A smart building-oriented cross-subsystem event-driven linkage control method and system

Abstract

The application relates to the technical field of intelligent building control and discloses a cross-subsystem event-driven linkage control method and system for intelligent buildings. The method comprises the following steps: deeply detecting and time-series extracting original operation data to obtain a two-dimensional time-series data set; constructing a life cycle attenuation model and encapsulating to obtain a standardized data sequence; constructing a backtracking buffer pool, performing reverse interpolation and retention operations to generate virtual data points, and performing discretization segmentation to generate a time-series alignment event slice sequence; performing a logic replay matching operation to obtain an event set, performing state arbitration and instruction generation operations, and generating adaptive collaborative control instructions. The application eliminates time-series logic errors caused by heterogeneous network transmission uncertainty and sampling frequency aliasing, solves the problems of causal inversion and decision lag in cross-system linkage, and realizes accurate research and reliable control of global perception.

Description

Technical Field

[0001] This invention relates to the field of intelligent building, and more specifically, to a cross-subsystem event-driven linkage control method and system for intelligent buildings. Background Technology

[0002] As smart buildings evolve towards comprehensive perception and deep collaboration, the demand for real-time response to complex events across subsystems is surging. However, traditional linkage control methods, such as relying on network time protocols for clock synchronization or based on simple first-in-first-out (FIFO) queue buffers, generally face bottlenecks in decision reliability. The accuracy and timeliness of linkage commands are difficult to balance, and the system is highly susceptible to oscillations under network fluctuations. Existing technologies mostly rely on linear sorting of data packet arrival times or static threshold judgments based on a single time dimension. This ignores the nonlinear decay differences in physical timeliness in heterogeneous network environments and the spatiotemporal logical tearing effect caused by sampling frequency aliasing. Specifically, the instantaneous value of high-frequency streaming media data and the long-tail value of low-frequency sensor data become misaligned under random interference in the transmission path, resulting in "false linkage" or "causal reversal." This makes it impossible for rule engines to distinguish between "state not occurring" and "data lag," causing the control center to oscillate between waiting and misjudging. Therefore, how to shift from mechanical time-point alignment to quantifying the value of the entire lifecycle of an event, and transform passive delay and waiting into proactive virtual reconstruction and trust game, thereby breaking through the obscuring of physical causality by network uncertainty, is a technical challenge to be solved in this field.

[0003] In the prior art, Chinese Patent No. CN106992884B discloses a method for multi-subsystem linkage in an intelligent building business system. This method includes five core modules: subsystem API registration, call chain orchestration, anomaly information collection, call chain analysis and anomaly view generation, and multi-subsystem linkage processing. By aggregating control anomaly information from each subsystem to an anomaly processing center, it reconstructs the business process under anomaly scenarios and automatically calls forward compensation or reverse reversal APIs to ensure data consistency between subsystems. Its core advantage lies in solving the data conflict problem when subsystems are abnormal, improving the stability of linkage control, and is suitable for the collaborative scheduling of multiple business systems in intelligent buildings. Chinese Patent No. CN116819995A discloses a method and system for building equipment linkage based on an intelligent building controller. With an intelligent building controller at its core, the system achieves hierarchical linkage of building devices by receiving linkage control commands, splitting target control command segments, sending commands to target devices, providing feedback on execution results, and modifying flag bits. At the same time, it adopts a controller-first, controller-later command transmission mode to reduce network latency caused by direct device interaction, improve linkage success rate, and provide a lightweight solution for device collaboration in small and medium-sized buildings.

[0004] However, while the two existing technologies mentioned above have some value in terms of multi-subsystem collaborative stability and device linkage latency optimization, they fail to address the core pain points of event slice timing alignment and causal reconstruction in heterogeneous networks of smart buildings. Specifically, the patent with authorization announcement number CN106992884B focuses on passive compensation after subsystem anomalies, without addressing timing calibration during data acquisition and transmission. It lacks a two-dimensional reference framework of physically generated timestamps and control center arrival timestamps, and cannot quantify the timeliness decay differences of different types of data, making it prone to false linkages due to data timing misalignment. The patent with publication number CN116819995A reduces latency through hierarchical transmission of commands by the controller, but it does not solve the data gap problem under heterogeneous protocols. It lacks reverse interpolation filling and virtual data point generation mechanisms, and the rule engine still cannot distinguish between "state not occurring" and "data lag." Furthermore, it does not introduce trust-based game-theoretic adjudication logic, making it difficult to handle the precise real-time response requirements of high-priority complex events, such as fires combined with dense populations, and it cannot overcome the obscuring of physical causal relationships by network uncertainty. Summary of the Invention

[0005] This invention is applicable to heterogeneous network environments in smart buildings, meeting the deep collaboration needs of multiple subsystems. It constructs a dual-time reference framework by using physically generated timestamps and control center arrival timestamps, combined with standardized data sequences generated by a lifecycle decay model, to achieve spatiotemporal decoupling and trust quantification of the original operational data. Reverse interpolation and hold operations generate virtual data points to fill data gaps in the backtracking buffer pool, eliminating state tearing caused by sampling frequency aliasing. Time-aligned event slice sequences transform streaming data into static logical units, and, in conjunction with trust-based game-theoretic state adjudication and instruction generation operations, filter valid states from contradictory event sets and generate adaptive collaborative control instructions with time compensation. This invention effectively solves the problems of causal inversion and decision lag caused by network transmission uncertainties, improving the accuracy of global perception and the reliability of collaborative control.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A cross-subsystem event-driven linkage control method for smart buildings includes:

[0008] The raw running data is acquired, and deep detection and time series extraction are performed on the raw running data to obtain a two-dimensional time series dataset. A life cycle decay model for quantifying the timeliness of data is constructed for the two-dimensional time series dataset. The two-dimensional time series dataset and the life cycle decay model are expanded and reconstructed to obtain a standardized data sequence for data closed-loop management.

[0009] A backtracking buffer is constructed based on standardized data sequences to recreate real events in the physical world. Inverse interpolation and hold operations are performed on the backtracking buffer to generate virtual data points that can fill gaps. Based on the virtual data points, the backtracking buffer is discretized to generate a logically complete time-aligned event slice sequence.

[0010] A logical replay matching operation is performed on the time-aligned event slice sequence to obtain an event set covering real-time response and historical tracing. Based on the event set, state adjudication and instruction generation operations are performed to generate adaptive collaborative control instructions.

[0011] Furthermore, the method for obtaining the two-dimensional time-series dataset includes:

[0012] The raw operational data includes three data elements: streaming media data, sensor data, and status alarm data;

[0013] Analyze the underlying data frame structure of the raw running data, and extract the raw values ​​or status codes and the unique device identifier from the underlying data frame structure.

[0014] For streaming media data, the absolute time code written by the camera hardware encoder is extracted as the physical timestamp; for sensor data, the local clock time of the gateway is extracted as the physical timestamp; for status alarm data, the trigger time recorded by the alarm controller is extracted as the physical timestamp.

[0015] When the raw operating data arrives at the network interface of the control center, the system time of the control center is read as the arrival timestamp of the control center.

[0016] The original numerical values ​​or status codes, unique device identifiers, physical generation timestamps, and control center arrival timestamps from the same underlying data frame structure are logically bound to obtain the two-dimensional time-series dataset.

[0017] Furthermore, the lifecycle decay model includes:

[0018] By collecting continuous time series raw operating data within a preset period, calculating the autocorrelation, extracting the time span required for the autocorrelation to decay to a preset threshold, defining the reciprocal of the time span as the benchmark decay rate, setting the effectiveness decay coefficient based on the benchmark decay rate for the raw operating data, and obtaining a pre-constructed sensor characteristic database.

[0019] Based on the device's unique identifier, a pre-built sensor characteristic database is retrieved to match the effectiveness attenuation coefficient;

[0020] A calculation formula is constructed based on the time difference between the current moment and the physically generated timestamp to calculate the trust level of each data element in the two-dimensional time series dataset.

[0021] Furthermore, the method for obtaining the standardized data sequence includes:

[0022] The network transmission delay is obtained by calculating the difference between the arrival timestamp of the control center and the physical generation timestamp, and then compared with a preset delay tolerance threshold.

[0023] If the network transmission delay is greater than the preset delay tolerance threshold, the trust level is multiplied by the preset penalty factor to obtain the corrected trust level, and a hysteresis suspicious state flag is added and set to True; otherwise, the trust level remains unchanged and the hysteresis suspicious state flag is set to False.

[0024] The original numerical value or status code, the device's unique identifier, the physical generation timestamp, the control center's arrival timestamp, the validity decay coefficient, the corrected trust level, and the hysteresis status mark are encapsulated into a standardized data sequence.

[0025] Furthermore, the backtracking buffer pool includes:

[0026] Physical memory space is allocated in the random access memory of the control center. The physical memory space is divided into several physical memory units of fixed size, and each physical memory unit corresponds to a fixed time granularity. A backtracking buffer pool is constructed using the physical memory space, and a circular buffer is used as the underlying physical storage structure of the backtracking buffer pool.

[0027] Starting from the current system time, it extends backward to the current system time minus the preset buffer window length, forming a continuous logical timeline;

[0028] The effective range of the backtracking buffer is determined based on the logical timeline, and it is determined whether the physically generated timestamps in the standardized data sequence fall within the effective range of the backtracking buffer.

[0029] If it falls within the effective range, calculate the time difference between the current system time and the physically generated timestamp, and calculate the storage index address in combination with the time granularity;

[0030] Based on the storage index address, the standardized data sequence is inserted into the physical memory unit.

[0031] Furthermore, the method for acquiring the virtual data points includes:

[0032] Traverse the physical memory units in the backtracking buffer in descending order of time. For each known device unique identifier, check whether there is a standardized data sequence containing the device unique identifier in the currently traversed physical memory unit.

[0033] If the current physical memory cell lacks a standardized data sequence with a unique device identifier, backtrack backward until a physical memory cell containing a standardized data sequence with a unique device identifier is found, and that physical memory cell is used as the source data point.

[0034] A virtual data point is generated based on the source data point. The virtual data point inherits the original value or state code of the source data point. The trust level of the virtual data point is obtained through the source data point construction formula, and the virtual data point is filled into the physical memory unit. Filling is not performed only when the trust level of the virtual data point is less than the preset validity truncation threshold.

[0035] Furthermore, the physical memory units of the backtracking buffer pool are accessed in ascending time order at preset logical step intervals.

[0036] Construct a timing-aligned event slice containing several state slots, where each state slot corresponds to a known unique device identifier;

[0037] Extract standardized data sequences and virtual data points from physical memory units, fill the original values ​​or status codes into the corresponding status slots as status values, and fill the trust level into the corresponding status slots as status weights.

[0038] If the device unique identifier corresponding to the state slot in the timing alignment event slice has neither a corresponding normalized data sequence nor a corresponding virtual data point in the currently processed physical memory unit, then the state slot is marked as invalid. Finally, a series of consecutive timing alignment event slices are generated in chronological order, forming a timing alignment event slice sequence.

[0039] Furthermore, the event set includes:

[0040] Traverse the timing-aligned event slices and extract the original values ​​or status codes from the status slots as measured values ​​based on the preset cross-subsystem linkage trigger conditions.

[0041] The measured values ​​are substituted into the state constraints and logical topology defined by the preset cross-subsystem linkage triggering conditions for calculation. If the state constraints are met and the logical topology operation is true, it is determined that a linkage event to be confirmed has occurred.

[0042] Calculate event lag duration by aligning event slices with time sequence;

[0043] If the delay time of an event is less than or equal to the preset real-time determination threshold, it is marked as a real-time event; otherwise, it is marked as a delayed confirmation event. The real-time event and the delayed confirmation event are combined to obtain the event set.

[0044] Furthermore, the method for obtaining the adaptive cooperative control command includes:

[0045] For real-time events, each real-time event and delayed confirmation event in the event set is traversed. When multiple real-time events or delayed confirmation events involving the same preset cross-subsystem linkage triggering condition but with mutually exclusive states are detected, the trust level is compared. The real-time event or delayed confirmation event with the highest trust level or the highest risk level is determined as the winning event and confirmed as a valid state.

[0046] By using preset cross-subsystem linkage triggering conditions, the linkage execution strategy corresponding to the effective state is retrieved. The linkage execution strategy indicates the execution device identifier, action type, and preset control duration to be invoked when the cross-subsystem linkage triggering conditions are met.

[0047] By employing a coordinated execution strategy, instruction generation operations are performed on both real-time events and delayed acknowledgment events within the event set to obtain adaptive collaborative control instructions.

[0048] A cross-subsystem event-driven linkage control system for smart buildings, used to implement the above method, the system comprising:

[0049] Spatiotemporal quantization module: used to acquire raw running data, perform deep detection and time series extraction on the raw running data to obtain a two-dimensional time series dataset, construct a life cycle decay model for quantifying the timeliness of the two-dimensional time series dataset, expand and reconstruct the two-dimensional time series dataset and the life cycle decay model to obtain a standardized data sequence for data closed-loop management;

[0050] The global restoration module is used to construct a backtracking buffer pool based on standardized data sequences, restore real events in the physical world, perform inverse interpolation and hold operations on the backtracking buffer pool to generate virtual data points that can fill gaps, and perform discretization segmentation on the backtracking buffer pool based on the virtual data points to generate a logically complete time-aligned event slice sequence.

[0051] Cross-system adjudication module: Used to perform logical replay matching operations on time-aligned event slice sequences to obtain an event set covering real-time response and historical tracing. Based on the event set, state adjudication and instruction generation operations are performed to generate adaptive collaborative control instructions.

[0052] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0053] This invention constructs a dual time reference framework by physically generating timestamps and control center arrival timestamps. Combined with a lifecycle decay model that quantifies data timeliness, it generates standardized data sequences, achieving spatiotemporal decoupling and trust quantification of the original operational data. This addresses the pain points of transmission delay uncertainty and cognitive bias caused by random interference factors in the network transmission path. A backtracking buffer pool, along with reverse interpolation and hold operations, generates virtual data points with dynamic trust levels, logically filling data gaps between heterogeneous subsystems and generating a logically complete time-aligned event slice sequence, eliminating logical state fragmentation caused by sampling frequency aliasing. A logical replay matching operation, combined with trust-based game-theoretic state adjudication and instruction generation operations, transforms contradictory event sets into adaptive collaborative control instructions with time compensation corrections. This thoroughly solves the problems of causal inversion and decision lag in cross-system linkage, improving the reliability of smart building's overall perception and control. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 A flowchart illustrating a cross-subsystem event-driven linkage control method for smart buildings, provided by an embodiment of the present invention;

[0056] Figure 2 This is a flowchart illustrating the judgment process for a two-dimensional time-series dataset provided in an embodiment of the present invention.

[0057] Figure 3 A flowchart illustrating the judgment process of the lifecycle decay model provided in this embodiment of the invention;

[0058] Figure 4 This is a functional block diagram of a cross-subsystem event-driven linkage control method for smart buildings, provided in an embodiment of the present invention. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] Example 1

[0061] Please see Figure 1 As shown, this embodiment provides a cross-subsystem event-driven linkage control method for smart buildings, including:

[0062] Step S10: Obtain the original running data, perform deep detection and time series extraction on the original running data to obtain a two-dimensional time series dataset, construct a life cycle decay model for quantifying the timeliness of the data for the two-dimensional time series dataset, expand and reconstruct the two-dimensional time series dataset and the life cycle decay model to obtain a standardized data sequence for data closed-loop management.

[0063] Further, step S10 includes:

[0064] Step S11: Obtain the raw running data, perform deep detection and time series extraction on the raw running data to obtain a two-dimensional time series dataset.

[0065] In the operational environment of a smart building, comprehensive perception and control rely on a heterogeneous network environment comprised of a video surveillance subsystem, an environmental monitoring subsystem, and a security alarm subsystem. Due to different acquisition mechanisms, these subsystems collect and output streaming media data, sensor data, and status alarm data, respectively. These output streaming media data, sensor data, and status alarm data are collectively referred to as raw operational data. When this raw operational data is transmitted from the front-end acquisition devices to the control center, it travels through a network transmission path consisting of switches, routers, and wireless gateways. Because of random interference factors such as network congestion, data packet queuing, and packet loss and retransmission along this network transmission path, the transmission delay of the raw operational data is uncertain and asymmetrical. If the order of events is determined solely by the time the data arrives at the control center, later events may be processed first due to faster transmission, while earlier events may be processed later due to slower transmission, resulting in timing logic errors. Therefore, it is necessary to establish a dual time reference framework that simultaneously anchors the actual physical occurrence time and the logical reception time of the control system to provide a basis for reconstructing the true event sequence.

[0066] Through multi-protocol parsing adaptation layer, such as Figure 2As shown, deep detection and timing extraction are performed on the raw operating data. The multi-protocol parsing and adaptation layer is a standardized processing middleware deployed at the data entry point of the control center, configured with multiple communication protocol parsing drivers to shield the differences in the underlying transmission protocols of the raw operating data. For streaming media data collected by the video surveillance subsystem, the absolute time code written to the header by the camera hardware encoder at the moment of image acquisition is extracted, and the absolute time code is defined as a physical timestamp. The physical timestamp represents the physical moment when the optical signal is converted into an electrical signal, that is, the source time of the event, and is not affected by the subsequent network transmission status; wherein, the streaming media data consists of a series of continuous image frames; for sensor data collected by the environmental monitoring subsystem, the gateway local clock time recorded at the moment when the sensor data completes the analog signal to digital signal conversion is extracted, and the gateway local clock time is defined as a physical timestamp; for status alarm data collected by the security alarm subsystem, the trigger time recorded by the hardware clock at the moment when the alarm controller detects a circuit level jump or interruption signal is extracted, and the trigger time is defined as a physical timestamp. During the deep packet inspection process of the original runtime data, the multi-protocol parsing and adaptation layer first parses the underlying data frame structure of the original runtime data. Because the original runtime data conforms to network communication protocol specifications, the underlying data frame structure is physically divided into a transport protocol header field and a service payload field. The transport protocol header field is the control information area at the beginning of the data frame, used to store communication addressing information, protocol control flags, and checksum sequences; the service payload field is the effective information area in the data frame that follows the transport protocol header field and carries the actual application layer service data. Based on the underlying data frame structure, the multi-protocol parsing and adaptation layer extracts the original numerical values ​​or status codes and the device unique identifier from the service payload field and the transport protocol header field, respectively. The raw numerical values ​​or status codes are extracted from the service payload field and represent the substantive monitoring content expressed by the raw operational data. In the video surveillance subsystem, this is digitally encoded and compressed video image frame data; in the environmental monitoring subsystem, it is a decimal value representing physical quantities such as temperature, humidity, or air quality; and in the security alarm subsystem, it is a Boolean logic value representing the trigger or reset status of the detector. The unique device identifier is extracted from the transmission protocol header field and is a hardware address code used to uniquely index and trace the source of data transmission in a heterogeneous network environment. For example, it may be the MAC address of the network interface controller, the IP address of the network interconnection protocol, or the device's factory-preset ID registration code.

[0067] While acquiring the physical generation timestamp, when the data packet of the original operating data finally arrives at the network interface of the control center after passing through the network transmission path, the operating system kernel of the control center will respond to the arrival interrupt of the data packet. The multi-protocol parsing and adaptation layer reads the system time of the operating system kernel clock at the moment of responding to the arrival interrupt, and defines the system time as the control center arrival timestamp. The control center arrival timestamp represents the moment when the control system logically perceives the data and reflects the end time of network transmission. In addition to extracting the physical generation timestamp and the control center arrival timestamp, the multi-protocol parsing and adaptation layer simultaneously parses the service payload field and transmission protocol header field of the original operating data, and extracts the original values ​​or status codes and unique device identifiers from them.

[0068] The original numerical or status codes, unique device identifiers, and physical generation timestamps and control center arrival timestamps from the same underlying data frame structure are logically bound and encapsulated into a standardized tuple structure to obtain a two-dimensional time-series dataset. The purpose of this two-dimensional time-series dataset is to decouple and explicitly associate the business content information represented by the original numerical or status codes with the spatiotemporal attribute information represented by the physical generation timestamps and control center arrival timestamps. This allows for accurate reconstruction of the true physical sequence of events based on the physical generation timestamps, and the evaluation of network transmission status using the control center arrival timestamps and physical generation timestamps, thereby completely eliminating timing perception biases caused by network transmission uncertainties.

[0069] Step S12: Construct a lifecycle decay model for quantifying the timeliness of two-dimensional time series datasets.

[0070] In the operational environment of smart buildings, different types of raw operational data in two-dimensional time-series datasets possess drastically different physical time-sensitivity characteristics. To transform the spatiotemporal attributes contained in the two-dimensional time-series datasets into quantifiable logical decision-making criteria and address the challenge of inconsistent effectiveness decay rates across subsystems over time, a lifecycle decay model is developed. For example, status alarm data collected by the security alarm subsystem is high-frequency, rapidly changing data with effectiveness lasting only a few seconds after triggering. Directly using alarm records from 5 minutes ago for current decision-making would lead to "false linkage" of completed intrusion events. Conversely, sensor data collected by the environmental monitoring subsystem, such as indoor temperature values, is low-frequency, gradually changing data with significant thermal inertia. Even temperature readings from 5 minutes ago still have high reference value at the current moment. Discarding them due to their age would cause the system to lose control during temperature polling intervals. Without a quantitative description of these differentiated time-sensitivity characteristics, it is difficult to accurately assess the reference value of historical data at the current moment. Therefore, constructing a lifecycle decay model aims to assign dynamic lifecycle attributes to each data element in the two-dimensional time-series dataset, providing a second dimension of evidence beyond the numerical value itself for subsequent precise logical judgments.

[0071] Specifically, the lifecycle decay model utilizes a pre-built sensor characteristic database and sets corresponding effectiveness decay coefficients for different data types in the two-dimensional time-series dataset. The sensor characteristic database is constructed through statistical analysis of massive amounts of historical raw operational data. Specifically, continuous time-series raw operational data from the built-in sensors of each subsystem are collected within a preset period, and autocorrelation is calculated. The preset period is based on a typical operating cycle covering the entire smart building to ensure the statistical sample is sufficiently representative. For example, the preset period is set to 30 calendar days. The formula for calculating autocorrelation is: Where N represents the total number of raw values ​​or state codes contained in the continuous time series raw operating data participating in the calculation, which is obtained by multiplying the sampling frequency of the built-in sensors of each subsystem with the preset period, and X represents the continuous time series raw operating data. This represents the original value or state code at the i-th sampling time in the raw running data of a continuous time series. This represents the original value or state code at the (i+k)th sampling time in the raw running data of the continuous time series, where k is the time lag step size, representing the time interval between two sampling times. This represents the arithmetic mean of all raw values ​​or state codes in the continuous time series raw operating data, used to eliminate the DC component offset of the data. The physical meaning of the calculation formula lies in quantifying the similarity between data collected by the same sensor at different time points.

[0072] The time span required for the correlation to decay to a preset threshold is extracted, and the reciprocal of this time span is defined as the baseline decay rate. The preset threshold is set based on the definition of the time constant for signal decorrelation in physics. For example, setting the preset threshold to 0.368 indicates that the signal correlation has decayed to 36.8% of its initial state. Based on the baseline attenuation rate, an effectiveness attenuation coefficient A is assigned to the raw operating data collected by each subsystem. Specifically, for high-frequency, abrupt state alarm data, due to the rapid state change, the autocorrelation drops below the threshold in a short time, resulting in a larger calculated baseline attenuation rate, which is assigned a larger effectiveness attenuation coefficient. For example, for infrared detection data with an effective window period of approximately 2 seconds, the effectiveness attenuation coefficient is set to 0.5. For low-frequency, gradual sensing data, due to the large inertia of physical quantities and the long autocorrelation duration, the calculated baseline attenuation rate is smaller, which is assigned a smaller effectiveness attenuation coefficient. For example, for temperature data with a thermal inertia time constant of approximately 1250 seconds, the effectiveness attenuation coefficient is set to 0.0008. For streaming media data, due to dynamic scene changes causing the image content correlation to decay at a moderate rate over time, the calculated baseline attenuation rate is between the above two values, and a medium-value effectiveness attenuation coefficient is assigned. For example, the effectiveness attenuation coefficient is set to 0.05.

[0073] In the real-time processing phase, the lifecycle degradation model is constructed by first retrieving the pre-built sensor characteristic database based on the unique device identifier in the two-dimensional time-series dataset, matching the corresponding effectiveness degradation coefficient, and calculating the confidence level W(t) for each data element in the two-dimensional time-series dataset. The formula for calculating the confidence level is: ,in, Indicates the current moment. The term represents the physically generated timestamp, and e is a natural constant. The reason for constructing the calculation formula is that the dissipation effect of physical state information in nature usually follows an exponential decay law, and the exponential function has the mathematical characteristics of memorylessness and time translation invariance, which can smoothly and monotonically describe the nonlinear decline process of data value over time. Through calculation, the lifecycle decay model adds a dynamically changing confidence level to each original value or state code in the two-dimensional time series dataset. The lifecycle decay model establishes a continuous, probability-based data validity evaluation mechanism.

[0074] Step S13: Expand and reconstruct the two-dimensional time series dataset and the life cycle decay model to obtain a standardized data sequence for data closed-loop management.

[0075] In order to logically integrate the trust level of the two-dimensional time series dataset with the life cycle decay model and construct the smallest information unit that can be uniformly identified and processed by the subsequent rule engine, we expand and reconstruct to generate a standardized data sequence. The purpose is to aggregate all attributes describing the same event into an indivisible standardized data sequence, thereby realizing closed-loop management of the data's state throughout its entire life cycle.

[0076] The process of generating a standardized data sequence involves expanding and reconstructing each data element in a two-dimensional time-series dataset. For example... Figure 3 Specifically, the process involves reading the original numerical values ​​or state codes, device unique identifiers, physical generation timestamps, and control center arrival timestamps from a two-dimensional time-series dataset. Simultaneously, the validity decay coefficient of the lifecycle decay model is invoked. Incremental calculations are performed to calculate network transmission delay, which is equal to the control center arrival timestamp minus the physical generation timestamp. This network transmission delay quantifies the degree of data lag in the network. The network transmission delay is compared with a preset delay tolerance threshold, which is set based on the maximum allowable response time and maximum network jitter time explicitly defined in the smart building control system. The purpose is to ensure that only data arriving within an acceptable time range is considered fully valid; for example, this threshold is set to 3000. If the network transmission delay exceeds the preset delay tolerance threshold, a severe lag is determined in the original numerical values ​​or state codes. A penalty correction operation is performed during encapsulation, multiplying the trust level corresponding to the original numerical values ​​or state codes by the penalty factor. The system obtains the corrected trust level and adds a Boolean-type hysteresis suspicious status flag, setting the value of the hysteresis suspicious status flag to True to explicitly warn subsequent processing modules that the original value or status code may be severely distorted. Here, the penalty factor is a conversion coefficient between 0 and 1, used to proactively reduce the trust weight of the data when severe network congestion occurs. The basis for setting it is that network congestion is often accompanied by packet loss and retransmission, which leads to a non-linear increase in the risk of data packet integrity and timing consistency. Therefore, it is necessary to prevent severely delayed data from dominating key decisions. For example, it is set to 0.5. If the network transmission delay is less than or equal to the preset delay tolerance threshold, the original trust level remains unchanged, and the hysteresis suspicious status flag is set to False.

[0077] Through encapsulation, a standardized data sequence is generated, which includes the following core fields: original numerical value or status code, unique device identifier, physical generation timestamp, control center arrival timestamp, validity decay coefficient, trust level, and hysteresis status marker. Eliminating the heterogeneity of the underlying hardware protocol, the standardized data sequence provides a full-dimensional information carrier, enabling data to possess self-description and self-evaluation capabilities before entering complex logical judgment processes. This provides an objective basis for adjudication when handling data conflicts, ensuring the scientific and accurate nature of decision-making.

[0078] Step S10 addresses the technical challenges of timing and logic errors caused by transmission path congestion and random interference in heterogeneous network environments of smart buildings, as well as the difficulty in quantifying the value of historical data due to inconsistent physical timeliness decay rates of data from different subsystems. This is achieved through a two-dimensional time-series dataset, a lifecycle decay model, and standardized data sequences, enabling end-to-end spatiotemporal anchoring of raw operational data from physical generation to logical reception, and closed-loop management based on trustworthiness. Specifically, the two-dimensional time-series dataset establishes a dual reference framework of physical generation timestamps and control center arrival times, eliminating cognitive biases caused by network transmission uncertainties. The lifecycle decay model uses autocorrelation analysis to assign dynamic trust attributes to static values, providing a continuous evaluation mechanism for data validity. The standardized data sequences, as a full-dimensional information carrier, decouple and explicitly correlate business content with spatiotemporal attributes and hysteresis states, providing a self-evaluation basis for subsequent accurate decision-making.

[0079] Step S20: Construct a backtracking buffer pool based on the standardized data sequence to restore real events in the physical world. Perform reverse interpolation and hold operations on the backtracking buffer pool to generate virtual data points that can fill gaps. Based on the virtual data points, perform discretization segmentation on the backtracking buffer pool to generate a logically complete time-aligned event slice sequence.

[0080] Further, step S20 includes:

[0081] Step S21: Construct a backtracking buffer pool based on the standardized data sequence to reconstruct real events in the physical world.

[0082] To transform the physical generation timestamps carried by standardized data sequences into a basis for actual temporal arrangement and to achieve a logical reconstruction of real-world event sequences, a backtracking buffer is constructed. In the operating environment of a smart building, although standardized data sequences possess complete spatiotemporal attributes and trust descriptions, they still maintain a network arrival order rather than a physical occurrence order during streaming transmission. The purpose of constructing the backtracking buffer is to provide a dynamic reorganization space with a time dimension, using the physical generation timestamp as a unique sorting index to rearrange discretely arriving standardized data sequences into an ordered queue that conforms to the real causal relationships of the physical world.

[0083] Specifically, a contiguous physical memory space is allocated in the random access memory of the control center. This physical memory space is used to construct a backtracking buffer pool, with a circular buffer serving as the underlying physical storage structure. The contiguous physical memory space is divided into several fixed-size physical memory units, each corresponding to a fixed time granularity. The time granularity is the minimum resolution of the backtracking buffer pool on the time axis, set to balance memory usage with the precision of timing reconstruction; for example, it is set to 10. The circular buffer is configured as a circular storage container with a fixed time span. Through a logical address space that connects the beginning and end, it maintains a dynamic storage interval that progresses synchronously with the current system time. The backtracking buffer pool uses the current system time as the endpoint of the time axis, covering forward to the current system time minus the buffer window length, thus forming a continuous logical time axis within the circular buffer. The buffer window length is set based on the sum of the maximum network jitter time and the minimum historical backtracking time required by the business logic, ensuring that even severely delayed original values ​​or status codes fall within the effective window; for example, it is set to 5000.

[0084] During the operation of the backtracking buffer pool, standardized data sequences are continuously received, and time-series repositioning mapping operations are performed. Specifically, the physical generation timestamp is extracted from the standardized data sequence; the physical generation timestamp is compared with the current system time to determine whether it falls within the effective range of the backtracking buffer pool, i.e., the judgment condition is: whether the physical generation timestamp is greater than the current system time minus the buffer window length and less than or equal to the current system time. If the judgment condition is met, it means that the physical generation timestamp falls within the effective range; the storage index address of the standardized data sequence in the circular buffer is calculated based on the physical generation timestamp. Specifically, the time difference between the current system time and the physical generation timestamp is calculated, the time difference is divided by the time granularity and rounded down to obtain the logical offset. The logical offset represents the number of lag slots of the standardized data sequence relative to the reference memory address corresponding to the current system time. The logical offset is subtracted from the reference memory address corresponding to the current system time, and a modulo operation is performed on the total capacity of the circular buffer to obtain the physical memory unit address, which is used as the storage index address. The standardized data sequence is then precisely inserted into the physical memory unit pointed to by the calculated storage index address.

[0085] The insertion process utilizes the random addressing characteristic of the circular buffer. As long as the physical generation timestamp of a later-arriving normalized data sequence is earlier, it is allowed to be inserted into a storage location preceding the already arrived normalized data sequence. The backtracking buffer, leveraging the physical characteristics of the circular buffer, dynamically maintains an ordered timeline generated by rearranging normalized data sequences in the random access memory. This achieves a logical transformation of the data stream from network order to physical order, overcoming the uncertainties of network transmission and reconstructing the true chronological order of events. This ensures that subsequent logical deductions can be based on genuine causal relationships, rather than being limited by the order of data arrival.

[0086] Step S22: Perform reverse interpolation and hold operations on the backtracking buffer pool to generate virtual data points that can fill gaps.

[0087] To address the sparse data distribution in the backtracking buffer caused by differing acquisition mechanisms across subsystems—for example, high-frame-rate data from the video surveillance subsystem almost fills every physical memory cell, while low-frequency polling data from the environmental monitoring subsystem occupies only sparse physical memory cells—resulting in data gaps across numerous time granularities. Directly performing logical operations based on this sparse, discrete data leads to rule matching failures at these gaps. To eliminate the logical state fragmentation caused by the aliasing of sampling frequencies from built-in sensors in each subsystem, inverse interpolation and hold operations are performed. This leverages the time-sensitivity of low-frequency data to logically extend its effective time window, filling the empty physical memory cells in the backtracking buffer and thus obtaining a full state view at any time slice.

[0088] Specifically, each physical memory unit in the backtracking buffer is traversed and filled. Starting from the base memory address corresponding to the current system time, the physical memory units in the backtracking buffer are traversed one by one in descending order of time. For each known device unique identifier, it is checked whether a standardized data sequence containing the device unique identifier exists in the currently traversed physical memory unit. If the current physical memory unit is missing a standardized data sequence containing the device unique identifier, the backtracking continues in descending order of time until the nearest physical memory unit containing a standardized data sequence containing the device unique identifier is found. The standardized data sequence in the physical memory unit is defined as the source data point. The original value or status code, physical generation timestamp, and validity decay coefficient encapsulated in the source data point are extracted.

[0089] A virtual data point is generated based on the source data point, and this virtual data point is then filled into the physical memory unit of the currently missing data. Specifically: based on the zero-order hold principle, the virtual data point inherits the original value or state code of the source data point, and the trust level of the virtual data point is recalculated. The calculation formula is: ,in, A and E represent the trust level and validity decay coefficient encapsulated in the source data points, respectively, and e is the natural constant. This represents the time interval between the physical memory unit corresponding to the currently missing data and the timestamp generated by the physical data point. The reason for this calculation formula is that the trust level of the virtual data point should be the sum of the inherent trust level of the source data point and the additional attenuation effect caused by the passage of time. By multiplying by the trust level of the source data point, it ensures that the virtual data point inherits the initial weight reduction that the source data point may suffer due to network latency, etc. This further quantifies the dissipation of data value over time. The calculation formula ensures that the confidence level of virtual data points decreases exponentially as they move away from the source data point. Before filling, the calculated confidence level of the virtual data point is compared with a preset validity truncation threshold. This validity truncation threshold is the critical confidence level baseline for determining whether data still has value for participating in logical operations. The threshold is set to balance the system's anti-interference capability and data continuity requirements, avoiding excessively high thresholds that lead to the discarding of large amounts of historical data and discontinuous control, and excessively low thresholds that introduce too much noise and increase the risk of false alarms. For example, it is set to 0.1. If the calculated confidence level of the virtual data point is less than the preset validity truncation threshold, the filling operation on the current physical memory unit is stopped, indicating that the time corresponding to the current physical memory unit has exceeded the effective radiation range of the source data point, and the data has become completely invalid.

[0090] Ultimately, the backtracking buffer generated a series of virtual data points with dynamic trust levels, successfully filling the gaps on the timeline. The reverse interpolation and hold operation smoothed out the differences in hardware sampling frequencies, enabling low-frequency updated sensor data to logically resonate with high-frequency updated streaming media data. This eliminated state tearing caused by sampling frequency aliasing, improving the continuity and reliability of cross-system linkage logic.

[0091] Step S23: Based on virtual data points, perform discretization segmentation on the backtracking buffer pool to generate a logically complete time-aligned event slice sequence.

[0092] In the operating environment of a smart building, virtual data points fill the low-frequency data gaps in the backtracking buffer, enabling the buffer to possess full state data on the timeline. However, it is essentially still a streaming storage structure based on contiguous memory addresses. Performing complex Boolean logic queries directly on the contiguous backtracking buffer results in extremely high computational complexity and makes parallel processing difficult. To transform the contiguous physical memory units containing standardized data sequences and virtual data points in the backtracking buffer into discrete logic units that the rule engine can directly process, a discretization segmentation operation is performed. This aims to slice the backtracking buffer after data filling, discretizing the continuous time stream into a series of independent time slices containing the full state of the quantum system. This transforms the complex streaming computation problem into a simple static pattern matching problem.

[0093] The discretization segmentation operation is based on the time granularity in the backtracking buffer pool. Specifically, it traverses every physical memory unit within the time interval covered by the backtracking buffer pool. This time interval refers to the period from the start time (the current system time minus the buffer window length) to the end time (the current system time). Following an increasing time order, the corresponding physical memory units are accessed sequentially at preset logical step intervals. The logical step interval is the time interval for generating the time-aligned event slice sequence, maintaining consistency with the time granularity. For example, if the time granularity is 10 milliseconds, the logical step interval is also set to 10 milliseconds. For each accessed physical memory unit, the standardized data sequence and virtual data points stored in the physical memory unit are extracted.

[0094] For each physical memory unit, a time-aligned event slice is constructed based on the standardized data sequence and virtual data point extracted. The time-aligned event slice pre-configured several state slots, each uniquely corresponding to a known device identifier. These slots store the device's runtime state information and serve as a structured data container, encapsulating a snapshot of the entire system's state at a specific moment. This snapshot represents the instantaneous operational state of each subsystem within the smart building at the same physical moment, containing the original values ​​or state codes corresponding to each device, as well as a trust level representing the data's reliability. This constitutes a frozen, information-complete logical judgment plane in the time dimension. The construction process is as follows: For the currently processed physical memory unit, each standardized data sequence and each virtual data point stored within it is traversed. For each standardized data sequence, the device identifier is extracted. Based on the device identifier, the corresponding state slot in the time-aligned event slice is located. The original values ​​or state codes encapsulated in the standardized data sequence are extracted and filled into the located state slot. The trust level encapsulated in the standardized data sequence is extracted as a state weight and filled into the located state slot. For each virtual data point encountered, the system performs the same operation. If the device unique identifier corresponding to the status slot in the timing alignment event slice has neither a corresponding normalized data sequence nor a corresponding virtual data point in the currently processed physical memory unit, then the status slot is marked as invalid, and no pending confirmation linkage event is generated.

[0095] Finally, a series of consecutive time-aligned event slices were generated in chronological order, forming a time-aligned event slice sequence. Each time-aligned event slice in the sequence represents a complete state mirror of the physical world at a specific historical moment. The asynchronously arriving, varying-frequency raw runtime data was thoroughly organized into a synchronous, structured time-aligned event slice sequence. This allows subsequent logical matching across subsystems to be performed with the precision defined by the time granularity simply by scanning each time-aligned event slice in the sequence, eliminating the need to consider the sampling time difference or network latency of the underlying raw runtime data. This greatly simplifies the calculation logic of complex linkage rules and improves matching efficiency.

[0096] Step S20 addresses the technical challenges of data streams arriving in network order rather than physical order despite possessing spatiotemporal attributes, and the sparse data distribution and logical state fragmentation caused by the aliasing of sampling frequencies across subsystems. This is achieved by using a backtracking buffer, reverse interpolation and hold operations, and a time-aligned event slice sequence. Specifically, the backtracking buffer dynamically reassembles out-of-order data streams into an ordered queue conforming to real causal relationships; the reverse interpolation and hold operations utilize the zero-order hold principle to generate virtual data points with dynamic trust, filling the temporal gaps in low-frequency data and eliminating logical discontinuities caused by frequency differences; and the time-aligned event slice sequence discretizes continuous streaming storage into independent logical units containing snapshots of the entire system's state, greatly simplifying the computational logic of complex cross-system linkage rules.

[0097] Step S30: Perform a logical replay matching operation on the time-aligned event slice sequence to obtain an event set covering real-time response and historical tracing. Based on the event set, perform state adjudication and instruction generation operations to generate adaptive collaborative control instructions.

[0098] Further, step S30 includes:

[0099] Step S31: Perform a logical replay matching operation on the time-aligned event slice sequence to obtain an event set covering real-time response and historical tracing.

[0100] The timing-aligned event slice sequence has reconstructed the actual timing state of the physical world, but the actual timing state is only statically stored in memory and has not yet triggered any linkage control. In order to transform the timing-aligned event slice sequence into actual control trigger signals and solve the problem of event missed and false alarms caused by network latency, a logic replay matching operation is performed. The purpose is to use preset cross-subsystem linkage trigger conditions to perform a full logic state evaluation of the timing-aligned event slice sequence, ensuring the reliability of safe linkage.

[0101] The preset cross-subsystem linkage triggering conditions are a set of custom-configured logical rules based on the actual business needs of smart buildings. Due to the diversity of business scenarios, the set of logical rules contains multiple judgment logics, each designed to detect specific cross-subsystem abnormal states. The cross-subsystem abnormal state refers to at least two different devices belonging to the video surveillance subsystem, environmental monitoring subsystem, or security alarm subsystem, respectively, whose operating states simultaneously meet a specific combination of risk characteristics. This combination of risk characteristics indicates that a complex security event has occurred within the building that cannot be independently confirmed by a single subsystem. Each preset cross-subsystem linkage triggering condition clearly specifies which device unique identifiers must participate in the judgment simultaneously. Each preset cross-subsystem linkage triggering condition consists of three core elements: the set of participating devices, state constraints, and logical topology. The set of participating devices explicitly lists all unique identifiers of devices participating in the logical judgment; the state constraints define mathematical inequalities or equality relationships that the original values ​​or state codes must satisfy for each unique identifier in the set of participating devices; and the logical topology defines the Boolean operation relationships between all state constraints, including logical AND, logical OR, and logical NOT.

[0102] To illustrate the triggering mechanisms in different scenarios, three specific scenarios are given below: Scenario 1 (Fire Confirmation): The triggering conditions specify the participating devices as: smoke detectors in the security alarm subsystem and infrared thermal imaging cameras in the video surveillance subsystem; the state constraints are: constraint one requires the smoke detector's state code to be 1, and constraint two requires the infrared thermal imaging camera's temperature value to be greater than 80 degrees Celsius; the logical topology is as follows: the cross-subsystem linkage triggering condition is established only when both constraint one and constraint two are true, i.e., the logical AND operation result is true. Scenario 2 (Illegal Intrusion): The triggering conditions specify the participating devices as: access control controllers in the security alarm subsystem and infrared curtain detectors in the security alarm subsystem; the state constraints include: constraint one requires the access control controller's state code to be equal to 0, and constraint two requires the infrared curtain detector's state code to be equal to 1; the logical topology stipulates that the cross-subsystem linkage triggering condition is established only when both constraint one and constraint two are true. Example Scenario 3 (Environmental Anomaly): The triggering conditions specify the set of participating devices as: PM2.5 sensor in the environmental monitoring subsystem and personnel counting camera in the video surveillance subsystem; the state constraints include: constraint one requires the PM2.5 sensor value to be greater than 150, and constraint two requires the personnel counting camera value to be greater than 10; the logical topology stipulates that the cross-subsystem linkage triggering condition is established only when both constraint one and constraint two are true.

[0103] Traverse each timing alignment event slice in the timing alignment event slice sequence. For the currently traversed timing alignment event slice, locate the corresponding state slot in the timing alignment event slice according to the set of participating devices specified in the preset cross-subsystem linkage triggering conditions. Extract the stored raw value or state code from the located state slot. Use the extracted raw value or state code as the measured value and compare it one by one with the state constraints specified in the preset cross-subsystem linkage triggering conditions. If the measured values ​​of all involved devices meet the specified state constraints, and the final Boolean value after calculation according to the logical topology structure is true, then it is determined that at the time corresponding to the timing alignment event slice, the operating state of the set of participating devices successfully triggered the preset cross-subsystem linkage triggering conditions, and a linkage event to be confirmed has occurred. For confirmed linkage events, an event classification operation is performed. Specifically, the time point corresponding to the timing-aligned event slice is read. This time point refers to the logical time point mapped by the physical memory unit corresponding to the timing-aligned event slice in the backtracking buffer pool. The logical time point is equal to the current system time minus the slot offset of the physical memory unit relative to the base memory address multiplied by the time granularity. The difference between the current system time and the time point corresponding to the timing-aligned event slice is calculated, and this difference is defined as the event lag duration. The event lag duration is compared with a preset real-time judgment threshold. The real-time judgment threshold is the time boundary that distinguishes between real-time control and historical tracing; for example, it is set to 100. If the event lag duration is less than or equal to the preset real-time judgment threshold, the linkage event to be confirmed is marked as a real-time event. If the event lag duration is greater than the preset real-time judgment threshold, the linkage event to be confirmed is marked as a delayed confirmation event. The real-time events and delayed confirmation events are combined to obtain an event set.

[0104] Step S32: Perform state adjudication and instruction generation operations based on the event set to generate adaptive collaborative control instructions.

[0105] In the operating environment of a smart building, the event set may contain multiple events with contradictory state descriptions for the same physical scenario. For example, the video surveillance subsystem might determine there is no fire, while the security alarm subsystem might determine there is. Without resolving these contradictions, the control system cannot determine which linkage strategy to execute. To transform the event set into precise physical control actions, and to resolve the decision-making inconsistencies caused by conflicts in data collected from various subsystems, as well as the control strategy failures due to event delays, state adjudication and command generation operations are implemented. The aim is to establish a game-theoretic mechanism based on objective data quality to filter out valid states from contradictory real-time events or delayed confirmation events, and to generate control commands that can adapt to time deviations caused by network latency. This breaks down logical barriers between subsystems and enables cross-system automated collaboration.

[0106] The execution state adjudication and instruction generation operations utilize a trust-based game to adjudicate the state and generate adaptive collaborative control instructions. Specifically, it iterates through each real-time event and delayed confirmation event in the event set. When multiple real-time events or delayed confirmation events with mutually exclusive states, involving the same preset cross-subsystem linkage triggering condition, are detected at the same time point corresponding to the time slice of the time sequence aligned event, a trust-based game is initiated for state adjudication. The trust level contained in each mutually exclusive real-time event or delayed confirmation event is extracted. The trust levels of each mutually exclusive event are numerically compared, and the real-time event or delayed confirmation event with the highest trust level is determined to be the winning event. If multiple events have the same trust level, the event representing the higher risk level is determined to be the winning event according to a preset risk-oriented principle. The winning event is confirmed as the valid state at the time point corresponding to the time slice of the time sequence aligned event. The risk-oriented principle refers to prioritizing events that represent abnormal states using raw numerical values ​​or state codes when trust levels are equal. Specifically, it compares the raw numerical values ​​or state codes encapsulated in mutually exclusive events, determining whether the raw numerical value or state code indicates a trigger state. Events with a state code of 1 or a value greater than a threshold are considered winning events, while those indicating a reset state are considered losing events. For example, events with a state code of 0 or a value less than a threshold are considered failing events. Winning events are confirmed as the valid system state at the time point corresponding to the time-series aligned event slice.

[0107] Based on the valid state, the execution instruction generation operation is performed. A pre-defined linkage execution strategy corresponds one-to-one with each pre-defined cross-subsystem linkage trigger condition. The linkage execution strategy represents the execution device identifier, action type, and preset control duration to be invoked when the cross-subsystem linkage trigger condition is met. The execution device identifier is a hardware address code used to uniquely index the execution mechanism in the smart building's control network; the action type defines the specific physical operation that the execution mechanism needs to perform, including turning on, turning off, adjusting parameters, or switching modes. The setting of the execution device identifier and action type is based on the physical device deployment topology and security management linkage specifications of the video surveillance subsystem, environmental monitoring subsystem, and security alarm subsystem within the smart building; the preset control duration refers to the physical length of time the control action should be continuously executed. The linkage execution strategy corresponding to the valid state is retrieved. For real-time events in the event set, since their event lag is extremely short, they can be considered to occur simultaneously with the system processing time. Based on the execution device identifier and action type defined in the linkage execution strategy, an immediate adaptive collaborative control instruction is generated. For delayed confirmation events, since these events have physically occurred for some time, to ensure the physical execution window of the control action aligns with the effective lifecycle of the logical event, the adaptive collaborative control instruction is a time compensation correction for the execution control parameters. Specifically, the calculated event lag is read, and the event lag is subtracted from the preset control duration defined in the linkage execution strategy to obtain the compensated control duration. If the event lag is greater than or equal to the preset control duration, the control action is deemed invalid, and no instruction is generated. Finally, the adaptive collaborative control instruction is sent to the execution ends of the video surveillance subsystem, environmental monitoring subsystem, and security alarm subsystem.

[0108] Step S30, through logical replay matching, event set, state adjudication, and instruction generation operations, solves the technical challenges of missed and false alarms of cross-subsystem events caused by network latency, and decision-making instability and control strategy failure caused by multi-source data state conflicts. It achieves accurate state assessment based on objective data quality and adaptive physical control capable of withstanding transmission latency. Specifically, the logical replay matching operation utilizes a complete logical topology structure to fully evaluate historical and real-time states, ensuring reliable triggering of secure linkages; the event set covers various risk characteristics requiring real-time response and historical tracing; the state adjudication and instruction generation operations use a trust-based game mechanism to select winning events from contradictory states and perform time compensation correction on control parameters based on event lag duration, ensuring precise alignment between the physical execution window of control actions and the effective lifecycle of logical events.

[0109] Example 2

[0110] This embodiment, based on Embodiment 1, provides a cross-subsystem event-driven linkage control system for smart buildings, such as... Figure 4 As shown, it includes:

[0111] Spatiotemporal quantization module: used to acquire raw running data, perform deep detection and time series extraction on the raw running data to obtain a two-dimensional time series dataset, construct a life cycle decay model for quantifying the timeliness of the two-dimensional time series dataset, expand and reconstruct the two-dimensional time series dataset and the life cycle decay model to obtain a standardized data sequence for data closed-loop management;

[0112] The global restoration module is used to construct a backtracking buffer pool based on standardized data sequences, restore real events in the physical world, perform inverse interpolation and hold operations on the backtracking buffer pool to generate virtual data points that can fill gaps, and perform discretization segmentation on the backtracking buffer pool based on the virtual data points to generate a logically complete time-aligned event slice sequence.

[0113] Cross-system adjudication module: Used to perform logical replay matching operations on time-aligned event slice sequences to obtain an event set covering real-time response and historical tracing. Based on the event set, state adjudication and instruction generation operations are performed to generate adaptive collaborative control instructions.

[0114] In the spatiotemporal quantization module, the process involves acquiring raw operational data, performing deep detection and time-series extraction on the raw operational data to obtain a two-dimensional time-series dataset, constructing a lifecycle decay model for quantifying the timeliness of the two-dimensional time-series dataset, and expanding and reconstructing the two-dimensional time-series dataset and the lifecycle decay model to obtain a standardized data sequence for data closed-loop management, including:

[0115] Step S11: Obtain the raw running data, perform deep detection and time series extraction on the raw running data to obtain a two-dimensional time series dataset;

[0116] Step S12: Construct a lifecycle decay model for quantifying the timeliness of two-dimensional time series datasets;

[0117] Step S13: Expand and reconstruct the two-dimensional time series dataset and the life cycle decay model to obtain a standardized data sequence for data closed-loop management.

[0118] In the global restoration module, a backtracking buffer pool is constructed based on standardized data sequences to restore real events in the physical world. Inverse interpolation and hold operations are performed on the backtracking buffer pool to generate virtual data points that can fill gaps. Based on these virtual data points, the backtracking buffer pool is discretized to generate a logically complete time-aligned event slice sequence, including:

[0119] Step S21: Construct a backtracking buffer pool based on standardized data sequences to reconstruct real events in the physical world;

[0120] Step S22: Perform reverse interpolation and hold operations on the backtracking buffer pool to generate virtual data points that can fill gaps;

[0121] Step S23: Based on virtual data points, perform discretization segmentation on the backtracking buffer pool to generate a logically complete time-aligned event slice sequence.

[0122] In the cross-system adjudication module, the logical replay matching operation is performed on the time-aligned event slice sequence to obtain an event set covering real-time response and historical tracing. Based on the event set, state adjudication and instruction generation operations are performed to generate adaptive collaborative control instructions, including:

[0123] Step S31: Perform a logical replay matching operation on the time-aligned event slice sequence to obtain an event set covering real-time response and historical tracing;

[0124] Step S32: Perform state adjudication and instruction generation operations based on the event set to generate adaptive collaborative control instructions.

[0125] The methods and systems of this application may be implemented in many ways. For example, they may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above-described order of steps for the method is for illustrative purposes only, and the steps of the method of this application are not limited to the order specifically described above, unless otherwise specifically stated.

[0126] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0127] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A cross-subsystem event-driven linkage control method for smart buildings, characterized in that, The method includes: The raw running data is acquired, and deep detection and time series extraction are performed on the raw running data to obtain a two-dimensional time series dataset. A life cycle decay model for quantifying the timeliness of data is constructed for the two-dimensional time series dataset. The two-dimensional time series dataset and the life cycle decay model are expanded and reconstructed to obtain a standardized data sequence for data closed-loop management. The lifecycle decay model collects continuous time series raw operating data within a preset period, calculates autocorrelation, extracts the time span required for autocorrelation to decay to a preset threshold, defines the reciprocal of the time span as the benchmark decay rate, sets the effectiveness decay coefficient based on the benchmark decay rate for the raw operating data, and obtains a pre-constructed sensor characteristic database. Based on the device's unique identifier, a pre-built sensor characteristic database is retrieved to match the effectiveness attenuation coefficient; A calculation formula is constructed based on the time difference between the current moment and the physically generated timestamp to calculate the trust level of each data element in the two-dimensional time series dataset; A backtracking buffer is constructed based on standardized data sequences. The purpose of constructing the backtracking buffer is to provide a dynamic reorganization space with a time dimension. Using physically generated timestamps as unique sorting indexes, the discretely arriving standardized data sequences are rearranged into an ordered queue that conforms to the real causal relationship in the physical world. Inverse interpolation and hold operations are performed on the backtracking buffer to generate virtual data points that can fill gaps. Based on the virtual data points, the backtracking buffer is discretized and segmented to generate a logically complete time-aligned event slice sequence. Perform logical replay matching operation on the time-aligned event slice sequence to obtain an event set covering real-time response and historical tracing. Based on the event set, perform state adjudication and instruction generation operations to generate adaptive collaborative control instructions. The method for obtaining the two-dimensional time-series dataset includes: parsing the underlying data frame structure of the original running data, and extracting the original numerical values ​​or status codes and the unique device identifier from the underlying data frame structure. The raw operational data includes three data elements: streaming media data, sensor data, and status alarm data. For streaming media data, the absolute time code written by the camera hardware encoder is extracted as the physical timestamp; for sensor data, the local clock time of the gateway is extracted as the physical timestamp; for status alarm data, the trigger time recorded by the alarm controller is extracted as the physical timestamp. When the raw operating data arrives at the network interface of the control center, the system time of the control center is read as the arrival timestamp of the control center. The original numerical values ​​or status codes, unique device identifiers, physical generation timestamps, and control center arrival timestamps from the same underlying data frame structure are logically bound to obtain the two-dimensional time-series dataset.

2. The cross-subsystem event-driven linkage control method for smart buildings as described in claim 1, characterized in that, The method for obtaining the standardized data sequence includes: The network transmission delay is obtained by calculating the difference between the arrival timestamp of the control center and the physical generation timestamp, and then compared with a preset delay tolerance threshold. If the network transmission delay is greater than the preset delay tolerance threshold, the trust level is multiplied by the preset penalty factor to obtain the corrected trust level, and a hysteresis suspicious state flag is added and set to True; otherwise, the trust level remains unchanged and the hysteresis suspicious state flag is set to False. The original numerical value or status code, the device's unique identifier, the physical generation timestamp, the control center's arrival timestamp, the validity decay coefficient, the corrected trust level, and the hysteresis status mark are encapsulated into a standardized data sequence.

3. The cross-subsystem event-driven linkage control method for smart buildings as described in claim 2, characterized in that, The backtracking buffer pool includes: Physical memory space is allocated in the random access memory of the control center. This physical memory space is divided into several fixed-size physical memory units, each corresponding to a fixed time granularity. The physical memory is then utilized... The backtracking buffer pool is constructed in space, and a circular buffer is used as the underlying physical storage structure of the backtracking buffer pool; Starting from the current system time, it extends backward to the current system time minus the preset buffer window length, forming a continuous logical timeline; The effective range of the backtracking buffer is determined based on the logical timeline, and it is determined whether the physically generated timestamps in the standardized data sequence fall within the effective range of the backtracking buffer. If it falls within the effective range, calculate the time difference between the current system time and the physically generated timestamp, and calculate the storage index address in combination with the time granularity; Based on the storage index address, the standardized data sequence is inserted into the physical memory unit.

4. The cross-subsystem event-driven linkage control method for smart buildings as described in claim 3, characterized in that, The method for acquiring the virtual data points includes: Traverse the physical memory units in the backtracking buffer in descending order of time. For each known device unique identifier, check whether there is a standardized data sequence containing the device unique identifier in the currently traversed physical memory unit. If the current physical memory cell lacks a standardized data sequence with a unique device identifier, backtrack backward until a physical memory cell containing a standardized data sequence with a unique device identifier is found, and that physical memory cell is used as the source data point. A virtual data point is generated based on the source data point. The virtual data point inherits the original value or state code of the source data point. The trust level of the virtual data point is obtained through the source data point construction formula, and the virtual data point is filled into the physical memory unit. Filling is not performed only when the trust level of the virtual data point is less than the preset validity truncation threshold.

5. The cross-subsystem event-driven linkage control method for smart buildings as described in claim 4, characterized in that, The time-aligned event slice sequence includes: Access the physical memory units of the backtracking buffer pool in ascending order of time, with a preset logical step size as the interval; Construct a timing-aligned event slice containing several state slots, where each state slot corresponds to a known unique device identifier; Extract standardized data sequences and virtual data points from physical memory units, fill the original values ​​or status codes into the corresponding status slots as status values, and fill the trust level into the corresponding status slots as status weights. If the device unique identifier corresponding to the state slot in the timing alignment event slice has neither a corresponding normalized data sequence nor a corresponding virtual data point in the currently processed physical memory unit, then the state slot is marked as invalid. Finally, a series of consecutive timing alignment event slices are generated in chronological order, forming a timing alignment event slice sequence.

6. The cross-subsystem event-driven linkage control method for smart buildings as described in claim 5, characterized in that, The event set includes: Traverse the timing-aligned event slices and extract the original values ​​or status codes from the status slots as measured values ​​based on the preset cross-subsystem linkage trigger conditions. The measured values ​​are substituted into the state constraints and logical topology defined by the preset cross-subsystem linkage triggering conditions for calculation. If the state constraints are met and the logical topology operation is true, it is determined that a linkage event to be confirmed has occurred. Calculate event lag duration by aligning event slices with time sequence; If the delay time of an event is less than or equal to the preset real-time determination threshold, it is marked as a real-time event; otherwise, it is marked as a delayed confirmation event. The real-time event and the delayed confirmation event are combined to obtain the event set.

7. The cross-subsystem event-driven linkage control method for smart buildings as described in claim 6, characterized in that, The method for obtaining the adaptive cooperative control command includes: For real-time events, each real-time event and delayed confirmation event in the event set is traversed. When multiple real-time events or delayed confirmation events involving the same preset cross-subsystem linkage triggering condition but with mutually exclusive states are detected, the trust level is compared, and the real-time event or delayed confirmation event with the highest trust level value is determined as the winning event. If multiple events have the same trust level value, the state code encapsulated in the mutually exclusive events is compared, and the event whose preset state code indicates the triggering state is determined as the winning event. The winning event is then confirmed as a valid state. By using preset cross-subsystem linkage triggering conditions, the linkage execution strategy corresponding to the effective state is retrieved. The linkage execution strategy indicates the execution device identifier, action type, and preset control duration to be invoked when the cross-subsystem linkage triggering conditions are met. By employing a coordinated execution strategy, instruction generation operations are performed on both real-time events and delayed acknowledgment events within the event set to obtain adaptive collaborative control instructions.

8. A cross-subsystem event-driven linkage control system for smart buildings, characterized in that, The system includes: Spatiotemporal quantization module: used to acquire raw running data, perform deep detection and time series extraction on the raw running data to obtain a two-dimensional time series dataset, construct a life cycle decay model for quantifying the timeliness of the two-dimensional time series dataset, expand and reconstruct the two-dimensional time series dataset and the life cycle decay model to obtain a standardized data sequence for data closed-loop management; The lifecycle decay model collects continuous time series raw operating data within a preset period, calculates autocorrelation, extracts the time span required for autocorrelation to decay to a preset threshold, defines the reciprocal of the time span as the benchmark decay rate, sets the effectiveness decay coefficient based on the benchmark decay rate for the raw operating data, and obtains a pre-constructed sensor characteristic database. Based on the device's unique identifier, a pre-built sensor characteristic database is retrieved to match the effectiveness attenuation coefficient; A calculation formula is constructed based on the time difference between the current moment and the physically generated timestamp to calculate the trust level of each data element in the two-dimensional time series dataset; The method for obtaining the two-dimensional time-series dataset includes: parsing the underlying data frame structure of the original running data, and extracting the original numerical values ​​or status codes and the unique device identifier from the underlying data frame structure. The raw operational data includes three data elements: streaming media data, sensor data, and status alarm data. For streaming media data, the absolute time code written by the camera hardware encoder is extracted as the physical timestamp; for sensor data, the local clock time of the gateway is extracted as the physical timestamp; for status alarm data, the trigger time recorded by the alarm controller is extracted as the physical timestamp. When the raw operating data arrives at the network interface of the control center, the system time of the control center is read as the arrival timestamp of the control center. The original numerical or status codes, unique device identifiers, physical generation timestamps, and control center arrival timestamps from the same underlying data frame structure are logically bound together to obtain the two-dimensional time-series dataset. The global restoration module is used to construct a backtracking buffer pool based on standardized data sequences. The purpose of constructing the backtracking buffer pool is to provide a dynamic reorganization space with a time dimension. Using physically generated timestamps as unique sorting indexes, the discrete standardized data sequences are rearranged into an ordered queue that conforms to the real causal relationship in the physical world. Inverse interpolation and hold operations are performed on the backtracking buffer pool to generate virtual data points that can fill gaps. Based on the virtual data points, the backtracking buffer pool is discretized and segmented to generate a logically complete time-aligned event slice sequence. Cross-system adjudication module: Used to perform logical replay matching operations on time-aligned event slice sequences to obtain an event set covering real-time response and historical tracing. Based on the event set, state adjudication and instruction generation operations are performed to generate adaptive collaborative control instructions.

Images