Intelligent property management platform and management method combined with AI scene simulation

Abstract

The invention relates to the technical field of property management, and discloses an intelligent property management platform and management method combined with AI scene simulation, and the platform comprises a dynamic scene twinning engine which is used for constructing a community total-factor digital twinning body fusing the state of Internet of Things equipment, user behaviors and environment data in real time, and carrying out the dynamic simulation through a physical engine, the model comprises a scene slice and a self-feedback twinborn model; the multi-modal data fusion layer adopts an event-driven data pipeline and is used for correlation analysis of multi-source heterogeneous data; and the decision tree is used for carrying out self-evolution decision making based on a dynamic twin simulation result, and a decision confidence attenuation model is introduced into the decision tree. According to the invention, the bottleneck of passive response in traditional property management is changed, a dynamic scene twin engine is constructed in a foreseeing manner, the property management is converted from a passive operation and maintenance mode depending on manpower and regular inspection into a mode capable of predicting equipment faults in advance and actively providing services, and the intelligence and foreseeing level of the property management is improved.

Description

TECHNICAL FIELD

[0001] The application relates to an intelligent property management platform combined with AI scene simulation and a management method, and belongs to the technical field of smart communities and property management. BACKGROUND

[0002] As an important part of smart cities, the core of intelligent property management systems lies in realizing digital monitoring and management of physical world property facilities and environment through cyber-physical convergence systems (CPPS). However, the existing technology has inherent limitations when dealing with increasingly complex property management needs. For example:

[0003] 1. Lack of accurate modeling and prediction ability for dynamic evolution of the physical world: Existing systems mainly focus on static monitoring and post-response of physical state, and fail to effectively capture and predict the complexity of dynamic evolution of property facilities and environment over time. For example, for the aging process of equipment, existing systems can only assess through regular detection, and cannot make more refined degradation cycle prediction based on real-time operation data and environmental factors, resulting in the lagging nature and high cost of maintenance strategies. 2. Challenges of multi-source heterogeneous data fusion and knowledge extraction: Property management involves a wide variety of data types, including structured equipment operation data, semi-structured user behavior data, and unstructured environmental perception data. Existing technology has significant shortcomings in effectively fusing these heterogeneous data and extracting domain knowledge to support intelligent decision-making, resulting in the potential value of data not being fully realized. 3. Robustness and adaptability of intelligent decision-making: The decision logic in existing property management systems is often pre-set or based on simple rules, lacking the ability to adapt to complex dynamic scenarios and self-optimization. For example, in response to emergencies or resource optimization, the system is difficult to adjust flexibly according to real-time changes in the environment and user needs, resulting in limited efficiency and effectiveness of decision-making.

[0004] In view of the above problems, some explorations have been made in the field, such as using machine learning algorithms for fault prediction and energy optimization. However, these methods often rely on a large amount of labeled data, and the generalization ability of the model may not be sufficient when faced with new, unseen situations. In addition, existing research has paid little attention to building a unified platform that can dynamically model and deduce the entire community ecology, lacking a deep understanding and application of the "human-property-environment" interaction. SUMMARY

[0005] The application provides an intelligent property management platform combined with AI scene simulation and a management method, which mainly aims to solve the problems of passive response, data silos and decision-making experience dependence in existing property management.

[0006] To achieve the above object, the application provides an intelligent property management platform combined with AI scene simulation, which comprises:

[0007] a dynamic scene twin engine for constructing a full-factor digital twin of a community, which can fuse device state data from Internet of Things devices, user behavior data from users and environment data in real time, and dynamically simulate community operation scenes through a physics engine, wherein the dynamic scene twin engine comprises:

[0008] a scene slicing module for decomposing community operation states into multiple time-space slices that can be simulated independently;

[0009] a self-feedback twin model for automatically correcting simulation parameters of the digital twin based on actual operation data, wherein for any simulation parameter P t , a correction value P t+Δt of the simulation parameter at a next time t+Δt is calculated based on actual operation data D t at the current time according to the following formula:

[0010] P t+Δt =P t +α·(D t -S t ),

[0011] wherein S t is a simulation value of the digital twin corresponding to the actual operation data at the current time, and alpha is a preset self-feedback coefficient;

[0012] a multi-modal data fusion layer for receiving real-time data from the dynamic scene twin engine and automatically associating device state data, user behavior data and environment data for joint analysis when a specific event is triggered through an event-driven data pipeline;

[0013] a self-evolving decision tree for generating multiple decision options based on dynamic simulation results of the dynamic scene twin engine, and automatically optimizing weights of the decision options through feedback of actual execution effects to form a closed loop of simulation, execution and feedback, wherein the self-evolving decision tree introduces a decision confidence decay model to adjust strategies according to scene changes;

[0014] wherein the intelligent property management platform uses dynamic simulation capabilities of the dynamic scene twin engine to predict device failure cycles and generate maintenance work orders in advance, dynamically adjusts public area lighting and elevator scheduling strategies by fusing energy consumption data, people flow data and device load data, and quantitatively evaluates strategy risks based on simulation deduction to assist property management decisions.

[0015] In a preferred embodiment, the decision confidence decay model calculates the decay value C of the confidence C(d) of the decision option d after time t based on the following formula t (d):

[0016] C t (d)=C(d)·e -β·t ,

[0017] wherein C(d) is the initial confidence of the decision option d, β is a preset decay coefficient, and t is a time interval.

[0018] In a preferred embodiment, the user behavior data includes repair frequency and activity trajectory.

[0019] In a preferred embodiment, the environmental data includes temperature and humidity and energy consumption data.

[0020] In a preferred embodiment, the event-driven data pipeline automatically associates the device state data, user behavior data and environmental data of a specific area for joint analysis when the area experiences a sudden surge in people flow.

[0021] In a preferred embodiment, the decision options generated by the self-evolving decision tree include early activation of standby elevators.

[0022] In a preferred embodiment, the dynamic scenario twin engine generates maintenance work orders in advance by predicting the trend of air conditioner filter clogging.

[0023] In a preferred embodiment, the intelligent property management platform dynamically adjusts public area lighting and elevator scheduling strategies by integrating energy consumption data, people flow data and device load data to achieve a balance between energy efficiency and service.

[0024] In a preferred embodiment, the intelligent property management platform quantitatively evaluates the impact of maintenance time period selection on property owners based on simulation and deduction, assisting property management decisions; the intelligent property management platform is compatible with mainstream Internet of Things protocols, and the dynamic scenario twin engine supports lightweight simulation on the property local server and complex deduction in the cloud.

[0025] In a preferred embodiment, the intelligent property management platform further comprises an intelligent property management method based on the intelligent property management platform, comprising the following steps:

[0026] S1, using the dynamic scenario twin engine to construct a digital twin of the community full factor, and real-time integrating device state data from Internet of Things devices, user behavior data from users and environmental data, and dynamically simulating the community operation scenario through a physics engine;

[0027] S2, automatically correcting simulation parameters of the digital twin based on actual operation data by a self-feedback twin model in the dynamic scene twin engine, to realize continuous synchronization of the digital twin with the real community state;

[0028] S3, using an event-driven data pipeline, automatically associating device state data, user behavior data and environment data for joint analysis when a specific event is triggered by the multi-modal data fusion layer;

[0029] S4, generating multiple decision options based on the dynamic simulation results of the dynamic scene twin engine by the self-evolving decision tree, and automatically optimizing the weights of the decision options through feedback of actual execution effect, forming a closed loop of simulation, execution and feedback;

[0030] S5, using the intelligent property management platform to predict equipment failure period and generate maintenance work order in advance to realize preventive service; integrating energy consumption data, passenger flow data and equipment load data to dynamically adjust public area lighting and elevator scheduling strategy to realize global resource optimization;

[0031] S6, based on simulation and deduction, quantitatively evaluate the risk of strategy to assist property management decision-making and improve the objectivity of decision-making.

[0032] Compared with the problems in the background art, the beneficial effects of the present application are:

[0033] 1. The physical community is converted into a computable digital twin, which can realize real-time sensing, simulation and prediction of the running state of the community, thereby completely changing the passive operation and maintenance mode of traditional property management which relies on manual repair and regular inspection, realizing the preposition of service supply, and fundamentally improving the intelligent level of property management.

[0034] 2. The event-driven data pipeline is adopted to realize real-time association and fusion analysis of multi-source heterogeneous data such as equipment operation data, user behavior data and environment monitoring data, overcoming the problem of data dispersion in traditional property management system which is difficult to support global decision-making, providing comprehensive and three-dimensional data support for property management decision-making, making the decision more scientific and accurate.

[0035] 3. By constructing a self-evolving decision tree, the platform can generate decision options based on the simulation results of the dynamic scene, and optimize feedback according to the actual execution effect, which eliminates the dependence on subjective experience of property personnel, realizes the quantitative evaluation of decision-making and dynamic adjustment of strategy, and especially through the introduction of decision confidence decay model, ensures that the strategy can adapt to the changing community scene, thereby improving the intelligence and objectivity of decision-making.

[0036] 4. The application of digital twinning is extended from the traditional device level to the community ecological level, realizing the collaborative simulation and management of man, object and environment, providing a broader perspective and more refined management means for property management, and helping to improve the overall operation efficiency and service quality of the community.

[0037] 5. The physical community is transformed into a scene that can be calculated and intelligently analyzed, thereby changing the minimum unit of property management from the fragmented "device-work order" mode to a dynamic scene unit, realizing a fundamental change in management concepts and providing a new technical path for building a smart community. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 The system architecture diagram of the intelligent property management platform proposed by the application.

[0039] Figure 2 The interaction flow chart between the decision tree and the decision confidence model is shown.

[0040] Figure 3 The data processing flow chart of the multi-modal data fusion layer.

[0041] Figure 4 The workflow diagram of the self-evolving decision tree.

[0042] Figure 5 The internal structure diagram of the dynamic scene twinning engine.

[0043] The purpose of the application, functional features and advantages will be further described with reference to the accompanying drawings. DETAILED DESCRIPTION

[0044] It should be understood that the specific embodiments described herein are only used to explain the application and do not limit the application.

[0045] Reference Figure 1 — Figure 5 , Figure 1 The system architecture of the intelligent property management platform proposed by the application is shown, which is characterized by connecting the environment data from the environment, the device state data from the Internet of Things device and the user behavior data generated by the user in the physical world to the intelligent property management platform. In the platform, the received data is first processed by the multi-modal data fusion layer, which can provide real-time state data and output fusion analysis results. Further, the dynamic scene twinning engine receives data from the fusion layer and provides twinning model data to the self-evolving decision tree. At the same time, the intelligent property management platform can send control instructions and provide service notifications to the physical world based on the analysis results, forming a closed-loop management system. Figure 2The interaction process between the decision tree and the decision confidence model is illustrated in detail. First, the decision tree sends a request for the current confidence to the decision confidence model, and the model returns the current confidence to the decision tree after receiving the request. Then, the decision tree reports the decision execution result to the decision confidence model after executing the decision, and the model updates the confidence based on the result and the time, and returns the updated confidence to the decision tree, thereby forming a closed-loop feedback mechanism for dynamically updating the decision confidence based on the decision execution result and the time. Figure 3 The data processing flow of the multi-modal data fusion layer is shown, which starts with receiving multi-source heterogeneous data: Internet of Things devices, user behavior, and environmental data. The received data first undergoes event listening and triggering, and then data preprocessing and cleaning. For different types of data, the system performs data format conversion and standardization, followed by data correlation and analysis, and finally outputs the fused data, providing a data basis for the subsequent dynamic scene twin engine and self-evolving decision tree. Figure 4 The workflow of the self-evolving decision tree is shown, which takes the dynamic scene simulation result as input and evolves based on the simulation result and feedback. Specifically, it includes generating decision options, optimizing decision weights, and introducing a decision confidence decay model. The decision options output by the decision tree are executed, producing actual execution effect feedback, which is used for further evolution of the decision tree, forming a closed loop. At the same time, the decision tree also outputs the optimized decision weights. Figure 5 The internal structure and data flow of the dynamic scene twin engine are shown, which receives real-time data from the physical community: Internet of Things devices, user behavior, and environmental data. These data first enter the scene slicing module for processing, outputting input slicing data to the self-feedback twin model. The model interacts with the physical engine by sending simulation parameters and receiving simulation result feedback, and at the same time, the self-feedback twin model sends the generated twin model state data to the multi-modal data fusion layer.

[0046] The embodiment of the present application provides an intelligent property management platform combined with AI scene simulation, which includes a dynamic scene twin engine for building a digital twin of all factors in the community. The digital twin can real-time fuse device state data from Internet of Things devices, user behavior data from users, and environmental data, and dynamically simulate the community running scene through a physical engine. The dynamic scene twin engine includes:

[0047] A scene slicing module for decomposing the community running state into multiple time-space slices that can be independently simulated;

[0048] A self-feedback twin model for automatically correcting the simulation parameters of the digital twin based on actual running data, wherein for any simulation parameter P t , its correction value Pt+Δt actual running data D at the current time t According to the following formula:

[0049] P t+Δt = P t + a · (D t - S t ),

[0050] wherein S t is a simulation value of the digital twin corresponding to the actual running data at the current time, and a is a preset self-feedback coefficient;

[0051] a multi-modal data fusion layer for receiving real-time data from the dynamic scene twin engine and employing an event-driven data pipeline to automatically associate device state data, user behavior data and environmental data for joint analysis upon triggering of a specific event;

[0052] a self-evolving decision tree for generating multiple decision options based on dynamic simulation results of the dynamic scene twin engine and automatically optimizing weights of the decision options through feedback of actual execution effects to form a closed loop of simulation, execution and feedback, wherein the self-evolving decision tree introduces a decision confidence decay model to adjust strategies according to scene changes;

[0053] wherein the intelligent property management platform utilizes dynamic simulation capabilities of the dynamic scene twin engine to predict device failure cycles and generate maintenance work orders in advance, dynamically adjusts public area lighting and elevator scheduling strategies by fusing energy consumption data, people flow data and device load data, and quantitatively evaluates strategy risks based on simulation deduction to assist property management decisions.

[0054] In a preferred embodiment, the decision confidence decay model calculates a decay value C t (d) of the confidence C(d) of a decision option d after a time t based on the following formula:

[0055] C t (d) = C(d) · e -β·t ,

[0056] wherein C(d) is the initial confidence of the decision option d, β is a preset decay coefficient, and t is a time interval.

[0057] In a preferred embodiment, the user behavior data includes repair frequency and activity trajectory.

[0058] In a preferred embodiment, the environmental data includes temperature and humidity and energy consumption data.

[0059] In a preferred embodiment, the event-driven data pipeline automatically associates device status data, user behavior data, and environmental data in a specific area for joint analysis when there is a sudden surge in human flow in the area.

[0060] In a preferred embodiment, the decision options generated by the self-evolving decision tree include early activation of standby elevators.

[0061] In a preferred embodiment, the dynamic scenario twin engine generates maintenance work orders in advance by predicting the trend of air conditioner filter clogging.

[0062] In a preferred embodiment, the intelligent property management platform dynamically adjusts public area lighting and elevator scheduling strategies by integrating energy consumption data, human flow data, and device load data to achieve a balance between energy efficiency and service.

[0063] In a preferred embodiment, the intelligent property management platform quantitatively evaluates the impact of maintenance time selection on tenants based on simulation and deduction, assisting property management decisions; the intelligent property management platform is compatible with mainstream Internet of Things protocols, and the dynamic scenario twin engine supports lightweight simulation on the property local server and complex deduction in the cloud.

[0064] In a preferred embodiment, the intelligent property management platform further includes an intelligent property management method based on the intelligent property management platform, comprising the following steps:

[0065] S1, using the dynamic scenario twin engine to construct a digital twin of the community with all factors, and real-time integrating device status data from Internet of Things devices, user behavior data from users, and environmental data, and dynamically simulating the community operating scenario through a physics engine;

[0066] S2, automatically correcting the simulation parameters of the digital twin based on actual operating data through the self-feedback twin model in the dynamic scenario twin engine, to realize continuous synchronization between the digital twin and the real community state;

[0067] S3, through the multi-modal data fusion layer, using an event-driven data pipeline to automatically associate device status data, user behavior data, and environmental data for joint analysis when a specific event is triggered;

[0068] S4, through the self-evolving decision tree, generating multiple decision options based on the dynamic simulation results of the dynamic scenario twin engine, and automatically optimizing the weights of the decision options through feedback of actual execution results to form a closed loop of simulation, execution, and feedback;

[0069] S5, using the intelligent property management platform to predict equipment failure period, generating maintenance work order in advance, realizing preventive service; integrating energy consumption data, people flow data and equipment load data, dynamically adjusting public area lighting and elevator scheduling strategy, realizing global resource optimization;

[0070] S6, based on analog deduction quantitative evaluation strategy risk, auxiliary property management decision, improve the objectivity of decision.

[0071] Embodiment 1: This embodiment aims to illustrate how the intelligent property management platform of the present application applies the dynamic scene twin engine to predict the trend of air conditioner filter clogging in actual property management, and generates maintenance work order in advance.

[0072] In a typical residential community, the intelligent property management platform collects data in real time through the Internet of Things devices deployed in each household and public areas. These data include the running time of each air conditioning device, energy consumption data, particulate matter concentration feedback from air quality sensors, and other equipment status data, as well as user behavior data such as average clogging period of air conditioner filter analyzed according to user repair records and historical maintenance records. At the same time, the platform also receives environmental data such as temperature and humidity, outdoor air quality from environmental monitoring devices.

[0073] After receiving the above real-time data, the dynamic scene twin engine first divides the running state of the entire community into independent space-time slices with each air conditioning device as a unit through the scene slicing module. For each slice, the self-feedback twin model compares the actual running data such as air conditioning running time, energy consumption data and particulate matter concentration collected at the current time with the simulation running data of the air conditioner in the digital twin. For example, if the energy consumption of a certain air conditioner is significantly higher than the historical average under the same running time, and the particulate matter concentration feedback from the air quality sensor is accelerating, it indicates that the filter of the air conditioner may be clogging rapidly.

[0074] The self-feedback twin model corrects the clogging parameter of the air conditioner filter according to the formula P t+Δt =P t +α·(D t -S t ). In this embodiment, the simulation parameter P can be the air permeability of the filter, the actual running data D can be the particulate matter concentration rising rate, the simulation value S is the particulate matter concentration rising rate predicted by the digital twin based on the current filter air permeability, and α is a preset self-feedback coefficient for adjusting the influence of actual data on model parameters. By continuously correcting the filter clogging parameter based on actual running data, the digital twin can more accurately reflect the actual state of the air conditioner filter in the real world.

[0075] The multi-modal data fusion layer continuously monitors the simulation results from the dynamic scenario twin engine. When it detects that the clogging parameter of one or more air conditioning units reaches a pre-set threshold (which can be set based on empirical data or industry standards, such as air permeability falling below a certain value or predicted particulate matter concentration exceeding health standards within a certain time period), an event-driven data pipeline is triggered, automatically correlating the operating data of the air conditioning unit, the user's historical repair records, and the current environmental data (such as whether there is an external pollution intensification such as smog).

[0076] The self-evolving decision tree generates decision options based on the results of the fusion analysis, such as generating a maintenance work order for replacing the filter of the air conditioner in advance. The decision tree takes into account the current time, the scheduling of property personnel, and other factors, and combines a decision confidence decay model to evaluate the urgency and priority of the decision. For example, if the air quality is predicted to decrease significantly within the next 24 hours, the confidence of the maintenance work order will be higher, and the priority will be correspondingly improved.

[0077] Finally, the intelligent property management platform sends the generated maintenance work order to the property management personnel, prompting them to arrange for the replacement of the filter of the corresponding air conditioner in a timely manner, thereby realizing the early prediction and preventive maintenance of the clogging trend of the air conditioner filter, avoiding user complaints and energy waste caused by filter clogging, and improving the intelligent and proactive service level of property management.

[0078] In the process of constructing the community full-factor digital twin, in order to ensure that the digital twin can accurately reflect the running state of the real world, the self-feedback twin model is crucial. This model compares the simulation results of the digital twin with the actual operating data continuously, and uses the comparison results to adjust the parameters inside the digital twin, thereby improving the accuracy of the simulation. The formula P t+Δt = P t + α · (D t - S t ) is the core of the self-feedback twin model, P represents one or more key parameters in the digital twin, such as the air permeability of the air conditioner filter mentioned in embodiment 1. The subscript t represents the current time, and t+Δt represents the next time after a time interval. D t represents the real operating data collected from the actual physical world at the current time t, such as the particulate matter concentration rise rate of the air conditioner filter obtained through Internet of Things sensors. S t represents the value simulated by the digital twin for this parameter at the current time t, such as the predicted particulate matter concentration rise rate of the air conditioner filter.

[0079] The self-feedback coefficient α is a value between 0 and 1. It determines the degree to which the difference between actual operating data and simulated data affects the adjustment of model parameters. For example, the self-feedback coefficient α can be understood as the model's trust in actual data or its response speed. When the value of α is large, the model will adjust its parameters more quickly according to the actual data, making the simulation results approach the true value more quickly. When the value of α is small, the model's response to actual data will be slower, and the parameter adjustment will be more conservative. The specific value of the self-feedback coefficient α can be preset or dynamically adjusted according to different types of parameters and application scenarios. Presetting methods include: determining a suitable fixed value based on historical experience data, industry standards, or through offline training and optimization of the model. For example, for parameters that change relatively slowly, α can be set smaller; for parameters that change more frequently, it can be set larger. In Example 1, for the air permeability parameter of the air conditioning filter, a suitable α value can be preset based on historical data analysis of the average rate and fluctuation range of filter clogging. Furthermore, the value of α can be dynamically adjusted according to actual operating conditions. For example, when the difference between actual and simulated data remains large, α can be appropriately increased to accelerate the model's convergence. Conversely, when the difference is small, α can be decreased to prevent the model from fluctuating drastically due to isolated outliers. Dynamic adjustments can be achieved based on predefined rules or using more complex adaptive control algorithms.

[0080] Let's take the prediction of air conditioner filter clogging in Example 1 as an example. Assume that at a certain time t, the actual particulate matter concentration increase rate D collected by the sensor... t It is 0.05 mg / m³ 3 / h, while the digital twin predicts the rate of increase S based on the current filter permeability. t It is 0.03 mg / m³ 3 / h. If the preset self-feedback coefficient α is 0.2, then after the next time interval Δt, the parameter P representing the filter permeability in the digital twin will be adjusted as follows: P t+Δt =P t +0.2·(0.05-0.03)=P t +0.004 means that the digital twin will adjust the filter permeability based on the difference between the actual data and the simulated data, making its predictions closer to reality in the future. By continuously repeating this self-feedback adjustment process, the digital twin can more accurately reflect the actual state of the air conditioning filter in the real world, thus providing a more reliable basis for subsequent fault prediction and maintenance decisions.

[0081] And in this application, the dynamic simulation of the running scene of the community by the physical engine refers to the simulation of the behavior of physical entities: for example, the simulation of the flow of air in the building, the transfer of heat, the running track and speed change of the elevator, etc., to help understand the mutual influence between various elements in the community; predict the occurrence of physical events: based on the current physical state and environmental conditions, predict the physical events that may occur in the future, such as wear and tear of equipment components, trends in energy consumption changes, etc.; provide a visual simulation environment: the physical engine can usually present the simulation results in a visual manner to help property managers more intuitively understand the running state and potential risks of the community. Specifically, for the application scene of intelligent property management, the physical engine can include but not limited to the following models, such as the fluid dynamics model: used to simulate the flow of fluids such as air and water, for example, to analyze the indoor ventilation effect, predict pipe blockage, etc.; thermodynamic model: used to simulate heat transfer and temperature change, for example, to optimize building energy consumption, predict equipment overheating, etc.; mechanical dynamics model: used to simulate the movement and stress of mechanical components, for example, to analyze the running efficiency of the elevator and predict its failure; electromagnetic model: used to simulate the running state of the power system, for example, to optimize power distribution and predict power failure, on this basis, by selecting appropriate physical models and combining real-time data collected by Internet of Things devices, the dynamic scene twin engine can dynamically simulate the running scene of the community with high precision, providing strong data support for intelligent property management, all of which belong to the extended implementation mode known to those skilled in the art.

[0082] And predict equipment failure cycle, for example, can be achieved by the following mechanism, to establish a device health model: for different types of equipment (such as air conditioning, elevator, water pump, etc.), the corresponding health model is established. These models can be based on the operating parameters of the equipment (such as temperature, current, vibration frequency, etc.), historical failure data and industry standards; Real-time monitoring of equipment operation data: through the Internet of Things sensors to collect real-time operation data of the equipment, and synchronize these data to the dynamic scene twin; Comparison of actual data and model prediction: compare the actual collected operation data with the normal operation range predicted by the equipment health model. When the actual data deviates from the normal range and shows a deterioration trend, for example, the current of the air conditioner compressor continues to rise beyond the normal threshold, or the vibration frequency of the elevator gradually increases when it is running, which may indicate that the equipment is about to fail; Use AI algorithms for failure prediction: time series analysis, regression analysis, machine learning and other AI algorithms can be used to analyze the trend of the equipment operation data and predict failures. For example, by analyzing the change rule of the operating parameters before the equipment failure in the historical data, a failure prediction model is trained. When new real-time data is input into the model, the model can predict the probability of equipment failure in the future. When the probability of equipment failure is predicted to reach a certain threshold, and combined with the decision confidence decay model evaluation, it is considered necessary to maintain in advance, the platform will automatically generate a maintenance work order, and send the work order information to the property management personnel, reminding them to arrange repair or replace equipment in time, so as to realize preventive maintenance, which is an extension of the implementation mode known to those skilled in the art.

[0083] Example 3: The intelligent property management platform collects real-time operation data through Internet of Things devices deployed in the community, such as sensors of air conditioning systems. These data include but are not limited to: air conditioning operating current, fan speed, outlet temperature, indoor and outdoor temperature and humidity, etc. At the same time, the platform also receives feedback information about air conditioning operation abnormalities reported by users through the mobile application, such as "outlet volume becomes smaller", "poor refrigeration effect", etc. After receiving these heterogeneous data from different sources, the multi-modal data fusion layer will first perform data cleaning and standardization processing to remove noise and outliers, and convert different formats of data into a format that the platform can process. For text information of user feedback, semantic analysis will be performed through natural language processing (NLP) technology to extract key information, such as the specific problem type and occurrence time of user feedback.

[0084] The dynamic scenario twin engine creates one or more spatio-temporal slices for the air conditioning system based on the pre-built community full-factorial digital twin. These slices contain the physical model of the air conditioning equipment, the operating parameter model, and the environmental impact model; the self-feedback twin model plays a key role in this process. In the initial stage, the twin model is initialized based on the historical operating data of the equipment and the parameters provided by the manufacturer. With the continuous access of actual operating data, the self-feedback mechanism is started. For example, in the scenario of gradual clogging of the air conditioner filter, the actual monitored air conditioner operating current (D t ) may gradually increase, while the normal operating current (S t ) predicted by the twin model based on the initial parameters may remain at a low level.

[0085] At this time, the parameters related to filter clogging in the twin model are corrected using the formula P t+Δt = P t + a · (D t -S t ). For example, P t may represent the clogging degree parameter of the filter at the current time, Δt is the time interval of data update, and a is the self-feedback coefficient. The value of the self-feedback coefficient a can be pre-set according to the characteristics of different types of equipment and the analysis results of historical data. For example, for air conditioning systems, the value of a can be set within an empirical range (e.g. 0.01-0.1), indicating the sensitivity of the actual current and the simulated current deviation to the correction of the filter clogging parameter. If the actual current is consistently higher than the simulated current, the filter clogging parameter P t will gradually increase, so that the twin model can more accurately reflect the operating state of the actual equipment.

[0086] With the update of the filter clogging parameter in the twin model, the physical engine will re-simulate based on the updated model parameters. For example, the physical engine can simulate the flow of air in the filter, and when the filter clogging parameter increases, the simulation result will show that the air flow through the filter decreases and the fan load increases, which is consistent with the actual situation. At the same time, the simulation result can also predict how much the energy consumption of the air conditioner will increase and how much the refrigeration efficiency will decrease under the current clogging degree.

[0087] The self-evolution decision tree generates decision options according to preset rules and weights based on the simulation results of the dynamic scene twin engine and the user feedback information provided by the multi-modal data fusion layer. When the twin model predicts that the air conditioner energy consumption significantly increases or the refrigeration efficiency significantly decreases due to filter clogging, or when the user reports feedback such as "air volume becomes smaller" and "poor refrigeration effect" related to filter clogging, the decision tree will generate corresponding decision options according to preset rules and weights, such as "suggesting to arrange air conditioner filter inspection", "suggesting to arrange air conditioner filter cleaning or replacement", etc. At the same time, in order to avoid false positives due to short-term data fluctuations or occasional user feedback, the decision tree introduces a decision confidence decay model. For each decision option d generated, its confidence C(d) will decay over time, as shown in formula C t (d) = C(d) e -βt For filter clogging warnings that affect user comfort, the value of β can be relatively small, indicating that the confidence decay is slow. For warnings involving safety hazards, the value of β can be relatively large, indicating that the confidence decay is fast, so that timely measures can be taken. Only when the confidence C t (d) of a certain decision option exceeds a preset threshold, the platform considers that the decision is reliable, and triggers the corresponding maintenance work order. For example, when the confidence of the decision option "suggesting to arrange air conditioner filter cleaning or replacement" exceeds the set threshold, the platform will automatically generate a maintenance work order and send the work order information to the property management personnel, reminding them to arrange for processing in a timely manner.

[0088] After the property management personnel check, clean or replace the air conditioner filter according to the maintenance work order, they will feed back the processing results to the platform. After receiving the feedback information, the platform will update the relevant parameters in the twin model, such as restoring the filter clogging parameter to normal level. At the same time, the platform will also use the results of this warning and maintenance as new data samples to optimize the self-feedback coefficient α of the self-feedback twin model and the decision rules and weights of the self-evolution decision tree, as well as the decay coefficient β of the decision confidence decay model, so as to continuously improve the warning accuracy and decision intelligence level of the platform.

[0089] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application.

Claims

1. An intelligent property management platform combined with AI scene simulation, characterized in that: include: A dynamic scene twin engine is used to construct a digital twin of all elements of a community. This digital twin can integrate device status data from IoT devices, user behavior data from users, and environmental data in real time, and dynamically simulate the community's operating scenarios through a physical engine. The dynamic scene twin engine includes: A scene slicing module is used to decompose the community's operating state into multiple spatiotemporal slices that can be simulated independently; a self-feedback twin model for automatically correcting simulation parameters of the digital twin based on actual operational data, wherein for any simulation parameter p t its corrected value P t+Δt based on actual operational data D t is calculated according to the following formula: P t+Δt = P t + a · (D t - S t ), wherein S t is a simulation value of the digital twin corresponding to the actual operation data at the current moment, and a is a preset self-feedback coefficient. A multimodal data fusion layer is used to receive real-time data from the dynamic scene twin engine and adopt an event-driven data pipeline to automatically associate device status data, user behavior data and environmental data for joint analysis when a specific event is triggered. A self-evolving decision tree is used to generate multiple decision options based on the dynamic simulation results of the dynamic scene twin engine, and automatically optimize the weight of the decision options through feedback of actual execution effects, forming a closed loop of simulation, execution and feedback. The self-evolving decision tree introduces a decision confidence decay model to adjust the strategy according to scene changes. The intelligent property management platform utilizes the dynamic simulation capability of the dynamic scene twin engine to predict equipment failure cycles and generate maintenance work orders in advance. 2.The intelligent property management platform integrated with AI scene simulation of claim 1, wherein, The decision confidence decay model calculates a decay value C of the confidence C(d) of a decision option at time t after the following formula t (d): C t (d) = C(d) · e -β·t , Where C(d) is the initial confidence level of decision option d, β is the preset decay coefficient, and t is the time interval. 3.The intelligent property management platform integrated with AI scene simulation of claim 2, wherein, The user behavior data includes the frequency of repair requests and activity patterns. 4.The intelligent property management platform integrated with AI scene simulation of claim 3, wherein, The environmental data includes temperature, humidity, and energy consumption data. 5.The intelligent property management platform integrated with AI scene simulation of claim 1, wherein, When there is a sudden surge in pedestrian traffic in a specific area, the event-driven data pipeline automatically correlates the device status data, user behavior data, and environmental data of that area for joint analysis.

6. The intelligent property management platform integrated with AI scenario simulation of claim 5, wherein, The decision options generated by the self-evolving decision tree include activating the backup elevator in advance.

7. The intelligent property management platform integrated with AI scenario simulation of claim 6, wherein, The dynamic scene twin engine generates maintenance work orders in advance by predicting the trend of air conditioning filter blockage. 8.The intelligent property management platform integrated with AI scenario simulation of claim 1, wherein, The intelligent property management platform uses simulation to quantitatively assess the impact of maintenance time selection on homeowners, thus assisting in property management decisions. The platform is compatible with mainstream IoT protocols, and the dynamic scenario twin engine supports lightweight simulations on the property's local server, while complex simulations are conducted in the cloud. 9.The intelligent property management platform integrated with AI scenario simulation of claim 1, wherein, The intelligent property management platform also includes an intelligent property management method based on the intelligent property management platform, comprising the following steps: S1. Utilize the dynamic scene twin engine to construct a digital twin of all elements of the community, and integrate device status data from IoT devices, user behavior data from users, and environmental data in real time. Use the physical engine to dynamically simulate the community operation scenario. S2, through the self-feedback twin model in the dynamic scene twin engine, the simulation parameters of the digital twin are automatically corrected based on the actual running data, so as to achieve continuous synchronization between the digital twin and the real community status; S3, through the multimodal data fusion layer, an event-driven data pipeline is used to automatically associate device status data, user behavior data, and environmental data for joint analysis when a specific event is triggered; S4, through the self-evolution decision tree, a plurality of decision options are generated based on the dynamic simulation results of the dynamic scene twin-engine, and the weights of the decision options are automatically optimized through the feedback of actual execution effect, forming a closed loop of simulation, execution and feedback; S5, using the intelligent property management platform to predict equipment failure period, generating maintenance work order in advance to realize preventive service; integrating energy consumption data, passenger flow data and equipment load data, dynamically adjusting public area lighting and elevator scheduling strategy to realize global resource optimization; S6, based on simulation and deduction, the risk of the strategy is quantitatively evaluated to assist property management decision-making and improve the objectivity of decision-making.

Images