AI artificial intelligence machine learning system and intelligent gateway

Abstract

The invention provides an AI artificial intelligence machine learning system and an intelligent gateway, and relates to the technical field of AI artificial intelligence control, and the system comprises a communication module which is in bidirectional connection with an air conditioning unit, a security and protection sensor array and an intelligent illumination network in a building through a plurality of Internet of Things protocols, collects equipment operation data flow in real time, and sends the data flow to a server; comprising an air conditioner compressor current waveform, a security alarm code and lighting equipment light intensity time sequence data; and the edge calculation module is used for extracting an air conditioner energy consumption characteristic value, a security alarm signal and an environment temperature and humidity data set according to the equipment operation data flow, and storing the processed structured data set to a simulation object memory area through a simulation object classification algorithm. Through cooperation of multiple modules, real-time acquisition and processing of multi-source equipment data, analysis of personnel and environment information, intelligent generation of control instructions and adaptation to various equipment through protocol conversion are realized, so that the building environment comfort, safety and energy utilization efficiency are improved, and system compatibility and emergency response capability are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of AI artificial intelligence control, and particularly to an AI artificial intelligence machine learning system and an intelligent gateway. Background Art

[0002] Traditional Internet of Things gateways face multiple bottlenecks in building automation scenarios. The traditional architecture adopts a time-sharing multiplexing mode and relies on a centralized data center to collect and store data, resulting in delays in data collection from building devices (such as air-conditioning units and security sensors) to the issuance of control instructions.

[0003] For example, the air-conditioning system needs to dynamically adjust the temperature according to the personnel density, but the delay in the traditional link causes a lag in temperature response, affecting comfort and energy efficiency optimization; after a smoke alarm is triggered in the security system, due to the delay in cloud data center processing, the response speed of access control linkage cannot meet the requirements of building safety codes.

[0004] In addition, heterogeneous data such as air-conditioning current waveforms and security alarm codes cannot be fused and analyzed in real time. Digital twins rely on the transmission of all data, exacerbating network congestion in the building; device-level features (such as air-conditioning energy efficiency ratio and personnel thermal distribution) cannot be dynamically extracted, resulting in a disconnection between control instructions and the building environment status; sensitive data (such as video surveillance streams) need to be uploaded to the cloud for processing, increasing the risk of privacy leakage, and high bandwidth occupancy restricts energy efficiency optimization. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide an AI artificial intelligence machine learning system and an intelligent gateway, which achieve protocol adaptive conversion, efficient data compression, and real-time decision-making at the edge.

[0006] To solve the above technical problem, the technical solution of the present invention is as follows:

[0007] In a first aspect, an AI artificial intelligence machine learning system includes:

[0008] A communication module that establishes a two-way connection with air-conditioning units, security sensor arrays, and intelligent lighting networks in a building through multiple Internet of Things protocols, and real-time collects device operation data streams, including air-conditioning compressor current waveforms, security alarm codes, and lighting device light intensity time-series data;

[0009] An edge computing module for extracting air-conditioning energy consumption characteristic values, security alarm signals, and environmental temperature and humidity data sets according to the device operation data streams, and storing the processed structured data set into a virtual memory area through a virtual classification algorithm;

[0010] A virtual memory area module for performing timestamp difference encoding on the air-conditioning historical power data in the structured data set to form a spatio-temporal correlation data cube;

[0011] An engine module, which is used to identify the heat map pattern of personnel movement in the building according to the spatio-temporal association data cube in the iconic memory area and generate a dynamic temperature control parameter matrix;

[0012] A control logic module, which is used to generate control instructions according to environmental comfort indicators, personnel density area distribution, abnormal behavior pattern recognition, and security threat level; when receiving an alarm signal, according to the signal type, dynamically adjust the focus area of video monitoring and the access control strategy through a three-level priority response mechanism;

[0013] A protocol conversion module, which is used to dynamically convert control instructions into the target device protocol format.

[0014] Furthermore, according to the device operation data stream, extract the air-conditioning energy consumption characteristic values, security alarm signals, and environmental temperature and humidity data sets, and store the processed structured data set in the iconic memory area through an iconic classification algorithm, including:

[0015] Perform a sliding window process on the original device data stream, and divide the continuous data stream into multiple equal-length data segments;

[0016] Extract the air-conditioning energy efficiency characteristic set from multiple equal-length data segments, including real-time energy efficiency ratio, current harmonic distortion rate, and load fluctuation index, and generate an alarm frequency density matrix based on security sensor data to obtain a characteristic data set containing air-conditioning energy efficiency characteristics and spatio-temporal distribution information of security events;

[0017] Generate device-level classification labels for the characteristic data set through a device fingerprint coding algorithm, and map the classified data to the iconic memory area according to the device entity ID.

[0018] Furthermore, perform timestamp difference coding on the air-conditioning historical power data in the structured data set to form a spatio-temporal association data cube, including:

[0019] Extract the air-conditioning historical power data from the structured data set, group it by device ID and arrange it according to the timestamp to form a time-series power sequence;

[0020] For each adjacent pair of timestamps, calculate the power difference and time interval to form a power difference coding sequence;

[0021] Associate the difference coding sequence with the device geographical coordinates to construct a three-dimensional data cube including time dimension, space dimension, and feature dimension.

[0022] Furthermore, according to the spatio-temporal association data cube in the iconic memory area, identify the heat map pattern of personnel movement in the building and generate a dynamic temperature control parameter matrix, including:

[0023] Extract multi-source probe location data from the spatio-temporal correlation data cube in the iconic memory area, and use a long short-term memory network to process the probe location data to predict the personnel residence probability in each area in the future for a period of time, and generate a dynamic heat map;

[0024] Perform spatial overlay analysis on the heat map and the temperature and humidity gradient data stored in the iconic memory area to identify crowded areas and areas with abnormal environmental parameters, and obtain the coupling analysis result;

[0025] According to the coupling analysis result, combined with the heat response characteristics of the air conditioning unit, dynamically calculate the matrix of regional temperature set values.

[0026] Furthermore, extract probe location data from the spatio-temporal correlation data cube in the iconic memory area, and use a long short-term memory network to process the probe location data to predict the personnel residence probability in each area in the future for a period of time, and generate a dynamic heat map, including:

[0027] Extract multi-source probe location data from the spatio-temporal correlation data cube in the iconic memory area, including probe connection events, Bluetooth beacon scan records, and video analysis coordinate point clouds;

[0028] Group the multi-source probe location data by personnel ID and construct the spatio-temporal sequence of individual movement trajectories, and divide them into continuous trajectory segments through a sliding window;

[0029] Input the trajectory segments into a pre-trained long short-term memory network, and predict the personnel residence probability distribution in each area in the future target period according to the historical movement rules;

[0030] Optimize the spatial interpolation of the residence probability distribution, and generate a dynamic heat map in combination with the building topology structure.

[0031] Furthermore, according to the coupling analysis result, combined with the heat response characteristics of the air conditioning unit, dynamically calculate the matrix of regional temperature set values, including:

[0032] Determine the dynamic reference temperature according to the historical temperature data in the building, and compare the personnel residence probability in each area in the future target period with the actual personnel residence probability at the current time point to obtain the personnel density change rate;

[0033] According to the heat response characteristics of the air conditioning equipment, including response speed and adjustment range, and historical energy efficiency data, including energy consumption and cooling capacity, use an adaptive learning algorithm to optimize the personnel heat load adjustment factor;

[0034] Real-time collect the outdoor environmental humidity monitoring value and meteorological prediction data, and generate an external disturbance compensation factor in combination with the heat inertia parameters of the air conditioning unit, including heat capacity and heat conduction value;

[0035] Multidimensionally fuse the reference temperature, the human thermal load adjustment factor, and the external disturbance compensation factor to generate a temperature control parameter matrix for spatial partitioning.

[0036] Further, generate a control instruction based on the environmental comfort index, the distribution of densely populated areas, the recognition of abnormal behavior patterns, and the security threat level; when an alarm signal is received, dynamically adjust the focus area of video surveillance and the access control strategy according to the signal type through a three-level priority response mechanism, including:

[0037] Perform spatio-temporal alignment and confidence-weighted fusion on the environmental temperature and humidity comfort index, the heat map of densely populated areas, the abnormal behavior characteristics of video analysis, and the security sensor alarm signal to generate a comprehensive threat assessment score;

[0038] According to the comprehensive threat assessment score, trigger a three-level response mechanism, and generate corresponding hierarchical response strategies according to the triggered response level. The hierarchical response strategies include the coordinate set of the video surveillance focus area, the access control locking logic rule, and the air volume adjustment parameter of the air conditioner;

[0039] Convert the hierarchical response strategy into a control instruction set in the protocol format of the target device, and send the control instruction set to the corresponding device through the corresponding communication protocol to perform corresponding operations, including adjusting the air volume of the air conditioner, adjusting the viewing angle of the surveillance camera, and controlling the access control electromagnetic lock.

[0040] Further, the three-level response mechanism includes:

[0041] Level 1 response: When the comprehensive threat assessment score > 90, activate the access control locking of the target area and the laser tracking video focus;

[0042] Level 2 response: When 70 < comprehensive threat assessment score ≤ 90, activate the strobe lighting and directional sound wave warning of the evacuation passage;

[0043] Level 3 response: When the comprehensive threat assessment score ≤ 70, non-core production equipment, including the cooling water circulation pump group and the ventilation auxiliary unit, switches to the energy-saving mode.

[0044] In a second aspect, an intelligent gateway includes:

[0045] One or more processors;

[0046] A storage device for storing one or more programs, which when executed by the one or more processors cause the one or more processors to implement the system.

[0047] In a third aspect, a computer-readable storage medium stores a program that, when executed by a processor, implements the system.

[0048] The above solution of the present invention has at least the following beneficial effects:

[0049] The communication module supports multiple Internet of Things protocols and can quickly and stably establish two-way connections with various devices in the building. This enables the system to collect in real time multi-source device operation data streams such as the current waveform of the air-conditioning compressor, the security alarm code, and the light intensity time-series data of lighting equipment, avoiding data collection obstacles caused by protocol incompatibility and ensuring the timeliness and integrity of the data. The edge computing module uses a physical classification algorithm to deeply process the collected data, accurately extracts key information such as the air-conditioning energy consumption characteristic values, security alarm signals, and environmental temperature and humidity data sets, and converts them into structured data for storage in the physical memory area. At the same time, the physical memory area performs timestamp difference encoding on the historical power data of the air conditioner to form a spatio-temporal correlation data cube, which not only improves the data storage efficiency but also facilitates data retrieval and analysis.

[0050] Based on the spatio-temporal correlation data cube in the physical memory area, the engine module can accurately identify the heat map pattern of personnel movement in the building and generate a dynamic temperature control parameter matrix. This enables the air-conditioning system to intelligently adjust the temperature of each area according to the personnel distribution and activity patterns, effectively reducing energy consumption while improving personnel comfort. For example, it automatically strengthens cooling in crowded areas and appropriately raises the temperature in sparse areas. The control logic module generates control instructions based on environmental comfort indicators, the distribution of crowded areas, abnormal behavior pattern recognition, and security threat levels, and responds to different levels of security threats through a three-level priority response mechanism. When receiving an alarm signal, it can quickly adjust the video surveillance focus area and access control strategy dynamically according to the signal type. For example, the first-level response promptly locks the access control of the dangerous area and monitors precisely, and the second-level response guides evacuation through stroboscopic lighting and directional sound waves, enhancing the building's security protection ability and emergency handling efficiency and ensuring the safety of personnel's lives and property.

[0051] The protocol conversion module can dynamically convert control instructions into the protocol format of target devices, ensuring that the instructions generated by the system can be adapted to different brands and models of devices in the building. Whether it is an air-conditioning unit, security equipment, or intelligent lighting equipment, it can accurately receive and execute instructions, avoiding control failures caused by device protocol differences, improving the versatility and compatibility of the system, and reducing the cost and difficulty of device replacement and upgrade. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a schematic diagram of an AI artificial intelligence machine learning system provided by an embodiment of the present invention.

[0053] Figure 2It is a schematic flowchart of the process of implementing timestamp difference encoding on the air conditioner historical power data in a structured dataset by an AI artificial intelligence machine learning system provided by an embodiment of the present invention to form a spatio-temporal correlation data cube. Detailed implementation manners

[0054] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0055] As Figure 1 shown, an embodiment of the present invention provides an AI artificial intelligence machine learning system, including:

[0056] Communication module 1, which establishes a two-way connection with air conditioner units, security sensor arrays, and intelligent lighting networks in a building through multiple Internet of Things protocols, and real-time collects device operation data streams, including air conditioner compressor current waveforms, security alarm codes, and lighting device light intensity time series data;

[0057] Edge computing module 2, which is used to extract air conditioner energy consumption characteristic values, security alarm signals, and environmental temperature and humidity datasets according to the device operation data streams, and store the processed structured dataset in the analog memory area through an analog classification algorithm;

[0058] Analog memory area module 3, which is used to implement timestamp difference encoding on the air conditioner historical power data in the structured dataset to form a spatio-temporal correlation data cube;

[0059] Engine module 4, which is used to identify the heat map pattern of personnel movement in a building according to the spatio-temporal correlation data cube in the analog memory area and generate a dynamic temperature control parameter matrix;

[0060] Control logic module 5, which is used to generate control instructions according to environmental comfort indicators, personnel dense area distribution, abnormal behavior pattern recognition, and security threat levels; when receiving an alarm signal, according to the signal type, dynamically adjust the focus area of video surveillance and the access control strategy through a three-level priority response mechanism;

[0061] Protocol conversion module 6, which is used to dynamically convert control instructions into the protocol format of target devices.

[0062] In the embodiments of the present invention, through the localized data processing of the edge computing module and the spatio-temporal associated data aggregation of the object-like memory area, an end-to-end closed-loop control of building equipment data from collection to decision-making is realized, reducing the data link delay in the traditional architecture and ensuring the real-time response capabilities of air-conditioning temperature control and security linkage. Based on the object-like classification algorithm and the device fingerprint coding technology, multi-source heterogeneous data such as air-conditioning current waveforms and security alarm codes are classified and stored according to device entities, breaking through the low-efficiency bottleneck of traditional "point-style" storage and improving the data reading and writing efficiency and the multi-device collaborative analysis ability.

[0063] Through the pattern recognition of the human thermal map by the AI / ML engine, combined with the ambient temperature and humidity data, a sub-region temperature control parameter matrix is dynamically generated, realizing a differential temperature control strategy for densely populated areas and low-load areas in the building, taking into account both energy efficiency optimization and comfort requirements. The protocol conversion module supports multi-protocol dynamic adaptation (such as Modbus, Zigbee, DALI), eliminating the manual configuration cost; the control logic module triggers a three-level priority response mechanism based on the fusion of multi-dimensional data such as environmental comfort and security threat levels, realizing the intelligent linkage of security, lighting, and air-conditioning systems.

[0064] Sensitive data (such as security alarm signals) complete feature extraction and compression processing at the edge, and only the structured analysis results are uploaded to the cloud through the spatio-temporal associated data cube, reducing the risk of privacy leakage and the network bandwidth occupancy. The design of the spatio-temporal associated data cube and the object-like memory area supports the rapid backtracking and predictive analysis of the historical data of building equipment, providing a lightweight data foundation for digital twins and reducing the software and hardware costs of system expansion at the same time.

[0065] In a preferred embodiment of the present invention, extracting the air-conditioning energy consumption characteristic values, security alarm signals, and ambient temperature and humidity data sets, and storing the processed structured data sets in the object-like memory area through the object-like classification algorithm may include:

[0066] Performing a sliding window process on the original device data stream to divide the continuous data stream into multiple equal-length data segments;

[0067] Extracting an air-conditioning energy efficiency feature set from multiple equal-length data segments, including the real-time energy efficiency ratio, current harmonic distortion rate, and load fluctuation index, and generating an alarm frequency density matrix based on the security sensor data to obtain a feature data set containing air-conditioning energy efficiency features and the spatio-temporal distribution information of security events;

[0068] Generating device-level classification labels for the feature data set through the device fingerprint coding algorithm, and mapping the classified data to the object-like memory area according to the device entity ID.

[0069] In the embodiments of the present invention, the original data streams generated by air conditioners, security systems, and lighting devices (such as the continuous current waveform of an air conditioner compressor, the alarm code stream of a security sensor, and the time-sequential light intensity data of a lighting device) are dynamically segmented. A fixed time window (such as 5 seconds) or an adaptive event window (such as every 1000 data points) is used to cut the data stream to ensure that each data segment contains a complete event cycle (such as one start-stop cycle of an air conditioner compressor). Device IDs, timestamps, and spatial tags (such as floor location) are attached to each data segment to ensure the spatio-temporal alignment of multi-device data. For example, the air conditioner current segment is strictly synchronized with the temperature and humidity data in its area.

[0070] Based on the instantaneous power of the current segment (current × voltage) and a preset refrigerating capacity mapping table, calculate the EER value (refrigerating capacity / power) , reflecting the current energy efficiency status; where is the refrigerating capacity, obtained from the preset mapping table and associated with the current, voltage, and device model; is the current; is the voltage. Perform a fast Fourier transform (FFT) on the current waveform to extract the fundamental wave and each harmonic component, and calculate the total harmonic distortion rate , to evaluate the current quality; where is the highest harmonic order; is the harmonic order; is the effective value of the harmonic voltage of the

[0071] The effective value of the fundamental wave voltage (RMS), that is, the voltage amplitude of the main frequency (power frequency, such as 50Hz / 60Hz) component in the signal. Statistically calculate the difference between the peak and valley values of the current within the statistical window, and combine it with the standard deviation of historical load data to quantify the operating stability of the compressor. Count the frequency of alarm events according to the spatial area (such as rooms, corridors) and time window (such as every 10 minutes) to construct a two-dimensional matrix. For example, the rows of the matrix represent areas, the columns represent time windows, and the element values are the number of alarms in the corresponding area within the corresponding time window.

[0072] For example, for the optimization of the air conditioner and security systems in a commercial building, the air conditioner unit generates a current waveform data stream every 5 seconds, and the sliding window divides it into continuous segments, and each segment contains 20 current cycle data.

[0073] Security sensors report alarm codes in real time (such as smoke alarms, intrusion detections), and count the number of alarms on each floor within a 10-minute time window. Calculate the EER value from the current segment, and it is found that the EER value of the air conditioner on the 3rd floor is continuously lower than the threshold, triggering an energy efficiency warning. The security alarm matrix shows that the alarm frequency in area B abnormally increases during peak hours, and it is determined as a high false alarm area in combination with historical data. The THD feature of the No. 3 air conditioner is compared with historical data to generate a unique fingerprint, which is mapped to an independent unit in the iconic memory area, and its installation location and maintenance records are associated. The security alarm matrix and the air conditioner operation data are associated by time stamps to form a multi-dimensional analysis view of "area - time - device".

[0074] Through sliding window and edge feature extraction, achieve second-level detection and response to abnormal air conditioner energy efficiency (such as sudden drop in EER), avoiding the minute-level delay of traditional architectures. The device fingerprint coding maps discrete current waveforms and alarm codes into a semantic dataset centered on the device, supporting cross-device (such as air conditioner and security) correlation analysis. Based on the coupled analysis of the security alarm matrix and the air conditioner energy efficiency characteristics, automatically switch the air conditioner to the low-power mode in high-security areas to balance the energy efficiency and emergency response requirements. The spatio-temporal correlation data cube compresses historical data through difference coding, reducing storage redundancy; the device-level classification labels simplify the data query logic and reduce the database load. The device fingerprint coding is compatible with multi-protocol devices (such as Modbus air conditioners, Zigbee sensors), avoiding manual configuration of rules and adapting to the complex device environment of the building.

[0075] In a preferred embodiment of the present invention, implementing time stamp difference coding on the air conditioner historical power data in the structured dataset to form a spatio-temporal correlation data cube may include:

[0076] Extract the air conditioner historical power data from the structured dataset, group it by device ID and arrange it by time stamp to form a time series power sequence;

[0077] For each adjacent pair of time stamps, calculate the power difference and time interval to form a power difference coding sequence;

[0078] Associate the difference coding sequence with the device geographical coordinates to construct a three-dimensional data cube including a time dimension, a space dimension, and a feature dimension.

[0079] In an embodiment of the present invention, extract the air conditioner historical power data from the structured database, and the data fields include device ID, time stamp, power value, and device geographical coordinates (such as floor, room number). Group by device ID, and each group of data is arranged in ascending order of time stamp to form a time series power sequence for each device. For example, the time series of device ID - A is: ( , ),( , ),…,( , ). For the timing power sequence of each device, calculate the power difference between adjacent timestamp pairs ( ) and the time interval ( ).

[0080] Power difference = ( is the time point index);

[0081] Time interval = (unit: second).

[0082] Generate the difference coding sequence: ( , ),( , ),…,( , ). Only store the initial power value ( ) and the difference sequence, replacing the original full data to achieve data compression. Associate the difference coding sequence with the device geographical coordinates (such as longitude and latitude, floor coordinates) to form data units with spatial tags.

[0083] Time dimension: Divide by time windows (such as hours, days) to record the time distribution of the difference sequence;

[0084] Spatial dimension: Divide the area according to the device coordinates (such as rooms, floors), and aggregate the difference data of devices in the same area;

[0085] Feature dimension: Include power difference ( ), time interval ( ), device type (such as variable-frequency air conditioner, fixed-frequency air conditioner).

[0086] Adopt columnar storage (such as Parquet format) to support fast query by time, space or device ID range.

[0087] Suppose the air conditioner energy consumption analysis of an office building is as follows: The original power sequence of device ID - A (the air conditioner in the 3rd floor meeting room) is (09:00, 1500W), (09:15, 1800W), (09:30, 1650W).... The original power sequence of device ID - B (the air conditioner in the 5th floor corridor) is (09:00, 800W), (09:15, 750W), (09:30, 820W).... The difference sequence of device ID - A is ( = 300W, = 900s),( = - 150W, = 900 s).... The device ID - B difference sequence is ( = - 50 W, = 900 s), ( = 70 W, = 900 s)....

[0088] Time dimension: Divided by hour, storing the difference sequence per hour;

[0089] Space dimension: Conference room on the 3rd floor (ID - A), corridor on the 5th floor (ID - B);

[0090] Feature dimension: Power change trend (such as "rapid rise and fall" of ID - A, "stable fluctuation" of ID - B).

[0091] By replacing full - volume storage with difference coding, only the initial value and the change amount are retained, significantly reducing the redundancy of historical data storage. For example, the original 1000 power records can be compressed into 1 initial value + 999 differences. The data cube supports multi - dimensional joint queries. For example: Space dimension: Statistically analyze the power fluctuation patterns of all air conditioners on a certain floor; Time dimension: Analyze the energy efficiency change trend during peak hours (such as 12:00 - 14:00); Feature dimension: Identify abnormal fluctuations of specific equipment types (such as variable - frequency air conditioners). The difference coding allows for the quick restoration of power values at any time point, meeting the requirements of real - time monitoring and historical data backtracking. Through the power difference distribution in the spatio - temporal cube, abnormal areas (such as a certain floor continuously being negative) can be quickly located, and the fault source can be judged by combining the equipment coordinates. Spatio - temporal correlation analysis reveals the regional energy consumption pattern (such as the air conditioner in the conference room still running at high power during unoccupied periods), providing a data basis for automatic scheduling strategies (such as adjusting the temperature according to the number of people).

[0092] In a preferred embodiment of the present invention, according to the spatio - temporal correlation data cube in the anthropomorphic memory area, identifying the heat - map pattern of personnel movement in the building and generating a dynamic temperature control parameter matrix may include:

[0093] Extracting multi - source probe positioning data from the spatio - temporal correlation data cube in the anthropomorphic memory area, and using a long short - term memory network to process the probe positioning data to predict the probability of personnel staying in each area in the future for a period of time, generating a dynamic heat - distribution map;

[0094] Performing spatial superposition analysis on the heat - distribution map and the temperature - humidity gradient data stored in the anthropomorphic memory area to identify areas of high personnel density and areas with abnormal environmental parameters, obtaining a coupled analysis result;

[0095] According to the coupled analysis result, combined with the heat - response characteristics of the air - conditioning unit, dynamically calculating the matrix of sub - area temperature set values.

[0096] In the embodiments of the present invention, Wi-Fi probes (device connection time, signal strength), Bluetooth beacons (device scan records), and video coordinate point clouds (real-time positions of personnel) are extracted from the spatio-temporal correlation data cube, aligned by timestamp, and noise data (such as signal drift points) is cleaned. Multi-source data is associated through device IDs or anonymous identifiers (such as hashed MAC addresses) to generate an individual movement trajectory with timestamps (for example: the path and residence duration of a user from the lobby to the meeting room). The trajectory is segmented into segments by a sliding window (such as 15 minutes), features (residence point density, moving direction change rate) are extracted, and input into a pre-trained LSTM model to predict the residence probability of personnel in each area in the future period (such as the probability of 85% in the meeting room in the next hour).

[0097] The predicted probability is mapped to the building floor grid (1m×1m), the blank area is filled by Kriging interpolation, and the distribution is corrected by combining topological structures such as walls and access controls to generate a color gradient heat map (red for high density, blue for low density). The real-time temperature and humidity gradient data (sampled every 5 minutes) stored in the physical memory area is extracted and aligned with the heat map at the same grid granularity to form a multi-dimensional data layer (heat + temperature and humidity + air conditioning status). The geographic information system (GIS) technology is used to perform a spatial overlay operation on the dynamic heat distribution map and the temperature and humidity gradient data. During the overlay process, the residence probability of personnel in each area in the heat map is associated and matched with the temperature and humidity data at the corresponding position. By setting the residence probability threshold of personnel (such as 0.7) and the normal range of temperature and humidity (temperature 22 - 26°C, humidity 40% - 60%), the crowded areas of personnel (areas where the residence probability of personnel is higher than the threshold) and the areas with abnormal environmental parameters (areas where the temperature or humidity exceeds the normal range) are identified.

[0098] For each area in the building (such as offices, meeting rooms, corridors on different floors, etc.), the appropriate temperature set value is calculated using the formula (dynamic reference temperature × personnel heat load adjustment factor × external disturbance compensation factor). The temperature set values of all areas are sorted into a matrix form to form a sub-region temperature set value matrix, providing specific parameters for the precise temperature control of the air conditioning system.

[0099] Suppose there is a large shopping mall with three floors, and each floor has multiple stores and public areas.

[0100] On a weekend afternoon, a large amount of customer connection information was collected through Wi-Fi probes, and the movement trajectories of many customers were also scanned by Bluetooth beacons. The video surveillance system recorded the coordinate point clouds of customers in various areas. After cleaning and integrating these data, an individual movement trajectory spatio-temporal sequence was constructed by grouping according to customer ID. For example, between 2 pm and 3 pm, customer A entered from the entrance on the first floor, passed by several clothing stores, and then went to the restaurant on the second floor. The sliding window was used to segment this spatio-temporal sequence. For example, the movement trajectory from 2 pm to 2:15 pm was taken as a segment. These segments were input into the LSTM network, and it was predicted that within the next hour, the probability of people staying in the cinema on the third floor and the dessert area on the first floor was relatively high, while the probability of people staying in some stores on the second floor was relatively low.

[0101] Spatial interpolation optimization was carried out on the prediction results, and combined with the floor layout of the shopping mall, a dynamic heat map was generated. In the map, the cinema and the dessert area were shown in red, indicating a dense crowd; some stores were shown in blue, indicating a sparse crowd. The temperature and humidity gradient data in the shopping mall were obtained from the iconic memory area, and it was found that due to the dense crowd in the cinema, the temperature was 3°C higher than the surrounding area, and the humidity also increased; the dessert area also had a slightly higher temperature and larger humidity due to the frequent movement of people. Through spatial overlay analysis, the cinema and the dessert area were determined as areas with dense crowds and abnormal environmental parameters. The coupling analysis results showed that the temperature and humidity in these areas needed to be adjusted. According to the areas with dense crowds and abnormal environmental parameters determined by the coupling analysis, combined with the cooling, heating capacity and response speed and other thermal response characteristics of the air conditioning units in the shopping mall, considering the changing trend of the personnel density in different areas, the appropriate temperature setting values for each area were dynamically calculated, and finally a matrix of sub-region temperature setting values was formed, providing a basis for the precise control of the shopping mall air conditioning system to meet the comfort requirements of people in different areas and achieve the purpose of energy conservation.

[0102] By accurately predicting the probability of people staying, the temperature control parameters can be dynamically adjusted according to the degree of crowd density in different areas. For areas with a sparse crowd, the temperature setting value can be appropriately increased or decreased to reduce the energy consumption of the air conditioner; for areas with a dense crowd, a suitable temperature environment is provided to avoid waste of energy, thereby improving the energy utilization efficiency of the entire building. Combining the coupling analysis results of the areas with dense crowds and abnormal environmental parameters can adjust the temperature and humidity more precisely. In areas with a dense crowd, the temperature and humidity are adjusted in a timely manner to provide a comfortable environment for people, reduce health problems and discomfort caused by environmental discomfort, and improve people's satisfaction. Using the LSTM network and the adaptive learning algorithm, it can learn and adjust according to historical data and real-time data, and automatically adapt to different personnel activity patterns and environmental changes. For example, on weekdays and weekends, the distribution and activity patterns of people are different, and the temperature control strategy can be automatically adjusted to improve the intelligent level and adaptability of the system.

[0103] In another preferred embodiment of the present invention, probe location data is extracted from the spatio-temporal association data cube in the iconic memory area, and the long short-term memory network is used to process the probe location data to predict the personnel residence probability in each area within a certain period of time in the future, and a dynamic heat map is generated, which may include:

[0104] Extract multi-source probe location data from the spatio-temporal association data cube in the iconic memory area, including probe connection events, Bluetooth beacon scan records, and video analysis coordinate point clouds;

[0105] Group the multi-source probe location data by personnel ID and construct a spatio-temporal sequence of individual movement trajectories, and divide them into continuous trajectory segments through a sliding window;

[0106] Input the trajectory segments into a pre-trained long short-term memory network, and predict the personnel residence probability distribution in each area within the future target period according to the historical movement rules;

[0107] Optimize the spatial interpolation of the residence probability distribution, and generate a dynamic heat map in combination with the building topology structure.

[0108] In the embodiment of the present invention, the iconic memory area is an area for storing processed data, and the spatio-temporal association data cube integrates multi-dimensional information such as time, space, and devices. Probe connection events record the time and related information when a device (such as a mobile phone) establishes a connection with a probe, which can be used to judge the appearance of personnel within the probe coverage range; Bluetooth beacon scan records include Bluetooth signal strength, scan time, etc., which can roughly determine the relative position of personnel and Bluetooth beacons; video analysis coordinate point clouds are obtained by analyzing the monitoring video to obtain the precise coordinate information of personnel in the scene. The system filters out these multi-source probe location data within a specific time period from the spatio-temporal association data cube through preset data extraction rules and interfaces. For example, set to extract relevant data from 9 am to 11 am on a certain working day, and accurately obtain the required probe connection events, Bluetooth beacon scan records, and video analysis coordinate point cloud data according to the time range and data type identification.

[0109] After obtaining multi-source probe positioning data, the data is grouped according to the person ID (such as the unique identifier of the device) contained in the data. The ID of each person corresponds to a group of data, and these data are arranged in chronological order to construct the spatio-temporal sequence of the individual movement trajectory. This sequence details the position information of each person at different time points and reflects their movement trajectory. Then, the sliding window technique is used to segment the spatio-temporal sequence. The sliding window has a fixed time length (such as 10 minutes) and a sliding step size (such as 1 minute). Starting from the start time of the sequence, it moves one step at a time and intercepts the data within a fixed length of time period to form continuous trajectory segments. For example, for the spatio-temporal sequence of a certain person, segmented with a window length of 10 minutes and a step size of 1 minute, the first segment may be the movement trajectory data from 9 o'clock to 9:10, and the second segment is the data from 9:01 to 9:11, and so on.

[0110] The Long Short-Term Memory network (LSTM) is a neural network specifically designed for processing time series data, which can learn long-term dependencies and complex patterns in the data. Before use, a large amount of historical trajectory segment data is required to train the LSTM network. During the training process, the network learns the movement patterns and rules of people in different time periods and different regions. After the training is completed, the segmented trajectory segments are input into the pre-trained LSTM network. The network analyzes and predicts the input trajectory segments according to the learned historical movement rules, and outputs the probability distribution of people staying in each region within the future target time period (such as the next 30 minutes). For example, the network may predict that within the next 30 minutes, the probability of people staying at the elevator entrance area on a certain floor is 0.6, and the probability of people staying in the restaurant area is 0.3, etc.

[0111] The predicted probability distribution of people staying is discrete data points. To more intuitively and continuously display the personnel distribution, spatial interpolation optimization is required. Spatial interpolation algorithms (such as inverse distance weighted interpolation method, Kriging interpolation method, etc.) estimate the probability values at other positions within the entire space range based on the known discrete probability data points, making the probability distribution smoother and more continuous. At the same time, combined with the topological structure information of the building (such as floor layout, room location and size, passage direction, etc.), the probability values are mapped to the actual spatial positions. Different colors (such as red for high probability and blue for low probability) and the depth of the color are used to represent the high and low probability of people staying in different regions, thereby generating a dynamic heat map. In this way, through the heat map, it is possible to clearly see the density and distribution trend of people in different regions of the building at a certain future time period.

[0112] Suppose there is a large office building with 5 floors, and each floor has multiple offices, meeting rooms and public areas.

[0113] Extract data from the spatio-temporal correlation data cube of the iconic memory area from 10 am to 12 pm on a certain working day. Through probe connection events, connection information of numerous employee devices with probes distributed on each floor at different times is obtained; Bluetooth beacon scan records show the activities of employees near Bluetooth beacons; video analysis coordinate point clouds accurately record the positions of employees within the monitoring range. For example, Employee A's mobile phone established a connection with a probe near the elevator entrance on the 3rd floor at 10:15 am, and at the same time, the Bluetooth beacon scanned his presence in this area, and the video analysis coordinate point cloud determined his specific coordinate position. Group these data according to the employee number (as the personnel ID) to construct the spatio-temporal sequence of each employee's individual movement trajectory. For example, for Employee A, from 10 am to 10:30 am, his trajectory shows that he starts from the office on the 3rd floor, passes through the corridor, and arrives at the meeting room. Use a sliding window, set the window length to 10 minutes, and the step size to 1 minute to segment this spatio-temporal sequence into segments. The first segment is the trajectory from 10 am to 10:10 am, the second segment is the trajectory from 10:01 am to 10:11 am, etc.

[0114] Input these segmented trajectory segments into a pre-trained LSTM network. This network has been trained with a large amount of historical movement data of employees in this office building and has learned the movement patterns of employees. After network analysis, it is predicted that within the next 30 minutes, the probability of personnel staying in the pantry area on the 4th floor is 0.7, because this time period is usually the break time for employees and more people will go to the pantry; while the probability of personnel staying in a small meeting room on the 5th floor is 0.2, because there is no meeting scheduled in this meeting room on that day. Optimize the predicted personnel stay probability distribution through spatial interpolation to make the probability distribution smoother. Combine the floor layout of the office building and map the probability values to the corresponding areas. On the generated dynamic heat map, the pantry area on the 4th floor is shown in dark red, representing a high probability of personnel staying; the small meeting room on the 5th floor is shown in light blue, representing a low probability of personnel staying, and other areas also show different degrees of color depth according to their respective probability values, intuitively demonstrating the personnel distribution trend in each area of the office building within the next 30 minutes.

[0115] Through the dynamic heat map, managers can clearly understand the future personnel flow trends and density in different areas. For an office building, cleaning, maintenance and other working hours can be reasonably arranged according to the heat map to avoid affecting the normal work of employees during peak personnel periods; for a shopping mall, the promotion activity areas can be planned in advance, and the activities can be set in areas with high pedestrian flow to improve the activity effect and business benefits.

[0116] According to the personnel residence probability distribution, resources can be allocated more reasonably. Increase the power or the number of devices such as air conditioners and lighting in crowded areas to ensure a comfortable environment; appropriately reduce the operating power of devices in areas with few people to save energy. For example, in an office building, the air conditioner can be turned on in advance and adjusted to a suitable temperature in the meeting room where people are concentrated according to the heat map, while some lighting devices can be turned off on floors with fewer people. The dynamic heat map is helpful for safety management.

[0117] When it is found that the personnel residence probability in a certain area is abnormally high or low, inspections and handling can be carried out in a timely manner. For example, in an office building, if a prediction of a sudden concentration of people appears in an area that is usually sparsely populated, there may be potential safety hazards, and security personnel can go to check in advance; in a shopping mall, areas where congestion or abnormal gatherings may occur can also be detected in a timely manner, and diversion measures can be taken in advance to ensure the safety of personnel. Provide strong support for operation decisions. For example, a shopping mall can analyze the attractiveness of different areas and the customer flow pattern according to the heat map, adjust the store layout and product display, and improve the shopping experience and purchase desire of customers; the office building manager can reasonably plan new office areas or adjust the use method of the existing space according to the personnel distribution, improving the space utilization rate.

[0118] In another preferred embodiment of the present invention, according to the coupling analysis result, combined with the heat response characteristics of the air conditioning unit, the dynamic calculation of the temperature setting value matrix for each area may include:

[0119] Determine the dynamic reference temperature according to the historical temperature data in the building, and compare the personnel residence probability of each area in the future target period with the actual personnel residence probability at the current time point to obtain the personnel density change rate;

[0120] According to the heat response characteristics of the air conditioning equipment, including the response speed and adjustment range, and the historical energy efficiency data, including energy consumption and cooling capacity, an adaptive learning algorithm is used to optimize the personnel heat load adjustment factor;

[0121] Real-time collect the outdoor environmental humidity monitoring value and meteorological prediction data, and combine the heat inertia parameters of the air conditioning unit, including heat capacity and heat conduction value, to generate an external disturbance compensation factor;

[0122] Multidimensionally fuse the reference temperature, the personnel heat load adjustment factor, and the external disturbance compensation factor to generate a temperature control parameter matrix for spatial partitioning.

[0123] In the embodiments of the present invention, temperature data of a building in different seasons and at different time periods over a relatively long past period (such as several years) are collected. These data are classified by season, and the average temperature of the same time period (such as 14:00 - 15:00 every day) in each season is calculated. For example, in summer, the average temperature of this time period every day in multiple summers is calculated to obtain the benchmark temperature value for summer; the same method is used for winter, spring, and autumn. The benchmark temperatures of different seasons obtained in this way , which will be adjusted with the change of seasons and reflects the basic temperature in the building in different seasons.

[0124] From the predicted data of the dynamic thermal distribution map generated previously, obtain the occupancy probability of each area within the future target time period ), and at the same time, obtain the actual occupancy probability of each area at the current time point through the actual monitoring data at the current moment. The area ) here can be a specific spatial location divided according to floors, rooms, etc. Subtract the actual occupancy probability at the current time point from the occupancy probability in the future target time period, and then divide by the time interval , that is , to obtain the personnel density change rate. This change rate reflects the speed of change in the number of people in each area in the future period of time. That is , to obtain the personnel density change rate. This change rate reflects the speed of change in the number of people in each area in the future period of time.

[0125] According to the thermal response characteristics of the air conditioning equipment (such as response speed and adjustment range) and historical energy efficiency data (energy consumption and cooling capacity), an enhanced learning algorithm is used for optimization. In this algorithm, we set the state space as the historical operation data of the air conditioner (including energy consumption, cooling capacity, indoor and outdoor temperature difference, etc. at different time periods), the current indoor personnel density, and outdoor environmental parameters (such as humidity, temperature), etc.; the action space is the range of values of the personnel heat load adjustment coefficient , assumed to be [0.5, 2]. The reward function is set according to comfort and energy consumption. If the indoor temperature quickly reaches the comfortable range and the energy consumption is low, a higher reward is given; conversely, if the temperature adjustment effect is poor or the energy consumption is too high, a lower reward is given.

[0126] After a large number of iterative trainings, for a certain area A, assuming that its air conditioner has a fast response speed, and the response time from receiving the temperature adjustment instruction to reaching the set temperature is 5 minutes under ideal conditions, and the cooling capacity is sufficient, and the indoor temperature can be reduced by 5°C per hour. Through the enhanced learning algorithm, analyze and optimize the historical operation data (including data such as energy consumption and temperature adjustment effect under different personnel densities) of this area in the past year, and obtain that when the current personnel density increases, = 1.8. This indicates that when the personnel density increases in this area, in order to adjust the temperature more quickly to maintain comfort, it is necessary to significantly adjust the cooling capacity of the air conditioner. Therefore, the value is relatively large.

[0127] For another area B, the response speed of its air conditioner is slow, the response time is 15 minutes, the cooling capacity is relatively low, and it can only reduce the indoor temperature by 3°C per hour, and the energy consumption is high. Similarly, through the optimization calculation of the historical operation data by the reinforcement learning algorithm, when the personnel density increases, = 1.2. Due to the limitation of the air conditioner's own performance in this area, it cannot adjust the temperature as quickly and efficiently as area A. To balance comfort and energy consumption, the value is relatively smaller than that of area A, and more attention will be paid to the control of energy consumption when adjusting the temperature.

[0128] By connecting with a meteorological monitoring device or a meteorological data service platform, the outdoor environmental humidity monitoring value is collected in real time and the outdoor humidity prediction value for the future time period in the meteorological prediction data is obtained . Subtract the current outdoor humidity monitoring value from the outdoor humidity prediction value for the future time period, and then divide by the time interval , that is , to obtain the outdoor humidity change rate. This change rate reflects the change trend of the outdoor humidity in the future period of time. Combine the thermal inertia parameters (heat capacity, heat transfer value) of the air conditioner unit to determine the external disturbance compensation coefficient , where , , is the heat transfer value; is the heat capacity of the air conditioner unit. The heat capacity and heat transfer value determine the ability of the air conditioner unit to resist the influence of outdoor environmental changes on the indoor temperature. For example, if the heat capacity of the air conditioner unit is large and the heat transfer value is small, it means that it can better maintain the indoor temperature stability. At this time the value is relatively small; on the contrary, if the air conditioner unit is more sensitive to outdoor environmental changes, the value will increase accordingly. By comprehensively considering these factors, use the corresponding calculation model or empirical formula to determine the value.

[0129] Add the previously calculated dynamic reference temperature , the part related to the personnel density change rate and the part related to the external disturbance compensation factor to obtain the temperature setting value for each area ( ). Arrange the temperature setting values of all areas in a matrix form according to the layout of the areas to generate the temperature control parameter matrix for spatial zoning . This matrix provides specific parameter basis for the temperature regulation of the air conditioning system in different areas.

[0130] By considering the change rate of personnel density to adjust the temperature set value, the indoor temperature can be adjusted in a timely manner according to the actual change of the number of people. When the number of people in a certain area gradually increases, the temperature set value is appropriately reduced to ensure the comfort of people in a crowded environment; when the number of people decreases, the temperature set value is increased to avoid energy waste and keep the area at a suitable temperature. At the same time, adjusting the temperature set in combination with the change of outdoor humidity takes into account the influence of the outdoor environment on indoor comfort, further enhancing the overall comfort experience. Using the thermal response characteristics and historical energy efficiency data of air conditioning equipment to optimize the human thermal load adjustment factor can avoid excessive cooling or heating. If the air conditioner has a fast response speed, it can adjust the temperature more accurately when the personnel density changes, reducing energy consumption; generating an external disturbance compensation factor according to the change rate of outdoor humidity and the thermal inertia of the air conditioning unit enables the air conditioning system to better adapt to the change of the outdoor environment, avoiding excessive adjustment of the indoor temperature due to outdoor environment fluctuations, thereby reducing energy consumption and achieving efficient utilization of energy.

[0131] This calculation process synthesizes a variety of real-time data and historical data, enabling the temperature control parameter matrix to be dynamically adjusted with the changes of seasons, personnel activities, and outdoor environment. Whether it is the difference in personnel distribution between weekdays and weekends or the change of outdoor climate in different seasons, it can automatically adapt and give reasonable temperature set values, improving the adaptive ability and stability of the entire air conditioning system. Reasonable temperature regulation avoids frequent large-scale adjustment of the temperature of the air conditioning unit, reducing equipment wear. For example, when the personnel density changes little and the outdoor environment is relatively stable, the air conditioner does not need to frequently change its working state, reducing the fatigue degree of equipment components, thereby extending the service life of the air conditioning equipment and reducing equipment maintenance and replacement costs.

[0132] In a preferred embodiment of the present invention, control instructions are generated according to environmental comfort indicators, personnel dense area distribution, abnormal behavior pattern recognition, and security threat level; when an alarm signal is received, according to the signal type, the focus area of video surveillance and access control strategy are dynamically adjusted through a three-level priority response mechanism, which may include:

[0133] Align the environmental temperature and humidity comfort indicators, the heat map of personnel dense areas, the abnormal behavior characteristics of video analysis, and the security sensor alarm signals in space and time and perform confidence weighted fusion to generate a comprehensive threat assessment score;

[0134] According to the comprehensive threat assessment score, a three-level response mechanism is triggered, and corresponding hierarchical response strategies are generated according to the triggered response levels. The hierarchical response strategies include the set of coordinates of the video surveillance focus area, the access control locking logic rules, and the air conditioning air volume adjustment parameters; the three-level response mechanism includes:

[0135] Level 1 response: When the comprehensive threat assessment score > 90, activate the access control locking of the target area and the laser tracking video focus;

[0136] Level 2 response: When 70 < comprehensive threat assessment score ≤ 90, start the strobe lighting of the evacuation passage and the directional sound wave alarm;

[0137] Level 3 response: When the comprehensive threat assessment score ≤ 70, non-core production equipment, including the cooling water circulation pump group and the ventilation auxiliary unit, switches to the energy-saving mode;

[0138] Convert the hierarchical response strategy into a control instruction set in the protocol format of the target device, and send the control instruction set to the corresponding device through the corresponding communication protocol to perform corresponding operations, including adjusting the air volume of the air conditioner, adjusting the viewing angle of the monitoring camera, and controlling the access control electromagnetic lock.

[0139] In the embodiment of the present invention, temperature and humidity sensors are deployed in various key areas of the building (such as offices, corridors, meeting rooms, etc.). These sensors collect environmental temperature and humidity data at regular time intervals (for example, every minute), and the data is transmitted to the central control system in real time. For example, in an office, the temperature and humidity sensor records the current temperature as 26°C and the humidity as 50%. Achieved by means of a variety of positioning technologies, Wi-Fi probes estimate the personnel position by detecting the MAC address and signal strength of the device; Bluetooth beacons use Bluetooth signals for more accurate short-range positioning; the video surveillance system detects and counts the personnel in the surveillance video through image recognition algorithms. Integrate the personnel position information from these different sources and use a special algorithm to generate a heat map of crowded areas, intuitively showing the personnel distribution density in different areas. For example, in a shopping mall, through these technologies, it is found that the dining area on a certain floor has a high personnel density, and this area will be displayed in a darker color on the heat map.

[0140] Video surveillance systems use computer vision technology to analyze surveillance videos frame by frame. Through preset behavior models, abnormal behaviors such as people running, staying stationary in sensitive areas for a long time, and multiple people gathering are identified. For example, when it is detected that someone suddenly starts running in the counter area of a bank, the video analysis system will extract relevant features of the behavior, including information such as the time, location, and number of people involved in the behavior, and mark it as an abnormal behavior. Various security sensors, such as intrusion detection sensors, smoke sensors, emergency buttons, etc., are installed at important locations in the building (such as entrances and exits, computer rooms, warehouses, etc.). Once the sensor detects an abnormal situation, it will immediately send an alarm signal to the central control system. For example, when the intrusion detection sensor detects someone illegally entering, it will send a signal containing information such as the intrusion location and the type of alarm.

[0141] Since this data comes from different devices and systems, their collection times and frequencies may not be consistent, so spatio-temporal alignment is required. Based on a high-precision clock source (such as an atomic clock or a network time server), the timestamps of data from different data sources are calibrated to ensure the consistency of all data in terms of time. At the same time, according to the installation location of the device and the monitoring range, different types of data are matched with specific physical locations. For example, the temperature and humidity sensor data in a certain room on a certain floor, the abnormal behavior data captured by the surveillance camera in that room, and the alarm signals of the security sensors in that room are all associated with the specific coordinate location of that room, so that these data correspond to the same area in terms of time and space.

[0142] Determine the confidence level for each data source, which is based on factors such as the reliability of the data, the performance of the device, and the accuracy of historical data. For high-precision and stable temperature and humidity sensors, a higher confidence level is given, such as 0.9; while for wireless security sensors that are easily affected by the environment, the confidence level may be set to 0.7. If the video analysis system has been trained with a large number of samples and shows a high accuracy rate in actual applications, the confidence level can be set to 0.8. The confidence level of the heat map in crowded areas depends on the accuracy of the positioning technology and the timeliness of data update. If the positioning is accurate and updated in real time, the confidence level can be set to 0.85. The system will dynamically adjust the confidence level according to factors such as the operating state of the device and the accuracy of historical data.

[0143] According to the importance and reliability of each piece of data, assign corresponding weights to it. Assume that the weight of the environmental temperature and humidity comfort index is 0.2, the weight of the heat map in crowded areas is 0.3, the weight of the abnormal behavior characteristics in video analysis is 0.3, and the weight of the alarm signal of the security sensor is 0.2. Before performing the fusion calculation, first standardize each piece of data to convert it into a unified numerical range (such as 0 - 1) to eliminate the differences in dimensions among different data types. The standardization process can adopt the normalization method. For example, normalize the temperature data through the formula for normalization, where is the original data, and are the minimum and maximum values of this data type respectively. Then calculate the comprehensive threat assessment score according to the weighted formula: Comprehensive threat assessment score = 0.2 × standardized value of the environmental temperature and humidity comfort index + 0.3 × standardized value of the heat map in crowded areas + 0.3 × standardized value of the abnormal behavior characteristics in video analysis + 0.2 × standardized value of the alarm signal of the security sensor.

[0144] Threshold setting:

[0145] Level 1 response (>90): Corresponding to high-risk events (such as breaking in with weapons, fire alarm), immediate physical isolation and precise tracking are required.

[0146] Level 2 response (70 - 90): Corresponding to medium-risk events (such as personnel gathering and conflict, local smoke), evacuation guidance and warning are required.

[0147] Level 3 response (≤70): Low-risk or energy efficiency optimization scenarios (such as off-peak hours), switch the equipment to the energy-saving mode.

[0148] Level 1 response strategy:

[0149] Call the preset laser tracking algorithm to control the PTZ camera to focus on the target coordinates (such as the intruder's position), and update the focused area in real time through the SLAM algorithm.

[0150] Access control: Lock all electromagnetic access controls (such as exit doors) in the target area and send a "forced lock" command to the central control.

[0151] Air conditioning adjustment: Turn off the fresh air system to prevent the spread of smoke and increase the smoke exhaust air volume to the maximum gear.

[0152] Level 2 response strategy:

[0153] Evacuation guidance: Trigger the strobe lighting (500Hz flashing frequency) and directional sound waves (145dB directional alarm) in the evacuation passage.

[0154] Video surveillance: Adjust the camera to the preset evacuation route perspective (such as corridors, stairwells), and enable crowd density analysis.

[0155] Three-level response strategy:

[0156] Equipment energy saving: Switch non-core equipment (such as cooling water pumps) to the low-frequency operation mode (30% power), and the ventilation unit operates at the minimum air change rate.

[0157] Convert the focus coordinates (such as XYZ coordinates) into PTZ control commands in ONVIF protocol format ( <continuousmove>Command).

[0158] Access control system: Generate Wiegand protocol or IP commands.

[0159] Air conditioning system: Convert to Modbus RTU commands (e.g., write to register address 0x4001, set the air volume value to 80%).

[0160] Send commands asynchronously through the MQTT message queue to ensure that high-priority commands (such as first-level responses) preempt the transmission channel. Use transaction logs to record the command status and automatically retry in case of failure (e.g., 3 retries + manual alarm).

[0161] Suppose in a large shopping mall scenario:

[0162] During the peak shopping period on Saturday afternoon, the temperature and humidity sensors on the second floor of the shopping mall detect that the temperature in a certain area reaches 30°C and the humidity is only 30%, far exceeding the comfortable range for the human body. After standardization, the comfort index value of the environmental temperature and humidity is 0.8 (assuming the standardized value corresponding to the comfort range is 0 - 1, the closer to 1, the less comfortable). The personnel positioning system shows that the clothing area on the second floor is crowded, presenting a dark red color on the heat map, and the standardized value of the heat map in the crowded area is 0.9. The video analysis system discovers that a customer in the clothing area has abnormal behavior, continuously wandering in front of a certain store and looking into the store many times, which is determined to be abnormal behavior, and the standardized value of the abnormal behavior characteristics of the video analysis is 0.7. At this time, the anti-theft sensor in the jewelry store on the first floor suddenly emits an intrusion alarm signal, and the standardized value of the security sensor alarm signal is 1.

[0163] According to the pre-set confidence level and the weights of each data, the comprehensive threat assessment score is assumed to be 72.9.

[0164] Since 72.9 is in the range of 70 < comprehensive threat assessment score ≤ 90, the secondary response mechanism is triggered. After receiving the command, the lighting equipment in all evacuation channels of the shopping mall starts to flash at a specific frequency (e.g., 3 times per second), forming an obvious visual guidance signal. The shopping mall broadcast system starts directional acoustic warnings and sends voice prompts to the clothing area on the second floor and surrounding areas: "An abnormal situation has been detected in the shopping mall. Please stay calm, dear customers, and leave the area in an orderly manner according to the evacuation signs." Convert the secondary response strategy into protocol format commands for corresponding devices. For lighting equipment, use the DALI protocol to convert the stroboscopic command into a specific control code and send it; for the broadcast system, transmit the voice alarm command to the speaker devices in the corresponding area through the IP network protocol. After receiving the command, the lighting equipment starts to strobe, and the broadcast system plays the alarm voice to complete the response operation.

[0165] Integrating multi-source data for comprehensive threat assessment can identify various potential security risks in a timely and accurate manner. From environmental anomalies to abnormal behavior of personnel and then to security alarms, a three-level response mechanism ensures that threats of different levels can be appropriately handled. For example, the first-level response quickly locks the access control of the target area and monitors it precisely to prevent the spread of danger; the second-level response guides personnel evacuation through lighting and alarms to reduce the risk of casualties; the third-level response optimizes the operation of equipment at low risk, providing a stable foundation for security protection, comprehensively ensuring the safety of personnel and property within the building. Dynamically adjusting the equipment operation strategy according to the threat level realizes the reasonable allocation of resources. In the third-level response, the energy-saving mode is switched for non-core production equipment to avoid unnecessary energy consumption of the equipment in the low-risk state; under different response levels, video surveillance, access control, lighting and other equipment are adjusted specifically to avoid waste of resources, reduce the operation cost and improve the resource utilization efficiency on the premise of ensuring safety and comfort. The clear three-level response mechanism and the automated instruction generation and distribution process enable the system to respond quickly when facing alarm signals and threat situations. From threat assessment to the execution of response strategies, without excessive manual intervention, the emergency response time is greatly shortened. For example, in case of an emergency, the evacuation passage lighting and directional acoustic alarms can be quickly activated to guide personnel to evacuate orderly, effectively improving the overall emergency handling ability and efficiency of the building. In addition to security-related responses, this mechanism also takes into account environmental comfort indicators. In daily operation, by adjusting the air volume of the air conditioner and combining with the distribution of crowded areas, it ensures that people in different areas can be in a comfortable environment. At the same time, the refined control and management of equipment make the building operation and management more scientific and intelligent, improving the overall management level and service quality.

[0166] An embodiment of the present invention also provides an intelligent gateway, including: a processor and a memory storing a computer program. When the computer program is run by the processor, it executes the system as described above. All implementation manners in the above system embodiment are applicable to this embodiment and can achieve the same technical effects.

[0167] An embodiment of the present invention also provides a computer-readable storage medium storing instructions. When the instructions are run on a computer, the computer is made to execute the system as described above. All implementation manners in the above system embodiment are applicable to this embodiment and can achieve the same technical effects.

[0168] The above are the preferred implementation manners of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle described in the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.< / continuousmove>

Claims

1. An AI artificial intelligence machine learning system, characterized in that: include: The communication module establishes a two-way connection with the air conditioning units, security sensor arrays and smart lighting networks in the building through multiple IoT protocols, and collects equipment operation data streams in real time, including air conditioning compressor current waveforms, security alarm codes and lighting equipment light intensity timing data; The edge computing module is used to extract the air conditioning energy consumption characteristic value, security alarm signal and environmental temperature and humidity data set according to the equipment operation data stream, and store the processed structured data set into the pseudo-physical memory area through the pseudo-physical classification algorithm; The virtual memory area module is used to implement timestamp difference encoding on the historical power data of the air conditioner in the structured data set to form a spatiotemporal correlation data cube; The engine module is used to identify the thermal map pattern of people moving in the building based on the spatiotemporal correlation data cube in the simulacrum memory area and generate a dynamic temperature control parameter matrix; The control logic module is used to generate control instructions based on environmental comfort indicators, distribution of densely populated areas, recognition of abnormal behavior patterns and security threat levels; when an alarm signal is received, the focus area of ​​video surveillance and access control strategy are dynamically adjusted according to the signal type through a three-level priority response mechanism; The protocol conversion module is used to dynamically convert control instructions into the target device protocol format.

2. The AI ​​artificial intelligence machine learning system according to claim 1, characterized in that: According to the equipment operation data stream, the air conditioning energy consumption characteristic value, security alarm signal and environmental temperature and humidity data set are extracted, and the processed structured data set is stored in the simulated memory area through the simulated classification algorithm, including: Perform sliding window processing on the original device data stream to split the continuous data stream into multiple data segments of equal length; Extract air conditioner energy efficiency feature sets from multiple data segments of equal length, including real-time energy efficiency ratio, current harmonic distortion rate, and load fluctuation index, and generate an alarm frequency density matrix based on security sensor data to obtain a feature data set containing air conditioner energy efficiency features and security event spatiotemporal distribution information; The feature data set is converted into device-level classification labels through the device fingerprint encoding algorithm, and the classification data is mapped to the quasi-physical memory area according to the device entity ID.

3. The AI ​​artificial intelligence machine learning system according to claim 2, characterized in that: Implement timestamp difference encoding on the historical power data of air conditioners in the structured data set to form a spatiotemporal correlation data cube, including: Extract historical power data of air conditioners from structured data sets, group them by device ID and arrange them by timestamp to form a time series power sequence; For each adjacent time stamp pair, the power difference and the time interval are calculated to form a power difference encoding sequence; The difference code sequence is associated with the device geographic coordinates to construct a three-dimensional data cube containing time dimension, space dimension and feature dimension.

4. The AI ​​artificial intelligence machine learning system according to claim 3, characterized in that: According to the spatiotemporal correlation data cube in the virtual memory area, the thermal map pattern of personnel movement in the building is identified, and a dynamic temperature control parameter matrix is ​​generated, including: Extract multi-source probe location data from the spatiotemporal correlation data cube in the quasi-memory area, and process the probe location data using the long short-term memory network to predict the probability of personnel staying in each area in the future and generate a dynamic thermal distribution map; The thermal distribution map is spatially superimposed with the temperature and humidity gradient data stored in the simulated memory area to identify densely populated areas and areas with abnormal environmental parameters, and obtain coupling analysis results; According to the coupling analysis results and combined with the thermal response characteristics of the air-conditioning units, the regional temperature setpoint matrix is ​​dynamically calculated.

5. The AI ​​artificial intelligence machine learning system according to claim 4, characterized in that: The probe location data is extracted from the spatiotemporal correlation data cube in the simulated memory area, and the probe location data is processed using the long short-term memory network to predict the probability of personnel staying in each area in the future and generate a dynamic heat distribution map, including: Extract multi-source probe positioning data from the spatiotemporal correlation data cube of the quasi-memory area, including probe connection events, Bluetooth beacon scanning records, and video analysis coordinate point clouds; The multi-source probe positioning data are grouped by person ID and the spatiotemporal sequence of individual movement trajectories is constructed, and then segmented into continuous trajectory segments through a sliding window; The trajectory segments are input into the pre-trained long short-term memory network, and the probability distribution of personnel staying in each area in the future target period is predicted based on the historical movement rules; The residence probability distribution is spatially interpolated and optimized, and a dynamic thermal distribution map is generated based on the building topology.

6. The AI ​​artificial intelligence machine learning system according to claim 5, characterized in that: According to the coupling analysis results and combined with the thermal response characteristics of the air-conditioning unit, the regional temperature setpoint matrix is ​​dynamically calculated, including: According to the historical temperature data in the building, the dynamic reference temperature is determined, and the probability of occupancy in each area in the future target period is compared with the actual probability of occupancy at the current time point to obtain the change rate of occupancy density; Adopting adaptive learning algorithm to optimize personnel heat load adjustment factor based on thermal response characteristics of air conditioning equipment, including response speed and adjustment range and historical energy efficiency data, including energy consumption and cooling capacity; Collect outdoor environmental humidity monitoring values ​​and meteorological forecast data in real time, and combine them with the thermal inertia parameters of the air-conditioning unit, including heat capacity and heat conduction values, to generate external disturbance compensation factors; The reference temperature, personnel heat load adjustment factor and external disturbance compensation factor are integrated in multiple dimensions to generate a temperature control parameter matrix for spatial partitions.

7. The AI ​​artificial intelligence machine learning system according to claim 6, characterized in that: Generate control instructions based on environmental comfort indicators, distribution of densely populated areas, recognition of abnormal behavior patterns, and security threat levels; when an alarm signal is received, dynamically adjust the focus area of ​​video surveillance and access control strategies based on the signal type through a three-level priority response mechanism, including: The environmental temperature and humidity comfort index, the heat map of crowded areas, the abnormal behavior characteristics of video analysis and the alarm signals of security sensors are aligned in time and space and fused with confidence weighting to generate a comprehensive threat assessment score. According to the comprehensive threat assessment score, the three-level response mechanism is triggered, and the corresponding hierarchical response strategy is generated according to the triggered response level. The hierarchical response strategy includes the video surveillance focus area coordinate set, access control lock logic rules and air conditioning air volume adjustment parameters; The hierarchical response strategy is converted into a control instruction set in the target device protocol format, and the control instruction set is sent to the corresponding device through the corresponding communication protocol to perform corresponding operations, including adjusting the air volume of the air conditioner, adjusting the viewing angle of the surveillance camera, and controlling the electromagnetic lock of the access control.

8. The AI ​​artificial intelligence machine learning system according to claim 6, characterized in that: The three-level response mechanism includes: Level 1 response: When the comprehensive threat assessment score is greater than 90, the target area access control lock and laser tracking video focus are activated; Level 2 response: When 70<comprehensive threat assessment score≤90, activate strobe lighting and directional acoustic alarm in the evacuation passage; Level 3 response: When the comprehensive threat assessment score is ≤70, non-core production equipment, including cooling water circulation pump units and ventilation auxiliary units, switch to energy-saving mode.

9. An intelligent gateway, characterized in that: include: one or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the system according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, and when the program is executed by a processor, the system according to any one of claims 1 to 8 is implemented.

Images