area and its corresponding risk level, combined with the response priority, are pushed to the fire command system via API interface.

6. The urban fire hazard early warning and analysis method based on big data as described in claim 1, characterized in that: The process of collecting and preprocessing multimodal data to obtain feature vectors includes: Multimodal sensors are deployed to collect multimodal data, and urban building information, meteorological data, and GIS geographic information are obtained using API interfaces to generate structured datasets. Flammability coefficients, CO concentration fluctuation rates, electrical load fluctuations, and environmental risk factors are extracted and horizontally stacked to obtain feature vectors.

7. A big data-based urban fire hazard early warning and analysis system, based on the big data-based urban fire hazard early warning and analysis method according to any one of claims 1 to 6, characterized in that: include, The data acquisition and processing module is used to collect multimodal data and preprocess it to generate feature vectors; The preliminary identification module is used to calculate the simulated defect density, actual critical index, and standardized deviation to preliminarily identify potential risk points. The fine screening and optimization module is used to optimize odd symmetric outliers by optimizing the objective function. The spatiotemporal risk field generation module is used to construct the heat diffusion equation based on the optimized feature vector and outliers, generate the spatiotemporal risk field, and calculate the risk propagation velocity field. The early warning and response module is used to generate a three-dimensional streamline diagram based on the spatiotemporal risk field, calculate the dynamic risk threshold, trigger an early warning and calculate the response priority, and push it to the fire command system via API.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the big data-based urban fire hazard early warning and analysis method as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the big data-based urban fire hazard early warning and analysis method as described in any one of claims 1 to 6.

Images