A city fire remote monitoring and early warning method and system

Abstract

This invention discloses a remote monitoring and early warning method and system for urban fire protection, relating to the field of fire early warning technology. The method includes the following steps: acquiring the gas concentration time series and vehicle exhaust time series of the target object; performing parallel hierarchical detection on the gas concentration time series and the vehicle exhaust time series to generate anomaly marker sequences; constructing an objective function using the gas concentration time series, the vehicle exhaust time series, and the anomaly marker sequences as constraints, and solving it using spatiotemporal gradient inversion and vehicle coupling terms to generate implicit flow field estimates and their variance fields. This scheme uses Bayesian updates to dynamically adjust the risk estimate as a correction term, solving the early warning bias problem caused by the lack of dynamic coupling of vehicle movement influence in existing technologies. Furthermore, it utilizes a Markov model to capture the temporal evolution of the risk field, improving the prediction accuracy of the risk evolution path and providing a dynamically updated decision-making basis for fire early warning.

Description

Technical Field

[0001] This invention relates to the field of fire early warning technology, specifically to a method and system for remote monitoring and early warning of urban fires. Background Technology

[0002] With the acceleration of urbanization, urban waste transfer stations, landfills, and waste transfer links have gradually become an indispensable part of urban operations. In these places, due to the long-term accumulation of waste, the easy decomposition of waste to produce combustible gases, and the frequent entry and exit of transport vehicles, the fire hazard has increased significantly. In such environments, urban fire monitoring systems not only need to detect combustible gas concentrations in real time, but also need to combine dynamic vehicle emission information to judge potential fire risks and issue early warnings in order to achieve rapid response and effective prevention and control of fires.

[0003] However, in the process of implementing the technical solution of the invention in the embodiments of this application, it was found that the above-mentioned technology has at least the following technical problems: In existing technologies, fire risk assessment schemes based on multi-source data fusion are used for urban fire protection. These schemes involve inputting multi-dimensional information from combustible gas concentration sensors, temperature sensors, and smoke sensors, and generating risk indicators or risk maps through data fusion algorithms. The advantage of this approach is that it can comprehensively assess potential risks and improve the accuracy and reliability of early warnings. However, the multi-source data fusion methods mentioned above often rely on manually set weights or static models, making it difficult to reflect the dynamic diffusion process of combustible gas leaks in real time and the impact of emissions from moving garbage trucks. Furthermore, the lack of dynamic coupling and iteration in the leak diffusion calculation makes it difficult to accurately simulate the impact of vehicle movement on risk areas, resulting in insufficient accuracy of the final fire warning scheme. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for remote monitoring and early warning of urban fires, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows: In a first aspect, the present invention discloses a method for remote monitoring and early warning of urban fires, comprising the following steps: Obtain the gas concentration time series and vehicle exhaust time series of the target object; Parallel hierarchical detection is performed on the gas concentration time series and the vehicle exhaust time series to generate anomaly marker sequences; An objective function is constructed using the gas concentration time series, the vehicle exhaust time series, and the anomaly marker sequence as constraints. The spatiotemporal gradient inversion and vehicle coupling term are then solved to generate the implicit flow field estimate and its variance field. The initial leakage diffusion field is solved based on the gas concentration time series, the implicit flow field estimation, and the anomaly marker sequence. The vehicle influence field is calculated within the vehicle trajectory time window based on the vehicle row time series and anomaly marker sequence, and used as a moving source correction term to iteratively update the initial leakage diffusion field until a preset number of iterations is reached to generate the leakage diffusion field. Based on the leakage diffusion field, the vehicle impact field, the anomaly marker sequence, and the variance field, a linkage risk index matrix is ​​calculated and output based on Bayesian update, and mapped using the anomaly marker sequence to generate a risk evolution probability field. Based on the risk evolution probability field, the linkage risk index matrix, and the anomaly marker sequence, a priority response ranking table is generated through multidimensional scoring, and then matched with a preset set of on-site executable actions to generate a fire early warning scheme.

[0006] Secondly, the present invention discloses an urban fire remote monitoring and early warning system, comprising: The data acquisition module is used to acquire the gas concentration time series and vehicle exhaust time series of the target object; An anomaly detection module is used to perform parallel hierarchical detection on the gas concentration time series and the vehicle exhaust time series to generate an anomaly marker sequence; The data solving module is used to construct an objective function with the gas concentration time series, the vehicle exhaust time series and the anomaly marker sequence as constraints, and to solve the spatiotemporal gradient inversion and vehicle coupling term to generate implicit flow field estimation and its variance field. The leakage diffusion analysis module is used to solve the initial leakage diffusion field based on the gas concentration time series, the implicit flow field estimation, and the anomaly marker sequence. The vehicle influence field is calculated within the vehicle trajectory time window based on the vehicle row time series and anomaly marker sequence, and used as a moving source correction term to iteratively update the initial leakage diffusion field until a preset number of iterations is reached to generate the leakage diffusion field. The risk analysis module is used to calculate and output the linkage risk index matrix based on Bayesian update according to the leakage diffusion field, the vehicle impact field, the anomaly marker sequence and the variance field, and to generate a risk evolution probability field by mapping the anomaly marker sequence. The early warning scheme generation module is used to generate a priority response ranking table based on the risk evolution probability field, the linkage risk index matrix and the anomaly marker sequence through multi-dimensional scoring, and match it with a preset set of on-site executable actions to generate a fire early warning scheme.

[0007] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This solution uses Bayesian updates to dynamically adjust the risk estimate by treating the vehicle influence field as a correction term, thus solving the problem of early warning deviation caused by the lack of dynamic coupling of vehicle movement influence in existing technologies. Furthermore, it utilizes Markov models to capture the temporal evolution of the risk field, improving the prediction accuracy of the risk evolution path and providing a dynamically updated decision-making basis for fire early warning.

[0008] 2. This scheme establishes a dynamic correlation between the mobile source and the diffusion field directly during the inversion process by jointly solving the spatiotemporal gradient inversion and the vehicle coupling term. It can accurately reconstruct the dynamic diffusion process of combustible gas in complex urban environments, quantify the impact of vehicle mobile emissions on leakage diffusion in real time, and use anomaly marker sequences to constrain the inversion time window to effectively distinguish between normal emissions and abnormal leakage events, avoid misjudgment, and significantly improve the fire early warning scheme's response capability to dynamic risk sources. Attached Figure Description

[0009] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals refer to the same parts. Wherein: Figure 1 This is a flowchart illustrating the steps of a remote monitoring and early warning method for urban fire protection according to the present invention. Figure 2 A schematic diagram of the process for generating implicit flow field estimates and their variance fields provided by the present invention; Figure 3 A schematic diagram of the process for generating a leakage diffusion field provided by the present invention; Figure 4 This invention provides a schematic diagram of the module functions of a remote monitoring and early warning system for urban fire protection. Detailed Implementation

[0010] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0011] Application Overview: In existing technologies, fire monitoring at urban waste disposal sites largely relies on the fusion of static sensor data and risk assessment through preset weights or fixed models. Such methods are difficult to reflect the dynamic diffusion process of combustible gases in real time and are difficult to effectively capture the impact of transport vehicle movement on leakage diffusion, resulting in delayed early warning response or misjudgment.

[0012] To address the aforementioned issues, the study found that existing technologies lack a dynamic coupling mechanism for mobile pollution sources, leading to errors in the calculation of leakage diffusion fields. Through analysis, it was discovered that there is a spatiotemporal correlation between vehicle trajectories and combustible gas concentrations, necessitating the establishment of a vehicle influence field as a dynamic correction term. Furthermore, considering the mutual feedback between leakage diffusion and vehicle emissions, an alternating iterative update method was proposed. To address the confidence issue in risk prediction, a Bayesian framework was introduced to fuse variance field information, enabling dynamic evolution modeling of risk probabilities.

[0013] After introducing the basic concept of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0014] Example 1:

[0015] Please see Figures 1-3 A method for remote monitoring and early warning of urban fires includes the following steps: Obtain the gas concentration time series and vehicle exhaust time series of the target object; Parallel hierarchical detection is performed on the gas concentration time series and the vehicle exhaust time series to generate anomaly marker sequences; An objective function is constructed using the gas concentration time series, vehicle exhaust time series, and anomaly marker sequence as constraints. The spatiotemporal gradient inversion and vehicle coupling term are then solved to generate the implicit flow field estimate and its variance field. The initial leakage diffusion field is solved based on the gas concentration time series, implicit flow field estimation, and anomaly marker sequence. The vehicle influence field is calculated within the vehicle trajectory time window based on the vehicle row time series and anomaly marker sequence, and is used as a moving source correction term to iteratively update the initial leakage diffusion field until the preset number of iterations is reached to generate the leakage diffusion field. Based on the leakage diffusion field, vehicle impact field, anomaly marker sequence and variance field, a linkage risk index matrix is ​​calculated and output based on Bayesian update, and mapped by combining anomaly marker sequence to generate a risk evolution probability field. Based on the risk evolution probability field, the linkage risk index matrix, and the anomaly marker sequence, a priority response ranking table is generated through multi-dimensional scoring, and then matched with a preset set of on-site executable actions to generate a fire early warning plan.

[0016] Parallel stratified detection refers to calculating short-term mutation scores, persistent trend scores, and co-mutation scores for gas concentration and vehicle exhaust time series respectively, and identifying composite anomaly patterns through structured combination. Specifically, it can be achieved by using multi-scale decomposition combined with covariance analysis. Spatiotemporal gradient inversion refers to the inversion of the implicit flow field distribution based on spatial temporal gradient data through regularized least squares optimization. Specifically, it can be achieved by combining finite element discretization with gradient descent algorithm. The vehicle coupling term refers to the introduction of vehicle trajectory time window constraints during the inversion process, using the vehicle queue time series as a dynamic boundary constraint. Specifically, it can be implemented by embedding the objective function using the Lagrange multiplier method. The alternating iteration method refers to using the vehicle influence field and the leakage diffusion field as inputs for cyclic correction, which can be implemented using the Jacobi iteration method. Bayesian update calculation refers to using the leakage diffusion field as the likelihood estimate, the variance field as the confidence correction factor, and combining the anomaly marker sequence to update the risk probability. Specifically, it can be implemented using the Markov chain Monte Carlo sampling method.

[0017] This solution dynamically corrects the impact of mobile sources on leakage diffusion by using vehicle coupling terms and an alternating iteration mechanism. By introducing a Bayesian framework to fuse variance field information, it significantly improves the reliability of risk prediction. A priority ranking table is generated through multi-dimensional scoring, achieving precise matching between response strategies and real-time risk conditions. Through these technical solutions, this application solves the problem of dynamically coupling vehicle mobile emissions with leakage diffusion fields, improving the accuracy of leak source location in complex environments. The alternating iteration update mechanism effectively reduces the calculation deviation of the diffusion field caused by mobile source interference. The risk probability evolution model based on the Bayesian framework enhances the robustness of early warning decisions to uncertainties. The final priority response ranking table can dynamically adjust the response strategy according to the real-time risk situation, improving the timeliness and operability of fire early warning.

[0018] The following describes how to obtain the gas concentration time series and vehicle exhaust time series of the target object: Obtain the combustible gas concentration and vehicle emission status of the target object; Combustible gas concentration and vehicle emission status are decomposed into short, medium and long scales respectively. Based on the preset inter-scale mutual verification rules, missing values ​​are reconstructed and outliers are corrected by using each other's sequences as conditions, generating gas concentration time series and vehicle emission time series.

[0019] Among them, short, medium and long scale decomposition refers to decomposing time series data into components with different time resolutions. Specifically, it can be achieved by wavelet transform or empirical mode decomposition methods. By separating the fluctuation characteristics of different time scales, it is possible to capture sudden changes, periodic trends and long-term evolution patterns respectively. Inter-scale verification rules refer to using the statistical correlation between the decomposition results of different scales to verify data. Specifically, cross-validation or conditional probability models can be used to achieve dynamic filling of missing values ​​and collaborative correction of outliers by matching short-scale outliers with medium- and long-scale trends.

[0020] The above content will be described in detail below: Combustible gas concentration and vehicle emission status data are decomposed into short, medium and long time scale components. The short-scale component is used to detect sudden concentration fluctuations, the medium-scale component reflects periodic emission patterns, and the long-scale component characterizes long-term trends such as infrastructure aging. When any scale component has missing data or outliers, the system uses the statistical characteristics of other scale components to reconstruct the conditions through preset mutual verification rules. For example, when short-scale gas concentration data is missing, the system uses the periodic characteristics of medium-scale vehicle exhaust data to deduce the possible gas concentration range for that period, and combines the long-scale trend to impose boundary constraints, thereby achieving multi-dimensional compensation for missing values ​​and generating gas concentration time series and vehicle exhaust time series.

[0021] This solution establishes a correlation framework across different time dimensions through multi-scale decomposition. When correcting data, it retains the characteristics of short-term sudden fire warnings while avoiding trend distortion caused by misjudgment at a single scale, significantly improving the physical rationality of data reconstruction. Through the above technical solution, this application effectively solves the problem of data loss and anomalies caused by sensor failure or environmental interference in waste transfer scenarios. The multi-scale collaborative verification mechanism ensures the spatiotemporal consistency of gas concentration and vehicle exhaust data, providing a highly reliable input data foundation for the accurate inversion of the subsequent leakage diffusion field.

[0022] The following describes the parallel hierarchical detection performed on the gas concentration time series and vehicle exhaust time series to generate anomaly marker sequences, specifically including: Short-term abrupt change scores and persistent trend scores were calculated for the gas concentration time series and vehicle exhaust time series, respectively, using the following specific formulas: , ,

[0023] In the formula, Let represent the short-time mutation fraction at time t, and d represent the set containing both gas concentration time series and vehicle exhaust time series. This represents the value of the set at time t. This represents the value of the set at time t-1. The standard deviation of a set The value represents the persistence score at time t, W represents the size of the trend analysis window, and k represents time k within the trend analysis window. All the above data have been normalized during the calculation. The co-mutation fractions of gas concentration time series and vehicle exhaust time series within the same time window were detected, and the specific calculation formula is as follows: ,

[0024] In the formula, This represents the co-mutation score at time t. Indicates the gas concentration time series in Short-term mutation fraction, Indicates the time series of vehicles in The short-term mutation fractions at time points are all normalized during the calculation. Short-term mutation scores, persistent trend scores, and co-mutation scores are structurally combined to generate anomalous marker sequences.

[0025] Among them, the short-time mutation fraction refers to the instantaneous change in concentration or emission at adjacent time points calculated by the sliding window difference method. For example, the first derivative combined with the standard deviation threshold can be used to judge instantaneous abnormal fluctuations. The persistence score refers to the degree of long-term deviation of a time series analyzed by linear regression or moving average methods. For example, the slope of the trend line can be compared with the historical baseline to assess persistent anomalies. Comutation score refers to the strength of associated anomalies between two sequences within the same time window, calculated based on mutual information or covariance matrix. For example, dynamic time warping algorithm can be used to match the mutation phase difference between two sequences.

[0026] The above content will be described in detail below: In the parallel stratified detection process, short-term window scanning is first performed on the gas concentration time series and vehicle emission time series, for example, the concentration gradient and emission intensity change rate are calculated at 5-minute intervals. When the change rate exceeds a preset threshold, it is marked as a short-term mutation event. At the same time, trend decomposition is performed on the two series, for example, the trend term is separated by Hodrick-Prescott filtering. When the trend term deviates from the historical mean by more than 3 times the standard deviation, it is marked as a persistent anomaly. Furthermore, within the period when short-term mutations are detected, the mutation phase matching degree of the two series is calculated by dynamic time warping algorithm. When the matching degree is higher than 0.8, it is judged as a co-mutation event. Finally, the three types of detection results are weighted and fused to generate a binary anomaly label sequence, for example, short-term mutation accounts for 40% weight, persistent trend accounts for 30% weight, and co-mutation accounts for 30% weight. Periods with a comprehensive score of more than 0.6 are marked as anomalies.

[0027] This solution effectively distinguishes between transient interference and genuine leakage events through a structured combination of three detection dimensions: short-term, continuous, and collaborative. Furthermore, collaborative mutation detection addresses the lack of spatiotemporal correlation between vehicle movement and gas diffusion in traditional isolated detection. Through these technical solutions, this application resolves the problem of misjudgment of anomalies caused by insufficient dynamic coupling analysis in existing technologies. The multi-dimensional detection mechanism improves the accuracy of anomaly labeling. This method combines gas concentration trend stability with collaborative mutation detection results to accurately exclude non-leakage anomalies, avoiding false alarms. Simultaneously, the structured combination strategy enhances the ability to identify anomalies in complex environments, providing a reliable data foundation for the accurate calculation of the subsequent leakage diffusion field.

[0028] The following describes the process of constructing an objective function using gas concentration time series, vehicle exhaust time series, and anomaly marker sequences as constraints, and then solving it using spatiotemporal gradient inversion and vehicle coupling terms to generate an implicit flow field estimate and its variance field. Specifically, this includes: The spatial temporal gradient is calculated based on the gas concentration time series and the vehicle exhaust time series. The specific calculation formula is as follows: ,

[0029] In the formula, Represents the spatial-temporal gradient at time t. This represents the difference in the time series of gas concentration. This represents the difference in the time series of vehicle rows. All the above data have been normalized during the calculation. The spatial temporal gradient inversion method is used to invert the gradient, and anomaly marker sequences are used as time window constraints during the inversion process. Vehicle contribution coupling terms are also added. The implicit flow field estimate and corresponding variance field are obtained by regularized least squares optimization. The specific calculation formula is as follows: , ,

[0030] In the formula, This represents the implicit flow field estimation, where A represents the mapping operator between the gradient and the flow field, V represents the flow field variables, and R represents the regularization matrix. The time window defined by the anomaly marker sequence is represented, and C represents the vehicle contribution coupling term. This represents the regularization coefficient. All the data above have been normalized during the calculation.

[0031] Among them, the spatial temporal gradient refers to the rate of change of combustible gas concentration and vehicle emission status in the spatial and temporal dimensions. Specifically, the difference method or gradient operator can be used to calculate the time series point by point. The spatiotemporal gradient inversion method refers to reconstructing the implicit flow field distribution by inversely solving the gas diffusion equation and the vehicle emission coupling equation. Specifically, the finite element method or particle tracking algorithm can be used, combined with regularization constraints to suppress noise interference in the inversion process. The vehicle contribution coupling term refers to the mathematical expression that quantifies the impact of vehicle emissions on the gas diffusion process. Specifically, it can be achieved by using a trajectory-based emission intensity model or a dynamic source term superposition method to map the vehicle movement path and emission rate into a flow field correction term. Regularized least squares optimization refers to the least squares objective function that introduces a Tikhonov regularization term. Specifically, the smoothness of the flow field can be constrained by the L2 norm, balancing data fitting and model stability, and preventing overfitting.

[0032] The above content will be described in detail below: Spatial-temporal gradients extract the dynamic characteristics of gas diffusion rate and vehicle emissions by calculating the differences between gas concentration time series and vehicle emission time series at adjacent time points and on spatial grids. The spatiotemporal gradient inversion method uses gradient data as input to construct a joint inversion model that includes gas diffusion equations and vehicle emission equations. Anomaly marker sequences are used to limit the inversion time window range and exclude interference from non-leakage events. Vehicle contribution coupling terms are embedded in the inversion model, and the emission intensity corresponding to the vehicle trajectory is transformed into a flow field correction factor to achieve dynamic coupling between the mobile source and the static diffusion field. Regularized least squares optimization solves the objective function iteratively, while satisfying gradient matching and flow field smoothness requirements. Finally, the implicit flow field estimate and its variance field are output. The variance field is used to characterize the uncertainty of the inversion results.

[0033] This solution establishes a dynamic correlation between the mobile source and the diffusion field directly during the inversion process by jointly solving the spatiotemporal gradient inversion and the vehicle coupling term. Simultaneously, it utilizes anomaly marker sequences to constrain the inversion time window, effectively distinguishing between normal emissions and abnormal leakage events and avoiding misjudgments. Through these technical solutions, this application can accurately reconstruct the dynamic diffusion process of combustible gases in complex urban environments, quantify the impact of vehicle mobile emissions on leakage diffusion in real time, and assess the reliability of the inversion results through variance field evaluation. This provides a high-precision flow field data foundation for subsequent risk prediction and significantly improves the response capability of fire early warning schemes to dynamic risk sources.

[0034] The following describes how to calculate the vehicle influence field within the vehicle trajectory time window based on the vehicle queue time series and anomaly marker sequence, and how to use this as a moving source correction term to iteratively update the initial leakage diffusion field until a preset number of iterations is reached. The specific steps for generating the leakage diffusion field include: Calculate the vehicle influence field within the vehicle trajectory time window based on the vehicle time sequence; The vehicle influence field is fed back to the initial leakage diffusion field as a moving source correction term in an alternating iterative manner, and the local cumulative concentration of the initial leakage diffusion field is fed back to the vehicle influence field as a static basis, until the preset number of iterations is reached to generate the leakage diffusion field.

[0035] Among them, the vehicle trajectory time window refers to the time interval dynamically defined based on the vehicle's movement path and emission status, which can be implemented by using a sliding window algorithm combined with vehicle GPS trajectory data. The vehicle impact field refers to the local disturbance field generated by the vehicle's movement on the diffusion of combustible gas, which can be specifically realized by calculating the convolution of the vehicle's emission intensity and trajectory spatial distribution. The mobile source correction term refers to the intermediate variable that introduces the vehicle influence field into the leakage diffusion calculation in a dynamic superposition manner. Specifically, it can be embedded into the diffusion equation in the form of a weighted residual term. Alternating iteration refers to a calculation mode in which parameters are adjusted between the vehicle influence field and the leakage diffusion field through a two-way feedback mechanism. Specifically, it can be implemented by using a fixed-point iterative algorithm combined with convergence condition judgment.

[0036] The above content will be described in detail below: The vehicle trajectory time window is defined based on the timestamps in the vehicle emission time series and the high anomaly score segments in the anomaly marker sequence, ensuring that the vehicle influence field is calculated within the critical time period. During the calculation of the vehicle influence field, the emission intensity data in the vehicle emission time series is spatiotemporally interpolated with the spatial coordinates of the vehicle trajectory to generate a mobile source distribution matrix that varies with time. In each iteration, the initial leakage diffusion field is superimposed on the vehicle influence field as an external input term into the diffusion equation. At the same time, the region in the leakage diffusion field where the local cumulative concentration exceeds the preset screening threshold is fed back to the vehicle influence field as a static basis to suppress the concentration amplification in the overlapping area of ​​the vehicle trajectory. The iteration termination condition can be a preset number of iterations, such as 5, or it can automatically terminate based on the difference between the leakage diffusion fields of two adjacent iterations being less than a set threshold.

[0037] This scheme achieves bidirectional coupling between the vehicle influence field and the leakage diffusion field through an alternating iterative approach. This allows the dynamic interference of the mobile source on the leakage path and the constraint effect of the static base on vehicle emissions to be updated synchronously, significantly improving the spatiotemporal simulation accuracy of the leakage diffusion field. Through the above technical solution, this application solves the problem in the prior art that the static calculation model of vehicle mobile emissions and leakage diffusion field cannot be dynamically coupled. By embedding the influence of the mobile source into the diffusion field update process in real time through an iterative correction mechanism, the final generated leakage diffusion field can accurately reflect the interference of vehicle movement on the diffusion path of combustible gas, providing a reliable data foundation for subsequent risk analysis and early warning scheme generation.

[0038] The following describes how, based on the leakage diffusion field, vehicle impact field, anomaly marker sequence, and variance field, a linked risk index matrix is ​​calculated and output using Bayesian updates. This matrix is ​​then mapped using the anomaly marker sequence to generate a risk evolution probability field. Specifically, this includes: Based on the likelihood estimation of the leakage diffusion field and the vehicle impact field, and using the anomaly label sequence as the observation correction term and the variance field as the confidence correction factor, the linkage risk index matrix is ​​calculated in the Bayesian update form. A Markov time transition model is established based on the linked risk index matrix, and the transition probability is corrected by combining the anomaly marker sequence to output the risk evolution probability field at several future time points.

[0039] Among them, the linkage risk index matrix refers to a risk quantification index that is dynamically updated through a Bayesian probability framework, based on the likelihood estimation of the leakage diffusion field and the vehicle impact field, combined with the observation correction term of the anomaly label sequence and the confidence correction factor of the variance field. Specifically, it can be realized by using a hierarchical Bayesian network to jointly model the uncertainty of multi-source data. Bayesian update refers to using the leakage diffusion field and the vehicle influence field as prior distributions, and adjusting the posterior probability using real-time observation data provided by the anomaly label sequence. Specifically, it can be implemented by using the Markov chain Monte Carlo method for approximate inference. The Markov time transition model refers to establishing a state transition probability matrix based on the linkage risk index matrix at the current moment, and correcting the mutation probability in the transition process through anomaly labeling sequences. Specifically, it can be implemented by using a hidden Markov model combined with the Viterbi algorithm for state decoding.

[0040] The above content will be described in detail below: The leakage diffusion field and the vehicle impact field are respectively used as static basis and dynamic disturbance input into the Bayesian update framework. The variance field provides the confidence weight of each spatial point, and the anomaly label sequence is used as observation evidence to calibrate the prior distribution. The linkage risk index matrix is ​​generated by iteratively calculating the posterior probability distribution. Subsequently, the risk values ​​in the matrix are divided into discrete states according to time windows, and a state transition relationship is established using a Markov model. The mutation points in the anomaly label sequence are used to correct the abnormal jumps in the transition probability, and finally the risk evolution probability field for multiple future time steps is output.

[0041] This solution uses Bayesian updates to dynamically adjust risk estimates in real time by treating the vehicle impact field as a correction term, and utilizes Markov models to capture the temporal evolution of the risk field, enabling risk prediction to reflect the coupling effect between vehicle movement and gas diffusion. Through the above technical solution, this application solves the problem of early warning deviation caused by the lack of dynamic coupling of vehicle movement impact in existing technologies. By combining probabilistic modeling with time series prediction, it improves the prediction accuracy of risk evolution paths and provides a dynamically updated decision-making basis for fire early warning.

[0042] The following describes how a priority response ranking table is generated through multi-dimensional scoring based on the risk evolution probability field, the linked risk index matrix, and the anomaly marker sequence. This table is then matched with a preset set of on-site executable actions to generate a fire early warning plan. Specifically, this includes: A priority response ranking table is generated based on the cumulative risk value along the time axis of the risk evolution probability field, combined with the anomaly marker sequence and the linked risk index matrix. Using a priority response ranking table as the sole input, the system transforms the priority response ranking table into a fire early warning scheme based on a preset set of on-site executable actions.

[0043] Among them, the risk evolution probability field refers to the probability of future risk distribution based on time series prediction, which can be implemented by combining Markov transition model with Bayesian update method. The linkage risk index matrix refers to a set of dynamic risk indicators calculated by comprehensively considering the leakage diffusion field, the vehicle impact field, and the anomaly marker sequence. Specifically, it can be achieved by weighted fusion of spatiotemporal gradient inversion results and variance field confidence. Anomaly labeling sequences refer to time series that identify anomalous events generated through parallel hierarchical detection. Specifically, they can be implemented using a structured combination of short-term mutation scores, persistent trend scores, and co-mutation scores. The priority response ranking table is a list of action priorities generated based on multidimensional scoring rules. Specifically, it can be calculated by integrating the cumulative risk value along the time axis and combining it with the weight allocation of the anomaly marker sequence. The set of on-site executable actions refers to a pre-set library of fire emergency operation instructions, which may include vehicle evacuation route planning, gas leak suppression measures, and fire equipment activation strategies.

[0044] The above content will be described in detail below: The cumulative risk value of the risk evolution probability field along the time axis is scored in multiple dimensions. The anomaly label sequence is used as a correction factor to adjust the scoring weight. The linked risk index matrix provides a real-time risk intensity reference. By weighted fusion of the three data, a priority response ranking table is generated. The priority response ranking table is ranked according to the risk accumulation speed, spatial coverage and anomaly event density. The priority response sorting table is matched with a preset set of on-site executable actions. For example, high-priority instructions corresponding to high-risk areas are automatically mapped to specific operations such as closing leak source valves and starting the ventilation system, ultimately generating a dynamically adjusted fire early warning scheme.

[0045] This solution introduces the cumulative risk value of the risk evolution probability field as a time-dimensional scoring benchmark, and combines it with anomaly marker sequences to correct weight allocation, enabling the sorting table to reflect dynamic risk evolution. Simultaneously, through a matching mechanism of preset action sets, the abstract sorting results are directly converted into executable instructions, solving the problem of disconnect between early warning schemes and actual operations in existing technologies. Through the above technical solutions, this application achieves dynamic optimization sorting of fire early warning instructions, improving the matching degree between early warning schemes and real-time risk scenarios. Furthermore, through an automated instruction mapping mechanism, the time delay from risk identification to execution response is shortened, effectively improving the emergency response efficiency of urban fire protection systems in complex and dynamic environments.

[0046] Example 2: Please see Figure 4 A remote monitoring and early warning system for urban fire protection, comprising: The data acquisition module is used to acquire the gas concentration time series and vehicle exhaust time series of the target object; The anomaly detection module is used to perform parallel hierarchical detection on the gas concentration time series and the vehicle exhaust time series, and generate anomaly marker sequences; The data solving module is used to construct an objective function with gas concentration time series, vehicle exhaust time series and anomaly marker sequence as constraints, and to solve it with spatiotemporal gradient inversion and vehicle coupling terms to generate implicit flow field estimates and their variance fields. The leakage diffusion analysis module is used to solve the initial leakage diffusion field based on the gas concentration time series, implicit flow field estimation, and anomaly marker sequence. The vehicle influence field is calculated within the vehicle trajectory time window based on the vehicle row time series and anomaly marker sequence, and is used as a moving source correction term to iteratively update the initial leakage diffusion field until the preset number of iterations is reached to generate the leakage diffusion field. The risk analysis module is used to calculate and output the linkage risk index matrix based on Bayesian updates according to the leakage diffusion field, vehicle impact field, anomaly marker sequence and variance field, and to generate a risk evolution probability field by mapping with the anomaly marker sequence. The early warning scheme generation module is used to generate a priority response ranking table based on the risk evolution probability field, the linkage risk index matrix and the anomaly marker sequence through multi-dimensional scoring, and match it with the preset set of on-site executable actions to generate a fire early warning scheme.

[0047] This embodiment has the same technical effects as Embodiment 1.

[0048] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. All data mentioned in this application have undergone normalization and other preprocessing to achieve dimensional uniformity during calculation.

[0049] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for remote monitoring and early warning of urban fires, characterized in that, Includes the following steps: Obtain the gas concentration time series and vehicle exhaust time series of the target object; Parallel hierarchical detection is performed on the gas concentration time series and the vehicle exhaust time series to generate anomaly marker sequences; An objective function is constructed using the gas concentration time series, the vehicle exhaust time series, and the anomaly marker sequence as constraints. The spatiotemporal gradient inversion and vehicle coupling term are then solved to generate the implicit flow field estimate and its variance field. The initial leakage diffusion field is solved based on the gas concentration time series, the implicit flow field estimation, and the anomaly marker sequence. The vehicle influence field is calculated within the vehicle trajectory time window based on the vehicle row time series and anomaly marker sequence, and used as a moving source correction term to iteratively update the initial leakage diffusion field until a preset number of iterations is reached to generate the leakage diffusion field. Based on the leakage diffusion field, the vehicle impact field, the anomaly marker sequence, and the variance field, a linkage risk index matrix is ​​calculated and output based on Bayesian update, and mapped using the anomaly marker sequence to generate a risk evolution probability field. Based on the risk evolution probability field, the linkage risk index matrix, and the anomaly marker sequence, a priority response ranking table is generated through multidimensional scoring, and then matched with a preset set of on-site executable actions to generate a fire early warning scheme.

2. The urban fire remote monitoring and early warning method according to claim 1, characterized in that: Obtaining the gas concentration time series and vehicle exhaust time series of the target object specifically includes: Obtain the combustible gas concentration and vehicle emission status of the target object; The combustible gas concentration and the vehicle emission status are decomposed into short, medium and long scales respectively, and the missing values ​​are reconstructed and outliers are corrected by using each other's sequences as conditions based on preset inter-scale mutual verification rules, so as to generate gas concentration time series and vehicle emission time series.

3. The urban fire remote monitoring and early warning method according to claim 1, characterized in that: Parallel hierarchical detection is performed on the gas concentration time series and the vehicle exhaust time series to generate anomaly marker sequences, specifically including: Calculate the short-term abrupt change score and the persistent trend score for the gas concentration time series and the vehicle exhaust time series, respectively; The co-mutation scores of the gas concentration time series and the vehicle exhaust time series within the same time window are detected, and the short-term mutation scores, the persistent trend scores, and the co-mutation scores are structurally combined to generate anomaly marker sequences.

4. The urban fire remote monitoring and early warning method according to claim 1, characterized in that: Using the gas concentration time series, the vehicle exhaust time series, and the anomaly marker sequence as constraints, an objective function is constructed. This objective function is then solved by spatiotemporal gradient inversion and vehicle coupling terms to generate the implicit flow field estimate and its variance field. Specifically, this includes: Calculate the spatial temporal gradient based on the gas concentration time series and the vehicle exhaust time series; The spatial temporal gradient is inverted using the spatiotemporal gradient inversion method. During the inversion process, the anomaly marker sequence is used as a time window constraint and a vehicle contribution coupling term is added. The implicit flow field estimate and the corresponding variance field are obtained by regularized least squares optimization.

5. The urban fire remote monitoring and early warning method according to claim 1, characterized in that: Based on the vehicle queue time series and anomaly marker sequence, the vehicle influence field is calculated within the vehicle trajectory time window, and used as a moving source correction term to iteratively update the initial leakage diffusion field until a preset number of iterations is reached. The generation of the leakage diffusion field specifically includes: Calculate the vehicle influence field within the vehicle trajectory time window based on the vehicle time sequence; The vehicle influence field is fed back to the initial leakage diffusion field as a moving source correction term in an alternating iterative manner, and the local cumulative concentration of the initial leakage diffusion field is fed back to the vehicle influence field as a static basis, until a preset number of iterations is reached to generate the leakage diffusion field.

6. The urban fire remote monitoring and early warning method according to claim 1, characterized in that: Based on the leakage diffusion field, the vehicle impact field, the anomaly marker sequence, and the variance field, a linked risk index matrix is ​​calculated and output using Bayesian updates. This matrix is ​​then mapped using the anomaly marker sequence to generate a risk evolution probability field, specifically including: The linkage risk index matrix is ​​calculated using Bayesian update form based on the likelihood estimation of the leakage diffusion field and the vehicle influence field, with the anomaly label sequence as the observation correction term and the variance field as the confidence correction factor. A Markov time transition model is established based on the linked risk index matrix, and the transition probability is corrected by combining the anomaly marker sequence to output the risk evolution probability field at several future time points.

7. The urban fire remote monitoring and early warning method according to claim 1, characterized in that: Based on the risk evolution probability field, the linkage risk index matrix, and the anomaly marker sequence, a priority response ranking table is generated through multi-dimensional scoring, and then matched with a preset set of on-site executable actions to generate a fire early warning scheme, specifically including: A priority response ranking table is generated based on the cumulative risk value along the time axis of the risk evolution probability field, combined with the anomaly marker sequence and the linkage risk index matrix; Using the priority response ranking table as the sole input, the priority response ranking table is transformed into a fire early warning scheme based on a preset set of on-site executable actions.

8. A remote monitoring and early warning system for urban fire protection, characterized in that, include: The data acquisition module is used to acquire the gas concentration time series and vehicle exhaust time series of the target object; An anomaly detection module is used to perform parallel hierarchical detection on the gas concentration time series and the vehicle exhaust time series to generate an anomaly marker sequence; The data solving module is used to construct an objective function with the gas concentration time series, the vehicle exhaust time series and the anomaly marker sequence as constraints, and to solve the spatiotemporal gradient inversion and vehicle coupling term to generate implicit flow field estimation and its variance field. The leakage diffusion analysis module is used to solve the initial leakage diffusion field based on the gas concentration time series, the implicit flow field estimation, and the anomaly marker sequence. The vehicle influence field is calculated within the vehicle trajectory time window based on the vehicle row time series and anomaly marker sequence, and used as a moving source correction term to iteratively update the initial leakage diffusion field until a preset number of iterations is reached to generate the leakage diffusion field. The risk analysis module is used to calculate and output the linkage risk index matrix based on Bayesian update according to the leakage diffusion field, the vehicle impact field, the anomaly marker sequence and the variance field, and to generate a risk evolution probability field by mapping the anomaly marker sequence. The early warning scheme generation module is used to generate a priority response ranking table based on the risk evolution probability field, the linkage risk index matrix and the anomaly marker sequence through multi-dimensional scoring, and match it with a preset set of on-site executable actions to generate a fire early warning scheme.

Images