Property operation and resource intelligent deployment management method based on AI knowledge graph

Abstract

The present application relates to the technical field of property operation management and fire safety operation and maintenance, and particularly relates to a property operation and resource intelligent deployment management method based on an AI knowledge graph, which gathers property data into a unified data set and updates a property semantic cognitive graph; a fire compartment comprehensive risk quantity is obtained by calculating a function coverage loss index, a stoppage exposure cumulative index and a function graph minimum cut index in a time window, and a threshold set comparison triggers dynamic regulation; resource deployment instructions are solved and output under the constraints of personnel resource availability, material resource availability, operation window and reachable relationship, and work order distribution is written back to update and recalculate, the present application unifies the data set and updates the property semantic cognitive graph, solves the multiple target constraints to output the resource deployment instructions, and recalculate in a closed loop.

Description

Technical Field

[0001] This invention relates to the fields of property operation management and fire safety maintenance technology, and more specifically, it relates to a property operation and resource intelligent allocation management method based on AI knowledge graph. Background Technology

[0002] As the scale and complexity of properties in large commercial complexes, transportation hubs, industrial parks, and high-rise buildings increase, the sources of property operation data are becoming increasingly diverse. This typically includes equipment operation and alarm records, inspection and maintenance records, planned shutdowns or downgrades, personnel scheduling and skill qualifications, spare parts and material inventory, work window approvals, and information on zoning and crowd load. Currently, the industry generally uses multiple management platforms or ledgers to manage this information separately, and conducts risk alerts and work order processing through reports, rule thresholds, or experience-based assessments. While this approach can achieve basic alarm response, closed-loop inspection, and resource dispatch in daily operations, inconsistencies in data definitions, inconsistent entity coding, and difficulties in explicitly expressing cross-professional links often necessitate manual verification across multiple data sources. This impacts the timeliness and consistency of risk assessment and decision-making.

[0003] On the other hand, the security posture at the fire compartment level depends not only on the status of individual equipment, but also on the completeness of functional coverage, the dependency transmission capability of key functional links, and the cumulative impact of planned shutdowns or downgrades on population exposure at different time periods. In existing technologies, shutdown or downgrade management is mostly based on approval and notification, and risk assessment often remains at the level of static description of whether equipment is available or unavailable. It is difficult to quantify the combined effect of shutdown duration, time sensitivity, and population load fluctuations on cumulative exposure within a preset time window. At the same time, the identification of weak links in functional links relies heavily on the experience inference of professionals or post-event review, lacking reusable and calculable structured indicators to characterize the key bottlenecks between the triggering of source risks and the achievement of a safe state in the compartment.

[0004] In terms of dynamic control, existing work assignment and resource allocation typically adopt fixed shifts or nearby assignment principles, supplemented by manual coordination. While this can meet general response needs, it is difficult to simultaneously address multiple objectives such as risk reduction, reduced operational disruption, and resource control when there are concurrent risks across multiple zones, limited work windows, mismatched personnel skills, material storage locations affecting arrival time, and reachability constraints leading to infeasible paths. Furthermore, although execution receipts can be sent back, the returned results are often only used for work order closed-loop statistics and fail to further drive continuous updates of relationships and attributes, thus making it difficult to form an adaptive closed loop of "evaluation—control—execution—writeback—re-evaluation".

[0005] To address the aforementioned shortcomings, this application proposes a property operation and resource intelligent allocation management method based on AI knowledge graphs. Summary of the Invention

[0006] In view of the shortcomings of existing technologies, the purpose of this invention is to provide a property operation and resource intelligent allocation management method based on AI knowledge graph.

[0007] To achieve the above objectives, the present invention provides the following technical solution: AI knowledge graph-based methods for property operation and intelligent resource allocation management include: Step 1: Acquire property operation data and form a unified dataset; Step 2: Construct and update the property semantic cognitive graph based on the unified dataset. The property semantic cognitive graph includes an entity set, a relation set, and an attribute set. The relation set includes at least coverage relationships, dependency relationships, shutdown impact relationships, reachability relationships, and mapping relationships. The attribute set includes at least equipment availability status attributes, shutdown and degradation period attributes, personnel and material capacity attributes, and zoning population load attributes. The property semantic cognitive graph is then updated based on the unified dataset. Step 3: Within a preset time window, calculate the functional coverage missing index, the cumulative index of disuse exposure, and the minimum cut index of the functional graph based on the property semantic cognitive graph, and then calculate the comprehensive risk of the fire compartment. Compare the comprehensive risk of the fire compartment with a preset threshold set to trigger dynamic control. Step four: When dynamic control is triggered, multi-objective constraint solving is performed based on the property semantic cognitive graph reasoning. Under the constraints of personnel resource availability data, material resource availability data, work window data and reachability relationship, a dynamic control strategy with the goal of reducing the comprehensive risk of fire compartments is generated and resource allocation instructions are output. The dynamic control strategy is transformed into a work order through the property management platform and dispatched to the execution end. The execution receipt is received and the unified dataset is written back. The property semantic cognitive graph is updated based on the write-back results and the next round of comprehensive risk calculation of fire compartments is triggered, forming a closed-loop dynamic control.

[0008] In one embodiment, the property operation data includes at least property space topology data, fire prevention zone data, life safety system equipment ledger data, real-time status data of life safety system equipment, planned shutdown or downgrade event data, personnel resource availability data, material resource availability data, work window data, and zoned population load data.

[0009] In one embodiment, the entity set of the property semantic cognitive graph includes at least property space units, fire compartments, life safety system equipment, functional nodes, planned shutdown or downgrade events, personnel resources, and material resources.

[0010] In one embodiment, the functional coverage gap index is obtained as follows: within a preset time window, planned deactivation or downgrade events are associated with fire compartments based on mapping relationships and deactivation impact relationships; the functional node coverage corresponding to the fire compartment is obtained based on the property semantic cognitive graph; the functional node coverage is compared with a preset coverage threshold to obtain a node gap index, and the node gap index within the fire compartment is weighted and aggregated to obtain the functional coverage gap index.

[0011] In one embodiment, the functional node coverage is obtained as follows: based on coverage and dependency relationships, the functional nodes corresponding to the fire compartment and the life safety system equipment corresponding to each functional node are determined; based on the equipment availability status attribute and combined with the decommissioning and degradation period attribute, the equipment availability of the life safety system equipment is obtained; based on coverage and dependency relationships, the equipment availability is synthesized into functional node coverage.

[0012] In one embodiment, the cumulative exposure index for deactivation is obtained as follows: an exposure factor is determined, and a time period weight is applied to the exposure factor according to the deactivation downgrade period attribute; the exposure factor is weighted and accumulated within a preset time window according to a preset sampling interval to obtain the cumulative exposure index for deactivation.

[0013] In one embodiment, the determination of the exposure factor includes: obtaining the population load weight within a preset time window based on the population load attributes of the partition; calculating the criticality weight of the life safety system based on the equipment type and dependency relationship of the life safety system; and obtaining the exposure factor based on the population load weight, the criticality weight of the life safety system, and the node gap index.

[0014] In one embodiment, the minimum cut index of the functional graph is obtained as follows: within a preset time window, a life safety functional graph oriented towards the fire compartment is constructed based on functional nodes, mapping relationships and dependencies, and edge weights are set for the edges of the life safety functional graph; at least one source node and at least one sink node are selected from the life safety functional graph; minimum cut calculation is performed with the edge weights as input to obtain the minimum cut index of the functional graph, and the minimum cut edge set obtained by the minimum cut calculation is output as the key function transmission edge.

[0015] In one embodiment, setting edge weights for the life safety function graph edges includes: life safety function graph nodes are function nodes, and life safety function graph edges are function transfer edges with dependencies at the function node level; determining the life safety system device supporting each function transfer edge based on the mapping relationship; obtaining the device availability of the supporting device based on the device availability status attribute and combined with the deactivation / degradation period attribute; synthesizing the device availability of the supporting device into the function transfer edge availability based on the dependency relationship, and using the function transfer edge availability as the edge weight.

[0016] In one embodiment, the source node is a functional node that indicates the triggering of a risk event, and the sink node is a functional node that indicates that the fire compartment has reached a safe state.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention acquires property operation data to form a unified dataset, and constructs and updates a property semantic cognitive graph containing entity sets, relationship sets, and attribute sets. This enables the unified expression of coverage relationships, dependency relationships, shutdown impact relationships, reachability relationships, and mapping relationships. Furthermore, it incorporates equipment availability status attributes, shutdown and degradation period attributes, personnel and material capacity attributes, and zoning population load attributes into the same update chain, thereby improving data consistency, traceability, and continuous update capabilities. Within a preset time window, based on the property semantic cognitive graph, the missing function coverage index, the cumulative index of disuse exposure, and the minimum cut index of the functional graph are calculated synchronously. Furthermore, the comprehensive risk of the fire compartment is obtained and compared with the preset threshold set to achieve hierarchical triggering and dynamic control. This transforms the compartment risk from a static state description into a calculable, comparable, and triggerable quantitative basis, enhancing the time-time adaptability of risk identification and the interpretability of indicators. After dynamic control is triggered, the system reasons based on the property semantic cognitive graph and performs multi-objective constraint solving. It incorporates personnel resource availability data, material resource availability data, operation window data, and reachability constraints into the feasible domain, generates a dynamic control strategy aimed at reducing the overall risk of fire compartments, and outputs resource allocation instructions. Then, the dynamic control strategy is transformed into work orders and dispatched to the execution end through the property management platform. The system receives execution receipts and writes back to the unified dataset to update the property semantic cognitive graph and trigger the next round of overall risk calculation, forming a closed-loop and iterative dynamic control process, improving the implementation of allocation decisions and the quality of closed-loop handling. Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating the overall process and closed loop of an AI knowledge graph-based intelligent allocation and management method for property operation and resources. Figure 2 A schematic diagram illustrating the structure for constructing and updating a property semantic cognitive graph based on a unified dataset; Figure 3 A flowchart illustrating the process of comparing the overall risk level of a fire compartment with a set of preset thresholds to trigger dynamic control. Figure 4 A schematic diagram of the closed-loop process for updating the property semantic cognitive graph and triggering the next round of comprehensive risk calculation for fire compartments. Detailed Implementation

[0019] Reference Figure 1The property operation and resource intelligent allocation management method based on AI knowledge graph includes: Step one involves acquiring property operation data and forming a unified dataset. This involves bringing together property operation data scattered across different systems, formats, and time granularities into a single data foundation. By forming a unified dataset, consistent data definitions, coding, and time dimensions are achieved, providing an input basis for the subsequent construction, updating, and calculation of a property semantic cognitive graph based on the same source of facts. It also reserves data storage space for the closed-loop mechanism of writing back to the unified dataset, enabling the execution results to be continuously accumulated and reused.

[0020] In one specific implementation, using property space topology data as the framework, spatial units such as buildings, floors, corridors, and rooms are coded and their interconnections are sorted out. At the same time, fire compartment data is introduced to establish a one-to-one correspondence between each fire compartment and the corresponding set of spatial units, ensuring that subsequent calculations can locate the specific fire compartment range. Data on life safety system equipment is collected, including at least the unique identifier of the equipment, installation location, fire compartment to which it belongs, associated functional nodes, maintenance responsibilities, and rated capacity. Real-time status data of life safety system equipment is continuously collected within a preset time window to form a timestamped status sequence to reflect the equipment's available status attributes. Data on planned shutdown or downgrade events is obtained, recording the start and end times of the events, the scope of impact, and execution constraints to characterize the attributes of the shutdown or downgrade period. Simultaneously, personnel and material resource availability data are aggregated to describe the skills, shifts, and reach of available personnel, as well as the specifications, quantity, and storage location of spare parts and materials, to support subsequent personnel and material capacity attributes; further, operational window data is obtained to limit the time period during which operations are permitted, and zonal crowd load data is obtained to quantify the intensity of personnel gathering in different fire protection zones at different time periods, forming zonal crowd load attributes; the above data are standardized in terms of definitions, aligned in terms of labels, and aligned in terms of time sequence to form a unified dataset that can be consistently referenced; For example, taking a three-story commercial complex as an example, access control records and inspection registrations during opening hours are aggregated to generate zonal crowd load data. At the same time, the time periods during which maintenance is allowed at night are fixed as work window data. Furthermore, the data on planned shutdown or downgrade events of the day are synchronized with the real-time status data of the corresponding life safety system equipment and incorporated into a unified dataset. This provides a complete and traceable data foundation for subsequent risk quantification and control reasoning within the same preset time window.

[0021] Reference Figure 2Step 2: Construct and update the property semantic cognitive graph based on the unified dataset. The property semantic cognitive graph includes an entity set, a relation set, and an attribute set. The relation set includes at least coverage relationships, dependency relationships, shutdown impact relationships, reachability relationships, and mapping relationships. The attribute set includes at least equipment availability status attributes, shutdown and degradation period attributes, personnel and material capacity attributes, and zoning population load attributes. The property semantic cognitive graph is then updated based on the unified dataset. The objects, relationships, and attributes in the unified dataset are organized into a computable and reasonable structured representation: entity sets represent objects such as property space units, fire compartments, life safety system equipment, and functional nodes; coverage relationships, dependency relationships, shutdown impact relationships, reachability relationships, and mapping relationships characterize the interaction paths and constraint boundaries between objects; and equipment availability status attributes, shutdown and degradation period attributes, personnel and material capacity attributes, and zoning population load attributes characterize the status and capability parameters of objects. By constructing and updating the graph, it is made to evolve synchronously with changes in the unified dataset, thereby ensuring that subsequent indicator calculations and dynamic control reasoning within the preset time window are always based on the latest operational facts and constraints.

[0022] In one specific implementation, an entity set is extracted and solidified from a unified dataset. The entity set includes at least property space units, fire compartments, life safety system equipment, functional nodes, planned shutdown or downgrade events, personnel resources, and material resources. Property space units and fire compartments are each uniquely identified and maintain a consistent hierarchy. Life safety system equipment uses a unique identifier in the ledger as the primary key and binds the installation location to the fire compartment to which it belongs. Functional nodes are used to abstractly represent the life safety functions that need to be protected within the fire compartment. Planned shutdown or downgrade events are used to describe the planned change objects within a preset time window. Personnel resources and material resources are used to describe the objects of available disposal capacity. Construct a set of relationships and perform consistency checks on the relationships. The set of relationships includes at least coverage relationships, dependency relationships, deactivation impact relationships, reachability relationships, and mapping relationships. Among them, the coverage relationship is used to characterize the coverage boundary of life safety system equipment to functional nodes and functional nodes to fire compartments. The dependency relationship is used to characterize the functional transfer dependency links between functional nodes or between equipment. The deactivation impact relationship is used to establish the impact path between planned deactivation or degradation events and the affected life safety system equipment, functional nodes, and fire compartments. The reachability relationship is used to characterize the feasible path and constraints for personnel resources and material resources to reach the target fire compartment under the property space topology constraints. The mapping relationship is used to establish corresponding rules between objects of different granularities, such as mapping life safety system equipment to the functional nodes it supports and mapping functional nodes to the fire compartments it serves. Mapping relationship generation: for each life safety system device Configuration fields ,in This indicates the functional nodes supported by the device; for each functional node... Configuration fields The mapping relationships include: Support mapping, Service mapping. When the device ledger or mapping configuration table changes, the mapping relationship is incrementally updated by primary key, and the data source and timestamp are recorded; when the same... Corresponding to multiple When a conflict is resolved, a unique mapping is determined by a preset priority rule or the latest effective record, and the conflict resolution result is written into the mapping relationship attribute for traceability. Reachability and arrival time: Constructing an undirected graph using property spatial topology data ,in For a set of spatial nodes, Let be a set of connected edges, with edges With passage cost For any personnel resources With material resources Given its starting spatial node or ;Target fire compartment Take its entrance space node set The arrival time is calculated using the shortest path algorithm. , ,in To follow the shortest path The summation yields the travel time. When And if there are no forbidden edges on the shortest path, write the resources into the graph. The reachability relationship, and will The path node sequence and constraint reasons are used as reachability attributes, and are incrementally updated when the topology or access constraints change; Write the attribute set, which includes at least the equipment availability status attribute, the deactivation / degradation period attribute, the personnel and material capacity attribute, and the zoned population load attribute. The equipment availability status attribute is updated by timestamps from the real-time status data of the life safety system equipment. The deactivation / degradation period attribute is written by the planned deactivation or degradation event data and associated with the corresponding impact path. The personnel and material capacity attribute is written by the personnel resource availability data and the material resource availability data and constrains the available capacity together with the work window data. The zoned population load attribute is written by the zoned population load data and updated by time slice. The semantic cognitive graph of a property is updated based on a unified dataset. The update process includes incrementally merging new entities, correcting relationship conflicts, supplementing missing attributes, and eliminating expired events. Each update records the data source and timestamp to ensure traceability. For example, taking an office building as an example, when the unified dataset shows that a life safety system device has been added to a fire compartment and bound to a specific property space unit in the ledger, the graph adds a corresponding life safety system device entity and connects it to relevant functional nodes through overlay relationships, and then connects it to the fire compartment through mapping relationships. When a planned shutdown or downgrade event occurs on the same day and the shutdown / downgrade period attribute is marked, the graph adds a planned shutdown or downgrade event entity and connects the impact path to the affected life safety system device, functional nodes, and fire compartment through shutdown impact relationships. Simultaneously, it updates the personnel and material capacity attributes by combining personnel and material resource availability data, and updates reachability relationships using property space topology data, thereby ensuring that subsequent calculations and inferences are based on the same up-to-date and consistent semantic expression.

[0023] Reference Figure 3 Step three: Within a preset time window, calculate the functional coverage missing index, the cumulative index of deactivation exposure, and the minimum cut index of the functional graph based on the property semantic cognitive graph. Then, calculate the comprehensive risk level of the fire compartment and compare it with a preset threshold set to trigger dynamic control. Transform the system status, deactivation impact, and functional transfer structure within the preset time window into measurable and comparable risk quantification results. The calculation of the functional coverage gap index reflects the functional coverage gap at the fire compartment level; the calculation of the deactivation exposure accumulation index reflects the degree of exposure accumulated over time by deactivation or downgrading; and the calculation of the functional diagram minimum cut index reflects the vulnerability of the life safety functional diagram in critical transmission links. The above results are further integrated to form the comprehensive risk quantity of the fire compartment, which is then compared with a set of preset thresholds to determine whether dynamic control needs to be triggered. This ensures that the control triggering has a unified scale and auditable basis, avoiding mis-adjustment or omission caused by relying solely on experience or a single signal.

[0024] In one specific implementation, planned deactivation or degradation events are read within a preset time window. Each planned deactivation or degradation event is mapped to the life safety system equipment and functional nodes it affects based on mapping relationships. The affected functional nodes are then correlated to the corresponding fire compartments based on the deactivation impact relationships, thus obtaining the set of deactivation or degradation impacts for each fire compartment within the preset time window. The coverage of functional nodes corresponding to the fire compartment is obtained based on the property semantic cognitive graph. Specifically, the set of functional nodes required to be protected by the fire compartment is determined based on coverage and dependency relationships, and the set of life safety system equipment corresponding to each functional node is further determined. Based on this, the equipment availability of each life safety system equipment within the preset time window is calculated based on the equipment availability status attributes and the deactivation / degradation period attributes, so that the equipment availability can reflect the effective availability level after the real-time status changes and the superposition of planned deactivation or degradation events. Deactivation Impact Reasoning: For any planned deactivation or downgrade event Read its affected device set ,in The impact scope field or the list of associated devices in the event data is given; for each The functional nodes are obtained through mapping relationships. And further obtained fire compartmentation This forms an influence chain. The effective range of this impact chain is limited to the period of time during which the event is deactivated or downgraded. ;If the same Affecting multiple or multiple Then, the affected chain is deduplicated and merged, and a list of affected objects and merging rules are recorded for traceability; The preset time window is denoted as Window duration is For any life safety system device, its availability status is determined based on the device's available status attributes. The state sequence within the device is used to obtain the set of deactivation or degradation periods for the device based on the deactivation / degradation period attribute. Define valid and usable indicator functions When the device is unavailable or When a period falls within any set of deactivated or downgraded periods When the device is available and When falling into the relegation period When the device is available and not in a degraded period ;in The downgrade factor has a range of values. When the attribute for the deactivation period is not marked with a deactivation coefficient, the default value will be used. The availability of the life safety system equipment within a preset time window is defined as follows: And the implementation of multiple discrete sampling points is written as For a given functional node, determine the set of life safety system devices it supports based on coverage relationships. And synthesize coverage using a redundant coverage model When the dependency relationship indicates that the functional node depends on the set of preceding functional nodes, At that time, the dependency constraints will be reflected as For the availability of the function transfer edge, the set of life safety system devices supporting the function transfer edge is obtained according to the mapping relationship. And use the same redundant coverage model to calculate the function to transfer edge availability. and will This weight is written into the life safety function graph; Based on coverage and dependency relationships, equipment availability is synthesized into functional node coverage, so that functional node coverage can reflect the collaborative contribution of multiple supporting devices to the same functional node and the constraint of dependent links on coverage effect. The functional node coverage is compared with a preset coverage threshold to obtain the corresponding node gap index. The node gap index within the fire compartment is then weighted and aggregated to obtain the functional coverage missing index. The weighted aggregation can adopt a weight rule consistent with the importance of functional nodes to ensure that the contribution of different functional nodes to the functional coverage missing index of the fire compartment is interpretable. Pre-configure a preset coverage threshold for each functional node. The preset coverage threshold value range is: Furthermore, the importance of functional nodes can be determined based on the type of life safety system equipment and historical safety operation data. The importance of functional nodes can be represented by the aggregated results of the criticality weights of the life safety system at the functional node level, and the historical safety operation data is limited to a preset time window sample where dynamic control has not been triggered. In a reproducible value selection method, the functional node coverage distribution in the historical samples is set as follows: Then Let it be its quantile value ,in The range of values ​​is And the default value is ; To obtain the functional node coverage within the current preset time window Then, the nodal gap index is defined as ,make The range of values ​​is and The time node gap index is zero; for multiple functional nodes within the same fire compartment, the node gap index is weighted and aggregated according to the importance of the functional nodes, and the weight set is... and The missing function coverage index is defined as follows: When no weight set is configured, equal weights are used by default; For example, taking a fire compartment in a commercial building as an example, if a planned shutdown or downgrade event occurs within a preset time window, causing a life safety system device to be unavailable during the period defined by the shutdown / downgrade time attribute, the graph will associate this event with the fire compartment through mapping relationships and shutdown impact relationships. Subsequently, in the set of functional nodes corresponding to the fire compartment, a certain functional node was originally covered by two life safety system devices. One of them is affected by shutdown within the preset time window, and the other device's availability status shows intermittent alarms. The combined device availability is reduced, and then, through coverage and dependency relationships, it is synthesized into a lower functional node coverage. When the functional node coverage is lower than the preset coverage threshold, a higher node gap index is obtained, and it is aggregated with the node gap indices of other functional nodes in the fire compartment according to weights. Finally, a functional coverage gap index is formed that can reflect the degree of functional coverage gap in the fire compartment within the preset time window, providing quantitative input for subsequent calculation of the comprehensive risk of the fire compartment.

[0025] In one specific implementation, the exposure factor is determined and a time-based weight is applied to the exposure factor based on the deactivation and degradation period attribute. Specifically, the population load attribute of the fire compartment is read within a preset time window, the population load level is statistically analyzed by time slice and normalized to obtain the population load weight; at the same time, the life safety system equipment type is read and combined with the dependency relationship to calculate the criticality weight of the life safety system. The criticality weight of the life safety system is used to characterize the importance of different life safety system equipment types in supporting functional nodes in the dependency link. The basic weight of the type can be modified according to the hierarchical depth of the functional transmission link in the dependency relationship, the number of support paths, and the range of affected functional nodes. Furthermore, a node gap index within the same preset time window is introduced, enabling the exposure factor to simultaneously reflect the population load risk, the criticality of the life safety system, and the degree of functional coverage gap. The exposure factor can be synthesized by population load weight, life safety system criticality weight, and node gap index in a weighted summation or weighted product manner. The effective period of exposure factor within the preset time window is determined based on the attributes of the shutdown or downgrade period, and a time period weight is applied to the exposure factor within the shutdown or downgrade period, so that the exposure intensity of the same fire compartment in different time segments changes with the duration of shutdown or downgrade and the time period sensitivity. Within a preset time window, the exposure factor after applying time period weights is weighted and accumulated at a preset sampling interval to obtain the deactivation exposure accumulation index. Specifically, sampling is performed from the start of the preset time window at a preset sampling interval. At each sampling moment, the corresponding population load weight and equipment-related dependency status are read, the exposure factor is updated and multiplied by the time period weight of that sampling moment, and the weighted results of each sampling moment are accumulated. If necessary, the sampling interval length can be scaled to ensure that the results under different preset sampling intervals are comparable. To ensure consistency between the exposure factor, time period weight, and preset sampling interval, the sampling time is set within the preset time window for each sampling moment. Read the load attributes of the partitioned audience and obtain the audience load weight. ,in The value is obtained by normalizing the population load level of the fire compartment within a preset time window using minimum and maximum normalization, and the value range is [value missing]. Read the device types of the life safety system and combine them with dependencies to obtain the criticality weight of the life safety system. and normalize it to Read the node gap index corresponding to the same sampling time. The exposure factor is defined using a product synthesis method. The fire compartment is determined based on the attributes of the downgrade period during which it is deactivated. Collection of deactivated or downgraded periods within And define time period weights :when When not falling within any set of deactivated or downgraded periods ,when When falling into the set of periods of inactivity or downgrade ,in The range of values ​​is And the default value is , The time-sensitivity is defined as follows, and the value range is: It can also be set according to the data of the work window and the population load attributes of the area during the day's peak and valley periods; let the preset sampling interval be... And the number of sampling points is The definition of discontinuing the cumulative exposure indicator is... ,in Used to perform scale conversion to ensure that results under different preset sampling intervals are comparable; For example, in a commercial complex, a fire compartment experiences a significant increase in its population load weight during midday. Simultaneously, the life safety system equipment associated with key functional nodes within this fire compartment holds a critical support position in the dependency relationship, resulting in a high life safety system criticality weight. Furthermore, the coverage of the corresponding functional nodes in this fire compartment is below a preset coverage threshold, leading to an increased node gap index. Therefore, the synthesized exposure factor increases. If planned shutdown or downgrade events cover the midday period within the same preset time window, the time period weight corresponding to the shutdown / downgrade period attribute increases the exposure contribution of that period. Ultimately, after accumulating at preset sampling intervals within the preset time window, a higher cumulative shutdown exposure index is formed, used to characterize the cumulative exposure degree of the fire compartment under the influence of shutdown or downgrade.

[0026] In one specific implementation, within a preset time window, a life safety function graph oriented towards fire compartments is constructed based on functional nodes, mapping relationships, and dependencies. Edge weights are assigned to the edges of the life safety function graph, where nodes are functional nodes and edges are functional transmission edges representing dependencies at the functional node level. For each functional transmission edge, a set of life safety system devices supporting that edge is determined based on the mapping relationship. The availability of each supporting device within the preset time window is obtained based on the device availability status attribute and the deactivation / degradation period attribute, ensuring that the device availability reflects the effective availability level after real-time status changes and planned deactivation or degradation events. The device availability of the supporting devices is then combined with the functional transmission edge availability based on dependencies, and this functional transmission edge availability is written into the life safety function graph as the edge weight. This ensures that the edge weight directly corresponds to the functional transmission capability, effectively characterizing the vulnerability of the functional transmission link within the preset time window. Select at least one source node and at least one sink node from the life safety function graph, where the source node represents the function node that triggers a risk event and the sink node represents the function node that the fire compartment has reached a safe state. Perform minimum cut calculation with edge weights as input to obtain the minimum cut index of the function graph, and output the minimum cut edge set obtained by the minimum cut calculation as the key function transfer edge, so that the key function transfer edge can be used to identify the function transfer bottleneck that has a decisive impact on the safety state of the fire compartment. To ensure the deterministic construction and normalization of the minimum cut index of the functional graph, the life safety functional graph constructed for fire compartments is denoted as a directed graph. ,in This is the set of functional nodes corresponding to this fire compartment. This is the set of transitive edges representing dependencies at the functional node level; for each transitive edge... Based on the mapping relationship, the set of life safety system devices that support the transmission edge of this function is obtained. And calculate the availability of the function-transfer edge. and with As the edge weight and Regarding the selection of source and sink nodes, the set of source nodes... The selected functional nodes are those representing the triggering of risk events and include at least functional nodes associated with planned deactivation or downgrade events through deactivation impact relationships. The sink node set $T$ is selected as the functional nodes representing the fire compartment reaching a safe state and includes at least functional nodes semantically corresponding to the fire compartment reaching a safe state. With any Perform minimum cut calculations on the endpoints and take the minimum value as the minimum cut value. The minimum cut edge set is denoted as The output is the key function transmission edge; to complete the normalization, a reference graph is constructed. Make its edge set and Consistent and for all edge commands Calculate the reference minimum cut value The minimum cut index of the function graph is defined as follows: And the range of values ​​is And the vulnerability index can be defined as So that a larger value indicates a higher risk; For example, within a fire compartment of an underground commercial street, functional nodes representing the triggering of risk events are selected as source nodes, such as functional nodes related to fire detection triggering, and functional nodes representing the fire compartment reaching a safe state are selected as sink nodes, such as functional nodes related to smoke exhaust linkage and evacuation guidance. If a planned shutdown or degradation event occurs within a preset time window, causing a decrease in the availability of some life safety system equipment during the shutdown / degradation period, the availability of the functional transmission edges synthesized by the dependency relationship will decrease accordingly and be reflected in the edge weights. Minimum cut calculation will tend to identify these low-weight links as key locations for cutting the connectivity from the source node to the sink node. The final output functional graph minimum cut index is used to quantify the weakest transmission capability of the fire compartment from the triggering of a risk event to the achievement of a safe state within the preset time window. At the same time, the output key functional transmission edges are used to indicate functional transmission links that should be given priority and restored.

[0027] In one specific implementation, within the same preset time window, the functional coverage gap index, the cumulative deactivation exposure index, and the functional graph minimum cut index of the fire compartment are summarized. The three are then standardized in scale and direction. For example, the functional graph minimum cut index is converted into a vulnerability index equal to one minus the normalized minimum cut value, so that a larger value indicates a higher risk. The three types of indicators are weighted and fused according to a preset weight set to obtain the comprehensive risk of the fire compartment. The preset weight set can be adaptively fine-tuned according to the purpose of the fire compartment, the characteristics of the population, and the sensitivity of the time period. A confidence decay rule can be introduced to reduce the contribution of missing or highly fluctuating inputs, thereby improving the stability and interpretability of the comprehensive risk. To avoid frequent fluctuations in the overall risk level around the threshold, the overall risk level can be smoothed over time and a continuous over-limit criterion can be set. For example, it can be required that the same threshold is exceeded within multiple consecutive preset sampling intervals before triggering is determined. At the same time, a fallback hysteresis threshold can be set to support the restoration of the safe state and the trigger cancellation. The overall risk level of the fire compartment is compared with the preset threshold set level by level. The preset threshold set includes at least the attention threshold, alarm threshold and intervention threshold. The corresponding output is the hierarchical control intention and trigger mark, and the marginal contribution of each indicator in the trigger cause is recorded for traceability. Functional coverage gap index Discontinue use of cumulative exposure indicators Vulnerability corresponding to the minimum cut metric of the function graph To achieve standardization in both scale and direction, all three aspects should be normalized to the same standard. Furthermore, a larger numerical value indicates a higher risk; let the preset weight set be... and The comprehensive risk of a fire compartment is defined as follows: The preset weight set can be adaptively fine-tuned based on the purpose of the fire compartment, population characteristics, and time sensitivity. Specifically, it involves... Multiply by the corresponding scene coefficients respectively Then renormalize; and a confidence decay rule can be introduced, so that the confidence levels of the three types of inputs are respectively And the range of values ​​is The overall risk level will then be revised to... The confidence level can be determined based on the proportion of missing inputs and the variance after time smoothing; to avoid frequent fluctuations in the overall risk value around the threshold, [further details are needed]. Perform time smoothing, and let the smoothing coefficient be... And the range of values ​​is The smoothed overall risk is Define the preset threshold set as follows: And satisfy ,when In continuous When the value exceeds the same threshold within a preset sampling interval, the corresponding level of dynamic adjustment is triggered, and a fallback hysteresis threshold is set. Make the trigger release condition as In continuous Within a preset sampling interval, if the value is less than the threshold, subtract... ; For example, during the afternoon peak hours when people gather, the functional coverage gap index of a fire compartment in an office building is normalized to 0.32, the cumulative exposure index after deactivation is scaled to 0.58, and the minimum cut index of the functional graph is converted to a vulnerability of 0.46. These are then combined with weights of 0.4, 0.4, and 0.2 to obtain a comprehensive risk of 0.45. When the preset threshold set has a concern threshold of 0.30, an alarm threshold of 0.40, and an intervention threshold of 0.55, if this comprehensive risk exceeds the alarm threshold for three consecutive sampling intervals but does not reach the intervention threshold, an alarm-level dynamic control is triggered, and a list of fire compartments requiring priority handling is generated. At the same time, the key function transmission edges and the deactivation impact paths are retained as the basis for control.

[0028] Reference Figure 4 Step four: When dynamic control is triggered, multi-objective constraint solving is performed based on the property semantic cognitive graph reasoning. Under the constraints of personnel resource availability data, material resource availability data, work window data, and accessibility relationships, a dynamic control strategy aimed at reducing the overall risk of fire compartments is generated, and resource allocation instructions are output. The dynamic control strategy is transformed into a work order through the property management platform and dispatched to the execution end. Execution receipts are received and the unified dataset is written back. The property semantic cognitive graph is updated based on the write-back results, and the next round of fire compartment overall risk calculation is triggered, forming a closed-loop dynamic control. The risk state that triggers dynamic control is transformed into an action plan that is executable, implementable, and verifiable in a closed loop: Based on the property semantic cognitive graph reasoning, combined with personnel resource availability data, material resource availability data, work window data, and accessibility relationships, multi-objective constraint solving is performed to ensure that the generated dynamic control strategy aims at reducing the overall risk of fire compartments while also meeting realistic constraints such as resource supply, work time, and accessibility, thereby avoiding unexecutable instructions or instructions that conflict with on-site conditions. The property management platform transforms dynamic control strategies into work orders and dispatches them to the execution end, giving the strategies a standardized execution vehicle and a closed-loop responsibility system. By receiving execution receipts and writing them back to the unified dataset, the execution results are solidified into data facts, and the property semantic cognitive graph is updated accordingly, triggering the next round of comprehensive risk calculation for fire compartments. This achieves a closed-loop dynamic control process of calculation—triggering—solving—execution—writing back—updating—recalculating, enabling continuous tracking, verification, and iterative optimization of risk changes and control effects.

[0029] In one specific implementation, based on the trigger record, the fire compartment and its contribution factor for dynamic control are locked. The functional coverage missing index, the cumulative index of deactivation exposure, the minimum cut index of the functional graph and its corresponding node gap index, the impact path of planned deactivation or downgrade events and the set of key function transmission edges of the fire compartment are read. The path constraint of the reachability relationship is expanded in the property semantic cognitive graph. The personnel resource availability data, material resource availability data and operation window data are transformed into feasible solution boundaries. A multi-objective constraint problem is constructed with the main objective of reducing the comprehensive risk of the fire compartment and the secondary objectives of minimizing resource occupation and minimizing business disturbance. The constraints include at least personnel skill matching, shift availability, material specification matching, storage location arrival time, allowed operation window time period, concurrent operation limit and safety preconditions limited by the deactivation impact relationship. Let the set of optional control actions be denoted as Each control action corresponds to one or more combinations of the following: disabling event scheduling strategy, partition load reduction strategy, critical edge recovery strategy, or coverage loss compensation strategy; let binary decision variables... Indicates whether to select a regulatory action. And the allocation variables of human resources and material resources are respectively and To represent personnel resources Whether it is assigned to a regulatory action and material resources In regulatory actions The amount of consumption in the process; regarding the calculation of the reduction in risk, based on the impact of the control actions on the equipment availability status attributes, the attributes of the shutdown and degradation period, the attributes of the load of the population in different areas, and the availability of the functional transfer edge, the predictive indicators after the control actions are implemented are derived. And calculate the predicted comprehensive risk amount Therefore, the decrease in risk quantity is defined as Construct a multi-objective constrained problem with the primary objective of reducing the overall risk of fire compartments and secondary objectives of minimizing resource consumption and minimizing business disruptions. Specifically, minimize the objective function. ,in For the cost of personnel resources, For business disruption costs, and The weights are defined as follows and their values ​​range from 1 to 2. Constraints are set based on personnel resource availability data, material resource availability data, work window data, and reachability constraints: personnel skill matching, shift availability, material specification matching, storage location arrival time, allowed work window time period, concurrent work limit, and safety preconditions defined by the impact of shutdown must all be met; in one implementation, the above problem is represented as a mixed integer programming problem and solved using a standard solver. When the scale is too large, greedy search or heuristic iteration is used. Improvement as an iteration stopping criterion; Recalculating control actions to indicators: The set of available control actions is denoted as... For each regulatory action Define its update rules for graph attributes. The equipment repair action will correspond to Update the device availability status attribute to "Available" or increase the device availability to a preset value. The spare parts replacement action will deduct the material resource capacity attribute according to the consumption quantity and update it synchronously. Status; Disable event scheduling action modification The attribute of the deactivation and downgrade period The partitioned load reduction action will adjust the load attributes of the user group in a partition according to a coefficient within a specified time slice. Downward adjustment. For any set of candidate actions. implement Then, recalculate the missing function coverage index according to step three. Discontinue use of cumulative exposure indicators Vulnerability corresponding to the minimum cut metric of the function graph To obtain the predictive indicators And calculate the predicted comprehensive risk amount Therefore, the decrease in risk quantity is defined as ; This problem-solving process allows for the generation and combination of multiple dynamic control strategies. These strategies include at least the following: a deactivation event scheduling strategy to adjust the concurrent number, execution order, or execution time of planned deactivation or degradation events to reduce the cumulative deactivation exposure index; a zonal load reduction strategy to implement property operation load reduction control on fire zones that trigger dynamic control to reduce the population load weight corresponding to the zonal population load attribute; a critical edge recovery strategy to generate priority recovery instructions based on critical function transmission edges and allocate maintenance personnel resources, spare parts resources, and work windows to prioritize the recovery of life safety system equipment supported by critical function transmission edges, thereby improving the minimum cut index of the functional graph; and a coverage gap compensation strategy to generate compensation resource allocation instructions based on the functional coverage gap index and allocate fire patrol personnel resources, security personnel resources, maintenance personnel resources, temporary fire-fighting equipment resources, and emergency lighting resources to reduce the effective exposure of the node gap index. The dynamic control strategy obtained from solving multi-objective constraints is output as a resource allocation instruction and executed in a closed loop. Specifically, the resource allocation instruction is structured according to fire compartment, execution time period, personnel and material resource list, arrival path and operation window. The dynamic control strategy is then converted into work orders and dispatched to the execution end through the property management platform. At the same time, each work order is bound to a corresponding planned shutdown or downgrade event, key function transmission edge or node gap index source to ensure traceability. After receiving the execution receipt, the arrival time, handling action, replacement parts, retest results and recovery status are written back to the unified dataset. Based on this, the equipment availability status attribute, shutdown and downgrade time period attribute, personnel and material capacity attribute and reachability relationship in the property semantic cognitive graph are updated, and then the next round of fire compartment comprehensive risk calculation is triggered to form a closed loop dynamic control. For example, in a commercial complex, a fire compartment experiences an increase in the midday crowd load weight and a cumulative increase in the exposure index due to planned shutdowns or downgrades. Simultaneously, the key function transmission edge corresponding to the minimum cut shows a low weight for the smoke exhaust linkage link edge. Under the constraint that hot work is only permitted at night, the multi-objective constraint solution prioritizes a shutdown event scheduling strategy to postpone unnecessary shutdowns and reduce concurrency. It also implements a partition load reduction strategy to limit flow and provide time-based guidance at the entrance of the fire compartment to immediately reduce the crowd load weight. Furthermore, it provides a key edge recovery strategy to allocate maintenance personnel with smoke exhaust linkage repair skills and corresponding spare parts and materials to handle the situation at night via the nearest accessible path. Additionally, it employs a coverage gap compensation strategy to temporarily deploy fire patrol personnel and emergency lighting materials for transitional compensation. The execution terminal feedback shows that the equipment's availability attribute has improved after the key equipment is restored, the graph is updated, and the next round of comprehensive risk calculation is triggered. If the comprehensive risk falls below the preset threshold set, the control level is automatically lifted, and a complete closed-loop record is formed.

[0030] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A property operation and resource intelligent allocation management method based on AI knowledge graph, characterized in that, include: Step 1: Acquire property operation data and form a unified dataset; Step 2: Construct and update the property semantic cognitive graph based on the unified dataset. The property semantic cognitive graph includes an entity set, a relation set, and an attribute set. The relation set includes at least coverage relationships, dependency relationships, shutdown impact relationships, reachability relationships, and mapping relationships. The attribute set includes at least equipment availability status attributes, shutdown and degradation period attributes, personnel and material capacity attributes, and zoning population load attributes. The property semantic cognitive graph is then updated based on the unified dataset. Step 3: Within a preset time window, calculate the functional coverage missing index, the cumulative index of disuse exposure, and the minimum cut index of the functional graph based on the property semantic cognitive graph, and then calculate the comprehensive risk of the fire compartment. Compare the comprehensive risk of the fire compartment with a preset threshold set to trigger dynamic control. Step 4: When dynamic regulation is triggered, multi-objective constraint solving is performed based on the property semantic cognitive graph reasoning. Under the constraints of personnel resource availability data, material resource availability data, operation window data, and reachability relationships, a dynamic control strategy aimed at reducing the overall risk of fire compartments is generated and resource allocation instructions are output. The property management platform transforms dynamic control strategies into work orders and dispatches them to the execution end. It receives execution receipts and writes back to the unified dataset. Based on the write-back results, it updates the property semantic cognitive graph and triggers the next round of comprehensive risk calculation for fire compartments, forming a closed-loop dynamic control.

2. The method for intelligent allocation and management of property operation and resources based on AI knowledge graphs according to claim 1, characterized in that, Property operation data should include at least the following: property space topology data, fire prevention and control zone data, life safety system equipment ledger data, real-time status data of life safety system equipment, planned shutdown or downgrade event data, personnel resource availability data, material resource availability data, work window data, and zoning population load data.

3. The property operation and resource intelligent allocation management method based on AI knowledge graph as described in claim 2, characterized in that, The entity set of the property semantic cognitive graph includes at least property space units, fire compartments, life safety system equipment, functional nodes, planned shutdown or downgrade events, personnel resources, and material resources.

4. The property operation and resource intelligent allocation management method based on AI knowledge graph as described in claim 1, characterized in that, The functional coverage gap index is obtained as follows: within a preset time window, planned shutdown or downgrade events are associated with fire compartments based on mapping relationships and shutdown impact relationships; the functional node coverage corresponding to the fire compartment is obtained based on the property semantic cognitive graph; the functional node coverage is compared with a preset coverage threshold to obtain the node gap index, and the node gap index within the fire compartment is weighted and aggregated to obtain the functional coverage gap index.

5. The property operation and resource intelligent allocation management method based on AI knowledge graph according to claim 4, characterized in that, The functional node coverage is obtained as follows: Based on coverage and dependency relationships, the functional nodes corresponding to the fire compartments and the life safety system equipment corresponding to each functional node are determined; based on the equipment availability status attributes and combined with the decommissioning and degradation period attributes, the equipment availability of the life safety system equipment is obtained; based on coverage and dependency relationships, the equipment availability is synthesized into functional node coverage.

6. The method for intelligent allocation and management of property operation and resources based on AI knowledge graph as described in claim 1, characterized in that, The cumulative exposure index for discontinuation is obtained as follows: the exposure factor is determined, and a time period weight is applied to the exposure factor according to the attributes of the discontinuation and downgrade period; the exposure factor is weighted and accumulated within a preset time window according to a preset sampling interval to obtain the cumulative exposure index for discontinuation.

7. The method for intelligent allocation and management of property operation and resources based on AI knowledge graphs according to claim 6, characterized in that, The determination of the exposure factor includes: obtaining the population load weight within a preset time window based on the population load attributes of the partition; calculating the criticality weight of the life safety system based on the equipment type and dependency relationship of the life safety system; and obtaining the exposure factor based on the population load weight, the criticality weight of the life safety system, and the node gap index.

8. The method for intelligent allocation and management of property operation and resources based on AI knowledge graphs according to claim 1, characterized in that, The minimum cut index of the functional graph is obtained as follows: within a preset time window, a life safety functional graph oriented towards fire compartments is constructed based on functional nodes, mapping relationships and dependencies, and edge weights are set for the edges of the life safety functional graph. Select at least one source node and at least one sink node from the life safety function graph; perform minimum cut calculation with edge weights as input to obtain the minimum cut index of the function graph, and output the minimum cut edge set obtained by the minimum cut calculation as the critical function transmission edge.

9. The method for intelligent allocation and management of property operation and resources based on AI knowledge graphs according to claim 8, characterized in that, Setting edge weights for the life safety function graph includes: life safety function graph nodes are function nodes, and life safety function graph edges are function transmission edges with dependencies at the function node level; based on the mapping relationship, determine the life safety system equipment supporting each function transmission edge; based on the equipment availability status attribute and combined with the deactivation / degradation period attribute, obtain the equipment availability of the supporting equipment; based on the dependencies, synthesize the equipment availability of the supporting equipment into the availability of the function transmission edge, and use the availability of the function transmission edge as the edge weight.

10. The property operation and resource intelligent allocation management method based on AI knowledge graph according to claim 8, characterized in that, The source node represents the functional node that triggers a risk event, and the sink node represents the functional node that indicates that the fire compartment has reached a safe state.

Images