Real-time monitoring method and device fusing environment perception and energy consumption information

Abstract

The embodiment of the invention provides a real-time monitoring method and device fusing environmental perception and energy consumption information, and high-precision data acquisition and adaptive sampling are realized by constructing a heterogeneous edge computing architecture based on a dual-core processor and an FPGA (Field Programmable Gate Array). The system constructs a scene adaptability prediction model through transfer learning, realizes abnormal mode recognition in combination with deep learning, and deploys the model to a neural network accelerator to improve the operation efficiency. Multi-objective optimization and a heuristic algorithm are adopted to realize balance control of environmental comfort and energy consumption efficiency, and intelligent adjustment is realized through a self-adaptive controller and a strategy library. According to the method, the limitation of traditional monitoring analysis is broken through, and an intelligent solution is provided for building environment and energy management.

Description

Technical Field

[0001] This application relates to the field of data processing, and specifically relates to a real-time monitoring method and device that integrates environmental perception and energy consumption information. Background Art

[0002] Traditional environmental energy consumption monitoring methods mainly rely on single data collection and simple control strategies, making it difficult to achieve intelligent management in complex scenarios. Existing technologies have obvious deficiencies in data collection accuracy and real-time performance, and lack in-depth analysis of the correlation between the environment and energy consumption.

[0003] At the same time, existing systems also have limitations in prediction model construction and anomaly recognition. Traditional methods often adopt fixed models, fail to fully consider the particularities of different scenarios, and lack an effective transfer learning mechanism. The system is relatively simple in multi-objective optimization and closed-loop control, making it difficult to achieve the optimal balance between environmental comfort and energy efficiency.

[0004] In addition, existing technologies also need to be improved in terms of edge computing architecture and control strategy adaptability. There is a lack of efficient heterogeneous computing solutions and dynamic adjustment mechanisms, and the intelligent accumulation and reuse of control strategies have not been achieved. Solving these problems is of great significance for improving the intelligent level of environmental energy consumption management. Summary of the Invention

[0005] In view of the problems in the prior art, this application provides a real-time monitoring method and device that integrates environmental perception and energy consumption information, which can break through the limitations of traditional monitoring and analysis and provide an intelligent solution for building environment and energy management.

[0006] To solve at least one of the above problems, this application provides the following technical solutions: In a first aspect, this application provides a real-time monitoring method that integrates environmental perception and energy consumption information, including: Construct an edge computing hardware architecture, including a heterogeneous computing unit composed of a dual-core processor and a field programmable gate array coprocessor, configure a high-precision analog-to-digital conversion circuit and a communication interface in the heterogeneous computing unit, and connect the communication interface to a distributed data collection node; collect environmental and energy consumption data based on the edge computing hardware architecture, establish an adaptive sampling mechanism for the data and store it in a multi-level data cache; Construct an environmental energy consumption prediction model and an anomaly recognition model, establish an energy consumption prediction equation based on environmental impact factors, use the transfer learning method to select and train a scenario adaptation model from a pre-trained model library, and deploy the scenario adaptation model and the anomaly pattern recognition model to a neural network accelerator; A closed-loop feedback control system is established based on the scenario adaptation model. A multi-objective optimization function is constructed with temperature comfort, air quality, and energy consumption indicators, and the balanced control parameters are obtained by solving the function. A control strategy library is established to automatically control the environmental adjustment equipment.

[0007] Further, it includes: Construct an edge computing hardware architecture. A heterogeneous computing unit is composed of a dual-core processor and a field-programmable gate array co-processor. A high-precision analog-to-digital conversion circuit and a programmable gain amplifier are configured in the heterogeneous computing unit. The high-precision analog-to-digital conversion circuit is signal-isolated and conditioned through a digital isolator. An Ethernet communication interface, a controller area network interface, and a serial communication interface are integrated in the heterogeneous computing unit, and the communication interfaces are connected to distributed data acquisition nodes. Based on the edge computing hardware architecture, environmental and energy consumption data are collected to obtain temperature, humidity, air quality, and power parameters. An adaptive sampling mechanism is established for the environmental and energy consumption data, and the sampling frequency is dynamically adjusted according to the working conditions. The sampled data are respectively stored in a multi-level data cache. Construct an environmental energy consumption prediction model and an anomaly recognition model. The environmental and energy consumption data are input into the prediction model for environmental impact factor analysis. An energy consumption prediction equation is established based on the environmental impact factors. A pre-training model library is constructed using transfer learning methods. The model with the highest similarity to the target scenario is selected from the pre-training model library for parameter transfer. The transferred model and the on-site collected data are incrementally trained to obtain a scenario adaptation model. The anomaly pattern recognition model is trained using a deep learning algorithm. The scenario adaptation model and the anomaly pattern recognition model are deployed to a neural network accelerator. A closed-loop feedback control system is established based on the prediction results of the scenario adaptation model. A multi-objective optimization function is constructed with temperature comfort indicators, air quality indicators, and energy consumption indicators. The multi-objective optimization function is solved using a heuristic algorithm to obtain balanced control parameters. The balanced control parameters are input into an adaptive controller, and the control parameters are dynamically adjusted according to real-time environmental and energy consumption data. A control strategy library is established to store the mapping relationship between the control parameters and the corresponding scenarios. Based on the control strategy library, the environmental adjustment equipment is automatically controlled.

[0008] In a second aspect, the present application provides a real-time monitoring device that integrates environmental perception and energy consumption information, including: An architecture construction module for constructing an edge computing hardware architecture, including a heterogeneous computing unit composed of a dual-core processor and a field-programmable gate array co-processor. A high-precision analog-to-digital conversion circuit and a communication interface are configured in the heterogeneous computing unit, and the communication interface is connected to a distributed data acquisition node. Based on the edge computing hardware architecture, environmental and energy consumption data are collected, an adaptive sampling mechanism is established for the data, and the data are stored in a multi-level data cache. A model construction module, configured to construct an environmental energy consumption prediction model and an anomaly recognition model, establish an energy consumption prediction equation based on environmental impact factors, select and train a scenario adaptation model from a pre-trained model library by using a transfer learning method, and deploy the scenario adaptation model and the anomaly pattern recognition model to a neural network accelerator; A monitoring and analysis module, configured to establish a closed-loop feedback control system based on the scenario adaptation model, construct a multi-objective optimization function with temperature comfort, air quality, and energy consumption indicators and solve to obtain balance control parameters, and establish a control strategy library to automatically control environmental regulation devices.

[0009] In a third aspect, the present application provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the real-time monitoring method for fusing environmental perception and energy consumption information are implemented.

[0010] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the real-time monitoring method for fusing environmental perception and energy consumption information are implemented.

[0011] In a fifth aspect, the present application provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the steps of the real-time monitoring method for fusing environmental perception and energy consumption information are implemented.

[0012] As can be seen from the above technical solutions, the present application provides a real-time monitoring method and device for fusing environmental perception and energy consumption information. By constructing a heterogeneous edge computing architecture based on a dual-core processor and an FPGA, high-precision data acquisition and adaptive sampling are realized. The system constructs a scenario adaptation prediction model through transfer learning, combines deep learning to realize anomaly pattern recognition, and deploys the model to a neural network accelerator to improve the operation efficiency. Multi-objective optimization and heuristic algorithms are used to achieve balanced control of environmental comfort and energy consumption efficiency, and intelligent adjustment is realized through an adaptive controller and a strategy library. This method breaks through the limitations of traditional monitoring and analysis and provides an intelligent solution for building environment and energy management. Description of the Drawings

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0014] Figure 1Schematic flowchart of the real-time monitoring method for integrating environmental perception and energy consumption information in the embodiments of the present application; Figure 2 Structural diagram of the real-time monitoring device for integrating environmental perception and energy consumption information in the embodiments of the present application; Figure 3 Schematic structural diagram of an electronic device in the embodiments of the present application.

[0015] Reference numerals: Electronic device 9600, central processing unit 9100, memory 9140, communication module 9110, input unit 9120, audio processor 9130, display 9160, power supply 9170, buffer memory 9141, application / function storage unit 9142, data storage unit 9143, driver program storage unit 9144, antenna 9111, speaker 9131, microphone 9132. Detailed implementation manners

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0017] The acquisition, storage, use, processing, etc. of data in the technical solutions of the present application all comply with the relevant provisions of national laws and regulations.

[0018] Considering the problems existing in the prior art, the present application provides a real-time monitoring method and device for integrating environmental perception and energy consumption information. By constructing a heterogeneous edge computing architecture based on a dual-core processor and an FPGA, high-precision data acquisition and adaptive sampling are achieved. The system constructs a scene adaptability prediction model through transfer learning, combines deep learning to realize abnormal pattern recognition, and deploys the model to a neural network accelerator to improve the operation efficiency. Multi-objective optimization and heuristic algorithms are used to achieve the balanced control of environmental comfort and energy consumption efficiency, and intelligent adjustment is realized through an adaptive controller and a policy library. This method breaks through the limitations of traditional monitoring and analysis and provides an intelligent solution for building environment and energy management.

[0019] In order to break through the limitations of traditional monitoring and analysis and provide an intelligent solution for building environment and energy management, the present application provides an embodiment of a real-time monitoring method for integrating environmental perception and energy consumption information. Refer to Figure 1 , the real-time monitoring method for integrating environmental perception and energy consumption information specifically includes the following content: Step S101: Construct an edge computing hardware architecture, including a heterogeneous computing unit composed of a dual-core processor and a field-programmable gate array co-processor. Configure a high-precision analog-to-digital conversion circuit and a communication interface in the heterogeneous computing unit, and connect the communication interface to a distributed data acquisition node; collect environmental and energy consumption data based on the edge computing hardware architecture, establish an adaptive sampling mechanism for the data and store it in a multi-level data cache; Specifically, construct an edge computing hardware architecture, form a heterogeneous computing unit with a dual-core processor and a field-programmable gate array co-processor. Configure a high-precision analog-to-digital conversion circuit and a programmable gain amplifier in the heterogeneous computing unit, perform signal isolation and conditioning on the high-precision analog-to-digital conversion circuit through a digital isolator. Integrate an Ethernet communication interface, a controller area network interface, and a serial communication interface in the heterogeneous computing unit, and connect the communication interface to a distributed data acquisition node; collect environmental and energy consumption data based on the edge computing hardware architecture, obtain temperature, humidity, air quality, and power parameters, establish an adaptive sampling mechanism for the environmental and energy consumption data, dynamically adjust the sampling frequency according to the working condition status, and store the sampled data in a multi-level data cache respectively; Optionally, in the design of the edge computing hardware architecture of this embodiment, a dual-core ARM Cortex-A72 processor is used as the main control unit, and a heterogeneous computing platform is constructed by connecting with a Xilinx Artix-7 series FPGA co-processor through an AXI bus. The two processing cores adopt an asymmetric multi-processing architecture, where one core runs a real-time operating system and is specifically responsible for data acquisition and signal processing tasks, and the working frequency is set to 1.5 GHz to ensure real-time performance; the other core runs a Linux operating system and is responsible for communication protocol processing and data storage management, and dynamically adjusts the working frequency between 0.6 - 1.5 GHz to balance performance and power consumption.

[0020] In the design process of the analog-to-digital conversion circuit of this embodiment, a Texas Instruments ADS1256 sigma-delta analog-to-digital converter is selected. This model has a 24-bit resolution and a sampling rate of up to 32 kSPS, and can achieve microvolt-level measurement accuracy within the ±2.5V input range. It is combined with a PGA309 programmable gain amplifier to construct a high-precision data acquisition front end. The PGA309 supports precise gain adjustment from 0.125 to 128 times, and has a built-in temperature compensation circuit and linear calibration function. For the output characteristics of different types of sensors, such as the weak voltage signal of a temperature sensor requires a larger gain, while the large signal of a power sensor requires a smaller gain. The gain parameters of the PGA309 are configured in real time through the I2C bus to ensure that the signal always works within the optimal range of the ADC.

[0021] In this embodiment, in the signal isolation and conditioning section, an ADuM7441 four-channel digital isolator is used to provide an electrical isolation capability of 2.5 kV. This isolator adopts iCoupler digital isolation technology, which has better temperature stability and aging characteristics compared to optocouplers. A second-order Butterworth low-pass filter is integrated in the signal conditioning circuit, and the cut-off frequency can be configured through an RC network to suppress high-frequency interference in the industrial field. At the same time, transient voltage suppression diodes and a multi-stage LC filter circuit are configured at the power supply end to effectively prevent common-mode interference and surge impacts in the industrial field.

[0022] In the implementation of the multi-protocol communication interface in this embodiment, the Ethernet interface uses an Intel I210 controller, which supports Gigabit Ethernet communication and the IEEE 1588 time synchronization protocol, integrates jumbo frame support and a TCP offload engine, and is specifically used for high-volume data transmission; the controller area network interface is based on a TI SN65HVD230 transceiver, supports the CAN2.0B protocol, has a communication rate of up to 1 Mbps and excellent anti-interference capabilities, and is mainly used for real-time control instruction transmission; the serial communication interface uses a MAX3485 transceiver, supports the RS-485 bus, and uses differential signal transmission to improve anti-interference capabilities for data interaction with traditional industrial devices.

[0023] In terms of the sensor system configuration in this embodiment, the temperature sensor uses a PT100 platinum resistance, and the lead resistance effect is eliminated through a four-wire connection method, and high-precision temperature measurement is achieved in cooperation with a constant current source excitation circuit; the humidity sensor selects a Honeywell HIH8000 series capacitive sensor, which has a digital output and an internal temperature compensation function; the air quality sensor uses a Bosch BME680 multi-functional environmental sensor, which integrates temperature, humidity, air pressure, and gas sensing units; the power parameter acquisition uses a Hall current sensor and a high-precision voltage transformer from LEM Company to achieve real-time monitoring of three-phase power parameters.

[0024] In the data acquisition strategy of this embodiment, a sampling control mechanism adaptive to the working conditions is developed. By calculating the data change rate (first derivative) and change acceleration (second derivative) in real time, combined with sliding variance analysis, the working state is dynamically divided into a steady state, a fluctuating state, and a mutation state. In the steady state (both the change rate and change acceleration are less than the threshold), a basic sampling frequency of 1 Hz is used; in the fluctuating state (the change rate exceeds the threshold), the sampling frequency is dynamically increased to 10 Hz; in the mutation state (the change acceleration exceeds the threshold), the highest sampling frequency of 100 Hz is enabled to capture the transient process.

[0025] In the design of the data cache system in this embodiment, a three-level storage architecture is constructed. The first-level cache is located in the on-chip Block RAM of the FPGA, with a capacity of 1MB, and a true dual-port configuration is adopted to achieve simultaneous reading and writing; the second-level cache uses 4GB of DDR4 system memory, and high-speed data transmission is achieved through a direct memory access (DMA) controller; the third-level cache uses a 256GB industrial-grade solid-state drive, which supports the power-off protection function and uses a write balancing algorithm to extend the service life. In addition, an automatic storage management strategy based on data timeliness is implemented, and data with high real-time requirements is preferentially stored in the high-speed cache.

[0026] Through the above technical solutions, this embodiment realizes the high-precision and high-reliability acquisition of industrial field environment and energy consumption data. In practical applications, this solution successfully solves the problems of large signal interference, low data accuracy, and poor real-time performance of traditional acquisition systems in complex industrial environments, laying a solid foundation for subsequent data analysis and intelligent control.

[0027] Step S102: Construct an environmental energy consumption prediction model and an anomaly recognition model, establish an energy consumption prediction equation based on environmental impact factors, select and train a scenario adaptation model from a pre-trained model library using transfer learning, and deploy the scenario adaptation model and the anomaly pattern recognition model to a neural network accelerator; Specifically, construct an environmental energy consumption prediction model and an anomaly recognition model, input the environmental and energy consumption data into the prediction model for environmental impact factor analysis, establish an energy consumption prediction equation based on the environmental impact factors, construct a pre-trained model library using transfer learning, select the model with the highest target scenario similarity from the pre-trained model library for parameter migration, perform incremental training on the migrated model and the on-site collected data to obtain a scenario adaptation model, train the anomaly pattern recognition model using a deep learning algorithm, and deploy the scenario adaptation model and the anomaly pattern recognition model to a neural network accelerator; Optionally, in the process of constructing the environmental energy consumption prediction model in this embodiment, multi-dimensional correlation analysis is first performed on the collected environmental and energy consumption data. By calculating the Pearson correlation coefficient and the time series mutual information between temperature, humidity, air quality indicators and energy consumption data, a parameter correlation network is constructed. For the air conditioning system, the correlation between temperature and energy consumption is usually the strongest, followed by humidity; for the fresh air system, the correlation between air quality indicators and energy consumption is relatively high. Based on this correlation characteristic, a feature fusion mechanism with adaptive weights is designed.

[0028] In this embodiment, in the analysis of environmental impact factors, a feature importance evaluation method based on XGBoost is adopted. By constructing multiple decision trees, the information gain of each feature at the splitting node is calculated, and the importance score of the feature is cumulatively obtained. For the data features in different time periods, such as weekdays and rest days, day and night, feature importance models are established respectively to capture the time-varying characteristics of the impact of environmental factors.

[0029] In this embodiment, in the construction of the energy consumption prediction equation, a multi-level prediction framework is developed. The autoregressive integrated moving average model is used in the basic layer to capture the periodic change law of energy consumption; the support vector regression model is used in the middle layer to process the non-linear impact of environmental factors; the deep neural network is used in the top layer to integrate multi-source information for final prediction. The models at each level are adaptively fused through the attention mechanism, enhancing the robustness of the prediction.

[0030] In the process of constructing the pre-trained model library in this embodiment, historical operation data is collected for different types of building scenarios (such as office buildings, shopping malls, factories, etc.) to train the scene feature extractor. The contrastive learning method is adopted to enable the model to learn the common features and individual features of different scenarios. The ResNet50 is used as the backbone network for the pre-trained model, and the energy consumption trend and abnormal patterns are predicted simultaneously through multi-task learning.

[0031] In the model migration link of this embodiment, a model selection strategy based on scene similarity is designed. First, the time series feature vector of the target scene is extracted, including the statistical features, periodic features, and fluctuation features of the energy consumption curve. Then, the cosine similarity with each scene in the pre-trained model library is calculated, and the model with the highest similarity is selected for migration. When migrating the parameters, a progressive fine-tuning strategy is adopted, first fixing the low-level feature extraction parameters and only training the parameters related to the high-level tasks.

[0032] In the incremental training process of this embodiment, a dynamic sample weight adjustment mechanism is implemented. Higher weights are assigned to the latest collected data samples, and the weights of historical data samples decay over time, enabling the model to adapt to scene changes in a timely manner. At the same time, by introducing the knowledge distillation technology, it is ensured that the model does not forget the original knowledge when adapting to new scenarios.

[0033] In the design of the abnormal pattern recognition model in this embodiment, a long short-term memory network (LSTM) is used to construct a time series anomaly detector. The input features include the original sensor data and the calculated statistical features, and the time series dependence relationship is extracted through multiple LSTM units. During the training process, an adversarial generative network (GAN) is used to generate diverse abnormal samples to enhance the generalization ability of the model.

[0034] In this embodiment, during the model deployment phase, the TensorRT framework is used to optimize the trained model. Through techniques such as operator fusion, weight quantization, and dynamic batching, the inference speed of the model is significantly improved. The deployment platform selects the NVIDIA Jetson Xavier NX embedded AI computing module, which has powerful neural network acceleration capabilities.

[0035] Through the above technical solutions, this embodiment realizes the accurate prediction and abnormal monitoring of environmental energy consumption. In practical applications, this solution can accurately predict the energy consumption change trends in different scenarios, detect abnormal energy consumption patterns in advance, and provide a reliable basis for energy management decisions. By combining transfer learning and incremental training, the data requirements for the model to adapt to new scenarios are significantly reduced, and the deployment cycle is accelerated. At the same time, the optimized model can achieve millisecond-level inference responses on edge devices, meeting the requirements of real-time monitoring.

[0036] Step S103: Based on the scenario adaptation model, establish a closed-loop feedback control system, construct a multi-objective optimization function with temperature comfort, air quality, and energy consumption indicators, solve it to obtain balance control parameters, and establish a control strategy library to automatically control the environmental adjustment equipment.

[0037] Specifically, based on the prediction results of the scenario adaptation model, establish a closed-loop feedback control system, construct a multi-objective optimization function with temperature comfort index, air quality index, and energy consumption index, use a heuristic algorithm to solve the multi-objective optimization function to obtain balance control parameters, input the balance control parameters into an adaptive controller, dynamically adjust the control parameters according to real-time environment and energy consumption data, establish a control strategy library to store the mapping relationship between the control parameters and the corresponding scenarios, and automatically control the environmental adjustment equipment based on the control strategy library.

[0038] Optionally, during the construction of the closed-loop feedback control system in this embodiment, a hierarchical control architecture design is adopted. Multiple proportional-integral-derivative (PID) controllers are configured at the bottom layer, which are respectively responsible for the basic adjustment of temperature, humidity, and air quality; the model predictive control (MPC) algorithm is implemented in the middle layer, and feedforward compensation is performed based on the prediction results of the scenario adaptation model; an intelligent coordination controller is deployed at the top layer, which is responsible for multi-objective optimization and strategy scheduling.

[0039] In the design of the temperature comfort evaluation index in this embodiment, an improved Predicted Mean Vote (PMV) model is adopted. This model comprehensively considers factors such as air temperature, mean radiant temperature, relative humidity, air velocity, human metabolic rate, and clothing thermal resistance. For different functional areas, such as office areas, meeting rooms, corridors, etc., different comfort target ranges are set. The metabolic rate parameters are dynamically adjusted according to the real-time personnel density detection results to improve the evaluation accuracy.

[0040] In the construction of air quality indicators in this embodiment, an evaluation system integrating multiple parameters is developed. In addition to the conventional PM2.5 and CO2 concentration indicators, characteristic parameters such as TVOC concentration and formaldehyde concentration are also introduced. The fuzzy comprehensive evaluation method is used to establish an analytic hierarchy process model to determine the weight coefficients of various indicators. According to the indoor personnel distribution characteristics and activity patterns, the evaluation criteria are dynamically adjusted.

[0041] In the design of energy consumption indicators in this embodiment, a comprehensive evaluation mechanism based on unit area is realized. An energy consumption sub-item model including air conditioning systems, fresh air systems, and lighting systems is established, and the total energy consumption indicator is obtained through weighted summation. The weight coefficients are dynamically adjusted based on equipment operation efficiency and environmental impact degree to ensure that energy-saving measures do not significantly affect indoor environmental quality.

[0042] In the construction of the multi-objective optimization function in this embodiment, a solution framework based on Pareto optimality is adopted. The temperature comfort index, air quality index, and energy consumption index are normalized to a unified dimension, and the balance between objectives is achieved by setting dynamic weight coefficients. In different time periods and scenarios, the dynamic conversion of optimization objectives is realized by adjusting the weight matrix.

[0043] In the design of the heuristic algorithm in this embodiment, an improved ant colony optimization algorithm is developed. By introducing an adaptive pheromone update strategy, the convergence speed and solution quality of the algorithm are improved. In the process of path construction, a simulated annealing mechanism is combined to avoid falling into local optimal solutions. A smooth transition parameter adjustment mechanism is designed for the continuity requirements of control parameters.

[0044] In the implementation of the adaptive controller in this embodiment, a parameter identification method based on least squares recursion is adopted. By analyzing the deviation between the control effect and the target value in real time, the controller parameters are dynamically adjusted. Dead zone compensation and feedforward compensation mechanisms are introduced to improve control accuracy and response speed. For parameters with large time lag characteristics such as temperature, a Smith predictor structure is adopted to eliminate the time delay effect.

[0045] In the construction of the control strategy library in this embodiment, a hierarchical storage mechanism based on scenario characteristics is realized. Each control strategy includes a scenario descriptor, environmental parameter range, control parameter configuration, and execution effect evaluation. The mapping relationship between scenario characteristics and control strategies is established through a decision tree algorithm to support fast strategy matching and online update.

[0046] In the control process of environmental conditioning equipment in this embodiment, a device collaborative optimization mechanism is developed. By establishing a device group control model, the operating states of equipment such as air conditioning hosts, fan coils, and fresh air units are coordinated. The start-stop optimization and frequency conversion adjustment of equipment based on load prediction are realized, significantly improving the system operation efficiency.

[0047] Through the above technical solutions, this embodiment realizes the intelligent control of the building environment. This solution can automatically adjust the control strategy according to the scenario requirements and external environment changes, achieving efficient utilization of energy while ensuring the comfort of the indoor environment. Through multi-objective optimization and strategy adaptation, it effectively solves the problem that it is difficult to balance comfort and energy conservation in traditional control methods. In practical applications, this solution significantly improves the intelligent level of building environment control and provides reliable technical support for building energy-saving renovation.

[0048] As can be seen from the above description, the real-time monitoring method for integrating environmental perception and energy consumption information provided by the embodiments of this application can achieve high-precision data acquisition and adaptive sampling by constructing a heterogeneous edge computing architecture based on a dual-core processor and an FPGA. The system constructs a scenario-adaptive prediction model through transfer learning, combines deep learning to achieve abnormal pattern recognition, and deploys the model to a neural network accelerator to improve the operation efficiency. It adopts multi-objective optimization and heuristic algorithms to achieve balanced control of environmental comfort and energy consumption efficiency, and realizes intelligent adjustment through an adaptive controller and a strategy library. This method breaks through the limitations of traditional monitoring and analysis and provides an intelligent solution for building environment and energy management.

[0049] In an embodiment of the real-time monitoring method for integrating environmental perception and energy consumption information of this application, it may specifically include the following content: Step S201: Use a dual-core ARM processor as the main controller, configure a field-programmable gate array as a coprocessor to construct a heterogeneous computing unit, and set a high-precision analog-to-digital conversion module and a signal conditioning module in the heterogeneous computing unit. The high-precision analog-to-digital conversion module consists of a 24-bit sigma-delta analog-to-digital converter and a low-noise programmable gain amplifier, and the signal conditioning module consists of a digital isolator and a low-pass filter. Connect the output end of the signal conditioning module to the input end of the high-precision analog-to-digital conversion module; Step S202: Based on industrial field bus technology, construct a multi-protocol communication interface in the heterogeneous computing unit. The multi-protocol communication interface includes a gigabit Ethernet interface, a controller area network interface, and a serial communication interface. Connect the data bus of the multi-protocol communication interface to the on-chip memory of the heterogeneous computing unit, establish a communication buffer in the on-chip memory, and connect the signal end of the multi-protocol communication interface to the communication module of the distributed data acquisition node.

[0050] Optionally, in the hardware architecture design of the heterogeneous computing unit in this embodiment, a dual-core ARM Cortex-A72 processor is used as the main controller, with a maximum operating frequency of up to 1.8 GHz. The two processing cores adopt a big.LITTLE architecture, where the big core is responsible for complex signal processing and protocol conversion tasks, and the little core is responsible for real-time data acquisition and basic control functions. The processor core is connected to on-chip peripherals through the AMBA bus matrix to achieve high-speed data interaction.

[0051] In the co-processor configuration section of this embodiment, an Xilinx Artix-7 series FPGA is selected, equipped with 100K logic cells and 240 DSP computing units. By configuring dedicated hardware acceleration modules, hardware-level acceleration of algorithms such as digital filtering and FFT transformation is achieved. Multiple clock domains are divided inside the FPGA, and the clock frequency is independently configured for different functional modules to optimize power consumption efficiency.

[0052] In the design of the analog-to-digital conversion module in this embodiment, Texas Instruments ADS1256 is used as the core device. This chip uses sigma-delta modulation technology and can achieve a conversion accuracy of 24 bits under a 5V reference voltage. The integration time of the converter is programmable, supporting a data output rate from 7.5 Hz to 30 kHz to meet the sampling requirements in different scenarios. A low-noise LTC6910-1 programmable gain amplifier is configured at the input end, providing an accurate amplification factor from 1 to 1000 times, and the gain parameter is adjusted in real time through the SPI interface.

[0053] In the implementation of the signal conditioning module in this embodiment, an ADuM7441 digital isolator is used to provide an electrical isolation ability of 3 kV. This isolator uses high-speed magnetic coupling technology, has excellent common-mode transient rejection ability, and the signal transmission delay is less than 50 ns. The low-pass filter adopts a second-order active RC structure, and the cut-off frequency can be adjusted online through a resistor array to effectively suppress high-frequency interference and switching noise in the industrial field.

[0054] In the signal link design of this embodiment, special attention is paid to anti-interference performance. When laying out the printed circuit board, differential trace design is adopted, analog signals and digital signals are routed in separate areas, and shielding layers are added to key signal lines. The power supply system adopts a multi-stage filtering structure, including LC filters and ferrite beads, to ensure that the power supply ripple is less than 1 mV.

[0055] In the implementation of the industrial fieldbus interface in this embodiment, a hierarchical design strategy is adopted. The gigabit Ethernet interface is based on the Intel I210 controller, supports the IEEE 1588 precise time protocol, and realizes network synchronization at the microsecond level. The controller area network interface uses the TI SN65HVD230 transceiver, supports a communication rate of up to 1 Mbps, and has an internal ESD protection circuit. The serial communication interface is based on the MAX3485 chip, supports the RS-485 bus, and the communication rate can reach 10 Mbps.

[0056] In the implementation process of the communication protocol stack in this embodiment, a modular software architecture is adopted. The physical layer and data link layer protocols are implemented in the FPGA, supporting automatic recognition of data frames and CRC check. The network layer and application layer protocols are implemented on the processor side, including industrial protocol stacks such as Modbus TCP, Profinet, and OPC UA. Heterogeneous network interconnection is realized through a protocol conversion gateway.

[0057] In the on-chip memory management of this embodiment, a multi-level cache mechanism is realized. Dual-port RAM is configured inside the FPGA as a communication data buffer, supporting high-speed data throughput. The system memory uses DDR4 technology, and efficient data transmission is realized through the DMA controller. The memory management unit realizes virtual address mapping and access permission control, improving system stability.

[0058] In the design of the distributed node communication module in this embodiment, a star topology structure is adopted. Each acquisition node is equipped with an independent communication processor, supporting the hot plug function. Automatic recognition and configuration of nodes are realized through the network discovery protocol, supporting online device replacement and system expansion. Redundant link design is adopted for network communication to ensure that the system can still work normally in case of a single point of failure.

[0059] Through the above technical solutions, this embodiment constructs a stable and reliable industrial data acquisition platform. This solution provides powerful computing capabilities through a heterogeneous computing architecture, a high-precision data acquisition link ensures signal quality, and multi-protocol communication interfaces meet the access requirements of different devices. In practical applications, this solution successfully solves the data acquisition problem in the complex industrial field environment and provides a reliable data basis for upper-layer applications.

[0060] In an embodiment of the real-time monitoring method for fusing environmental perception and energy consumption information in this application, the following specific contents may also be included: Step S301: Connect the temperature sensor, humidity sensor, air quality sensor, and power sensor to the acquisition channels of the high-precision analog-to-digital conversion module. Based on the signal conditioning module, suppress interference and amplify the sensor input signals. Adjust the gain coefficients of each acquisition channel through the programmable gain amplifier. Use the analog-to-digital converter to perform digital conversion on the conditioned sensor signals, and transmit the converted digital quantities to the on-chip memory through the data bus of the heterogeneous computing unit. Step S302: Establish a working condition judgment model for the environmental and energy consumption data in the on-chip memory. The working condition judgment model divides the working condition states into steady state, wave state, and mutation state based on the data change rate and fluctuation amplitude. Configure corresponding sampling frequency parameters for different working condition states, and store the collected data in a multi-level data cache composed of a field-programmable gate array cache, system memory, and solid-state memory according to time priority.

[0061] Optionally, in this embodiment, a multi-level distributed layout strategy is adopted in the sensor configuration link. The temperature sensor selects a PT100 platinum resistance, and a four-wire connection method is used to eliminate the influence of lead resistance. A constant current source excitation circuit is configured to provide a 1 mA precision excitation current. 3-5 temperature sensors are deployed in each monitoring area according to the area size, and the regional temperature characteristic value is obtained through a weighted average algorithm. The humidity sensor adopts a Honeywell HIH8000 series capacitive sensor, which has digital calibration and temperature compensation functions and maintains good linearity in the range of 0-100%RH.

[0062] In this embodiment, a variety of detection units are integrated in the air quality sensor system. The particulate matter detection adopts the laser scattering principle, and an intake air duct and a laminar flow shaping structure are configured to ensure the measurement accuracy. The CO2 concentration detection adopts non-dispersive infrared technology and supports a measurement range of 0-5000 ppm. The TVOC detection adopts a metal oxide semiconductor sensor, which has a fast response characteristic. The analog output signals of each detection unit are processed by the signal conditioning module and then connected to the high-precision ADC channel.

[0063] In the design of the power parameter acquisition circuit in this embodiment, a two-way sampling scheme is adopted. The voltage sampling is realized through a high-precision voltage transformer to achieve electrical isolation, and a precision voltage dividing network is configured to adjust the signal to the ADC input range. The current sampling adopts a Hall current sensor, which has zero hysteresis characteristics and maintains excellent linearity in the range of -50 A to +50 A. The power calculation is realized in the FPGA, supporting the real-time calculation of active power, reactive power, and power factor.

[0064] In this embodiment, an adaptive gain control mechanism is implemented in the signal conditioning section. By monitoring the signal amplitude in real time and dynamically adjusting the PGA gain coefficient, it is ensured that the signal always operates within the optimal range of the ADC. For temperature signals, the typical gain setting is 128 times; for humidity signals, it is 64 times; for air quality signals, it is set to 16 - 32 times according to the characteristics of different sensors; for power signals, an adjustable gain of 2 - 8 times is adopted.

[0065] In the design of the digital filtering algorithm in this embodiment, a cascaded structure is adopted to achieve multi - stage filtering. The first stage is a hardware low - pass filter, and the cut - off frequency can be adjusted online through an RC network; the second stage implements a FIR filter in the FPGA, supporting a configurable decimation ratio; the third stage implements a Kalman filter in the processor to adaptively estimate the true value of the signal.

[0066] In the process of constructing the working condition judgment model in this embodiment, a state recognition algorithm based on multi - feature fusion is developed. The change rate feature is obtained by calculating the first - order derivative of the data sequence, and the fluctuation amplitude feature is obtained by sliding variance analysis. For temperature parameters, the change rate threshold is set to 0.1 °C / minute, and the fluctuation amplitude threshold is 0.5 °C; for humidity parameters, the corresponding thresholds are 0.5%RH / minute and 2%RH respectively; the thresholds for air quality parameters and power parameters are dynamically adjusted according to the actual application scenario.

[0067] In the sampling frequency control strategy of this embodiment, a hierarchical scheduling mechanism is implemented. Under steady - state working conditions, the temperature and humidity parameters adopt a basic sampling frequency of 1 Hz, the air quality parameters adopt a sampling frequency of 0.2 Hz, and the power parameters adopt a sampling frequency of 10 Hz. When a fluctuating working condition is detected, the sampling frequency of the corresponding parameters is increased by 5 - 10 times. Under sudden - change working conditions, the highest sampling frequency is enabled to achieve precise capture of the mutation process.

[0068] In the multi - level data cache management of this embodiment, a time - priority storage strategy is adopted. The high - frequency sampling data in the last 5 minutes is stored in the on - chip RAM of the FPGA, supporting real - time access; the data in the last 24 hours is stored in the system DDR4 memory, and fast read - write is achieved through the DMA controller; the historical data is stored in an industrial - grade solid - state drive, and a compression storage algorithm is adopted to reduce the storage space occupancy.

[0069] In terms of data integrity protection in this embodiment, a multiple - backup mechanism is implemented. Critical data is stored with double copies, and the data integrity is ensured through CRC verification. In the case of power failure, a super - capacitor is used to provide backup power to ensure that the cached data is safely written to the non - volatile memory. The system also supports the function of resuming data transfer from the breakpoint, and can automatically re - transmit the lost data after a communication interruption.

[0070] Through the above technical solutions, this embodiment realizes the highly reliable acquisition of environmental parameters and energy consumption data in an industrial environment. This solution improves the data credibility through multi-sensor collaborative measurement, ensures the acquisition accuracy through adaptive signal conditioning, and optimizes the system resource utilization efficiency through the sampling strategy adaptable to working conditions. In practical applications, this solution can accurately reflect the dynamic change process of environmental parameters and provide reliable data support for subsequent data analysis and control optimization.

[0071] In an embodiment of the real-time monitoring method for fusing environmental perception and energy consumption information in this application, the following content may also be specifically included: Step S401: Conduct a correlation analysis on the environmental and energy consumption data in the multi-level data cache, calculate the Pearson correlation coefficients between the temperature parameter, humidity parameter, air quality parameter, and energy consumption data, establish an environmental impact factor weight matrix based on the correlation coefficients, input the weight matrix into a gradient boosting decision tree to construct an energy consumption prediction benchmark model, and use the benchmark model to fit the historical data to obtain an energy consumption prediction equation; Step S402: Construct a pre-trained model for different application scenarios based on a convolutional neural network, store the pre-trained model in the pre-trained model library according to the application scenario type, calculate the cosine similarity of the feature vectors between the target scenario and each scenario in the pre-trained model library, select the pre-trained model with the highest similarity, transfer the convolutional layer parameters of the pre-trained model to the target scenario, re-train the parameters of the fully connected layer, and import the re-trained model parameters into the energy consumption prediction equation.

[0072] Optionally, in the process of correlation analysis of environmental and energy consumption data in this embodiment, a multi-time scale feature extraction method is adopted. First, perform time alignment and outlier processing on the original data, and calculate the statistical features at the hourly, daily, and weekly levels through the sliding window technique. For the temperature parameter, extract features such as the average value, maximum value, minimum value, and fluctuation range; for the humidity parameter, pay attention to the change trend of relative humidity and the saturation index; for the air quality parameter, calculate the cumulative exposure and over-standard duration of each pollutant concentration.

[0073] In this embodiment, a dynamic time lag analysis mechanism is realized in the calculation of the Pearson correlation coefficient. Considering the time delay in the impact of environmental parameters on energy consumption, the optimal time lag value is determined through sliding correlation analysis. The temperature parameter usually shows a lag effect of 15 - 30 minutes, the lag time of the humidity parameter is about 30 - 45 minutes, and the air quality parameter takes 1 - 2 hours to fully reflect in the energy consumption change.

[0074] In the construction process of the environmental impact factor weight matrix in this embodiment, an adaptive weight allocation strategy is adopted. Based on the significance test results of the correlation coefficient, the weight coefficients of each environmental factor are dynamically adjusted. Under the summer cooling condition, the weight of the temperature factor is usually the highest, followed by the humidity factor; under the winter heating condition, the temperature factor still dominates, but the influence of the outdoor humidity is relatively small; in the transitional season, the weight of the air quality factor will increase accordingly.

[0075] In the design of the gradient boosting decision tree model in this embodiment, the XGBoost framework is used to implement it. The input features of the model include the original values, statistical features, and time series features of environmental parameters, and the output is the energy consumption prediction values at multiple future time points. The hyperparameters of the model are optimized by the grid search method, including the maximum depth of the tree, the number of samples in the minimum leaf node, and the learning rate, etc. A regularization term is introduced to control the model complexity and prevent overfitting.

[0076] In the construction process of the energy consumption prediction equation in this embodiment, a piecewise linear correction mechanism is implemented. By analyzing the distribution characteristics of the prediction residuals, different correction coefficients are adopted in different load intervals. For the partial load condition, the equipment efficiency curve is introduced as a correction factor; for the maximum load condition, the influence of equipment performance attenuation is considered for correction.

[0077] In the construction link of the pre-trained model in this embodiment, a multi-branch convolutional neural network structure is designed. The time series feature branch uses a one-dimensional convolutional layer to extract the local features of the time series; the environmental feature branch uses a two-dimensional convolutional layer to process the spatial distribution features; the equipment status branch integrates the discrete features through a fully connected layer. The features of each branch are adaptively fused through the attention mechanism.

[0078] In the construction process of the scene feature vector in this embodiment, a multi-dimensional feature extractor is developed. The building physical features include area, orientation, insulation performance, etc.; the usage pattern features include personnel density, equipment operation time, etc.; the environmental features include season features, weather patterns, etc. The feature dimension is reduced by principal component analysis, and the most representative feature combination is retained.

[0079] In the model migration process of this embodiment, a progressive fine-tuning strategy is implemented. First, the underlying convolutional parameters of the pre-trained model are fixed, and these parameters are mainly responsible for extracting general low-level features; then the high-level convolutional parameters are unfrozen, and the target scene data is used for fine-tuning to make the model gradually adapt to the feature distribution of the new scene; finally, the fully connected layer is retrained to establish the mapping relationship between the features and the prediction target.

[0080] In this embodiment, during the parameter retraining process, knowledge distillation technology is adopted. The output of the pre-trained model is used as the soft label to guide the training process of the target model. The smoothness of the soft label is adjusted through the temperature parameter to balance the relationship between knowledge transfer and scenario adaptation. At the same time, a contrastive learning loss function is introduced to enhance the discriminative ability of the model for scenario features.

[0081] Through the above technical solutions, this embodiment realizes the accurate prediction of environmental energy consumption. This solution reveals the internal relationship between environmental factors and energy consumption through multi-dimensional correlation analysis, and quickly constructs a prediction model adapted to the new scenario based on the transfer learning method. In practical applications, this solution can accurately predict the energy consumption change trend in different scenarios and provide decision-making support for energy optimization management.

[0082] In an embodiment of the real-time monitoring method for fusing environmental perception and energy consumption information in this application, the following content may also be specifically included: Step S501: Set the migrated model as the basic network structure, construct a multi-branch convolutional network including a temperature feature extraction layer, a humidity feature extraction layer, an air quality feature extraction layer, and an energy consumption feature extraction layer, perform incremental training on the multi-branch convolutional network based on the on-site collected data in the multi-level data cache, evaluate the training results using the cross-validation method, and determine the model with the training error meeting the preset threshold as the scenario adaptation model; Step S502: Construct an anomaly pattern recognition model based on the long short-term memory network, input the environmental and energy consumption data into the long short-term memory network according to the time series, optimize the network parameters using the backpropagation algorithm, quantize the network structure and weight parameters of the scenario adaptation model and the anomaly pattern recognition model into fixed-point number forms, and deploy the quantized models to the neural network accelerator composed of tensor processing units.

[0083] Optionally, in the process of constructing the multi-branch convolutional network in this embodiment, a modular feature extraction architecture is adopted. The temperature feature extraction layer consists of three parallel one-dimensional convolutional branches, which respectively extract the temperature change features in the short term (within 1 hour), medium term (within 24 hours), and long term (within 7 days). The size of the convolutional kernel is dynamically adjusted according to the time span. Small-sized convolutional kernels are used to capture detailed changes in shorter time scales, and large-sized convolutional kernels are used to extract periodic features in longer time scales.

[0084] In the design of the humidity feature extraction layer in this embodiment, an adaptive receptive field mechanism is realized. Through the dilated convolution technology, the receptive range is expanded without increasing the number of parameters. The dilation rate parameter is dynamically adjusted according to the humidity change characteristics in different seasons. At the same time, a residual connection structure is introduced to ensure the gradient propagation effect of the deep network and improve the convergence performance of the model.

[0085] In this embodiment, a multi-scale feature fusion module is developed in the air quality feature extraction layer. The inception structure (Inception Module proposed by Google) parallel convolution branches (parallel convolution branches) are adopted to process air quality indicators with different granularities simultaneously. The channel attention mechanism is used to adaptively weight the features of different pollutant indicators to highlight the influence of major pollutants. A cross-channel pooling layer is designed to capture the synergistic effect between pollutants.

[0086] In the implementation process of the energy consumption feature extraction layer of this embodiment, a hierarchical feature learning network is constructed. The bottom layer uses a feedforward neural network to process the base load features; the middle layer uses a bidirectional LSTM network to model the temporal dependence relationship; the top layer uses a transformer structure (transformer structure) to achieve long-range dependence modeling. The features of each layer are adaptively fused through a gating mechanism.

[0087] In the design of the incremental training strategy of this embodiment, a progressive learning method is adopted. First, the model is pre-trained using historical data to establish a basic feature representation. Then, the newly collected data is used for online fine-tuning to update the model parameters. To prevent catastrophic forgetting, an elastic weight consolidation algorithm is introduced to protect the key parameters from being over-updated.

[0088] In the cross-validation evaluation link of this embodiment, a stratified sampling verification mechanism is implemented. The data set is stratified according to the time span and working condition type to ensure that the verification set contains various typical scenarios. A weighted loss function based on the working condition is designed to assign higher weights to important working conditions. The confidence interval of the model prediction is estimated by the bootstrap method.

[0089] In the design of the long short-term memory network of this embodiment, a stacked LSTM structure is adopted. The first layer of LSTM is responsible for extracting the basic temporal features; the second layer of LSTM selectively focuses on the time steps related to abnormal patterns through the attention mechanism; the third layer of LSTM integrates multi-scale features for anomaly recognition. A skip connection is introduced into the network to alleviate the problem of gradient disappearance.

[0090] In the process of backpropagation optimization of this embodiment, an adaptive learning rate adjustment mechanism is implemented. The Adam (cosine annealing learning rate strategy) optimizer is adopted, combined with the cosine annealing learning rate strategy. A larger learning rate is used at the beginning of training for rapid convergence, and a smaller learning rate is used in the later stage for fine adjustment. Gradient clipping technology is used to prevent gradient explosion and improve the training stability.

[0091] In this embodiment, during the model quantization process, a mixed-precision quantization strategy is adopted. For computationally intensive convolutional layers, 8-bit fixed-point numbers are used for representation; for fully connected layers sensitive to precision, 16-bit fixed-point number precision is retained. Through quantization-aware training, the impact of quantization errors on model performance is compensated. A lookup table is designed to accelerate fixed-point multiplication operations.

[0092] In the neural network accelerator deployment phase of this embodiment, the optimization of the computational pipeline is achieved. Intermediate data transmission is reduced through operator fusion, and matrix operations are accelerated by leveraging tensor core parallelization. A double-buffering mechanism is designed to achieve pipeline parallelism between computation and data loading. Through dynamic batch processing technology, the inference throughput is optimized.

[0093] Through the above technical solutions, this embodiment realizes the intelligent analysis and anomaly detection of environmental energy consumption data. This solution captures the feature patterns of different environmental parameters through a multi-branch feature extraction network, and realizes the continuous optimization of the model based on the incremental learning method. In the model deployment phase, through quantization optimization and accelerator adaptation, the inference performance is ensured. In practical applications, this solution can accurately identify abnormal energy consumption patterns and provide early warning support for building energy conservation management.

[0094] In an embodiment of the real-time monitoring method for fusing environmental perception and energy consumption information in this application, the following specific content may also be included: Step S601: Use the prediction result of the scenario adaptation model as a feedforward signal, and use the measured value of the environmental and energy consumption data as a feedback signal. Based on a proportional-integral-derivative (PID) controller, construct a closed-loop feedback control system. Set the integral time constant and derivative time constant of the PID controller, and connect the output end of the controller to the signal input end of the actuator. Step S602: Construct a temperature comfort evaluation model based on the prediction average voting method, and construct an air quality evaluation model based on the fuzzy comprehensive evaluation method. Combine the output results of the evaluation models with the energy consumption index per unit area to construct a weighted normalized multi-objective optimization function. Use the particle swarm optimization algorithm to perform iterative calculations on the multi-objective optimization function, and use the optimal solution obtained by iteration as the balance control parameter to input into the closed-loop feedback control system.

[0095] Optionally, during the construction of the closed-loop feedback control system in this embodiment, a hierarchical control architecture is adopted. The prediction result of the scenario adaptation model is processed through standardization and then input into the control system as a feedforward compensation signal. The measured data undergoes filtering and scale transformation through a signal conditioning module to form a feedback loop. For different types of environmental parameters, independent PID control channels are designed.

[0096] In this embodiment, an adaptive adjustment mechanism is implemented in the PID controller parameter tuning. The initial parameters are obtained by the Ziegler-Nichols method, and then the control parameters are fine-tuned online based on fuzzy rules. For the temperature control channel, the integral time constant is set relatively large to avoid overshoot, and the derivative time constant is relatively small to improve the response speed; for the humidity control channel, a relatively small integral time constant is adopted to accelerate the elimination of the steady-state error.

[0097] In this embodiment, a seamless switching mechanism is developed in the actuator control strategy design. The variable-frequency device uses a ramp function to achieve soft start and soft stop, avoiding the impact of sudden operating conditions on the device. For multiple parallel devices, the running time is balanced through a polling scheduling algorithm to maximize the device life. A fault switching strategy is designed to automatically switch to the standby device when a single device fails.

[0098] In this embodiment, an improved PMV-PPD (Predicted Mean Vote - Predicted Percentage of Dissatisfaction) algorithm is adopted in the construction of the temperature comfort evaluation model. Parameters such as the human metabolic rate and clothing thermal resistance are configured differently according to different functional areas. The average radiant temperature is obtained through an infrared array sensor, and combined with parameters such as air temperature, relative humidity, and air velocity, the comprehensive comfort index is calculated. The adaptive thermal comfort theory is introduced to consider the influence of seasonal changes on human thermal adaptation.

[0099] In this embodiment, a multi-level index system is constructed in the design of the air quality evaluation model. The first layer includes basic indexes such as particulate matter concentration, gaseous pollutant concentration, and bioaerosol; the second layer introduces derived indexes such as air age and ventilation efficiency; the third layer integrates weight factors such as human exposure time and the distribution of sensitive populations. Each index is mapped to a unified evaluation scale through a fuzzy membership function.

[0100] In this embodiment, an adaptive weight allocation strategy is adopted in the construction of the multi-objective optimization function. The weight coefficients of the comfort index and the air quality index are dynamically adjusted according to the spatial function and usage period, and the weight coefficient of the energy consumption index is related to the energy price and the peak-valley electricity price policy. Normalization processing is performed to ensure the comparability of indexes with different dimensions.

[0101] In this embodiment, a hybrid coding mechanism is designed in the implementation of the particle swarm optimization algorithm. Continuous variables such as the temperature set value are encoded using real numbers, and discrete variables such as the device start-stop state are encoded using binary numbers. By introducing a chaotic map to generate the initial particle swarm, the coverage rate of the search space is improved. An adaptive inertia weight strategy is designed to maintain a large exploration ability in the initial stage of the search and gradually strengthen the local fine search in the later stage.

[0102] In the iterative optimization process of this embodiment, a constraint handling mechanism is implemented. The hard constraints include the operating range and safety limits of the equipment, and the penalty function method is used to ensure the feasibility of the solution; the soft constraints include the energy consumption target and the requirement of smooth regulation, and the Lagrange multiplier method is used to achieve multi-objective balance. An early stopping mechanism is designed to end the iteration in advance when there is no significant improvement in the optimal solution for multiple consecutive generations.

[0103] In the application link of the balance control parameters of this embodiment, a parameter mapping mechanism is constructed. The optimized control parameters are converted into the actual control quantities of the equipment through linear interpolation. For temperature control, the comfort interval is mapped into a combined control strategy of supply air temperature and air volume; for fresh air control, the fresh air ratio is determined based on the enthalpy difference between indoors and outdoors and the air quality requirement.

[0104] Through the above technical solutions, this embodiment realizes the intelligent regulation of the building environment. This solution ensures the stability of the system through closed-loop feedback control, and the multi-objective optimization algorithm realizes the dynamic balance of comfort, air quality and energy conservation. In practical applications, this solution can automatically adjust the control strategy according to the scenario requirements, significantly improving the intelligent level of building environment regulation.

[0105] In an embodiment of the real-time monitoring method for fusing environmental perception and energy consumption information of this application, the following content may also be specifically included: Step S701: Set the balance control parameters as the initial values of the adaptive controller, construct a parameter identification module based on the least squares recursive algorithm, input the environmental and energy consumption data into the parameter identification module to calculate the system model parameters, use the model reference adaptive control method to correct the control parameters online, and transmit the corrected control parameters to the actuator through the digital output interface; Step S702: Establish an association data table including the scene feature vector and the control parameters, store the association data table in the control strategy library, classify the scenes of the data in the control strategy library based on the decision tree algorithm, establish a mapping function between the scene classification result and the control parameters, use the mapping function to quickly match the control parameters of the new scene, and send the obtained control instructions to the control unit of the environmental regulation equipment through the fieldbus protocol.

[0106] Optionally, in the construction process of the parameter identification module of this embodiment, the recursive least squares method is used to realize the online identification of the system model. First, preprocess the environmental and energy consumption data, including removing outliers and data normalization. Extract the dynamic characteristics of the system through the sliding time window technology, and the window length is adaptively adjusted according to the system response characteristics. A forgetting factor is introduced into the identification algorithm to enhance the tracking ability of the system's time-varying characteristics.

[0107] In the calculation of system model parameters in this embodiment, a multi-model switching mechanism is implemented. Local linear models are established for different working conditions, and smooth switching between models is achieved through weight functions. The model structure adopts the ARX form, and the model order is automatically determined by the AIC criterion. A parameter constraint mechanism is designed to ensure that the identified parameters meet the physical meaning.

[0108] In the implementation process of model reference adaptive control in this embodiment, a two-layer control architecture is constructed. The inner layer uses a feedback controller based on model prediction, and the outer layer dynamically adjusts the controller parameters through a parameter adaptation law. The reference model selects a second-order system, and its dynamic characteristics reflect the desired closed-loop response. The adaptation law is designed through the Lyapunov stability theory to ensure the asymptotic stability of the system.

[0109] In the control parameter correction strategy of this embodiment, a progressive adjustment mechanism is developed. The parameter update step size is dynamically adjusted according to the magnitude of the tracking error, using a larger step size for rapid convergence when the error is large and a smaller step size for fine adjustment when the error is small. A parameter normalization algorithm is designed to solve the problem of inconsistent dimensions of different parameters.

[0110] In the design of the digital output interface in this embodiment, a multi-protocol adaptation function is implemented. Standard industrial protocols such as Modbus RTU and BACnet MS / TP are supported, and interconnection between different communication interfaces is achieved through a protocol conversion gateway. A data frame verification mechanism is designed to ensure the reliable transmission of control instructions. A queue buffering strategy is adopted to optimize the communication timing.

[0111] In the process of constructing the scene feature vector in this embodiment, a multi-dimensional feature extraction method is adopted. Physical features include static attributes such as spatial dimensions, orientations, and building materials; operation features include dynamic attributes such as personnel density and equipment load; environmental features include external factors such as meteorological conditions and outdoor pollution. The feature dimension is reduced through principal component analysis, and the most representative feature combination is retained.

[0112] In the management of the control strategy library in this embodiment, a hierarchical storage mechanism is implemented. The core strategies are stored in a relational database, supporting efficient data retrieval and update; the historical strategies are stored in a time-series database to optimize the access performance of large-scale time-series data. A data synchronization mechanism is designed to ensure data consistency between distributed nodes.

[0113] In the implementation of the decision tree algorithm in this embodiment, a random forest model is used for scene classification. Multiple decision trees are generated through bootstrap sampling, and each tree is trained using a subset of features to improve the generalization ability of the model. The Gini coefficient is used as the feature splitting criterion to optimize the growth process of the tree. An online learning mechanism is implemented to support the incremental update of the decision tree.

[0114] In the process of constructing the mapping function in this embodiment, a fuzzy inference mechanism is designed. A fuzzy rule base from scenario features to control parameters is established, and smooth switching of control strategies is achieved through fuzzy inference. A confidence evaluation mechanism is introduced, and when the confidence of the mapping result is lower than the threshold, manual intervention is triggered.

[0115] In the control instruction generation link of this embodiment, a priority scheduling strategy is implemented. Instruction priorities are set according to the importance of control objectives, and the instruction sending frequency is controlled through the token bucket algorithm. An instruction merging mechanism is designed to reduce communication overhead. The instruction rollback function is implemented to support fast recovery in case of exceptions.

[0116] Through the above technical solutions, this embodiment realizes the intelligence and self - adaptability of building environment control. This solution realizes the real - time optimization of the system through online parameter identification and adaptive control, and realizes the rapid reuse of control experience based on scenario classification and policy mapping. In practical applications, this solution can dynamically adjust control strategies according to environmental changes and usage requirements, significantly improving the intelligence level and operation efficiency of building environment control.

[0117] In order to break through the limitations of traditional monitoring and analysis and provide an intelligent solution for building environment and energy management, this application provides an embodiment of a real - time monitoring device for fusion environment perception and energy consumption information, which is used to implement all or part of the content of the real - time monitoring method for fusion environment perception and energy consumption information. Refer to Figure 2 , the real - time monitoring device for fusion environment perception and energy consumption information specifically includes the following: An architecture construction module 10, used to construct an edge - computing hardware architecture, including a heterogeneous computing unit composed of a dual - core processor and a field - programmable gate array coprocessor. A high - precision analog - to - digital conversion circuit and a communication interface are configured in the heterogeneous computing unit, and the communication interface is connected to a distributed data acquisition node; based on the edge - computing hardware architecture, environment and energy consumption data are collected, an adaptive sampling mechanism is established for the data and stored in a multi - level data cache; A model construction module 20, used to construct an environment energy consumption prediction model and an anomaly recognition model, establish an energy consumption prediction equation based on environmental impact factors, select and train a scenario adaptability model from a pre - trained model library using transfer learning methods, and deploy the scenario adaptability model and the anomaly pattern recognition model to a neural network accelerator; A monitoring and analysis module 30, used to establish a closed - loop feedback control system based on the scenario adaptability model, construct a multi - objective optimization function with temperature comfort, air quality, and energy consumption indicators and solve to obtain balanced control parameters, and establish a control strategy library to automatically control environmental regulation equipment.

[0118] As can be seen from the above description, the real-time monitoring device that integrates environmental perception and energy consumption information provided by the embodiments of the present application can achieve high-precision data collection and adaptive sampling by constructing a heterogeneous edge computing architecture based on a dual-core processor and an FPGA. The system constructs a scene adaptability prediction model through transfer learning, combines deep learning to achieve abnormal pattern recognition, and deploys the model to a neural network accelerator to improve the operation efficiency. The multi-objective optimization and heuristic algorithm are used to achieve the balanced control of environmental comfort and energy consumption efficiency, and the intelligent adjustment is realized through the adaptive controller and the policy library. This method breaks through the limitations of traditional monitoring and analysis and provides an intelligent solution for building environment and energy management.

[0119] From a hardware perspective, in order to break through the limitations of traditional monitoring and analysis and provide an intelligent solution for building environment and energy management, the present application provides an embodiment of an electronic device for implementing all or part of the content in the real-time monitoring method of the integrated environmental perception and energy consumption information. The electronic device specifically includes the following: A processor, a memory, a communication interface, and a bus; wherein, the processor, the memory, and the communication interface complete communication with each other through the bus; the communication interface is used to realize the information transmission between the real-time monitoring device that integrates environmental perception and energy consumption information and related devices such as the core business system, the user terminal, and the relevant database. The logic controller can be a desktop computer, a tablet computer, a mobile terminal, etc., and this embodiment is not limited thereto. In this embodiment, the logic controller can be implemented with reference to the embodiments of the real-time monitoring method of the integrated environmental perception and energy consumption information and the embodiments of the real-time monitoring device that integrates environmental perception and energy consumption information, and the content is incorporated herein, and the repeated parts will not be described again.

[0120] It can be understood that the user terminal may include a smart phone, a tablet electronic device, a network set-top box, a portable computer, a desktop computer, a personal digital assistant (PDA), a vehicle-mounted device, a smart wearable device, etc. Among them, the smart wearable device may include smart glasses, a smart watch, a smart bracelet, etc.

[0121] In practical applications, part of the real-time monitoring method of the integrated environmental perception and energy consumption information can be executed on the electronic device side as described above, or all operations can be completed in the client device. Specifically, it can be selected according to the processing capacity of the client device and the limitations of the user usage scenario, etc. The present application does not make any limitation in this regard. If all operations are completed in the client device, the client device may further include a processor.

[0122] The above-mentioned client device may have a communication module (i.e., a communication unit), which can be communicatively connected to a remote server to achieve data transmission with the server. The server may include a server on the task scheduling center side, and in other implementation scenarios, it may also include a server of an intermediate platform, such as a server of a third-party server platform having a communication link with the task scheduling center server. The server may include a single computer device, or a server cluster composed of multiple servers, or a server structure of a distributed device.

[0123] Figure 3 It is a schematic block diagram of the system composition of the electronic device 9600 according to an embodiment of the present application. As Figure 3 shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It should be noted that this Figure 3 is exemplary; other types of structures may also be used to supplement or replace this structure to implement telecommunication functions or other functions.

[0124] In one embodiment, the function of the real-time monitoring method for integrating environmental perception and energy consumption information may be integrated into the central processing unit 9100. Among them, the central processing unit 9100 may be configured to perform the following controls: Step S101: Build an edge computing hardware architecture, including a heterogeneous computing unit composed of a dual-core processor and a field programmable gate array co-processor. Configure a high-precision analog-to-digital conversion circuit and a communication interface in the heterogeneous computing unit, and connect the communication interface to a distributed data acquisition node; collect environmental and energy consumption data based on the edge computing hardware architecture, establish an adaptive sampling mechanism for the data, and store it in a multi-level data cache; Step S102: Build an environmental energy consumption prediction model and an anomaly recognition model, establish an energy consumption prediction equation based on environmental impact factors, select and train a scenario adaptation model from a pre-trained model library using transfer learning methods, and deploy the scenario adaptation model and the anomaly pattern recognition model to a neural network accelerator; Step S103: Build a closed-loop feedback control system based on the scenario adaptation model, construct a multi-objective optimization function with temperature comfort, air quality, and energy consumption indicators, solve to obtain balance control parameters, and establish a control strategy library to automatically control environmental adjustment devices.

[0125] As can be seen from the above description, the electronic device provided by the embodiments of the present application constructs a heterogeneous edge computing architecture based on a dual-core processor and an FPGA to achieve high-precision data acquisition and adaptive sampling. The system constructs a scene adaptability prediction model through transfer learning, combines deep learning to achieve abnormal pattern recognition, and deploys the model to a neural network accelerator to improve the operation efficiency. Multi-objective optimization and heuristic algorithms are used to achieve balanced control of environmental comfort and energy consumption efficiency, and intelligent adjustment is achieved through an adaptive controller and a policy library. This method breaks through the limitations of traditional monitoring and analysis and provides an intelligent solution for building environment and energy management.

[0126] In another embodiment, the real-time monitoring device that fuses environmental perception and energy consumption information can be separately configured from the central processor 9100. For example, the real-time monitoring device that fuses environmental perception and energy consumption information can be configured as a chip connected to the central processor 9100, and the functions of the real-time monitoring method of fusing environmental perception and energy consumption information are realized through the control of the central processor.

[0127] As Figure 3 shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It should be noted that the electronic device 9600 does not necessarily have to include Figure 3 all the components shown in Figure 3 ; in addition, the electronic device 9600 may further include

[0128] components not shown in Figure 3 ; reference may be made to the prior art.

[0129] Among them, the memory 9140 may be, for example, one or more of a buffer, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory, or other suitable devices. The above information related to failures can be stored, and in addition, programs for executing relevant information can also be stored. And the central processor 9100 can execute the program stored in the memory 9140 to achieve information storage or processing, etc.

[0130] The input unit 9120 provides input to the central processor 9100. The input unit 9120 is, for example, a key or a touch input device. The power supply 9170 is used to supply power to the electronic device 9600. The display 9160 is used to display display objects such as images and texts. The display may be, for example, an LCD display, but is not limited thereto.

[0131] The memory 9140 can be a solid-state memory, such as a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It can also be a memory that stores information even when the power is off, can be selectively erased and has more data. Examples of such a memory are sometimes referred to as EPROMs, etc. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 can include an application / function storage unit 9142 that is used to store application programs and function programs or the processes for operating the electronic device 9600 by the central processing unit 9100.

[0132] The memory 9140 can also include a data storage unit 9143 that is used to store data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 can include various drivers of the electronic device for communication functions and / or for performing other functions of the electronic device (such as a messaging application, an address book application, etc.).

[0133] The communication module 9110 is a transmitter / receiver that transmits and receives signals via the antenna 9111. The communication module 9110 (transmitter / receiver) is coupled to the central processing unit 9100 to provide input signals and receive output signals, which can be the same as in the case of a conventional mobile communication terminal.

[0134] Based on different communication technologies, multiple communication modules 9110 can be provided in the same electronic device, such as a cellular network module, a Bluetooth module, and / or a wireless local area network module, etc. The communication module 9110 (transmitter / receiver) is also coupled to the speaker 9131 and the microphone 9132 via the audio processor 9130 to provide an audio output via the speaker 9131 and receive an audio input from the microphone 9132, thereby implementing normal telecommunication functions. The audio processor 9130 can include any suitable buffer, decoder, amplifier, etc. Additionally, the audio processor 9130 is also coupled to the central processing unit 9100, so that recording can be performed on the local machine through the microphone 9132 and the sound stored on the local machine can be played through the speaker 9131.

[0135] Embodiments of the present application also provide a computer-readable storage medium capable of implementing all steps in the real-time monitoring method of fusion environment perception and energy consumption information where the execution entity in the above embodiments is a server or a client. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, it implements all steps of the real-time monitoring method of fusion environment perception and energy consumption information where the execution entity in the above embodiments is a server or a client. For example, when the processor executes the computer program, the following steps are implemented: Step S101: Construct an edge computing hardware architecture, including a heterogeneous computing unit composed of a dual-core processor and a field-programmable gate array co-processor. Configure a high-precision analog-to-digital conversion circuit and a communication interface in the heterogeneous computing unit, and connect the communication interface to a distributed data acquisition node; collect environment and energy consumption data based on the edge computing hardware architecture, establish an adaptive sampling mechanism for the data, and store it in a multi-level data cache; Step S102: Construct an environment energy consumption prediction model and an anomaly recognition model, establish an energy consumption prediction equation based on environmental impact factors, select and train a scenario adaptation model from a pre-trained model library using transfer learning, and deploy the scenario adaptation model and the anomaly pattern recognition model to a neural network accelerator; Step S103: Establish a closed-loop feedback control system based on the scenario adaptation model, construct a multi-objective optimization function with temperature comfort, air quality, and energy consumption indicators, solve to obtain balance control parameters, and establish a control strategy library to automatically control environmental adjustment devices.

[0136] As can be seen from the above description, the computer-readable storage medium provided by the embodiments of the present application realizes high-precision data acquisition and adaptive sampling by constructing a heterogeneous edge computing architecture based on a dual-core processor and an FPGA. The system constructs a scenario adaptation prediction model through transfer learning, combines deep learning to realize anomaly pattern recognition, and deploys the model to a neural network accelerator to improve the operation efficiency. The method adopts multi-objective optimization and heuristic algorithms to achieve the balance control of environmental comfort and energy consumption efficiency, and realizes intelligent adjustment through an adaptive controller and a strategy library. This method breaks through the limitations of traditional monitoring and analysis and provides an intelligent solution for building environment and energy management.

[0137] Embodiments of the present application also provide a computer program product capable of implementing all steps in the real-time monitoring method of fusion environment perception and energy consumption information where the execution entity in the above embodiments is a server or a client. When the computer program / instructions are executed by a processor, they implement the steps of the real-time monitoring method of fusion environment perception and energy consumption information. For example, the computer program / instructions implement the following steps: Step S101: Construct an edge computing hardware architecture, including a heterogeneous computing unit composed of a dual-core processor and a field-programmable gate array co-processor. Configure a high-precision analog-to-digital conversion circuit and a communication interface in the heterogeneous computing unit, and connect the communication interface to a distributed data acquisition node; collect environment and energy consumption data based on the edge computing hardware architecture, establish an adaptive sampling mechanism for the data and store it in a multi-level data cache; Step S102: Construct an environment energy consumption prediction model and an anomaly recognition model, establish an energy consumption prediction equation based on environmental impact factors, select and train a scenario adaptation model from a pre-trained model library using transfer learning, and deploy the scenario adaptation model and the anomaly pattern recognition model to a neural network accelerator; Step S103: Establish a closed-loop feedback control system based on the scenario adaptation model, construct a multi-objective optimization function with temperature comfort, air quality, and energy consumption indicators and solve to obtain balance control parameters, and establish a control strategy library to automatically control environmental adjustment devices.

[0138] As can be seen from the above description, the computer program product provided by the embodiment of the present application realizes high-precision data collection and adaptive sampling by constructing a heterogeneous edge computing architecture based on a dual-core processor and an FPGA. The system constructs a scenario adaptation prediction model through transfer learning, combines deep learning to realize anomaly pattern recognition, and deploys the model to a neural network accelerator to improve the operation efficiency. The multi-objective optimization and heuristic algorithm are used to achieve the balance control of environmental comfort and energy efficiency, and the intelligent adjustment is realized through an adaptive controller and a strategy library. This method breaks through the limitations of traditional monitoring and analysis and provides an intelligent solution for building environment and energy management.

[0139] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, an apparatus, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0140] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (devices), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in the flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks.

[0141] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in the flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks.

[0142] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks.

[0143] Specific embodiments are applied in the present invention to elaborate on the principles and implementation manners of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, based on the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A real-time monitoring method integrating environmental perception and energy consumption information, characterized in that: The method comprises: Construct an edge computing hardware architecture, including a heterogeneous computing unit composed of a dual-core processor and a field programmable gate array coprocessor, configure a high-precision analog-to-digital conversion circuit and a communication interface in the heterogeneous computing unit, and connect the communication interface to a distributed data acquisition node; based on the edge computing hardware architecture, collect environment and energy consumption data, establish an adaptive sampling mechanism for the data, and store it in a multi-level data cache; Construct an environmental energy consumption prediction model and an abnormality recognition model, establish an energy consumption prediction equation based on environmental influencing factors, use a transfer learning method to select and train a scene adaptability model from a pre-trained model library, and deploy the scene adaptability model and abnormal pattern recognition model to a neural network accelerator; A closed-loop feedback control system is established based on the scenario adaptability model, a multi-objective optimization function is constructed for temperature comfort, air quality and energy consumption indicators, and the balance control parameters are solved, and a control strategy library is established to automatically control the environmental conditioning equipment.

2. The real-time monitoring method integrating environmental perception and energy consumption information according to claim 1 is characterized in that: The edge computing hardware architecture is constructed, including a heterogeneous computing unit composed of a dual-core processor and a field programmable gate array coprocessor, a high-precision analog-to-digital conversion circuit and a communication interface are configured in the heterogeneous computing unit, and the communication interface is connected to a distributed data acquisition node, including: Construct an edge computing hardware architecture, combine a dual-core processor and a field programmable gate array coprocessor into a heterogeneous computing unit, and configure a high-precision analog-to-digital conversion circuit and a programmable gain amplifier in the heterogeneous computing unit; The high-precision analog-to-digital conversion circuit is signal isolated and conditioned by a digital isolator, an Ethernet communication interface, a controller area network interface and a serial communication interface are integrated in the heterogeneous computing unit, and the communication interface is connected to a distributed data acquisition node.

3. The real-time monitoring method integrating environmental perception and energy consumption information according to claim 1 is characterized in that: The collecting environment and energy consumption data based on the edge computing hardware architecture, establishing an adaptive sampling mechanism for the data and storing them in a multi-level data cache, includes: Collect environmental and energy consumption data based on the edge computing hardware architecture to obtain temperature, humidity, air quality and power parameters; An adaptive sampling mechanism is established for the environment and energy consumption data, the sampling frequency is dynamically adjusted according to the working condition, and the sampled data is stored in multi-level data caches respectively.

4. The real-time monitoring method integrating environmental perception and energy consumption information according to claim 1 is characterized in that: The construction of the environmental energy consumption prediction model and the abnormality recognition model, and the establishment of the energy consumption prediction equation based on the environmental influencing factors, include: An environmental energy consumption prediction model and an abnormality recognition model are constructed, the environmental and energy consumption data are input into the prediction model to analyze environmental impact factors, and an energy consumption prediction equation is established based on the environmental impact factors.

5. The real-time monitoring method integrating environmental perception and energy consumption information according to claim 1 is characterized in that: The method of adopting the transfer learning method to select and train a scene adaptability model from a pre-trained model library, and deploying the scene adaptability model and the abnormal pattern recognition model to the neural network accelerator includes: An energy consumption prediction equation is established based on the environmental influencing factors, a pre-trained model library is constructed using a transfer learning method, and a model with the highest similarity to the target scene is selected from the pre-trained model library for parameter migration; The migrated model is incrementally trained with the on-site collected data to obtain a scene adaptability model, a deep learning algorithm is used to train the abnormal pattern recognition model, and the scene adaptability model and the abnormal pattern recognition model are deployed in a neural network accelerator.

6. The real-time monitoring method integrating environmental perception and energy consumption information according to claim 1 is characterized in that: The closed-loop feedback control system is established based on the scenario adaptability model, and the temperature comfort, air quality and energy consumption indicators are used to construct a multi-objective optimization function and solve the balance control parameters, including: A closed-loop feedback control system is established based on the prediction results of the scenario adaptability model, and a multi-objective optimization function is constructed by combining the temperature comfort index, the air quality index and the energy consumption index; A heuristic algorithm is used to solve the multi-objective optimization function to obtain the balance control parameters.

7. The real-time monitoring method integrating environmental perception and energy consumption information according to claim 1 is characterized in that: The control strategy library is established to automatically control the environmental conditioning equipment, including: The balance control parameters are input into an adaptive controller to dynamically adjust the control parameters according to real-time environment and energy consumption data; A control strategy library is established to store the mapping relationship between the control parameters and the corresponding scenes, and the environmental adjustment equipment is automatically controlled based on the control strategy library.

8. A real-time monitoring device integrating environmental perception and energy consumption information, characterized in that: The device comprises: An architecture building module is used to build an edge computing hardware architecture, including a heterogeneous computing unit composed of a dual-core processor and a field programmable gate array coprocessor, in which a high-precision analog-to-digital conversion circuit and a communication interface are configured, and the communication interface is connected to a distributed data acquisition node; based on the edge computing hardware architecture, the environment and energy consumption data are collected, an adaptive sampling mechanism is established for the data, and the data is stored in a multi-level data cache; A model building module is used to build an environmental energy consumption prediction model and an abnormality recognition model, establish an energy consumption prediction equation based on environmental influencing factors, use a transfer learning method to select and train a scene adaptability model from a pre-trained model library, and deploy the scene adaptability model and abnormal pattern recognition model to a neural network accelerator; The monitoring and analysis module is used to establish a closed-loop feedback control system based on the scenario adaptability model, construct a multi-objective optimization function for temperature comfort, air quality and energy consumption indicators and solve them to obtain balanced control parameters, and establish a control strategy library to automatically control environmental conditioning equipment.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the real-time monitoring method integrating environmental perception and energy consumption information as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the real-time monitoring method integrating environmental perception and energy consumption information as described in any one of claims 1 to 7 are implemented.

Images