Long-term house renting monitoring management system and method

Abstract

The invention relates to the technical field of house renting management, and discloses a long-term house renting monitoring management system and method.The method comprises the steps that monitoring units in the same physical space are divided into equipment groups, state jump timestamps of the equipment groups are recorded, behavior logic templates are automatically precipitated based on a historical timestamp sequence, and the equipment groups are stored; through logic self-consistency comparison of a real-time sequence and a template, risk early warning is generated, the system has the capability of understanding equipment collaborative behavior logic, the behavior template is constructed by multiplexing a basic switch signal, cognitive transition from physical quantity monitoring to behavior logic verification is realized, and the risk early warning effect is improved. And meanwhile, an anti-interference data quality firewall and system-level resource coordination capability is naturally formed under a centerless architecture by utilizing a state credibility evaluation and distributed priority arbitration mechanism, so that the economy and the reliability of intelligent reconstruction of the old house are remarkably improved.

Description

Technical Field

[0001] The present invention relates to a long-term rental monitoring and management system and method, belonging to the technical field of rental management data processing. Background Art

[0002] In the field of long-term rental housing management, for vulnerable groups such as nursing homes and care for elderly people living alone, existing technologies generally use a single-point sensor threshold alarm method. That is, by deploying independent sensors such as temperature, humidity, smoke, and current in the house, monitoring whether specific physical quantities exceed preset thresholds and triggering alarms. This solution is considered a standard management method. Its core assumption is that the risk is only caused by the exceeding of a single physical quantity. However, when the system faces chronic risks caused by the coupling of cross-device behavioral logic, this method has some fundamental flaws.

[0003] For example, when a tenant turns on a high-power heating device for a long time but keeps the doors and windows closed, the traditional system remains silent because the values of each sensor do not exceed the threshold. It is unable to identify the potential fire hazards caused by the increased carbon dioxide concentration in the closed space due to the combination of continuous heating and zero ventilation, such as hypoxia or the continuous use of heating equipment. Specifically: 1. The system can only perceive the instantaneous state of physical quantities and cannot understand the behavioral coordination relationship between device groups, such as the correlation between turning on the heater and closing doors and windows, resulting in a complete oversight of the risk of behavioral combinations. For example, human breathing in a closed space will continuously consume oxygen and produce carbon dioxide. Long-term accumulation will cause the indoor oxygen concentration to decrease and the carbon dioxide concentration to increase, thereby causing the risk of hypoxia or deterioration of air quality; 2. Relying on the original physical quantity values to judge risks, ignoring the behavioral logic information contained in the equipment state jump sequence, makes the system respond to chronic and cross-domain risks with a lag.

[0004] The industry has tried to alleviate the problem by increasing sensor density or optimizing threshold settings, but such improvements are still limited by the following: the increase in physical quantity increases the deployment cost of sensors, making it difficult to adapt to the renovation of old houses; threshold optimization requires frequent manual calibration and cannot adapt to dynamic tenant behavior; it has failed to break through the shackles of thinking about isolated data point analysis and cannot deconstruct the device collaboration logic from a system level. The above bottlenecks essentially stem from the systematic neglect of the management dimension of the collaborative state of device group behavior. The resulting risk recognition blind spots have become a common problem that restricts the improvement of the efficiency of smart rental management. Therefore, how to build a management system that does not rely on physical quantity thresholds and realizes active perception of hidden risks through self-consistency verification of device behavior logic, and then solves the problem of accurate identification of cross-device collaboration failures, has become the technical problem to be solved by the present invention. Summary of the Invention

[0005] The present invention provides a long-term rental monitoring and management system, the main purpose of which is to solve the problem that the existing technology cannot actively identify the risk of logical rupture of device group behavior.

[0006] To achieve the above objectives, the present invention provides a long-term rental monitoring and management system, comprising:

[0007] a device group definition module configured to logically define at least two monitoring units deployed in the same physical space as a device group;

[0008] a state transition recording module configured to record the unique identification code of the monitoring unit and the precise timestamp of the state transition when the operating state of the monitoring unit transitions from one state to another;

[0009] The behavioral logic template precipitation module is configured to automatically generate one or more high-frequency behavioral logic templates through an unsupervised learning algorithm based on the precise timestamp sequence recorded by the device group during a preset initial learning cycle. The high-frequency behavioral logic templates represent the stable timing associations between the state jumps of each monitoring unit in the device group. Among them, the preset initial learning cycle is a time period that can be pre-configured by the property manager according to the actual application scenario. For example, it can be set to the first two weeks after the tenant moves in, or the first 14 days of system operation to ensure that data sufficient to analyze the user's stable living habits can be collected.

[0010] The logic self-consistency verification module is configured to continuously compare the precise timestamp sequence output in real time by the state jump recording module with the high-frequency behavior logic template, and generate a risk warning signal when there is a behavior logic deviation between the real-time timestamp sequence and the high-frequency behavior logic template; the determination of the behavior logic deviation is based on the following logic: when the absolute time difference between the actual state jump time of a monitoring unit and the expected state jump time indicated by the high-frequency behavior logic template exceeds the logic allowable deviation threshold preset by the template, it is determined that a behavior logic deviation exists.

[0011] Preferably, the operating state recorded by the state jump recording module is two binary states: the on state and the off state; the high-frequency behavior logic template generated by the behavior logic template precipitation module includes the typical timing between the state jumps of different monitoring units in the group and the time interval range corresponding to each typical timing.

[0012] Preferably, the system also includes a state credibility assessment module, which is configured to: dynamically calculate and update the historical average stable operating time of each monitoring unit in the device group based on the precise timestamp recorded by the state jump recording module; and before the logic self-consistency verification module performs comparison, when the state jump recording module records that the state of a certain monitoring unit has jumped, the state credibility assessment module calculates the actual time interval between the state jump and the previous state jump of the monitoring unit; if the actual time interval is less than ten percent of the historical average stable operating time of the monitoring unit, the state credibility assessment module enters a delayed confirmation mode; only when the state of the monitoring unit has not recovered to the state before the jump after the delayed confirmation mode continues for a preset period of time, the state credibility assessment module recognizes the validity of the state jump, and submits the timestamp of the state jump to the behavior logic template precipitation module for template generation within the initial learning cycle; after the initial learning cycle ends, the timestamp of the state jump is submitted to the logic self-consistency verification module for real-time comparison.

[0013] Preferably, the system further comprises a priority label configuration module configured to configure different priority digital labels for at least two monitoring units belonging to at least two different device groups and consuming common resources.

[0014] Preferably, the system also includes a resource competition logic arbitration module, which is configured to: before the state of a monitoring unit configured with a lower priority digital tag is about to switch from a non-operating state to a running state, determine whether there is a monitoring unit configured with a higher priority digital tag that is currently in a running state based on the latest precise timestamp of other monitoring units recorded by the state jump recording module; if so, the resource competition logic arbitration module delays the state switching of the lower priority unit until the higher priority monitoring unit is restored to a non-operating state.

[0015] Preferably, the common resource includes the total electrical load of the apartment or the total water pressure of the apartment.

[0016] Preferably, the system further comprises a warning signal sending module configured to send the risk warning signal generated by the logic self-consistency verification module to a remote management server via a wireless communication network.

[0017] Preferably, the monitoring unit includes a smart electrical socket, a smart door magnetic sensor or a smart water flow switch.

[0018] Preferably, the behavior logic template precipitation module counts the occurrence frequencies of state transition sequences of each monitoring unit in the device group during the initial learning cycle, and identifies sequences with an occurrence frequency higher than 80% as high-frequency behavior logic templates.

[0019] A long-term rental risk warning method is characterized by comprising the following steps: step a, logically defining at least two monitoring units deployed in the same physical space as a device group; step b, recording the unique identification code of the monitoring unit and the precise timestamp when the state jump occurs when the operating state of the monitoring unit jumps from one state to another; step c, automatically generating one or more high-frequency behavior logic templates through an unsupervised learning algorithm based on the precise timestamp sequence recorded by the device group within a preset initial learning cycle, the high-frequency behavior logic template representing the stable temporal association between the state jumps of each monitoring unit in the device group; step d, continuously comparing the precise timestamp sequence output in real time with the high-frequency behavior logic template, and generating a risk warning signal when there is a behavioral logic deviation between the real-time timestamp sequence and the high-frequency behavior logic template; the determination of the behavioral logic deviation is based on the following logic: when the absolute time difference between the actual state jump time of a monitoring unit and the expected state jump time indicated by the high-frequency behavior logic template exceeds the logical allowable deviation threshold preset in the template, it is determined that a behavioral logic deviation exists.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] 1. For vulnerable groups, such as nursing homes and care for elderly people living alone, the system converts the timestamp sequence of device state transitions into a behavioral logic template, enabling it to independently understand the coordination patterns of devices. When the real-time behavior sequence deviates from the template's preset temporal logic, for example, the exhaust fan fails to start within a reasonable time window after the water heater is turned off, the system automatically identifies the logical omission of such a key step. This cognitive upgrade from physical quantity monitoring to behavioral logic verification enables the system to capture cross-device coordination failure risks that are invisible to traditional threshold mechanisms, providing higher-dimensional information support for management decisions. The system also reuses the binary state signals of existing devices, such as the on / off records of smart sockets, to construct behavioral logic templates, eliminating the need for high-precision sensors. When the water heater continues to run but the door and window sensors are unresponsive, the system can determine a ventilation logic break simply by comparing the temporal correlation between the two state transitions. This logical reconstruction of low-dimensional data eliminates the reliance on dedicated hardware for the renovation of old houses, significantly reducing the technical barriers and marginal costs of large-scale implementation.

[0022] 2. The state credibility assessment module analyzes the distribution characteristics of the device's historical stable operating time. For example, it triggers delayed confirmation of abnormal jumps that are shorter than 10% of the historical average, naturally filtering out transient signal interference. The resource contention arbitration module uses preset priority tags and global timestamp visibility to automatically delay the start of low-priority devices (such as bathroom water heaters) when high-priority devices (such as kitchen ovens) are detected to be running. These two mechanisms form a closed loop with the core behavioral logic verification. Credibility verification ensures the quality of input data, priority arbitration expands system-level collaborative capabilities, and behavioral template analysis provides a basis for decision-making. The three are coupled through the timestamp data stream to form a self-consistent and robust system that can resist signal disturbances and resource conflicts without the need for external compensation mechanisms.

[0023] 3. Low-priority devices autonomously avoid resource conflicts by parsing the global timestamp bulletin board. For example, an instant water heater will pause when the total water pressure is insufficient. This process does not rely on a centralized controller or real-time communication. Instead, it relies on the distributed interpretation of public information by the devices and the local execution of static rules. While maintaining the simplicity of the edge computing architecture, it naturally emerges with system-level resource coordination capabilities. The early warning signal sending module maps behavioral logic inconsistent events (such as continuous water use without ventilation) into structured alarms, directly pointing to the collaborative failure scenarios of specific equipment groups. This means that the remote management server's handling instructions are no longer based on vague physical quantity exceeding alarms, but are precisely linked to executable equipment control actions, such as forcibly shutting off the water heater and starting the exhaust fan, forming a closed-loop management path of risk identification-decision output-execution feedback.

[0024] 4. In scenarios such as nursing homes and care for elderly people living alone, potential safety risks and resource usage conflicts are proactively identified by analyzing the sequential logic of device operation rather than relying on isolated physical quantity thresholds. This is because the life patterns of this specific population are generally highly regular, providing an ideal data foundation for the system to automatically generate accurate, stable, and high-frequency behavioral logic templates based on unsupervised learning algorithms. Therefore, whether it is an interruption in the behavioral sequence caused by an acute event such as a fall, or a chronic risk such as forgetting to turn off electrical appliances due to cognitive decline, these can be accurately identified as behavioral logic deviations from existing safety templates through the logical self-consistency verification module, thereby generating a risk warning signal. More importantly, this method only relies on binary state data from monitoring units such as smart electrical sockets and door sensors, achieving non-invasive passive monitoring while protecting user privacy and life dignity. This addresses the limitations of traditional invasive monitoring solutions and provides an important passive safety net for people requiring special care. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a flow chart of state transition records and behavior logic verification in the long-term rental monitoring and management system of the present invention;

[0026] Figure 2 This is a learning curve diagram of the behavioral logic template stability and matching accuracy of the long-term rental monitoring and management system of the present invention;

[0027] Figure 3 This is a schematic diagram of the modular structure of the long-term rental monitoring and management system of the present invention;

[0028] Figure 4 This is a graph showing changes in template indicators and recognition quantity during the learning cycle of the present invention.

[0029] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0030] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0031] An embodiment of the present application provides a long-term rental housing monitoring and management system and method. For vulnerable groups, such as nursing homes and care for elderly people living alone, the purpose is to proactively identify potential security risks and resource usage conflicts by analyzing the timing logic of device operation rather than relying on isolated physical quantity thresholds. The monitoring units deployed in the apartment may include a smart electric socket for powering the bathroom water heater, a smart door magnetic sensor installed on the bathroom door, and another smart electric socket configured for high-power appliances such as electric ovens in the kitchen. These monitoring units together constitute a collection that needs to be collaboratively monitored in terms of physical space and functional logic. The initialization operation of the system begins with the functional implementation of the device group definition module. A specific implementation of this module is a software configuration interface for use by house managers. This interface can be embodied as an application installed on a smartphone or personal computer. Through this interface, managers can configure devices deployed in the same physical space and with behavioral At least two monitoring units that are inherently associated are logically defined as a device group. Taking this apartment as an example, the manager can define the smart electrical socket and smart door magnetic sensor in the bathroom as a bathroom behavior group. The inherent logic is that showering behavior will inevitably cause orderly changes in the status of these two devices. The essence of this definition process is to create an association record in the configuration database of the system background. This record binds a unique group identification code to the unique identification codes of a group of monitoring units. After the group definition is completed, the state jump recording module begins to work continuously on each monitoring unit. This module is usually a program solidified in the hardware firmware of the monitoring unit. When the operating state of the monitoring unit changes, such as when the smart electrical socket jumps from the off state to the on state, the module immediately captures this event and records two core information: the unique identification code of the monitoring unit, which is usually its MAC address or a globally unique factory serial number;And the precise timestamp when the state jump occurs. To ensure the uniformity and comparability of cross-device timestamps, all monitoring units are periodically synchronized with the same time server through the network time protocol, so that the precise timestamp can achieve millisecond-level synchronization accuracy. In this embodiment, the operating state is limited to two basic and reliable binary states: the on state and the off state. This design significantly reduces the requirements for the complexity of the sensor hardware. The system then enters its core intelligent learning phase, which is executed by the behavioral logic template precipitation module. This module performs a preset initial learning cycle on the specified device, such as the first two weeks after the tenant moves in. All state transition records within the group are collected and analyzed. The ultimate goal is to automatically mine high-frequency behavioral logic templates that can characterize the tenants' stable living habits from these original time-series data streams through unsupervised learning algorithms. In terms of specific implementation, the module first aggregates the collected timestamp sequences by device group. For the aforementioned bathroom behavior group, the system may accumulate a large number of sequence fragments in the form of [Time T1, door sensor, closed] > [Time T2, socket, open] > [Time T3, socket, closed] > [Time T4, door sensor, open]. The algorithm will perform in-depth pattern matching and frequency statistics on these sequences. A feasible unsupervised learning algorithm is a sequence pattern analysis technology based on time series data mining. This technology can identify recurring sequence patterns with stable time intervals between each step. For example, the algorithm will find that after the door closing event, the socket opening event is likely to occur within 5 to 60 seconds, and after the socket opening event, the socket closing event is likely to occur within 15 to 30 minutes. When the frequency of occurrence of a specific time series sequence exceeds a preset threshold in all recorded sequences during the initial learning cycle, which is set to 80%, the sequence is identified and solidified as a high-frequency behavior. The technical essence of setting an 80% threshold for logical templates lies in the trade-off between universality and specificity during template generation. A threshold that is too high, such as 95%, may prevent the formation of an effective template due to occasional variations in normal tenant behavior, resulting in insufficient coverage. Conversely, a threshold that is too low, such as 50%, may mistakenly incorporate occasional, irregular behaviors into the template, leading to excessive false alarms during subsequent verification. Therefore, 80% is considered a robust engineering choice that ensures the template represents core behavioral patterns while effectively filtering out random noise.

[0032] The high-frequency behavior logic template that is finally formed has a data structure that not only includes the typical timing of device state jumps, such as door sensor. Close, socket. Open, socket. Close, door sensor. Open, but more importantly, it also accurately records the stable time interval range between each two consecutive state jumps in the sequence, such as 5 seconds to 60 seconds, 15 minutes to 30 minutes, and 2 seconds to 45 seconds. This complete structure including timing and intervals constitutes the fundamental basis for subsequent logical self-consistency verification. After the initial learning cycle is over, the system seamlessly switches to continuous monitoring mode, and the logical self-consistency verification module undertakes the real-time analysis task. The module receives the precise timestamp sequence output in real time by the state jump recording module, and compares it with all the generated high-frequency behavior logic The template performs continuous pattern matching. When a behavior sequence that matches the initial part of the template is started, for example, when the door sensor of the bathroom behavior group is detected to be closed, the verification module will predict the next expected state jump in the sequence, that is, the time window for the opening of the smart socket, based on the corresponding time interval range recorded in the template. If the absolute time difference between the actual state jump time of a monitoring unit and the expected state jump time indicated by the high-frequency behavior logic template based on the previous event exceeds the logic tolerance deviation threshold preset by the template, the system will determine that there is a behavior logic deviation. The logic tolerance deviation threshold here is an additional buffer zone for the inherent time interval range in the template. Its setting purpose is to accommodate that Some reasonable immediate fluctuations that do not affect the essence of behavioral logic. For example, if the template expects the water heater to turn off within 15 to 30 minutes, a 2-minute logic tolerance threshold means that as long as the water heater is turned off within the wider window of 13 to 32 minutes, it is considered logically self-consistent. However, if the water heater is turned on for 40 minutes and its shutdown time is significantly outside this window, the logical self-consistency is judged to be destroyed. At this time, the verification module will immediately generate a structured risk warning signal, which is sent to the remote management server through the warning signal sending module via the wireless communication network deployed in the apartment for management personnel to conduct timely analysis and disposal, in order to improve the overall quality of input data and prevent transient disturbances in the circuit or unintentional user errors. The rapid operation of recognition leads to misjudgment. The system also integrates a state credibility assessment module. Before submitting any state jump data to the core logic analysis engine, the module performs a pre-validity query. It dynamically calculates and continuously updates the historical average stable operation time for each monitoring unit. When the state jump recording module captures a new state jump, the state credibility assessment module calculates the actual time interval between this jump and the previous state jump of the monitoring unit. If this actual time interval is significantly smaller than the normal behavior pattern of the device, for example, less than 10% of its historical average stable operation time, the module determines that the jump is likely to be a signal noise or invalid operation. At this time, it will enter a short delay confirmation mode.Only when the status of the monitoring unit has not returned to its original state before the jump after the preset delay confirmation time, such as 5 seconds, has passed, the module will finally recognize the business validity of the state jump and formally submit its timestamp to the behavioral logic template precipitation module or the logic self-consistency verification module. This mechanism builds a barrier at the front end of the data stream, effectively filtering out invalid data glitches caused by physical reasons or human error. In addition, to deal with the systemic risks that may be caused by multiple devices competing for limited public resources, this system also includes a distributed resource coordination mechanism. First, the priority label configuration module allows the administrator to configure different priority digital labels for monitoring units that consume public resources. For example, in a scenario where the total electricity load of an apartment is limited, the smart socket associated with the kitchen electric oven that guarantees basic cooking needs can be configured with a high priority label, while the smart socket for the water heater used for bathing can be configured with a lower priority label. The public resource here can be the total electricity load of the apartment or the total water pressure of the apartment. Subsequently, the resource competition logic arbitration module makes autonomous decisions based on these labels. The function of this module Embedded in the firmware of each relevant monitoring unit, when a monitoring unit assigned a lower-priority digital tag, such as a water heater, is about to switch from an inoperative state to an operational state, its built-in arbitration logic proactively queries the operational status of other devices on the network. This process is achieved by monitoring globally broadcast state transition records, each of which contains a precise timestamp and device identity information. If it determines that a monitoring unit assigned a higher-priority digital tag, such as an oven, is currently operational, the resource contention logic arbitration module proactively delays the state transition request of the lower-priority unit. It continues to monitor the network until it detects a broadcast indicating that the higher-priority monitoring unit has returned to an inoperative state, allowing its own startup process to proceed. This decentralized, distributed arbitration mechanism, through autonomous information sensing and local rule enforcement between devices, cleverly avoids issues such as circuit overloads or water pressure drops that can result from the simultaneous startup of high-power appliances. It achieves system-level resource coordination and safe operation without increasing the hardware cost of a centralized controller.

[0033] Example 1: In the scenario of nursing homes and elderly people living alone, the monitoring units deployed there include a smart electric socket connected to a high-power electric heater and a smart door magnetic sensor installed on the bathroom door. During the initial learning cycle, the system uses the behavioral logic template precipitation module to automatically identify and solidify the event sequence of the electric heater turning on a short time after the bathroom door is closed and the electric heater turning off after a relatively fixed longer period of time into a high-frequency behavioral logic template. This template not only defines the sequence of device actions, but more importantly, sets a stable time interval based on historical data for the duration of the electric heater's on time. The effectiveness of this learning process is premised on the continuous operation of the pre-set state credibility assessment module: when the instantaneous fluctuation of the grid voltage causes the smart electric socket of the electric heater to produce continuous on-off jumps within milliseconds, the module calculates that the actual time interval is far less than 10% of the historical average stable operation of the device, and then enters the delayed confirmation mode. After confirming that the state is not continuously valid, such signal noise is filtered out. This purification mechanism ensures the high fidelity of the data stream used as input to the behavioral logic template precipitation module, so that the finally generated behavioral logic template can accurately represent the user's real and stable habits.

[0034] In the subsequent daily operation, the logic self-consistency verification module performs continuous comparison based on the aforementioned high-fidelity behavior logic template. When it detects that the actual on-time of the electric heater has exceeded the sum of the upper limit of the stable time interval range embedded in the template and the logic allowable deviation threshold, the system generates a risk warning signal. The generation of this warning is not based on whether the power or ambient temperature of the electric heater exceeds a certain isolated physical threshold, but on the behavioral feature of its duration of operation, which has logically deviated from the behavior template that has been learned as a safe mode. This realizes the active identification of potential fire risks that are slowly accumulated due to abnormal behavior and cannot be perceived by traditional threshold alarm mechanisms. Furthermore, a smart electric socket associated with the electric oven is deployed in the kitchen of the apartment. The socket is given a higher priority digital label through the priority label configuration module. When If a tenant is using an electric oven and enters the bathroom, triggering the shower-related behavior logic, the system's internal resource competition logic arbitration mechanism is activated. The arbitration logic located in the bathroom water heater smart socket will parse the globally broadcast state transition records before starting the water heater. It will determine that the electric oven with a higher priority is in operation. Based on this, the arbitration logic will deterministically delay the water heater startup until it detects that the electric oven has returned to non-operational status. This process demonstrates an inherent ability to resolve contradictions: within a single architecture, the system makes distributed, local decisions based on static priority rules, giving priority to system-level requirements for ensuring global power safety over local, non-immediate comfort needs. This resolves potential resource competition conflicts among multiple high-power devices without relying on any central controller intervention.

[0035] Example 2: In order to verify the actual effectiveness of the system of the present invention in identifying chronic risks caused by cross-device behavioral logic coupling, the following experiment was performed. The purpose of the experiment was to accurately compare the response differences between the system of the present invention and the traditional single-point physical quantity threshold alarm system when dealing with the same continuous heating and continuous sealing risk scenarios. The test platform was built in a software simulation environment, which integrated a thermodynamic and aerodynamic model of a standard single room. The following key virtual components were deployed in the platform: a virtual electric heater with a rated power of 1500W, whose start and stop status was recorded and controlled by a simulated smart electric socket; a virtual window, whose open and closed status was recorded and controlled by a simulated smart door magnetic sensor; a traditional carbon dioxide concentration sensor as a control group, with its alarm threshold set at 1500ppm according to indoor environmental safety standards; and a monitoring and management system deployed on a virtual edge computing node that integrates all the functional modules of the present invention. The setting of key parameters in the experiment follows a certain engineering logic: First, the total simulation time is set to 120 minutes. The technical consideration of this decision is to balance the integrity of the test process with the effective use of computing resources, ensuring that the time is sufficient to fully cover a typical behavior sequence and provide a sufficient time window for observing the cumulative effects of chronic risks. Second, the logic tolerance threshold of the system core is essentially set to balance the sensitivity of the system response with the effective use of computing resources. A trade-off is made between the tolerance for normal behavior fluctuations. In order to avoid false alarms due to minor changes in user habits and ensure the timely discovery of significant anomalies, the value of the threshold is associated with the average duration of the corresponding steps in the learned behavior template. In this experiment, it is set to 10% of the heating time in the template. The test process is divided into two stages. The first stage is the precipitation of the behavior logic template. In this stage, the platform performs 10 simulations of normal heating behavior sequences. Each sequence includes a series of state jumps, such as the window closing, the electric heater turning on, the electric heater turning off after 28 to 32 minutes of continuous operation, and then the window opening for ventilation. The state jump recording module in the system of the present invention accurately records the time of each jump. Based on these sequence data, the behavioral logic template precipitation module outputs a high-frequency behavioral logic template after counting sequences with a frequency higher than 80%. This template clearly defines the normal operating time range of the electric heater as 28 minutes and 32 minutes. The second stage is risk scenario verification. The platform initiates an abnormal heating sequence simulation: after the window is closed, the electric heater turns on and continues to run. However, after the expected time point of 32 minutes, the window and the electric heater remain in their original state until the end of the 120-minute simulation. During this process, the test platform synchronously records the status output of the system of the present invention and the traditional CO2 sensor. When the simulation reaches different key time points, the core data comparison is shown in the following table:

[0036] Table 1: Schematic diagram of CO2 concentration and risk warning signals in the long-term rental monitoring and management system.

[0037]

[0038] The data shows that at the 36th minute of the simulation, the reading of the traditional CO2 sensor has increased significantly, but has not yet reached its 1500 The alarm threshold of ppm was exceeded, thus maintaining normal operation. However, the system of the present invention generated a risk warning signal at this moment. This differentiated response was directly caused by the operation of the logic self-consistency verification module within the system of the present invention: this module compared the actual state transition time of the electric heater with the high-frequency behavioral logic template and determined that the absolute time difference exceeded the preset logic tolerance threshold, indicating a behavioral logic deviation. Until the late stages of the simulation, the readings of the traditional sensor remained near the threshold and failed to effectively alarm, while the warning status of the system of the present invention remained stable. The results of this experiment demonstrate that the technical solution proposed by the present invention can effectively identify potential safety risks caused by the coordinated failure of the behavioral logic of multiple devices, which develop slowly and cannot be captured by a single physical quantity threshold. This experimental conclusion forms a strict logical closed loop with the core technical concept of the present invention: by converting the timestamp sequence of device state transitions into a behavioral logic template, the system achieves a transition from physical quantity monitoring to behavioral logic verification. This enables the system to issue accurate and reliable warnings at the early stages of risk accumulation based solely on the deviation of the behavioral sequence from the safe mode, thus fundamentally resolving the inherent technical shortcomings of existing technologies in dealing with such chronic, cross-domain risks.

[0039] Example 3: This example combines Figures 1 to 4 , a long-term rental monitoring and management system and method are described, such as Figure 1 As shown in the figure, Scenario 1 and Scenario 2 describe the operating procedures of the system under normal operation and abnormal conditions (outside the template range) respectively. In Scenario 1, when the state of the smart electric socket jumps from on to off, the jump time is recorded by the state jump recording module. The system further performs a logical check of the query template, and the returned template time is 15-30 minutes. At this time, the behavioral logic template library will generate a high-frequency behavioral logic template that matches the state jump of the device group, and perform continuous comparison to ensure that it is consistent with the expected time window, thereby verifying its logical consistency. If it meets expectations, the system maintains normal operation. , the early warning sending module will not trigger any alarm; in scenario 2, the state jump of the smart socket occurs at an abnormal time point, which exceeds the expected time window of the behavior logic template. Specifically, the state change record of the smart socket occurs at 14:00:00, and according to the behavior logic template, the expected jump time should be 14:28-14:32, and the actual jump time is 14:36. Therefore, the system triggers the behavior logic verification module for verification and generates a risk early warning signal. This signal is sent to the remote management server through the early warning sending module so that the management personnel can respond in time.

[0040] like Figure 2 As shown in the figure, by recording the matching accuracy and stability index of the model, the changing trend of the template matching accuracy and template stability index during the learning cycle is shown. The horizontal axis in the figure is the learning cycle (number of days), from the 1st day to the 14th day, and the vertical axis is the percentage (%), which represents the percentage of model matching accuracy and template stability index. Curve 1 represents the template matching accuracy. As the learning cycle prolongs, the accuracy gradually increases, from the initial approximately 50% to gradually increase, and finally reaches a matching accuracy of nearly 90%. Curve 2 represents the template stability index. The stability is low in the early stage of learning, but with the passage of time, the stability gradually improves, and tends to stabilize in the later stage of the learning cycle, achieving a higher stability. This shows that through continuous learning, the system can continuously optimize and improve the matching effect of the behavioral logic template, thereby improving the intelligence level of the long-term rental monitoring and management system.

[0041] like Figure 3 As shown in the figure, four typical monitoring units are marked, namely: smart electric sockets for water heater control, smart door magnets installed on bathroom doors, smart electric sockets for electric ovens, and smart water flow switches for water pipe monitoring. The system first logically divides the relevant monitoring units deployed in the same physical space into a device group through the device group definition module. The state jump recording module is used to record the unique identification code and the precise timestamp of the state jump when the operating status of the monitoring unit changes. The state credibility evaluation module dynamically calculates the historical stable operating time of the monitoring unit based on the above timestamp, which is used to filter abnormal data jumps before real-time comparison. For multiple monitoring units that consume public resources (such as water, electricity, etc.), the system configures different priority digital labels for them through the priority label configuration module so that For subsequent resource coordination, the above modules jointly support the operation of the core modules, including: behavioral logic template precipitation module (unsupervised learning), which is used to automatically generate high-frequency behavioral logic templates based on state jump time series; logical self-consistency verification module (real-time comparison), which is used to perform logical consistency verification on the real-time state sequence and the template to identify potential risks; resource competition logic arbitration module (distributed decision-making), which is used to dynamically judge resource usage conflicts and control equipment operation based on priority labels. The behavioral deviations and logical anomalies identified by all core modules in the system are finally transmitted to the remote management server in the form of structured warning signals through the early warning signal sending module (wireless communication network) for the back-end to handle and analyze. In the figure, solid lines represent data flows, dotted lines represent configuration relationships, and black boxes mark the core modules of the system.

[0042] like Figure 4As shown in the figure, the horizontal axis indicates the consecutive dates in the system learning cycle from the 1st day to the 14th day, the left side of the vertical axis indicates the numerical range of the template matching accuracy (%) and the template stability index (%) in percentage (%), and the right side of the vertical axis indicates the changing trend of the number of recognized templates in the form of graphical numbers; the solid dots in the figure indicate the growth trajectory of the template matching accuracy (%), the dotted diamonds indicate the changes in the template stability index (%), and the dotted triangle curve represents the cumulative growth of the number of recognized templates.

[0043] Example 4: In the scenario of nursing homes and care for elderly people living alone, a long-term rental apartment scenario is based on a smart bathroom lighting, a smart electrical socket linked to a water heater, and an independent smart exhaust fan switch. In this scenario, a complete shower behavior that complies with safety regulations has a behavioral logic chain of: lighting on, water heater started, water heater turned off, exhaust fan started, exhaust fan turned off, and lighting turned off; the core task of the system in the initial learning cycle is to execute a three-stage unsupervised learning process aimed at building a high-frequency behavioral logic template. The first stage is session segmentation. The system uses the state jump of the bathroom smart lighting on as the starting mark of the behavioral session and the state jump of the bathroom smart lighting off as the end mark. The system accurately timestamps the state jump sequence of all monitoring units that occurred during this period and divides it into independent session segments from the continuous data stream; the second stage is sequence vectorization and clustering. The system uses a sequence similarity measurement algorithm based on edit distance to calculate the sequence similarity of each session. The similarity scores between the conversation segments are calculated, thereby converting the matching problem of time series patterns into the calculation problem of vector distance in high-dimensional space; then, a density clustering algorithm DBSCAN suitable for time series data is executed. The algorithm automatically aggregates conversation sequences with sufficiently high similarity into the same cluster. The system then verifies and confirms the cluster with the highest frequency. The number of conversation sequences contained in it accounts for more than 80% of the total number of conversation sequences in the initial learning cycle. This cluster is confirmed as the basic data source for building high-frequency behavior logic templates. Those isolated sequences that cannot be classified into this high-frequency cluster are identified as low-frequency or irregular behaviors; the third stage is the template structure solidification. Based on the identified high-frequency behavior clusters, the system constructs a probabilistic directed acyclic graph as the final high-frequency behavior logic template. Each node represents a unique device state jump event, one of which is the water heater. Turn on. The directed edge connecting the two nodes represents a commonly observed behavioral step with a stable time series association. Each edge , that is, from the node Pointing to a node The edges are all given a set of precise statistical properties, the core of which is the conditional probability , average time interval , and the standard deviation of the time interval The set of structured templates not only defines the behavioral steps, but also accurately defines the expected time relationship and regularity between the steps in statistical language.

[0044] After the learning is completed and the continuous monitoring phase is entered, the logic self-consistency verification module verifies the real-time state jump sequence based on the probabilistic directed acyclic graph generated above. When an event that matches the template start node occurs, the system makes a probabilistic prediction of the occurrence time of the next expected event based on the corresponding edge in the graph. At this time, the judgment of behavioral logic deviation no longer relies on a global fixed logic tolerance threshold, but is completed through a dynamic, statistically significant test logic. If the previous event The actual time of occurrence , to the current event The actual time of occurrence The actual time interval that has elapsed , the average time interval between the time interval and the time interval recorded in the template The degree of deviation exceeds one standard deviation If the dynamic boundary of the decision is exceeded, it is determined that there is a behavioral logic deviation. The judgment logic is described by the following inequality: , in this inequality, It is a dimensionless sensitivity coefficient, which is a certain value used to define the tolerable fluctuation range equivalent to several standard deviations. It is an engineering parameter that can be pre-configured by the system administrator according to the management accuracy requirements. The value can be 2 or 3, which correspond to the confidence intervals of about 95% and 99.7% in statistics respectively. When the water heater is turned off, the exhaust fan does not run at the expected time as expected by the template. Once activated within the time window, the system generates a structured risk warning signal pointing to a lack of ventilation logic based on the statistical fact that it deviates from the learned high-confidence behavior pattern. This method of binding parameter settings with the intrinsic statistical distribution characteristics of the data improves the sensitivity and robustness of risk identification.

[0045] Example 5: In the scenario of nursing homes and care for elderly people living alone, this example is based on a rental apartment that contains both high-security risk areas and regular living areas. In this apartment, the monitoring units associated with the kitchen gas valve smart switch and carbon monoxide sensor are defined as the kitchen safety group, and the monitoring units associated with the living room electric kettle smart socket and smart curtain controller are defined as the living room convenience group. Given the differences in the management objectives of the two groups, the former requires the highest recognition sensitivity for any potential risks, while the latter focuses on avoiding nuisance false alarms due to normal fluctuations in user habits. Therefore, a systematic setting procedure is adopted for the setting of multiple core parameters in the system, rather than a globally unified fixed value.

[0046] First, for the frequency threshold used to identify high-frequency behaviors in the behavioral logic template precipitation module, the specific value determination process is as follows: After the initial learning cycle, the system independently executes a threshold optimization procedure for each device group. Taking the living room convenience group as an example, the system administrator first sets a higher initial frequency threshold, such as 95%. The system then analyzes all conversation sequences within the learning cycle and presents non-high-frequency sequence patterns that are excluded due to the current threshold. Based on the understanding of the expected functions and operating methods of the device group, the administrator determines whether the excluded sequences contain key behaviors that should be identified as templates. If the current threshold is too high, resulting in insufficient template coverage, the administrator will lower the frequency threshold of the group by a preset step, such as 5%, through the management interface, and repeat the above analysis and judgment process. This iteration continues until the administrator confirms that the current threshold can ensure all key and regular behaviors. While incorporating regular behavior patterns into the template, extremely low-frequency, accidental operations are effectively avoided. Through this procedure, different equipment groups can obtain customized thresholds that meet their respective management objectives. Secondly, the judgment criteria used to filter signal noise in the state credibility assessment module are the percentage of the actual time interval between state jumps and its historical average stable operation time. This percentage is also determined according to empirical optimization logic. The system provides analysis tools to generate a distribution histogram of all state jump intervals for a specific monitoring unit within the learning cycle. Managers can use this to observe the main distribution of time intervals formed by normal operation, as well as the distribution of transient jumps on extremely short time scales, which are caused by instantaneous circuit disturbances or invalid physical operations and do not represent stable device state migration. The goal of this procedure is to set an appropriate percentage so that the absolute time threshold converted to it can completely cover and filter the transient jump distribution area.Similarly, the preset duration that the module needs to maintain after entering the delayed confirmation mode is set based on the signal transmission and stability characteristics of the network where the physical device is located. The duration is set to be slightly longer than the maximum signal jitter or rebound time observed by the system in the deployment environment to ensure that the system only responds to the fact that the device enters a stable new state, rather than misjudging the transient process; furthermore, after the initial learning cycle is completed and the system automatically generates a set of candidate high-frequency behavioral logic templates, a mandatory template review and confirmation link involving managers is set up. The system will not automatically apply any learning results directly to risk verification. Instead, the system will submit each candidate template in a human-readable natural language format through the remote management service. The interface of the device is presented to the manager. For example, a candidate template will be described as Kitchen Safety Group Template 1: The gas valve is detected to be open, and the carbon monoxide sensor reading rises slightly and steadily in the following thirty seconds to three minutes. This behavior pattern occurs at a frequency of 92% during the learning cycle. The manager must review each candidate template and make a clear operation of approval or rejection. Only approved templates will be moved into the system's formal behavior logic template library for subsequent real-time logic consistency verification. If the manager determines that a learned template itself represents a behavior that does not comply with safety regulations, it can be discarded through a rejection operation, thereby avoiding the system solidifying the dangerous behavior sequence as a benchmark for generating risk warnings.

[0047] In the specific implementation using a probabilistic directed acyclic graph as a template, the conditional probability attribute contained in the directed edge connecting two event nodes in the graph is calculated using the following logic: among all the data in the learning cycle, the actual number of successful transfers from the previous event node to the current event node is counted, and then this number is divided by the total number of occurrences of the previous event node. This quantitative probability value, together with the average time interval of the transfer and its standard deviation, constitutes a complete mathematical characterization of the stability of the behavioral step, providing a rigorous data foundation for the subsequent dynamic risk judgment logic based on statistical significance.

[0048] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0049] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention 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 invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A long-term rental monitoring and management system, characterized in that: The system comprises: a device group definition module configured to logically define at least two monitoring units deployed in the same physical space as a device group; a state transition recording module configured to record the unique identification code of the monitoring unit and the precise timestamp of the state transition when the operating state of the monitoring unit transitions from one state to another; A behavioral logic template precipitation module is configured to automatically generate one or more high-frequency behavioral logic templates through an unsupervised learning algorithm based on the precise timestamp sequence recorded by the device group during a preset initial learning cycle. The high-frequency behavioral logic templates represent the stable temporal associations between the state transitions of each monitoring unit in the device group. The logic self-consistency verification module is configured to continuously compare the precise timestamp sequence output in real time by the state jump recording module with the high-frequency behavior logic template, and generate a risk warning signal when there is a behavior logic deviation between the real-time timestamp sequence and the high-frequency behavior logic template; the determination of the behavior logic deviation is based on the following logic: when the absolute time difference between the actual state jump time of a monitoring unit and the expected state jump time indicated by the high-frequency behavior logic template exceeds the logic allowable deviation threshold preset by the template, it is determined that a behavior logic deviation exists.

2. The long-term rental monitoring and management system according to claim 1 is characterized in that: The operating status recorded by the state jump recording module is a binary state of on and off. The high-frequency behavior logic template generated by the behavior logic template precipitation module includes the typical timing between the state jumps of different monitoring units in the group and the time interval range corresponding to each typical timing.

3. The long-term rental monitoring and management system according to claim 1 is characterized in that: The system also includes a state credibility assessment module configured to: dynamically calculate and update the historical average stable operating time of each monitoring unit in the device group based on the precise timestamp recorded by the state jump recording module; And before the logic self-consistency verification module performs comparison, when the state jump recording module records that the state of a certain monitoring unit has jumped, the state credibility assessment module calculates the actual time interval between the state jump and the previous state jump of the monitoring unit; if the actual time interval is less than ten percent of the historical average stable operation time of the monitoring unit, the state credibility assessment module enters the delayed confirmation mode; only when the state of the monitoring unit has not recovered to the state before the jump after the delayed confirmation mode continues for a preset period of time, the state credibility assessment module recognizes the validity of the state jump, and submits the timestamp of the state jump to the behavior logic template precipitation module for template generation within the initial learning cycle; after the initial learning cycle ends, the timestamp of the state jump is submitted to the logic self-consistency verification module for real-time comparison.

4. The long-term rental monitoring and management system according to claim 1, characterized in that: The system also includes a priority label configuration module configured to configure different priority digital labels for at least two monitoring units that belong to at least two different equipment groups and consume common resources.

5. The long-term rental monitoring and management system according to claim 4 is characterized in that: The system further includes a resource contention logic arbitration module configured to: before the state of a monitoring unit configured with a lower priority digital tag is about to switch from a non-operating state to an operating state, determine, based on the latest accurate timestamps of other monitoring units recorded by the state transition recording module, whether there is a monitoring unit configured with a higher priority digital tag currently in an operating state; If so, the resource contention logic arbitration module delays the state switching of the lower priority unit until the higher priority monitoring unit recovers to the non-operating state.

6. The long-term rental monitoring and management system according to claim 5, characterized in that: Common resources include the total electrical load of the apartment or the total water pressure of the apartment.

7. The long-term rental monitoring and management system according to claim 1, characterized in that: The system also includes a warning signal sending module configured to send the risk warning signal generated by the logic self-consistency verification module to a remote management server via a wireless communication network.

8. The long-term rental monitoring and management system according to claim 1 is characterized in that: The monitoring unit includes a smart electrical socket, a smart door magnetic sensor or a smart water flow switch.

9. The long-term rental monitoring and management system according to claim 1, characterized in that: During the initial learning cycle, the behavior logic template precipitation module counts the occurrence frequency of the state jump sequence of each monitoring unit in the device group, and identifies the sequence with an occurrence frequency higher than 80% as a high-frequency behavior logic template.

10. A long-term rental risk warning method, characterized in that: The method comprises the following steps: Step a, logically defining at least two monitoring units deployed in the same physical space as a device group; Step b, when the operating state of the monitoring unit changes from one state to another, recording the unique identification code of the monitoring unit and the precise timestamp when the state change occurs; Step c: Based on the precise timestamp sequence recorded by the device group during a preset initial learning cycle, one or more high-frequency behavior logic templates are automatically generated through an unsupervised learning algorithm. The high-frequency behavior logic templates represent the stable temporal associations between the state transitions of each monitoring unit in the device group. Step d: Continuously compare the precise timestamp sequence output in real time with the high-frequency behavior logic template, and generate a risk warning signal when there is a behavior logic deviation between the real-time timestamp sequence and the high-frequency behavior logic template; the determination of the behavior logic deviation is based on the following logic: when the absolute time difference between the actual state jump time of a monitoring unit and the expected state jump time indicated by the high-frequency behavior logic template exceeds the logic allowable deviation threshold preset in the template, it is determined that a behavior logic deviation exists.

Images