Park environment and energy consumption coordinated regulation and control terminal and method based on edge intelligence

Abstract

The invention discloses a park environment and energy consumption coordinated regulation and control terminal and method based on edge intelligence. According to terminal hardware, an edge AI calculation unit, a high-precision clock source and a multi-protocol interface are physically integrated on the same printed circuit board of a main control module, a unified time reference is provided for multi-source data from the hardware construction level, and millisecond timestamp synchronization of environment and energy consumption data is achieved. Based on this, the edge AI calculation unit executes a trained machine learning model, carries out fusion and intelligent reasoning on synchronized data, directly generates an optimization control instruction or a graded early warning signal locally, and drives an actuator or triggers early warning, thereby realizing local closed-loop control of hundreds of milliseconds. Meanwhile, the terminal maintains reliable connection with the cloud through the dual-mode communication module supporting dynamic switching between the TS-LoRa and the NB-IoT. According to the invention, the technical problems of data dislocation of a heterogeneous system and high cloud response delay are fundamentally solved, and deep fusion and intelligent closed-loop regulation and control of the park environment and energy consumption are realized.

Description

Technical Field

[0001] This invention relates to the fields of smart parks, the Internet of Things and edge computing technology, and in particular to a terminal and method for coordinated control of park environment and energy consumption that integrates multi-source sensing, local intelligent decision-making and dual-mode reliable communication. Background Technology

[0002] The current park environment and energy consumption monitoring system faces two major technological bottlenecks, which severely restrict its intelligent operation and maintenance and safety response capabilities.

[0003] First, environmental monitoring systems generally adopt communication architectures based on Modbus or BACnet protocols, while energy consumption metering systems mainly use communication architectures based on DL / T645 or M-Bus protocols. These two systems are independent in their communication standards and sampling clocks, resulting in a significant time misalignment between the collected environmental parameters (including temperature, humidity, and carbon dioxide concentration) and energy consumption data (including electricity, water, and gas consumption). This time asynchrony prevents effective alignment of multi-source data, hindering accurate correlation analysis and energy efficiency attribution, creating "data silos," and severely impeding the realization of park-level collaborative optimization control. For example, Chinese patent CN 116192906A discloses a "method for integrated digital monitoring of energy in industrial parks based on cloud-edge-device collaboration," which, while achieving energy consumption data collection and cloud analysis, does not involve millisecond-level synchronous collection and fusion with environmental data, failing to solve the data time misalignment problem.

[0004] Second, existing terminals generally lack local intelligent processing capabilities, and all data analysis and decision-making rely on remote cloud processing. In the event of a sudden safety incident (including gas leaks, electrical overload, and a sudden drop in air quality), the round-trip communication delay in the cloud is usually in the range of hundreds of milliseconds to several seconds, far exceeding the safety response window (which is usually required to be less than 100 milliseconds), resulting in the inability to trigger local linkage control in a timely manner, posing a significant safety hazard.

[0005] Furthermore, the high deployment cost, poor scalability, and rudimentary power consumption management of the system further limit its large-scale application in green and low-carbon industrial parks. Therefore, there is an urgent need for a collaborative control terminal and method that can achieve real-time synchronization of multi-source data, local intelligent decision-making, and millisecond-level emergency response at the edge. Summary of the Invention

[0006] The purpose of this invention is to solve the technical problems in the prior art, such as data time misalignment caused by heterogeneous protocols and high response latency caused by cloud decision-making, and to provide a terminal and method for coordinated control of campus environment and energy consumption based on edge intelligence.

[0007] To achieve the above objectives, the present invention provides the following technical solution: On the one hand, a campus environment and energy consumption coordinated control terminal based on edge intelligence is provided, including a main control module (101), a sensor group (102), an energy consumption metering module (103), a dual-mode communication module (104), an actuator interface module (105), an early warning module (106), and a power supply module (107); the main control module (101) is connected to the sensor group (102), the energy consumption metering module (103), the dual-mode communication module (104), the actuator interface module (105), and the early warning module (106) respectively, and the power supply module (107) supplies power to each module; wherein, the main control module (101) integrates an edge AI computing unit (101a) and a high-precision clock source (101b); the edge AI computing unit (101a) is a hardware acceleration unit with a lightweight machine learning model embedded and configured to execute data fusion, conflict resolution, and intelligent inference programs.

[0008] On the other hand, a control method for a campus environment and energy consumption collaborative regulation terminal based on edge intelligence is provided, including the following steps: based on the I²C bus interface and RS-485 interface directly driven by the high-precision clock source (101b) in the main control module (101), environmental data and energy consumption data are collected synchronously, and a unified timestamp generated by the high-precision clock source (101b) is added to each frame of data to completely eliminate time drift caused by heterogeneous protocols; the edge AI computing unit (101a) performs confidence-weighted data fusion and conflict resolution on multi-source data; the edge AI computing unit (101a) runs its internally fixed lightweight machine learning model to perform energy efficiency and heterogeneous data regulation. Preliminary judgment of normal state; the preliminary judgment result is optimized by the edge AI computing unit (101a) in combination with the preset scene template and energy efficiency index to generate the final strategy; according to the final strategy, if there is room for optimization, an optimization control instruction is generated and executed locally through the actuator interface module (105); if there is a safety risk, a graded warning is triggered according to the risk level: when a first-level warning is triggered, the warning module (106) is driven to perform a local sound and light alarm; when a second-level warning is triggered, a remote alarm message is generated and sent through the dual-mode communication module (104); when a third-level warning is triggered, the building management system (BMS) is linked through the actuator interface module (105) to perform emergency control.

[0009] Furthermore, the sensor group (102) includes at least one of a temperature and humidity sensor, a CO2 sensor, a PM2.5 sensor, and a light sensor, and is connected via I 2 The C bus is connected to the main control module (101) to realize hardware-level synchronization of data acquisition.

[0010] Furthermore, the energy consumption metering module (103) is used to collect data on electricity, water consumption and gas consumption, and supports pulse or RS-485 interfaces.

[0011] Furthermore, the dual-mode communication module (104) supports two wireless communication protocols: TS-LoRa self-organizing network and NB-IoT cellular network, and is equipped with a signal strength detection unit; when the TS-LoRa signal strength is lower than a preset threshold, it automatically switches to NB-IoT mode; regardless of the current communication protocol used, the terminal periodically sends heartbeat packets through a timed task to maintain the connection status.

[0012] Furthermore, the I²C interface of the edge AI computing unit (101a), the high-precision clock source (101b), the sensor group (102), and the RS-485 interface of the energy consumption metering module (103) are integrated on the same hardware board and connected through unified circuit board wiring to form the hardware foundation for synchronous data acquisition.

[0013] Furthermore, the early warning module (106) is an audible and visual alarm; the actuator interface module (105) supports the industrial standard protocol for linkage with the building management system (BMS).

[0014] Furthermore, the power module (107) adopts a hybrid power supply system consisting of a solar power supply unit and a rechargeable lithium battery.

[0015] Furthermore, the data fusion and conflict resolution steps include: assigning dynamic confidence weights to each sensor data and performing data fusion using a weighted average algorithm.

[0016] Furthermore, the lightweight machine learning model pre-installed in the edge AI computing unit (101a) is a hybrid regression and classification model trained by the gradient boosting decision tree (GBDT) algorithm.

[0017] Furthermore, the device status data is sent through the dual-mode communication module (104), and the TS-LoRa and NB-IoT communication modes are dynamically switched according to the signal strength. In either communication mode, a heartbeat packet is periodically sent to maintain the network connection status.

[0018] Furthermore, the terminal housing adopts a modular snap-fit ​​structure with a protection level of no less than IP65, facilitating on-site installation and maintenance. Beneficial effects

[0019] Compared with the prior art, the present invention has the following significant advantages: This solution fundamentally addresses the time misalignment issue of environmental and energy consumption data caused by heterogeneous protocols: By physically integrating edge AI computing units, high-precision clock sources, and multi-protocol interfaces onto the same printed circuit board, the high-precision clock source directly drives the data acquisition timing of each interface and assigns a unified timestamp to all data frames. This eliminates time drift caused by multiple independent clock sources at the hardware construction level, achieving millisecond-level alignment of multi-source data and laying a solid foundation for accurate energy efficiency attribution and collaborative optimization.

[0020] Achieve millisecond-level local closed-loop response to park safety incidents: Since control commands are generated by the edge AI computing unit on the terminal and sent directly to the controlled device through the local actuator interface, the second-level delay caused by round-trip communication to the cloud is completely avoided, which can effectively deal with sudden risks such as gas leaks and power overloads.

[0021] Combining high reliability and intelligence in communication: The dual-mode communication module supports dynamic switching between TS-LoRa and NB-IoT, and combined with a heartbeat mechanism, ensures stable uploading of critical status information even in weak network or interference environments. Simultaneously, the terminal possesses local intelligence based on a two-step decision-making mechanism of "model inference - scenario optimization," enabling dynamic and precise optimization of energy consumption strategies.

[0022] Low deployment and maintenance costs and sustainability: The integrated design reduces cabling length and the number of devices, while the modular structure supports rapid replacement and expansion, significantly reducing construction and maintenance costs. The hybrid power supply strategy of solar energy and lithium batteries effectively extends the operating time of equipment on a single charge, meeting the needs of green park construction. Attached Figure Description

[0023] Figure 1 This is a block diagram of the terminal structure of the present invention; Figure 2 This is a schematic diagram of the hardware connection between the dual-mode communication module and the main control module of the present invention; Figure 3 This is a flowchart of the dual-mode communication switching process of the present invention; Figure 4 This is a detailed flowchart of the edge AI algorithm of the present invention. Detailed Implementation

[0024] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be understood that the embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0025] It should be noted that the core protection of this invention lies in the hardware structure of the terminal itself and the control method based thereon. (Appendix) Figure 1The system configuration of the terminal in a typical application scenario is illustrated. As shown in the figure, the terminal establishes a wireless data link with a remote "cloud" through its dual-mode communication module (104) to conduct all data interactions. The "cloud" refers to all remote servers that the terminal can connect to, and its specific internal architecture and implementation are not within the scope of protection of this invention.

[0026] First, the hardware configuration of the terminal device of the present invention is described as follows: like Figure 1 As shown, the park environmental energy consumption monitoring terminal in this embodiment includes a main control module (101), a sensor group (102), an energy consumption metering module (103), a dual-mode communication module (104), an actuator interface module (105), an early warning module (106), and a power supply module (107). The main control module (101) uses an STM32H743VIT6 microcontroller, which has a built-in AI acceleration unit to form the core of the edge AI computing unit (101a), and integrates a high-precision clock source (101b). The sensor group (102) includes an SHT35 temperature and humidity sensor, an MH-Z19B CO2 sensor, a PMS5003 PM2.5 sensor, and a BH1750 light sensor. The energy consumption metering module (103) uses an ADE7953 power metering chip and supports water / gas pulse input. The dual-mode communication module (104) integrates an SX1262 TS-LoRa chip and a BC95 NB-IoT module.

[0027] The sensor group (102) is connected to the main control module (101) via an I²C bus; the energy consumption metering module (103) is connected to the main control module (101) via an RS-485 interface. Crucially, the edge AI computing unit (101a), the high-precision clock source (101b), the I²C bus interface for connecting the sensor group (102), and the RS-485 interface for connecting the energy consumption metering module (103) are all physically integrated on the same printed circuit board of the main control module (101), and connected through a unified wiring layer on this printed circuit board. Furthermore, the data acquisition timing of both the I²C bus interface and the RS-485 interface is directly driven by the high-precision clock source (101b). This hardware configuration ensures the timing consistency of multi-source data acquisition at the physical level, providing a unified hardware foundation for millisecond-level synchronous acquisition of multi-source data, and ensuring the uniformity and high precision of all input data timestamps.

[0028] During operation, the main control module (101) synchronously collects data from various sensors and meters at a fixed frequency. After the data enters the edge AI computing unit (101a), the following core processes are executed sequentially by its internally embedded program: first, confidence-weighted data fusion and conflict resolution are performed; then, a lightweight machine learning model is run for inference; finally, the inference results are optimized by combining the scene template to generate the final strategy.

[0029] Specifically, the lightweight machine learning model is a hybrid regression and classification model trained using the Gradient Boosting Decision Tree (GBDT) algorithm. The model's input features include: synchronized temperature, humidity, CO2 concentration, PM2.5 concentration, illuminance, instantaneous power consumption, historical hourly water consumption, gas flow rate, and an estimated population density calculated by back-calculating the CO2 change rate and regional spatial volume. The model's outputs include: the adjustment amount of the air conditioning set temperature, the on / off status of lighting circuits, and the probability of anomaly risk levels. This model is trained using one year's worth of historical operating data from multiple parks for supervised learning, with the training objective being to minimize energy consumption prediction error and anomaly detection false negative rate.

[0030] If the final strategy determines that the temperature in a certain area is high but the population density is low, an optimized control command is generated and sent to the actuator interface module (105). If a safety risk is determined (such as an abnormal increase in CO2 concentration accompanied by abnormal gas meter readings), the main control module (101) will generate and execute graded warning signals according to the strategy: when a level 1 warning is triggered, the audible and visual alarm of the warning module (106) is activated; when a level 2 warning is triggered, a remote alarm message is generated and sent through the dual-mode communication module (104); when a level 3 warning is triggered, an emergency control command is sent to the building management system (BMS) through the actuator interface module (105). The entire local closed-loop response process can be completed within hundreds of milliseconds.

[0031] like Figure 2As shown, the specific hardware connection between the dual-mode communication module (104) and the main control module (101) is further detailed. The dual-mode communication module (104) internally includes a TS-LoRa communication unit (104a) and an NB-IoT communication unit (104b). The TS-LoRa communication unit (104a) uses an SX1262 chip and is connected to the main control module (101) via an SPI bus. The main control module (101) can periodically read the RSSI register inside this unit to monitor the TS-LoRa signal strength in real time. The NB-IoT communication unit (104b) uses a BC95-G module and exchanges data with the main control module (101) via a UART interface. Furthermore, the main control module (101) provides two independent enable signals (EN_LoRa and EN_NB) to control the operating status of the TS-LoRa communication unit (104a) and the NB-IoT communication unit (104b), respectively. This hardware design provides... Figure 3 The dual-mode communication automatic switching mechanism shown provides a solid hardware foundation.

[0032] In the accompanying figures of this article, Figure 3 and Figure 4 The process steps are prefixed with "COM" and "AI" respectively. This numbering method is only used for clear distinction and reference and does not constitute a limitation on the process itself.

[0033] like Figure 3 As shown, the dual-mode communication switching process of the present invention includes the following steps: COM101: Continuously monitors LoRa signal strength (RSSI).

[0034] COM102: Determine if RSSI is greater than or equal to a preset threshold? (If the determination is yes) then execute COM103: send data and heartbeat packets using TS-LoRa transmission mode.

[0035] (If the determination is negative) then execute COM104: switch to NB-IoT transmission mode to send data and heartbeat packets.

[0036] After sending is completed, wait for one cycle T, and the process returns to COM101 to start a new round of detection and sending.

[0037] like Figure 4 As shown, the edge AI algorithm flow of this invention includes the following detailed steps: AI101: The data input layer receives synchronized environmental, energy consumption, and calculated crowd density data.

[0038] AI102: The data fusion and feature construction layer completes timestamp alignment and performs conflict resolution and feature extraction based on a confidence-weighted method.

[0039] AI103: The model inference layer inputs the fused feature vectors into the pre-trained lightweight machine learning model for inference computation.

[0040] AI104: The decision optimization layer combines preset scenario templates and energy efficiency indicators to optimize the model output and generate the final strategy.

[0041] AI105: Generates optimized control commands and issues them through the actuator interface module (105).

[0042] AI106: Generates graded early warning signals that include risk levels.

[0043] The power module (107) adopts a hybrid power supply scheme of solar panels and rechargeable lithium batteries to provide a sustainable energy supply for the system.

[0044] In summary, this invention achieves deep integration and closed-loop optimization of park environment and energy consumption through the collaborative design of the above-mentioned hardware structure and software process.

Claims

1. An edge intelligence-based park environment and energy consumption collaborative regulation terminal, characterized in that, The application relates to an energy consumption monitoring system, which comprises a main control module (101), a sensor group (102), an energy consumption metering module (103), a dual-mode communication module (104), an actuator interface module (105), a pre-warning module (106) and a power module (107); the main control module (101) is connected with the sensor group (102), the energy consumption metering module (103), the dual-mode communication module (104), the actuator interface module (105) and the pre-warning module (106) respectively; the power module (107) supplies power to each module; the main control module (101) is integrated with an edge AI computing unit (101a) and a high-precision clock source (101b); the edge AI computing unit (101a) is a hardware acceleration unit which is solidified with a lightweight machine learning model and is configured to execute data fusion, conflict resolution and intelligent reasoning programs; the hardware structure of the main control module (101) is configured to support synchronous collection of environmental data and energy consumption data.

2. The terminal according to claim 1, characterized by The sensor group (102) comprises at least one of a temperature and humidity sensor, a CO2 sensor, a PM2.5 sensor and an illumination sensor, and is connected with the main control module (101) through an I2C bus.

3. The terminal according to claim 1, characterized by The energy consumption metering module (103) is used for collecting electric energy, water consumption and gas consumption data and supports an RS-485 interface or a pulse interface.

4. The terminal according to claim 1, characterized by The dual-mode communication module (104) comprises a TS-LoRa communication unit (104a) and an NB-IoT communication unit (104b) and is configured with a signal strength detection unit; when the TS-LoRa signal strength is lower than a preset threshold value, the system is automatically switched to the NB-IoT mode; and no matter which communication protocol is adopted, a heartbeat packet is periodically sent to maintain the connection state.

5. The terminal according to claim 1, wherein The edge AI computing unit (101a), the high-precision clock source (101b), an I2C bus interface for connecting the sensor group (102) and an RS-485 interface for connecting the energy consumption metering module (103) are physically integrated on the same printed circuit board of the main control module (101) and are connected through a unified wiring layer on the printed circuit board, and the data collection time sequences of the I2C bus interface and the RS-485 interface are directly driven by the high-precision clock source (101b).

6. The terminal according to claim 1, wherein The pre-warning module (106) is an audible and visual alarm; and the actuator interface module (105) supports an industrial standard protocol which is linked with a building management system (BMS).

7. The terminal according to claim 1, wherein The power module (107) adopts a hybrid power supply system composed of a solar power supply unit and a rechargeable lithium battery.

8. A control method based on the terminal according to any one of claims 1 to 7, characterized by, The application further discloses an energy consumption monitoring method, which comprises the following steps: Synchronous collection of environmental data and energy consumption data based on an I2C bus interface and an RS-485 interface which are directly driven by a high-precision clock source (101b) in a main control module (101) and stamping of a unified timestamp generated by the high-precision clock source (101b) on each frame of data; Data fusion and conflict resolution of multi-source data based on confidence weighting through an edge AI computing unit (101a). A lightweight machine learning model is run by the edge AI computing unit (101a) to make a preliminary judgment of energy efficiency and abnormal state; The edge AI computing unit (101a) optimizes the preliminary judgment result in combination with a preset scene template and an energy efficiency index to generate a final strategy; According to the final strategy, if there is an optimization space, an optimization control instruction is generated and executed locally through the actuator interface module (105); If there is a security risk, a hierarchical early warning is triggered according to the risk level: when a first-level early warning is triggered, the warning module (106) is driven to perform local sound and light alarm; when a second-level early warning is triggered, a remote alarm message is generated and sent through the dual-mode communication module (104); when a third-level early warning is triggered, the actuator interface module (105) is linked to the building management system BMS to execute emergency control.

9. The method of claim 8, wherein, The data fusion and conflict resolution step includes assigning dynamic confidence weights to each sensor data and using a weighted average algorithm for data fusion.

10. The method of claim 8, wherein, The lightweight machine learning model is a gradient boosting decision tree model.

11. The method of claim 8, wherein, The device state data is sent through the dual-mode communication module (104), and the TS-LoRa and NB-IoT communication modes are dynamically switched according to the signal strength, and a heartbeat packet is periodically sent in any communication mode to maintain the network connection state.

Images