Intelligent property service management method and management platform

Abstract

The application provides a kind of intelligent property service management method and management platform, it is related to intelligent property management technical field, the method includes: constructing multistage time granularity window carries out dynamic sampling, carries out early fire-fighting hidden danger feature capture to first granularity area monitoring set, constructs first fire-fighting hidden danger heat map, according to second granularity area monitoring set carries out multidimensional trend review correction, establishes second fire-fighting hidden danger heat map, introduces fire accident deduction hierarchical architecture combines third granularity area monitoring set carries out fire accident deduction optimization, generates third fire-fighting hidden danger heat map, according to first, second and third fire-fighting hidden danger heat map carries out multi-granularity synchronous fire-fighting management.The technical problem that the real-time and accuracy of hazard identification in the prior art of property fire management is insufficient, lacks dynamic prediction capability is solved.The technical effect that multi-granularity real-time monitoring, accurate hazard identification and dynamic prediction and early warning are achieved, the initiative and accuracy of property fire safety management are improved.

Description

Technical Field

[0001] This invention relates to the field of smart property management technology, specifically to a smart property service management method and management platform. Background Technology

[0002] With the acceleration of urbanization, the scale of property management areas is expanding, posing greater challenges to fire safety management. Current property fire safety services mainly rely on manual inspections and fixed-period testing, combined with basic fire alarm systems and video surveillance to monitor the environment, facilities, and personnel daily. However, this existing property management approach has several shortcomings: First, manual inspections are inefficient and have limited coverage, making it difficult to detect subtle changes in hazards or potential risks in a timely manner. Second, the fixed-period testing frequency cannot adapt to the dynamic changes in fire risks, resulting in monitoring blind spots and delayed response. Furthermore, existing technologies are mostly based on single data sources or single-dimensional analysis, lacking a comprehensive assessment of the three factors—environment, facilities, and personnel—and their coupling relationships. This makes it difficult to accurately predict the development trend of hazards and the possible accident paths, leading to a significant passivity in fire management and hindering timely warnings of potential risks, thus affecting decision-making efficiency and prevention effectiveness.

[0003] Existing technologies for property fire safety management suffer from technical problems such as insufficient real-time and accuracy in hazard identification and a lack of dynamic prediction capabilities. Summary of the Invention

[0004] The purpose of this application is to provide a smart property service management method and management platform to solve the technical problems of insufficient real-time and accuracy of hazard identification and lack of dynamic prediction capabilities in existing property fire management.

[0005] In view of the above problems, this application provides a smart property service management method and management platform.

[0006] The first aspect of this application provides a smart property service management method, the method comprising: performing multi-level time decomposition based on a historical fire accident event set of a property management area to construct a multi-level time granularity window; controlling an IoT device group to perform time-domain dynamic sampling of the property management area according to the multi-level time granularity window to obtain a first granularity area monitoring set, a second granularity area monitoring set, and a third granularity area monitoring set; introducing a fire hazard knowledge space to capture early fire hazard features of the first granularity area monitoring set to construct a first fire hazard heat map; performing multi-dimensional trend verification and correction on the first fire hazard heat map based on the second granularity area monitoring set to establish a second fire hazard heat map; introducing a fire accident inference hierarchy architecture, and combining the third granularity area monitoring set to optimize the second fire hazard heat map through fire accident inference to generate a third fire hazard heat map; and performing multi-granularity synchronous fire management based on the first fire hazard heat map, the second fire hazard heat map, and the third fire hazard heat map.

[0007] Optionally, the time characteristics of the accident process are identified based on the historical fire accident event set to obtain multiple accident time characteristic sequences; multimodal clustering is performed based on the multiple accident time characteristic sequences to establish short-term accident time modal class, medium-term accident time modal class and long-term accident time modal class; adaptive mutation interval identification and fusion is performed based on the short-term accident time modal class, the medium-term accident time modal class and the long-term accident time modal class to generate the multi-level time granularity window.

[0008] Optionally, multimodal characteristic identification is performed based on the first granularity area monitoring set to construct a first granularity environmental characteristic vector, a first granularity facility characteristic vector, and a first granularity personnel characteristic vector; the fire hazard knowledge space is activated, which includes an environmental fire hazard knowledge space, a facility fire hazard knowledge space, and a personnel fire hazard knowledge space; fire hazard capture is performed on the first granularity environmental characteristic vector based on the environmental fire hazard knowledge space to obtain a first granularity environmental fire hazard vector; fire hazard capture is performed on the first granularity facility characteristic vector based on the facility fire hazard knowledge space to obtain a first granularity facility fire hazard vector; fire hazard capture is performed on the first granularity personnel characteristic vector based on the personnel fire hazard knowledge space to obtain a first granularity personnel fire hazard vector; multidimensional hazard integration rendering is performed based on the first granularity environmental fire hazard vector, the first granularity facility fire hazard vector, and the first granularity personnel fire hazard vector to generate the first fire hazard heatmap.

[0009] Optionally, anomaly detection is performed based on the first granularity environmental characteristic vector to obtain a first granularity environmental anomaly feature group; based on the first granularity environmental anomaly feature group, the Kth short-term environmental anomaly feature is extracted, where K is a positive integer; the cosine similarity of the Kth short-term environmental anomaly feature with each environmental fire hazard knowledge feature in the environmental fire hazard knowledge space is calculated to construct a Kth environmental anomaly hazard twin sequence; the Kth environmental anomaly hazard twin sequence is iteratively compared and optimized to determine the Kth environmental anomaly hazard triggering coefficient; if the Kth environmental anomaly hazard triggering coefficient is greater than or equal to the hazard triggering tolerance threshold, the Kth short-term environmental anomaly feature is added to the first granularity environmental fire hazard vector; if the Kth environmental anomaly hazard triggering coefficient is less than the hazard triggering tolerance threshold, the Kth short-term environmental anomaly feature is filtered.

[0010] Optionally, multi-dimensional fire hazard detection is performed on the second granularity regional monitoring set based on the fire hazard knowledge space to establish a second granularity environmental fire hazard vector, a second granularity facility fire hazard vector, and a second granularity personnel fire hazard vector; environmental hazard fluctuation analysis is performed on the first fire hazard heatmap based on the second granularity environmental fire hazard vector to establish an environmental hazard fluctuation characteristic path; facility hazard fluctuation analysis is performed on the first fire hazard heatmap based on the second granularity facility fire hazard vector to establish a facility hazard fluctuation characteristic path; personnel hazard fluctuation analysis is performed on the first fire hazard heatmap based on the second granularity personnel fire hazard vector to establish a personnel hazard fluctuation characteristic path; adaptive verification and correction are performed on the first fire hazard heatmap based on the environmental hazard fluctuation characteristic path, the facility hazard fluctuation characteristic path, and the personnel hazard fluctuation characteristic path to generate the second fire hazard heatmap.

[0011] Optionally, multi-dimensional characteristic anomaly detection is performed based on the third-granularity regional monitoring set to obtain a third-granularity environmental anomaly feature group, a third-granularity facility anomaly feature group, and a third-granularity personnel anomaly feature group; fire accident simulation is performed on the third-granularity environmental anomaly feature group according to the fire accident simulation hierarchy architecture to obtain an environmental anomaly simulation path; fire accident simulation is performed on the third-granularity facility anomaly feature group according to the fire accident simulation hierarchy architecture to obtain a facility anomaly simulation path; fire accident simulation is performed on the third-granularity personnel anomaly feature group according to the fire accident simulation hierarchy architecture to obtain a personnel anomaly simulation path; multi-factor coupled fire accident simulation is performed based on the environmental anomaly simulation path, the facility anomaly simulation path, and the personnel anomaly simulation path to obtain a full-scale anomaly simulation path; and the second fire hazard heat map is integrated and optimized based on the environmental anomaly simulation path, the facility anomaly simulation path, the personnel anomaly simulation path, and the full-scale anomaly simulation path to obtain the third fire hazard heat map.

[0012] Optionally, based on the historical fire accident event set, multi-factor classification is performed to establish environmental factor fire accident zones, facility factor fire accident zones, and personnel factor fire accident zones; based on the environmental factor fire accident zones, accident path tracing and multi-scale perturbation injection are performed to establish an environment-coupled fire accident space; based on the environment-coupled fire accident space, a Bayesian network model is iteratively supervised and trained to generate an environment-coupled fire accident simulation first node; based on the environment-coupled fire accident space, a GRU model is iteratively supervised and trained to generate an environment-coupled fire accident simulation second node; based on the environment-coupled fire accident simulation first node and the environment-coupled fire accident simulation second node, federated aggregation and reinforcement are performed to generate an environment-coupled fire accident simulation layer; based on the facility factor fire accident zones, a facility-coupled fire accident simulation layer is trained; based on the personnel factor fire accident zones, a personnel-coupled fire accident simulation layer is trained; the environment-coupled fire accident simulation layer, the facility-coupled fire accident simulation layer, and the personnel-coupled fire accident simulation layer are encapsulated into the fire accident simulation hierarchical architecture.

[0013] Optionally, a multi-granularity fire hazard alarm can be generated based on the first fire hazard heat map, the second fire hazard heat map, and the third fire hazard heat map.

[0014] Optionally, the multi-level time granularity window can be adaptively optimized based on the first fire hazard heat map, the second fire hazard heat map, and the third fire hazard heat map.

[0015] A second aspect of this application provides a smart property service management platform, the platform comprising: a multi-level time decomposition module, used to perform multi-level time decomposition based on a set of historical fire accident events in the property management area, constructing a multi-level time granularity window; a time-domain dynamic sampling module, used to control an IoT device group to perform time-domain dynamic sampling of the property management area according to the multi-level time granularity window, obtaining a first granularity area monitoring set, a second granularity area monitoring set, and a third granularity area monitoring set; and a first fire hazard heat map construction module, used to introduce a fire hazard knowledge space to capture early fire hazard features in the first granularity area monitoring set. The system comprises three modules: a first fire hazard heat map; a second fire hazard heat map creation module, used to perform multi-dimensional trend verification and correction on the first fire hazard heat map based on the second granularity regional monitoring set, and to create a second fire hazard heat map; a third fire hazard heat map generation module, used to introduce a fire accident simulation hierarchy architecture, and to optimize the second fire hazard heat map based on the third granularity regional monitoring set, and to generate a third fire hazard heat map; and a fire management module, used to perform multi-granularity synchronous fire management based on the first fire hazard heat map, the second fire hazard heat map, and the third fire hazard heat map.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages: The method provided in this application embodiment decomposes historical fire incident events in a property management area into multiple time levels to construct multi-granularity windows. Based on these multi-granularity windows, it controls an IoT device group to perform dynamic temporal sampling of the property management area, obtaining a first-granularity regional monitoring set, a second-granularity regional monitoring set, and a third-granularity regional monitoring set. A fire hazard knowledge space is introduced to capture early fire hazard features in the first-granularity regional monitoring set, constructing a first fire hazard heatmap. The first fire hazard heatmap is then subjected to multi-dimensional trend verification and correction based on the second-granularity regional monitoring set, establishing a second fire hazard heatmap. A fire incident simulation hierarchy is introduced, and the second fire hazard heatmap is optimized using the third-granularity regional monitoring set to generate a third fire hazard heatmap. Multi-granularity synchronous fire management is then performed based on the first, second, and third fire hazard heatmaps. This achieves the technical effects of multi-granularity real-time monitoring, accurate hazard identification, and dynamic prediction and early warning, improving the initiative and accuracy of property fire safety management.

[0017] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 A flowchart illustrating the intelligent property service management method provided for this application.

[0020] Figure 2 A schematic diagram of the intelligent property service management platform provided for this application.

[0021] Figure labeling: Multi-level time decomposition module 11, time-domain dynamic sampling module 12, first fire hazard heat map construction module 13, second fire hazard heat map establishment module 14, third fire hazard heat map generation module 15, fire management module 16. Detailed Implementation

[0022] This application provides a smart property service management method and platform to address the technical problems in existing property fire safety management, such as insufficient real-time and accuracy of hazard identification and lack of dynamic prediction capabilities. It achieves multi-granularity real-time monitoring, accurate hazard identification, and dynamic prediction and early warning, thereby improving the initiative and accuracy of property fire safety management.

[0023] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. It should also be noted that, for ease of description, only the parts related to the present invention are shown in the accompanying drawings, not all of them.

[0024] Example 1 like Figure 1As shown, this application provides a smart property service management method, which includes: Based on the historical fire incident event set of the property management area, a multi-level time decomposition is performed to construct a multi-level time granularity window.

[0025] Furthermore, based on the historical fire accident event set of the property management area, a multi-level time decomposition is performed to construct a multi-level time granularity window, including: identifying the time characteristics of the accident process based on the historical fire accident event set to obtain multiple accident time characteristic sequences; performing multimodal clustering based on the multiple accident time characteristic sequences to establish short-term accident time modal class, medium-term accident time modal class, and long-term accident time modal class; and performing adaptive abrupt change interval identification and fusion based on the short-term accident time modal class, the medium-term accident time modal class, and the long-term accident time modal class to generate the multi-level time granularity window.

[0026] Specifically, based on information databases such as property fire logs, IoT device alarm records, fire inspection reports, and historical accident archives, a historical fire accident event set is obtained for the property management area. The historical fire accident event set is a summary of all fire-related event records that occurred in the property management area during previous management cycles, including at least: the time of the accident, the duration of the accident development stage, the factors that triggered the accident, the area affected by the event, the environmental state parameters at the time of the incident, the personnel activities, and a complete time record of the accident handling process. For example, a fire broke out in a residential community due to an electric vehicle charging. The incident occurred at 10:30 PM. The triggering factor was an aging charging line that short-circuited, generating an electrical spark that ignited the electric vehicle's plastic casing. The incident affected the public areas on the first to third floors of the building and some residents' homes. Environmental parameters at the time of the incident showed a temperature of 28°C and an air humidity of 60%. Some residents on lower floors quickly fled after discovering the fire. The complete timeline of the incident handling process is as follows: the fire broke out at 10:30 PM; a resident called 119 at 10:32 PM; the fire truck arrived at the community at 10:35 PM; the fire was initially brought under control at 10:50 PM; the fire was completely extinguished at 11:10 PM; and firefighters subsequently cleaned up and inspected the scene until 11:40 PM.

[0027] The acquired historical fire accident event set is subjected to accident process temporal characteristic identification. These characteristics refer to the temporal change patterns exhibited at different stages throughout the entire fire accident process, from triggering, development, and spread to eventual resolution. Examples include the latency period from hazard discovery to alarm activation, the response time from alarm activation to fire confirmation, the fire spread rate, and the duration of the response process. By analyzing the timeline structure of each accident sample in the historical fire accident event set, key stage nodes and corresponding time periods are extracted, thereby forming multiple accident temporal characteristic sequences describing the temporal evolution of the accidents.

[0028] Dynamic Time Warping (DTW) technology is used to process multiple accident time characteristic sequences, generating time alignment distances between sequences to address the time scale offset problem of different accidents. A hierarchical density clustering algorithm based on the DTW distance matrix is ​​introduced to perform multimodal clustering of multiple accident time characteristic sequences. For example, the HDBSCAN clustering algorithm can simultaneously process different time modal distributions and automatically identify the number of clusters through density peaks, achieving a natural division of short-term, medium-term, and long-term modes, forming short-term accident time modal classes, medium-term accident time modal classes, and long-term accident time modal classes. The short-term accident time modal class includes fire accidents that occur, develop, and end within a relatively short period, with a relatively small time span. The medium-term accident time modal class corresponds to fire accidents with a slightly longer time process and certain development stages. The long-term accident time modal class includes fire accidents with a long duration, possibly involving multiple stages and complex factors.

[0029] After obtaining the short-term, medium-term, and long-term accident time modalities, adaptive mutation interval identification is performed based on the corresponding accident time feature distributions. The mutation interval refers to the time period in the development of a fire accident that exhibits significant changes in the time series, reflecting the critical node where a hidden danger evolves from controllable to uncontrollable. For example, an adaptive mutation detection algorithm based on dynamic threshold adjustment of time gradient is used. By calculating the local gradient change rate, second-order difference volatility, and stage stability index of the time characteristic sequence, a comprehensive mutation score curve reflecting the speed of accident evolution is formed. Subsequently, based on the mean, standard deviation of volatility of the overall sequence, and the time scale characteristics of the target modality, a dynamic threshold is adaptively constructed: threshold T = mean + α × standard deviation of volatility, where α is automatically adjusted by the modal features. For example, short-term modes are more prone to drastic changes, so α is smaller, and vice versa. When the mutation score of any time period exceeds the threshold T, it is determined to be a mutation interval. Based on this, the mutation time intervals in the short-term, medium-term, and long-term modalities are identified.

[0030] After identifying the mutation intervals of the three modes, the weight ratios of the short-term, medium-term, and long-term modes are determined by statistically analyzing the frequency of occurrence of each mode in historical accident samples, the severity of the accidents, and the degree of impact on monitoring sensitivity. Combined with the weight ratios of each mode in the accident samples, multi-level time granularity windows suitable for IoT sampling are extracted, namely short-granularity time windows, medium-granularity time windows, and long-granularity time windows, which correspond to rapid monitoring, trend monitoring, and long-term change monitoring, respectively.

[0031] By constructing multi-level time granularity windows, sampling of IoT device groups can be controlled in a targeted manner according to the characteristics of different time granularities, thereby obtaining more representative regional monitoring data that better meets actual needs and achieves precise matching between the monitoring cycle and the actual time characteristics of the accident.

[0032] The IoT device group is controlled to perform time-domain dynamic sampling of the property management area according to the multi-level time granularity window to obtain a first granularity area monitoring set, a second granularity area monitoring set, and a third granularity area monitoring set.

[0033] Specifically, after constructing a multi-level time granularity window, the three time windows of short granularity, medium granularity, and long granularity are used as the basis for sampling and scheduling to implement a hierarchical and differentiated time-domain dynamic sampling strategy for the IoT device group deployed in the property management area.

[0034] Among them, short-granularity time windows correspond to the highest device sampling frequency, used to capture rapidly changing information such as environmental parameters, electrical load, smoke concentration, and personnel movement. Medium-granularity time windows allow devices to operate at a moderate frequency, focusing on recording medium-range stability data such as equipment status trends, energy consumption changes, and facility operating cycles. Long-granularity time windows are used for low-frequency sampling, primarily monitoring long-term accumulated signs of potential hazards, such as equipment aging characteristics, long-term environmental fluctuations, or changes in personnel behavior patterns. The IoT device group includes at least environmental monitoring devices, electrical and facility monitoring devices, and video and personnel behavior monitoring devices, among other types of terminals used to continuously collect fire safety-related data for the property area. Environmental monitoring devices include temperature and humidity sensors, smoke sensors, and combustible gas detectors; electrical and facility monitoring devices include current and voltage sensors and equipment operation status acquisition devices; and video and personnel behavior monitoring devices include smart cameras, access control systems, and personnel flow sensing devices.

[0035] By dynamically scheduling the sampling cycle, data upload interval, and real-time requirements of various devices, three granularity regional monitoring sets corresponding to the three types of time windows are formed, thereby achieving hierarchical time-series monitoring of the property area.

[0036] A fire hazard knowledge space is introduced to capture early fire hazard characteristics of the first granularity area monitoring set and construct a first fire hazard heat map.

[0037] Furthermore, a fire hazard knowledge space is introduced to capture early fire hazard features of the first granularity area monitoring set and construct a first fire hazard heat map. This includes: performing multimodal characteristic identification based on the first granularity area monitoring set to construct a first granularity environmental characteristic vector, a first granularity facility characteristic vector, and a first granularity personnel characteristic vector; activating the fire hazard knowledge space, which includes an environmental fire hazard knowledge space, a facility fire hazard knowledge space, and a personnel fire hazard knowledge space; capturing fire hazards from the first granularity environmental characteristic vector based on the environmental fire hazard knowledge space to obtain a first granularity environmental fire hazard vector; capturing fire hazards from the first granularity facility characteristic vector based on the facility fire hazard knowledge space to obtain a first granularity facility fire hazard vector; capturing fire hazards from the first granularity personnel characteristic vector based on the personnel fire hazard knowledge space to obtain a first granularity personnel fire hazard vector; and performing multi-dimensional hazard integration rendering based on the first granularity environmental fire hazard vector, the first granularity facility fire hazard vector, and the first granularity personnel fire hazard vector to generate the first fire hazard heat map.

[0038] Specifically, multimodal analysis is performed on the first-granularity regional monitoring set according to the data source, dividing the set into three major data streams: environment, facilities, and personnel. Feature standardization and noise filtering are then performed on each data stream. Feature standardization can employ normalization or z-score standardization to ensure comparability of features with different dimensions. During noise filtering, methods such as sliding window averaging, outlier pruning, and median filtering based on time continuity are used to remove transient jitter, equipment errors, and occasional interference. Key feature indicators are extracted from the standardized environment, facility, and personnel data streams. For example, environmental parameters such as temperature, humidity, smoke concentration, and air quality are extracted for the environment stream; equipment operating status parameters, start / stop frequency, and maintenance status are extracted for the fire protection facilities stream; and flow density, residence area, and activity trajectory are extracted for the personnel stream. The extracted features and the multimodal features after numerical vectorization are integrated according to categories to form a first-granularity environmental characteristic vector, a first-granularity facility characteristic vector, and a first-granularity personnel characteristic vector. The first-granularity environmental characteristic vector is used to reflect the environmental status characteristics of the property management area under a short-granularity time window. The first-granularity facility characteristic vector is used to reflect the overall status of fire protection facilities in the property management area under a short-granularity time window. The first-granularity personnel characteristic vector is used to reflect the potential impact of personnel factors on fire safety in the property management area under a short-granularity time window.

[0039] The fire hazard knowledge space is activated. This knowledge space is constructed through the analysis of numerous fire accident cases, fire safety standards and regulations, and the summary of expert experience. It includes three main categories: environmental fire hazard knowledge space, facility fire hazard knowledge space, and personnel fire hazard knowledge space. The environmental fire hazard knowledge space contains multiple environmental fire hazard knowledge features, such as high temperatures easily causing fires and humid environments potentially leading to electrical faults. The facility fire hazard knowledge space covers knowledge features such as aging, damage, and improper installation of fire protection facilities. The personnel fire hazard knowledge space includes knowledge features such as personnel violating operating procedures and lack of fire safety knowledge.

[0040] Then, based on the environmental fire hazard knowledge space, fire hazards are captured from the first-granularity environmental characteristic vector to obtain the first-granularity environmental fire hazard vector. Similarly, based on the facility fire hazard knowledge space and the personnel fire hazard knowledge space, fire hazards are captured from the first-granularity facility characteristic vector and the first-granularity personnel characteristic vector to obtain the first-granularity facility fire hazard vector and the first-granularity personnel fire hazard vector, respectively. The first-granularity environmental fire hazard vector, the first-granularity facility fire hazard vector, and the first-granularity personnel fire hazard vector are then integrated and rendered in a multi-dimensional hazard manner. For example, using Geographic Information System (GIS) technology, hazard information of different dimensions is mapped onto a regional map, and different colors and patterns are used to represent the severity and type of the hazard, generating a first-granularity fire hazard heat map. The heat map can intuitively display the distribution of fire hazards in different locations within the area; the darker the color, the more severe the fire hazard.

[0041] By introducing a fire hazard knowledge space and performing multimodal hazard capture on the first-level monitoring set, the limitations of single-factor judgment are avoided. This achieves high sensitivity and comprehensive identification of early fire hazards in the property area. The generated first-level fire hazard heat map presents the distribution and severity of fire hazards in the area in an intuitive and visual way, providing property managers with a clear basis for decision-making. This enables them to take timely and targeted prevention and rectification measures, effectively reduce the probability of fire accidents, and ensure fire safety in the property management area.

[0042] Furthermore, the environmental fire hazard knowledge space captures fire hazards from the first granular environmental characteristic vector to obtain a first granular environmental fire hazard vector, including: performing anomaly detection based on the first granular environmental characteristic vector to obtain a first granular environmental anomaly feature group; extracting the Kth short-term environmental anomaly feature based on the first granular environmental anomaly feature group, where K is a positive integer; calculating the cosine similarity between the Kth short-term environmental anomaly feature and each environmental fire hazard knowledge feature in the environmental fire hazard knowledge space to construct a Kth environmental anomaly hazard twin sequence; performing iterative comparison and optimization based on the Kth environmental anomaly hazard twin sequence to determine the Kth environmental anomaly hazard triggering coefficient; if the Kth environmental anomaly hazard triggering coefficient is greater than or equal to the hazard triggering tolerance threshold, adding the Kth short-term environmental anomaly feature to the first granular environmental fire hazard vector; if the Kth environmental anomaly hazard triggering coefficient is less than the hazard triggering tolerance threshold, filtering the Kth short-term environmental anomaly feature.

[0043] Specifically, the first-level environmental characteristic vector contains multiple environmental parameters, such as temperature, humidity, and smoke concentration. Anomaly detection is performed on each environmental parameter in the first-level environmental characteristic vector. For example, a normal distribution can be used for anomaly detection. The mean and standard deviation are calculated for each environmental characteristic dimension. The mean reflects the average level of the environmental characteristic in historical data, while the standard deviation measures the dispersion of the data. For each data point in the current first-level environmental characteristic vector, the difference between its value and the mean of the corresponding dimension is calculated, and then divided by the standard deviation of that dimension to obtain the standardized value of the data point. An anomaly detection threshold is set to ±2 times the standard deviation. If the standardized value of a data point is greater than or less than the anomaly detection threshold, it indicates that the data point is anomaly in that dimension. This calculation and judgment are performed on all data points in the first-level environmental characteristic vector. All data points identified as anomaly and their corresponding environmental parameters are combined to form a first-level environmental anomaly feature group. This first-level environmental anomaly feature group contains multiple short-term environmental anomaly features, reflecting the deviation of the current environmental state from normal conditions.

[0044] Based on the first-level environmental anomaly feature group, the Kth short-term environmental anomaly feature (K is a positive integer) is extracted sequentially. Cosine similarity is calculated between this feature and each environmental fire hazard knowledge feature in the environmental fire hazard knowledge space. Cosine similarity measures the degree of similarity between two vectors in a direction. The Kth environmental anomaly hazard twin sequence is obtained through similarity calculation. This Kth environmental anomaly hazard twin sequence contains multiple short-term environmental anomaly hazard twin coefficients, reflecting the similarity between the Kth short-term environmental anomaly feature and each environmental fire hazard knowledge feature.

[0045] The twin sequences of the Kth environmental anomaly hazard are iteratively compared and optimized to find the maximum similarity value, which is then determined as the trigger coefficient of the Kth environmental anomaly hazard. Based on the typical similarity distribution of different hazard types in historical accident samples, a hazard trigger tolerance threshold is set by statistically analyzing the average matching degree of known hazards in the knowledge feature space and combining it with safety level requirements. The trigger coefficient of the Kth environmental anomaly hazard is compared with the hazard trigger tolerance threshold. If the trigger coefficient of the Kth environmental anomaly hazard is greater than or equal to the hazard trigger tolerance threshold, it indicates that the short-term environmental anomaly feature has a high probability of causing a fire hazard, and it is added to the first-granularity environmental fire hazard vector. If the trigger coefficient of the Kth environmental anomaly hazard is less than the hazard trigger tolerance threshold, the short-term environmental anomaly feature is filtered out.

[0046] The process of capturing fire hazards from the first-granularity facility characteristic vector based on the facility fire hazard knowledge space and capturing fire hazards from the first-granularity personnel characteristic vector based on the personnel fire hazard knowledge space adopts the same logic as described above, generating the first-granularity facility fire hazard vector and the first-granularity personnel fire hazard vector respectively.

[0047] Based on the fire hazard knowledge space, fire hazards are captured by first-level environmental, facility, and personnel characteristic vectors. This allows for the rapid screening of abnormal features highly correlated with typical fire risks from environmental, facility, and personnel monitoring data within a short time scale, enabling accurate identification of early abnormal data and effectively improving the sensitivity and accuracy of fire hazard identification.

[0048] Based on the second granularity area monitoring set, the first fire hazard heat map is reviewed and corrected in multiple dimensions to establish a second fire hazard heat map.

[0049] Furthermore, based on the second granularity regional monitoring set, the first fire hazard heatmap is subjected to multi-dimensional trend verification and correction to establish a second fire hazard heatmap. This includes: performing multi-dimensional fire hazard detection on the second granularity regional monitoring set based on the fire hazard knowledge space to establish a second granularity environmental fire hazard vector, a second granularity facility fire hazard vector, and a second granularity personnel fire hazard vector; performing environmental hazard fluctuation analysis on the first fire hazard heatmap based on the second granularity environmental fire hazard vector to establish an environmental hazard fluctuation characteristic path; performing facility hazard fluctuation analysis on the first fire hazard heatmap based on the second granularity facility fire hazard vector to establish a facility hazard fluctuation characteristic path; performing personnel hazard fluctuation analysis on the first fire hazard heatmap based on the second granularity personnel fire hazard vector to establish a personnel hazard fluctuation characteristic path; and performing adaptive verification and correction on the first fire hazard heatmap based on the environmental hazard fluctuation characteristic path, the facility hazard fluctuation characteristic path, and the personnel hazard fluctuation characteristic path to generate the second fire hazard heatmap.

[0050] Specifically, based on the fire hazard knowledge space, multi-dimensional fire hazard detection is performed on the second-granularity regional monitoring set. The processing procedure is similar to that of the first-granularity regional monitoring set. First, multi-modal characteristic identification is performed on the second-granularity regional monitoring set to construct second-granularity environmental characteristic vector, second-granularity facility characteristic vector, and second-granularity personnel characteristic vector. Based on the environmental fire hazard knowledge space, facility fire hazard knowledge space, and personnel fire hazard knowledge space in the fire hazard knowledge space, fire hazard capture is performed on the second-granularity environmental characteristic vector, second-granularity facility characteristic vector, and second-granularity personnel characteristic vector, respectively. Through anomaly detection, second-granularity environmental, facility, and personnel anomaly feature groups are obtained, that is, multiple mid-term environmental, facility, and personnel anomaly features are obtained. Cosine similarity is calculated to obtain multiple mid-term environmental, facility, and personnel anomaly hazard twin coefficients. Iterative comparison and optimization are performed, and the results are compared with the hazard trigger tolerance threshold to obtain the second-granularity environmental fire hazard vector, second-granularity facility fire hazard vector, and second-granularity personnel fire hazard vector.

[0051] By using a unified geographic coordinate system and time benchmark, the second-granularity environmental fire hazard vector is precisely mapped onto the spatial area of ​​the first fire hazard heatmap according to its corresponding geographic coordinates. Simultaneously, based on the same time interval or timestamp, each data point in the second-granularity environmental fire hazard vector strictly corresponds to the corresponding time node in the first fire hazard heatmap in the time dimension, ensuring consistency in the geographical scope and time nodes covered by both. Time series analysis is performed on the data in the second-granularity environmental fire hazard vector. For example, the moving average method is used to calculate the average value of environmental hazards over different time periods to smooth the data and highlight long-term trends. Exponential smoothing is used to capture recent changes in the data, more sensitively reflecting the fluctuations in current environmental hazards.

[0052] The fluctuation amplitude of the second-granularity environmental fire hazard vector at different time points is calculated. By comparing the hazard value at the current time point with the average value at a reference time point, such as the average value of the previous time period, the fluctuation amplitude data of environmental parameters, such as temperature, humidity, and smoke concentration, are obtained. The fluctuation frequency is analyzed, and the number of times the environmental hazard value exceeds a set threshold per unit time is counted to measure the frequency of hazard fluctuations. Based on the analysis results, a curve of environmental hazard changing over time is plotted with time as the horizontal axis and environmental hazard value as the vertical axis, and key fluctuation points, such as maximum values, minimum values, and inflection points, are marked on the graph. Based on the shape characteristics of the curve, combined with the fluctuation amplitude and frequency data, the fluctuation characteristics of the environmental hazard are determined, such as periodic fluctuations or random fluctuations, and an environmental hazard fluctuation characteristic path is established. This path can intuitively display the dynamic change pattern of environmental hazards at different time stages.

[0053] Similarly, a similar analysis is performed on the second-level facility fire hazard vector and the second-level personnel fire hazard vector. Based on a unified geographical coordinate and time reference, the second-level facility fire hazard vector is mapped to the corresponding spatial area of ​​the first fire hazard heat map and aligned with the time nodes of the first fire hazard heat map. Time series analysis is performed on the characteristics such as equipment status and energy consumption in the second-level facility fire hazard vector, calculating the fluctuation amplitude and frequency, plotting the facility hazard change curve over time, and marking key fluctuation points, thereby establishing the facility hazard fluctuation characteristic path to reflect the dynamic change trend of facility risk in different time periods. The second-level personnel fire hazard vector is mapped to the first fire hazard heat map according to geographical coordinates and time, and the time series changes of personnel distribution and flow density are analyzed. By calculating the fluctuation amplitude, frequency, and key event points, a personnel hazard fluctuation curve is formed, and a personnel hazard fluctuation characteristic path is established based on this to reflect the spatiotemporal dynamic impact of personnel behavior on fire hazards.

[0054] The process involves obtaining fluctuation characteristic paths for environmental hazards, facility hazards, and personnel hazards, and assigning weights to each path. These weights can be determined based on the relative importance of each hazard's impact on the overall fire hazard situation, for example, through expert assessment or historical data analysis. For each area in the first fire hazard heatmap, the corresponding values ​​for each of these paths are obtained. A weighted summation method is used to calculate the comprehensive hazard fluctuation value for that area across the three dimensions, based on the weights and corresponding values ​​of each path. The calculated comprehensive hazard fluctuation value is then compared with the original hazard value for that area in the first fire hazard heatmap. If the comprehensive hazard fluctuation value indicates an increased hazard severity, the hazard display level for that area in the heatmap is increased proportionally; conversely, if the comprehensive hazard fluctuation value indicates a decreased hazard severity, the hazard display level is decreased accordingly. A certain percentage can be determined through statistical analysis of historical accident evolution data and the fluctuation range of potential hazards. For example, by calculating the average increase or decrease in actual risk caused by fluctuations in environmental, facility, and personnel hazards in similar areas in the past, an empirical coefficient can be obtained. Under certain conditions, the hazard value of that area in the first fire hazard heat map can be adjusted by 10%-30% of the average increase or decrease, thus reflecting the real-time changes in hazards while avoiding overly sensitive or volatile heat map displays. After adaptive load correction processing of all areas in the first fire hazard heat map, a second fire hazard heat map is generated. The second fire hazard heat map can more accurately reflect the actual situation and dynamic trend of fire hazards in the current area.

[0055] By utilizing a second-granularity regional monitoring set to perform multi-dimensional trend verification and correction on the first fire hazard heat map, the impact of environmental, facility, and personnel factors on fire hazards at different time granularities can be fully considered, making the identification of fire hazards more accurate and comprehensive. The established hazard fluctuation characteristic paths in various dimensions can dynamically reflect the changing patterns of fire hazards, providing property management personnel with more timely and accurate information. The generated second fire hazard heat map can more realistically display the distribution and severity of fire hazards within the area, effectively improving the efficiency and reliability of fire safety management in the property management area, thereby reducing the risk of fire accidents.

[0056] A hierarchical framework for fire accident simulation is introduced, and the second fire hazard heat map is optimized by combining the third-granularity regional monitoring set to generate a third fire hazard heat map.

[0057] Furthermore, the construction steps of the fire accident simulation hierarchy include: classifying historical fire accident events into multiple factors to establish environmental factor fire accident zones, facility factor fire accident zones, and personnel factor fire accident zones; performing accident path tracing and multi-scale perturbation injection based on the environmental factor fire accident zones to establish an environment-coupled fire accident space; performing iterative supervised training of a Bayesian network model based on the environment-coupled fire accident space to generate an environment-coupled fire accident simulation first node; performing iterative supervised training of a GRU model based on the environment-coupled fire accident space to generate an environment-coupled fire accident simulation second node; performing federated aggregation reinforcement based on the environment-coupled fire accident simulation first node and the environment-coupled fire accident simulation second node to generate an environment-coupled fire accident simulation layer; training a facility-coupled fire accident simulation layer based on the facility factor fire accident zones; training a personnel-coupled fire accident simulation layer based on the personnel factor fire accident zones; and encapsulating the environment-coupled fire accident simulation layer, the facility-coupled fire accident simulation layer, and the personnel-coupled fire accident simulation layer into the fire accident simulation hierarchy.

[0058] Specifically, the historical fire accident event set is classified into multiple factors, and the accident samples are divided into environmental factor fire accident area, facility factor fire accident area and personnel factor fire accident area according to environmental factors, facility factor fire accident area and personnel factor fire accident area. Among them, environmental factor fire accident area includes temperature and humidity, smoke concentration, facility factor fire accident area includes equipment operating status, abnormal energy consumption, aging characteristics, etc., and personnel factor fire accident area includes personnel flow density, residence trajectory, abnormal behavior, etc.

[0059] For fire accident zones with environmental factors, the development path of the accident is traced along the time and space dimensions based on historical accident event records and environmental parameter change sequences. The evolution process of the accident from initial triggering to spread is reconstructed. On this basis, multi-scale perturbations are injected into key environmental variables, such as temperature, humidity, and smoke concentration, including small fluctuations, short-term abnormal peaks, and long-term trend changes. In the simulation environment, possible evolution scenarios of the accident under different environmental conditions are simulated. By combining and spatially mapping the traced path with the perturbed multi-scenario data, an environmentally coupled fire accident space covering multiple time scales and environmental change scenarios is formed to simulate the development and evolution of the accident under different conditions.

[0060] Based on the environmentally coupled fire accident space, the Bayesian network model is iteratively trained under supervised supervision. Using the environmental characteristics and corresponding accident evolution results of each accident sample in the environmentally coupled fire accident space as training data, Bayesian network nodes and directed edges are constructed to represent the causal relationships between environmental factors and their probabilistic impact on accident evolution. During the iterative supervised training process, the node conditional probability table is continuously adjusted through maximum likelihood estimation or Bayesian parameter updates, enabling the Bayesian network model to accurately describe the dependency structure between environmental factors and the changes in accident triggering probability. After training, the generated Bayesian network model serves as the first node in the environmentally coupled fire accident simulation, used to predict the probability of accident occurrence and evolution path under different combinations of environmental conditions.

[0061] Iterative supervised training of the GRU model based on the environmentally coupled fire accident space refers to using environmental parameters and accident states arranged in a time series within the environmentally coupled fire accident space as input to construct a GRU (Gated Recurrent Unit) model to capture time-dependent features and long-term and short-term dynamic relationships. During iterative supervised training, the prediction error is calculated through forward propagation, and the gating parameters and weights of the GRU network are continuously adjusted using backpropagation combined with gradient descent. This allows the GRU model to learn the influence of environmental variables on accident evolution over time. After training, the generated GRU model serves as the second node in the environmentally coupled fire accident simulation, providing dynamic accident evolution prediction based on time series. A federated learning framework is used to weight and aggregate the parameters of the first and second nodes of the environmentally coupled fire accident simulation. The weights are dynamically adjusted according to the model accuracy to generate the environmentally coupled fire accident simulation layer. For example, if the Bayesian network accuracy is 92% and the GRU model accuracy is 88%, the weights are set to 0.52 and 0.48, respectively.

[0062] Repeat the above steps, coupling fire accident simulation layers with equipment-factor fire accident zones and personnel-factor fire accident zones, respectively. Finally, the environment-coupled fire accident simulation layer, facility-coupled fire accident simulation layer, and personnel-coupled fire accident simulation layer are encapsulated according to a unified data interface to form a hierarchical architecture for fire accident simulation, used for multi-factor, multi-level prediction of the paths through which fire hazards in property management areas may evolve into accidents.

[0063] By constructing a hierarchical framework for fire accident simulation, it is possible to analyze three types of factors: environment, facilities, and personnel. This allows for in-depth learning and simulation of historical accident patterns, enabling multi-dimensional prediction of the evolution path of fire hazards. This improves the accuracy and timeliness of accident prediction and enhances the proactive prevention and control capabilities of smart property fire management.

[0064] Furthermore, a hierarchical architecture for fire accident simulation is introduced. The second fire hazard heatmap is optimized for fire accident simulation based on the third-granularity regional monitoring set to generate a third fire hazard heatmap. This includes: performing multi-dimensional characteristic anomaly detection based on the third-granularity regional monitoring set to obtain third-granularity environmental anomaly feature groups, third-granularity facility anomaly feature groups, and third-granularity personnel anomaly feature groups; performing fire accident simulation on the third-granularity environmental anomaly feature groups based on the hierarchical architecture for fire accident simulation to obtain environmental anomaly simulation paths; and performing fire accident simulation on the third-granularity facility anomaly feature groups based on the hierarchical architecture for fire accident simulation. Obtain the facility anomaly accident simulation path; perform fire accident simulation on the third-granularity personnel anomaly feature group according to the fire accident simulation hierarchy architecture to obtain the personnel anomaly accident simulation path; perform multi-factor coupled fire accident simulation based on the environmental anomaly accident simulation path, the facility anomaly accident simulation path, and the personnel anomaly accident simulation path to obtain the full-scale anomaly accident simulation path; perform hazard integration optimization on the second fire hazard heat map based on the environmental anomaly accident simulation path, the facility anomaly accident simulation path, the personnel anomaly accident simulation path, and the full-scale anomaly accident simulation path to obtain the third fire hazard heat map.

[0065] Specifically, similarly, multimodal characteristic identification is performed based on the third-granularity regional monitoring set to construct a third-granularity environmental characteristic vector, a third-granularity facility characteristic vector, and a third-granularity personnel characteristic vector. Anomaly detection is performed based on the third-granularity environmental characteristic vector, the third-granularity facility characteristic vector, and the third-granularity personnel characteristic vector to obtain a third-granularity environmental anomaly feature group, a third-granularity facility anomaly feature group, and a third-granularity personnel anomaly feature group. The third-granularity environmental anomaly feature group includes multiple long-term environmental anomaly features, the third-granularity facility anomaly feature group includes multiple long-term facility anomaly features, and the third-granularity personnel anomaly feature group includes multiple long-term personnel anomaly features.

[0066] The third-level environmental anomaly feature group is input into the fire accident simulation hierarchy. Within this hierarchy, an environment-coupled fire accident simulation layer uses a Bayesian-GRU federated model to simulate fire accidents based on the input third-level environmental anomaly feature group, obtaining simulation paths for environmental anomalies. Simultaneously, a facility-coupled fire accident simulation layer within the same hierarchy performs fire accident simulation on the third-level facility anomaly feature group, obtaining simulation paths for facility anomalies. Finally, a personnel-coupled fire accident simulation layer performs fire accident simulation on the third-level personnel anomaly feature group, obtaining simulation paths for personnel anomalies. By aligning the simulation paths of environmental, facility, and personnel anomalies by spatial region and time node, and based on a weighted fusion strategy, the mutual influence weights of each factor on the evolution of the accident are calculated. The potential driving effect of environmental changes on facility operation and personnel behavior, as well as the feedback effect of facility status and personnel behavior on environmental risks, are analyzed. Through iterative superposition and coupled simulation, the interaction of the simulation paths of environmental, facility, and personnel anomalies is accumulated and calculated to generate a full set of anomaly simulation paths that reflect the overall risk coupling effect, thereby more accurately predicting the possible development and impact range of the accident.

[0067] The simulation paths for environmental anomalies, facility anomalies, personnel anomalies, and all anomalies are mapped to the corresponding spatial units of the heat map. Based on the risk intensity and regional distribution predicted by each path, a weighted fusion method is used to dynamically adjust the hazard display values ​​of each region in the heat map, including enhancing the hazard intensity of high-risk areas and weakening the hazard values ​​of low-risk or declining-trend areas. The second fire hazard heat map is optimized through iterative updates, and a third fire hazard heat map is generated.

[0068] By introducing a hierarchical framework for fire accident simulation and combining it with a third-level regional monitoring set, it is possible to proactively simulate and optimize the paths through which potential hazards in a region may evolve into accidents in the future. This significantly improves the dynamic prediction capability of hazard identification. The generated third-level fire hazard heat map not only reflects the current status of hazards but also predicts the development trend of potential risks, providing a more accurate and scientific basis for decision-making in multi-granularity fire management and enhancing the initiative and accuracy of overall property fire prevention and control.

[0069] Furthermore, the method also includes generating a multi-granularity fire hazard alarm based on the first fire hazard heat map, the second fire hazard heat map, and the third fire hazard heat map.

[0070] Specifically, the first, second, and third fire hazard heat maps are matched according to spatial regions and time nodes, and the hazard values ​​of each region are analyzed using multi-granularity fusion. For each region, based on the different granularities of hazard intensity information in the heat maps, its current risk level, trend changes, and potential development status are assessed, and corresponding alarm signals are generated by combining preset thresholds or dynamic threshold judgment conditions. The short-granularity first fire hazard heat map is used for rapid alarm capture of real-time sudden risks, the medium-granularity second fire hazard heat map is used for trend alarms indicating regional risk changes, and the long-granularity third fire hazard heat map is used for potential hazard warnings and long-term risk warnings. Alarms are classified according to risk level, and alarm information is pushed to management personnel in real time through the visualization interface of the heat maps, realizing multi-level, dynamic, and traceable fire risk warnings for the property area.

[0071] Furthermore, the method also includes: adaptively optimizing the multi-level time granularity window based on the first fire hazard heat map, the second fire hazard heat map, and the third fire hazard heat map.

[0072] Specifically, the intensity and trend of hazards in each area of ​​the first, second, and third fire hazard heatmaps are statistically analyzed to calculate the fluctuation amplitude and frequency of hazards within each time granularity window. Based on the monitoring needs corresponding to different time granularity windows, the start and end times, lengths, and sampling intervals of the original time windows are dynamically adjusted: for areas with large hazard fluctuation amplitudes or continuously increasing risks, the sampling interval can be shortened or the monitoring frequency increased to form a more refined time window; for areas with small hazard fluctuation amplitudes and stable risks, the sampling interval can be appropriately extended to save resources. Through iterative optimization of time windows at all levels, an adaptive multi-level time granularity window dynamically matches the characteristics of regional hazards, further guiding the time-domain dynamic sampling strategy of IoT device clusters, thereby improving the accuracy and response efficiency of fire hazard perception in property management areas and providing reliable and dynamic time support for multi-granularity hazard heatmaps and multi-level fire early warning systems. Example

[0073] Based on the same inventive concept as the smart property service management method in the foregoing embodiments, such as Figure 2 As shown, this application provides a smart property service management platform, wherein the smart property service management platform includes: The multi-level time decomposition module 11 is used to perform multi-level time decomposition based on the historical fire accident event set of the property management area, and construct a multi-level time granularity window; the time-domain dynamic sampling module 12 is used to control the IoT device group to perform time-domain dynamic sampling of the property management area according to the multi-level time granularity window, and obtain a first granularity area monitoring set, a second granularity area monitoring set, and a third granularity area monitoring set; the first fire hazard heat map construction module 13 is used to introduce a fire hazard knowledge space to capture early fire hazard features of the first granularity area monitoring set and construct a first fire hazard heat map; the second fire hazard... The fire hazard heat map establishment module 14 is used to perform multi-dimensional trend verification and correction on the first fire hazard heat map based on the second granularity regional monitoring set, and establish a second fire hazard heat map; the third fire hazard heat map generation module 15 is used to introduce a fire accident simulation hierarchy architecture, combine the third granularity regional monitoring set to optimize the second fire hazard heat map for fire accident simulation, and generate a third fire hazard heat map; the fire management module 16 is used to perform multi-granularity synchronous fire management based on the first fire hazard heat map, the second fire hazard heat map and the third fire hazard heat map.

[0074] Furthermore, the multi-level time decomposition module 11 is also used to: identify the time characteristics of the accident process based on the historical fire accident event set to obtain multiple accident time characteristic sequences; perform multimodal clustering based on the multiple accident time characteristic sequences to establish short-term accident time mode classes, medium-term accident time mode classes, and long-term accident time mode classes; and perform adaptive abrupt change interval identification and fusion based on the short-term accident time mode classes, the medium-term accident time mode classes, and the long-term accident time mode classes to generate the multi-level time granularity window.

[0075] Furthermore, the first fire hazard heat map construction module 13 is also used for: performing multimodal characteristic identification based on the first granularity area monitoring set to construct a first granularity environmental characteristic vector, a first granularity facility characteristic vector, and a first granularity personnel characteristic vector; activating the fire hazard knowledge space, which includes an environmental fire hazard knowledge space, a facility fire hazard knowledge space, and a personnel fire hazard knowledge space; capturing fire hazards from the first granularity environmental characteristic vector based on the environmental fire hazard knowledge space to obtain a first granularity environmental fire hazard vector; capturing fire hazards from the first granularity facility characteristic vector based on the facility fire hazard knowledge space to obtain a first granularity facility fire hazard vector; capturing fire hazards from the first granularity personnel characteristic vector based on the personnel fire hazard knowledge space to obtain a first granularity personnel fire hazard vector; and performing multidimensional hazard integration rendering based on the first granularity environmental fire hazard vector, the first granularity facility fire hazard vector, and the first granularity personnel fire hazard vector to generate the first fire hazard heat map.

[0076] Furthermore, the first fire hazard heatmap construction module 13 is also used for: performing anomaly detection based on the first granularity environmental characteristic vector to obtain a first granularity environmental anomaly feature group; extracting the Kth short-term environmental anomaly feature based on the first granularity environmental anomaly feature group, where K is a positive integer; calculating the cosine similarity between the Kth short-term environmental anomaly feature and each environmental fire hazard knowledge feature in the environmental fire hazard knowledge space to construct a Kth environmental anomaly hazard twin sequence; performing iterative comparison and optimization based on the Kth environmental anomaly hazard twin sequence to determine the Kth environmental anomaly hazard triggering coefficient; if the Kth environmental anomaly hazard triggering coefficient is greater than or equal to the hazard triggering tolerance threshold, adding the Kth short-term environmental anomaly feature to the first granularity environmental fire hazard vector; if the Kth environmental anomaly hazard triggering coefficient is less than the hazard triggering tolerance threshold, filtering the Kth short-term environmental anomaly feature.

[0077] Furthermore, the second fire hazard heat map establishment module 14 is also used for: performing multi-dimensional fire hazard detection on the second granularity area monitoring set according to the fire hazard knowledge space, and establishing a second granularity environmental fire hazard vector, a second granularity facility fire hazard vector, and a second granularity personnel fire hazard vector; performing environmental hazard fluctuation analysis on the first fire hazard heat map according to the second granularity environmental fire hazard vector, and establishing an environmental hazard fluctuation characteristic path; performing facility hazard fluctuation analysis on the first fire hazard heat map according to the second granularity facility fire hazard vector, and establishing a facility hazard fluctuation characteristic path; performing personnel hazard fluctuation analysis on the first fire hazard heat map according to the second granularity personnel fire hazard vector, and establishing a personnel hazard fluctuation characteristic path; and performing adaptive verification and correction on the first fire hazard heat map according to the environmental hazard fluctuation characteristic path, the facility hazard fluctuation characteristic path, and the personnel hazard fluctuation characteristic path to generate the second fire hazard heat map.

[0078] Furthermore, the third fire hazard heat map generation module 15 is also used for: performing multi-dimensional characteristic anomaly detection based on the third-granularity regional monitoring set to obtain a third-granularity environmental anomaly feature group, a third-granularity facility anomaly feature group, and a third-granularity personnel anomaly feature group; performing fire accident simulation on the third-granularity environmental anomaly feature group according to the fire accident simulation hierarchy architecture to obtain an environmental anomaly accident simulation path; performing fire accident simulation on the third-granularity facility anomaly feature group according to the fire accident simulation hierarchy architecture to obtain a facility anomaly accident simulation path; and performing fire accident simulation on the third-granularity facility anomaly feature group according to the fire accident simulation hierarchy architecture. The deduction hierarchy architecture performs fire accident deduction on the third-granularity personnel abnormal feature group to obtain personnel abnormal accident deduction paths; based on the environmental abnormal accident deduction paths, the facility abnormal accident deduction paths, and the personnel abnormal accident deduction paths, multi-factor coupled fire accident deduction is performed to obtain full-scale abnormal accident deduction paths; based on the environmental abnormal accident deduction paths, the facility abnormal accident deduction paths, the personnel abnormal accident deduction paths, and the full-scale abnormal accident deduction paths, the second fire hazard heat map is integrated and optimized to obtain the third fire hazard heat map.

[0079] Furthermore, the third fire hazard heatmap generation module 15 is also used for: classifying fire accidents into multiple factors based on the historical fire accident event set, establishing environmental factor fire accident zones, facility factor fire accident zones, and personnel factor fire accident zones; tracing accident paths and injecting multi-scale disturbances based on the environmental factor fire accident zones, establishing an environment-coupled fire accident space; iteratively supervising and training a Bayesian network model based on the environment-coupled fire accident space, generating a first node for environment-coupled fire accident inference; iteratively supervising and training a GRU model based on the environment-coupled fire accident space, generating a second node for environment-coupled fire accident inference; performing federated aggregation reinforcement based on the first and second nodes for environment-coupled fire accident inference, generating an environment-coupled fire accident inference layer; training a facility-coupled fire accident inference layer based on the facility factor fire accident zones; training a personnel-coupled fire accident inference layer based on the personnel factor fire accident zones; and encapsulating the environment-coupled fire accident inference layer, the facility-coupled fire accident inference layer, and the personnel-coupled fire accident inference layer into the fire accident inference hierarchical architecture.

[0080] Furthermore, the intelligent property service management platform also includes: generating multi-granularity fire hazard alarms based on the first fire hazard heat map, the second fire hazard heat map, and the third fire hazard heat map.

[0081] Furthermore, the intelligent property service management platform also includes: adaptively optimizing the multi-level time granularity window based on the first fire hazard heat map, the second fire hazard heat map, and the third fire hazard heat map.

[0082] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The smart property service management method and specific examples in the aforementioned embodiment one are also applicable to the smart property service management platform in this embodiment. Through the foregoing detailed description of a smart property service management method, those skilled in the art can clearly understand the smart property service management platform in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0083] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0084] Obviously, those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A smart property service management method, characterized in that, The method includes: Based on the historical fire accident event set of the property management area, a multi-level time decomposition is performed to construct a multi-level time granularity window; The IoT device group is controlled to perform time-domain dynamic sampling of the property management area according to the multi-level time granularity window to obtain a first granularity area monitoring set, a second granularity area monitoring set, and a third granularity area monitoring set; A fire hazard knowledge space is introduced to capture early fire hazard characteristics of the first granularity area monitoring set and construct a first fire hazard heat map. Based on the second granularity area monitoring set, the first fire hazard heat map is reviewed and corrected in multiple dimensions to establish a second fire hazard heat map; A hierarchical framework for fire accident simulation is introduced, and the second fire hazard heat map is optimized by combining the third-granularity regional monitoring set to generate a third fire hazard heat map. Multi-granularity synchronous fire management is carried out based on the first fire hazard heat map, the second fire hazard heat map, and the third fire hazard heat map.

2. The intelligent property service management method according to claim 1, characterized in that, Based on the historical fire incident event set of the property management area, a multi-level time decomposition is performed to construct a multi-level time granularity window, including: Based on the historical fire accident event set, the temporal characteristics of the accident process are identified to obtain multiple accident temporal characteristic sequences; Multimodal clustering is performed based on the multiple accident time characteristic sequences to establish short-term accident time modal class, medium-term accident time modal class and long-term accident time modal class; Adaptive mutation interval identification and fusion are performed based on the short-term accident time modality class, the medium-term accident time modality class, and the long-term accident time modality class to generate the multi-level time granularity window.

3. The intelligent property service management method according to claim 1, characterized in that, A fire hazard knowledge space is introduced to capture early fire hazard characteristics of the first granularity area monitoring set, and a first fire hazard heat map is constructed, including: Based on the first granularity area monitoring set, multimodal characteristic identification is performed to construct a first granularity environmental characteristic vector, a first granularity facility characteristic vector, and a first granularity personnel characteristic vector; Activate the fire hazard knowledge space, which includes an environmental fire hazard knowledge space, a facility fire hazard knowledge space, and a personnel fire hazard knowledge space; Fire hazard capture is performed on the first granularity environmental characteristic vector based on the environmental fire hazard knowledge space to obtain the first granularity environmental fire hazard vector; Fire hazard capture is performed on the first-granularity facility characteristic vector based on the facility fire hazard knowledge space to obtain the first-granularity facility fire hazard vector; Fire hazard capture is performed on the first-granularity personnel characteristic vector based on the personnel fire hazard knowledge space to obtain the first-granularity personnel fire hazard vector; Based on the first granularity environmental fire hazard vector, the first granularity facility fire hazard vector, and the first granularity personnel fire hazard vector, a multi-dimensional hazard integrated rendering is performed to generate the first fire hazard heat map.

4. The intelligent property service management method according to claim 3, characterized in that, Fire hazard capture is performed on the first granularity environmental characteristic vector based on the aforementioned environmental fire hazard knowledge space to obtain the first granularity environmental fire hazard vector, including: Anomaly detection is performed based on the first granularity environmental characteristic vector to obtain the first granularity environmental anomaly feature group; Based on the first granularity environmental anomaly feature group, extract the Kth short-term environmental anomaly feature, where K is a positive integer; The cosine similarity of the Kth short-term environmental anomaly feature with each environmental fire hazard knowledge feature in the environmental fire hazard knowledge space is calculated to construct the Kth environmental anomaly hazard twin sequence. The triggering coefficient of the Kth environmental anomaly hazard is determined by iterative comparison and optimization based on the twin sequence of the Kth environmental anomaly hazard. If the trigger coefficient of the Kth environmental anomaly hazard is greater than or equal to the hazard trigger tolerance threshold, the Kth short-term environmental anomaly feature is added to the first granularity environmental fire hazard vector; If the trigger coefficient of the Kth environmental anomaly hazard is less than the hazard trigger tolerance threshold, the Kth short-term environmental anomaly feature is filtered out.

5. The intelligent property service management method according to claim 1, characterized in that, Based on the second granularity area monitoring set, the first fire hazard heat map is subjected to multi-dimensional trend verification and correction to establish a second fire hazard heat map, including: Based on the fire hazard knowledge space, multi-dimensional fire hazard detection is performed on the second-granularity regional monitoring set to establish a second-granularity environmental fire hazard vector, a second-granularity facility fire hazard vector, and a second-granularity personnel fire hazard vector. Based on the second-granularity environmental fire hazard vector, the first fire hazard heatmap is analyzed for environmental hazard fluctuations, and a path for environmental hazard fluctuation characteristics is established. Based on the second-granularity facility fire hazard vector, the first fire hazard heat map is analyzed for facility hazard fluctuations, and a facility hazard fluctuation characteristic path is established. Based on the second-granularity personnel fire hazard vector, the first fire hazard heatmap is analyzed for personnel hazard fluctuations, and a personnel hazard fluctuation characteristic path is established. The first fire hazard heat map is adaptively reviewed and corrected based on the fluctuation characteristic paths of the environmental hazards, the facility hazards, and the personnel hazards to generate the second fire hazard heat map.

6. The intelligent property service management method according to claim 1, characterized in that, A hierarchical framework for fire accident simulation is introduced. The second fire hazard heatmap is optimized using the third-granularity regional monitoring set to generate a third fire hazard heatmap, including: Multidimensional characteristic anomaly detection is performed based on the third-granularity regional monitoring set to obtain the third-granularity environmental anomaly feature group, the third-granularity facility anomaly feature group, and the third-granularity personnel anomaly feature group; Based on the fire accident simulation hierarchy, fire accident simulation is performed on the third-granularity environmental anomaly feature group to obtain the environmental anomaly accident simulation path. Based on the fire accident simulation hierarchy, fire accident simulation is performed on the third-granularity facility anomaly feature group to obtain the facility anomaly accident simulation path. Based on the fire accident simulation hierarchy, fire accident simulation is performed on the third-level personnel abnormal feature group to obtain the personnel abnormal accident simulation path. Based on the environmental anomaly accident simulation path, the facility anomaly accident simulation path, and the personnel anomaly accident simulation path, a multi-factor coupled fire accident simulation is performed to obtain a full set of anomaly accident simulation paths. Based on the environmental anomaly simulation path, the facility anomaly simulation path, the personnel anomaly simulation path, and the total anomaly simulation path, the second fire hazard heat map is integrated and optimized to obtain the third fire hazard heat map.

7. The intelligent property service management method according to claim 1, characterized in that, The steps for constructing the hierarchical architecture for fire accident simulation include: Based on the historical fire accident event set, multi-factor classification was performed to establish fire accident zones for environmental factors, facility factors, and personnel factors. Based on the aforementioned environmental factors, accident path tracing and multi-scale disturbance injection are performed in the fire accident zone to establish an environmentally coupled fire accident space. Based on the aforementioned environmentally coupled fire accident space, the Bayesian network model is iteratively supervised and trained to generate the first node for environmentally coupled fire accident simulation. Based on the environmentally coupled fire accident space, the GRU model is iteratively supervised and trained to generate the second node of the environmentally coupled fire accident simulation. Based on the first node and the second node of the environmental coupled fire accident simulation, a federated aggregation and enhancement are performed to generate the environmental coupled fire accident simulation layer. Based on the aforementioned facility factors, fire accident zones are established, and training facilities are coupled with fire accident simulation layers. Based on the aforementioned personnel factors in fire accident zones, train personnel to couple fire accident simulation layers; The environment-coupled fire accident simulation layer, the facility-coupled fire accident simulation layer, and the personnel-coupled fire accident simulation layer are encapsulated into the fire accident simulation hierarchical architecture.

8. The intelligent property service management method according to claim 1, characterized in that, Based on the first fire hazard heat map, the second fire hazard heat map, and the third fire hazard heat map, a multi-granularity fire hazard alarm is generated.

9. The intelligent property service management method according to claim 1, characterized in that, The multi-level time granularity window is adaptively optimized based on the first fire hazard heat map, the second fire hazard heat map, and the third fire hazard heat map.

10. A smart property service management platform, characterized in that, The steps for implementing the smart property service management method according to any one of claims 1 to 9 include: The multi-level time decomposition module is used to perform multi-level time decomposition based on the historical fire accident event set of the property management area and construct multi-level time granularity windows. The time-domain dynamic sampling module is used to control the IoT device group to perform time-domain dynamic sampling on the property management area according to the multi-level time granularity window, and obtain the first granularity area monitoring set, the second granularity area monitoring set and the third granularity area monitoring set; The first fire hazard heat map construction module is used to introduce the fire hazard knowledge space to capture early fire hazard features of the first granular area monitoring set and construct the first fire hazard heat map. The second fire hazard heat map establishment module is used to perform multi-dimensional trend verification and correction on the first fire hazard heat map based on the second granularity area monitoring set, and establish the second fire hazard heat map. The third fire hazard heat map generation module is used to introduce a fire accident simulation hierarchy architecture, combine the third granularity area monitoring set to optimize the second fire hazard heat map for fire accident simulation, and generate the third fire hazard heat map. The fire management module is used to perform multi-granularity synchronous fire management based on the first fire hazard heat map, the second fire hazard heat map, and the third fire hazard heat map.

Images