Intelligent control equipment of central air-conditioning system

Abstract

The invention relates to the technical field of intelligent buildings, and discloses intelligent control equipment for a central air-conditioning system, and the equipment comprises a distributed sensor network which is used for collecting indoor and outdoor environment data in real time; the low-power-consumption edge computing unit is connected with the distributed sensor network and is used for primarily processing the environment data; the environment self-adaptive energy management system is provided with a solar energy collection panel and an energy storage unit and is connected with the edge computing unit through a converter; the cross-device cooperative control network is integrated with a wireless communication module; a self-adaptive learning algorithm is adopted, and a recurrent neural network model is adopted to process the time sequence data; and the multi-modal user feedback optimization mechanism is used for detecting temperature distribution and triggering control parameter adjustment based on a vision and thermal imaging feedback system. Multi-dimensional environment data are collected in a grid layout mode through the distributed sensor network, the real-time processing and caching functions of the low-power-consumption edge computing unit are combined, the system response speed is increased, and data collection uniformity and processing efficiency are guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent buildings, and in particular to intelligent control equipment for a central air-conditioning system. Background Art

[0002] With the continuous development of intelligent building technology, central air-conditioning systems, as the core equipment for building environment regulation and energy management, their intelligent control technology has become a key link in improving the overall energy efficiency and living comfort of buildings, involving the integrated application of multiple technical dimensions such as environmental perception, data processing, and equipment collaboration.

[0003] In the existing technology, environmental data collection of central air-conditioning systems mostly adopts a non-grid discrete sensor layout method. The number of sensors is limited and unevenly distributed. They mainly collect a few environmental parameters such as temperature and humidity. Data processing relies on centralized servers. The raw data collected by the sensors needs to be transmitted to the remote server through the communication network for analysis and calculation. The edge end only undertakes the data forwarding function and is not equipped with a computing unit with real-time processing capabilities. It also lacks a data caching mechanism to cope with data transmission fluctuations.

[0004] However, in the process of implementing the technical solution of this application, the inventors of this application found that the above technology has at least the following technical problems: the uniformity of environmental data collection is poor, and data processing has obvious delays due to reliance on remote transmission and centralized computing, which makes it difficult for the system to respond quickly to environmental changes. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides an intelligent control device for a central air-conditioning system, which solves the problem in the existing technology that the uniformity of environmental data collection is poor and the data processing has obvious delays due to reliance on remote transmission and centralized computing, making it difficult for the system to respond quickly to environmental changes.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A central air conditioning system intelligent control device, comprising:

[0007] Distributed sensor network to collect temperature, humidity, population density, light intensity and thermal imaging data in real time;

[0008] A low-power edge computing unit that connects to the sensor network and initially processes environmental data;

[0009] An environmentally adaptive energy management system, including solar energy collection panels and energy storage units, connected to an edge computing unit via a converter to monitor solar energy input and power demand;

[0010] Cross-device collaborative control network, integrated wireless communication module, to achieve multi-device data interaction;

[0011] Adaptive learning algorithm, using recursive neural network model to process time series data and dynamically adjust control parameters;

[0012] Multimodal user feedback optimization mechanism, based on visual and thermal imaging feedback, detects temperature distribution and adjusts control parameters;

[0013] Enhanced user interface supports voice, touch screen and multi-language operation, and integrates thermal imaging charts and energy statistics functions.

[0014] By adopting the above technical solution, the distributed sensor network deploys multiple sensors in a grid layout, which can collect multi-dimensional environmental information such as temperature, humidity, population density, light intensity and thermal imaging data in real time and uniformly, effectively improving the uniformity and comprehensiveness of environmental data collection. The low-power edge computing unit is directly connected to the distributed sensor network, performs local preliminary processing on the collected environmental data, and supports data caching function, avoiding the delay caused by data dependence on remote transmission and centralized computing, ensuring that the system can quickly respond to environmental changes, thereby solving the problems existing in the existing technology.

[0015] Preferably, the adaptive learning algorithm performs standardized preprocessing on temperature, humidity, population density, light intensity and thermal imaging data, uses the gradient descent method to optimize the weight matrix and bias vector, and updates the weights in real time according to environmental changes.

[0016] Preferably, the prediction output formula of the adaptive learning algorithm is:

[0017] ;

[0018] in, For the The control output at each moment, is the input feature vector, is the weight matrix, is the bias vector.

[0019] Preferably, the environmentally adaptive energy management system dynamically adjusts the proportion of solar power and battery power through an energy allocation formula, which is:

[0020] ;

[0021] in, is the total power, is the solar input power, For battery power supply, is the distribution coefficient.

[0022] Preferably, the multimodal user feedback optimization mechanism calculates the average comfort through a comfort evaluation formula, which is:

[0023] ;

[0024] in, For average comfort, For the Thermal imaging value of the area, The total number of regions.

[0025] Preferably, the cross-device collaborative control network coordinates the operation of multiple devices through a regional optimization target formula, which is:

[0026] ;

[0027] in, For regional optimization goals, For the The weight of the device, For the Device No. The control output at each moment, is the total number of devices.

[0028] Preferably, the distributed sensor network deploys humidity sensors, light sensors and thermal imaging sensors in a grid layout to ensure uniformity of data collection.

[0029] Preferably, the low-power edge computing unit adopts the embedded operating system FreeRTOS, communicates with the distributed sensor network through the serial port protocol UART, and supports data caching function.

[0030] Preferably, the enhanced user interaction interface integrates a D3.js library to generate thermal imaging charts and energy usage statistics.

[0031] Preferably, the test optimization step builds a test platform in a constant temperature and humidity laboratory, equipped with a solar simulator and a thermal imaging camera, to verify the stability of the adaptive learning algorithm, the environmental adaptive energy management system, the multimodal user feedback optimization mechanism and the cross-device collaborative control network.

[0032] The present invention provides an intelligent control device for a central air-conditioning system. It has the following beneficial effects:

[0033] 1. The distributed sensor network of the present invention collects multi-dimensional environmental data in a grid layout, and combines the real-time processing and caching functions of the low-power edge computing unit to improve the system response speed and ensure data collection uniformity and processing efficiency.

[0034] 2. The environmentally adaptive energy management system of the present invention coordinates the solar energy collection panels with the energy storage unit, dynamically adjusts the power supply ratio according to the formula, gives priority to the use of solar energy, reduces the consumption of traditional energy, achieves energy conservation and consumption reduction, reduces operating costs, and contributes to green environmental protection.

[0035] 3. The adaptive learning algorithm of the present invention undergoes data standardization preprocessing and uses recursive neural networks and gradient descent to optimize parameters. It can update weights in real time to adapt to environmental changes, improve control accuracy, make air conditioning operation more in line with actual needs, and improve system adaptability and efficiency.

[0036] 4. The multimodal user feedback optimization mechanism of the present invention calculates the average comfort level based on thermal imaging, triggers parameter adjustments in a timely manner, accurately improves temperature distribution, enhances environmental comfort, makes the control more in line with the user's physical needs, and optimizes the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a diagram showing the overall architecture of an intelligent control device for a central air-conditioning system according to the present invention;

[0038] Figure 2 This is a data flow diagram of an intelligent control device for a central air-conditioning system according to the present invention;

[0039] Figure 3 This is a structural diagram of an energy management system for intelligent control equipment of a central air-conditioning system of the present invention;

[0040] Figure 4 This is a diagram of a multimodal user feedback mechanism for an intelligent control device of a central air-conditioning system according to the present invention. DETAILED DESCRIPTION

[0041] The following will clearly and completely describe the technical solution of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0042] Please see the attached Figure 1 -Attached Figure 4 , an embodiment of the present invention provides an intelligent control device for a central air-conditioning system, comprising:

[0043] Distributed sensor network to collect temperature, humidity, population density, light intensity and thermal imaging data in real time;

[0044] A low-power edge computing unit that connects to the sensor network and initially processes environmental data;

[0045] An environmentally adaptive energy management system, including solar energy collection panels and energy storage units, connected to an edge computing unit via a converter to monitor solar energy input and power demand;

[0046] Cross-device collaborative control network, integrated wireless communication module, to achieve multi-device data interaction;

[0047] Adaptive learning algorithm, using recursive neural network model to process time series data and dynamically adjust control parameters;

[0048] Multimodal user feedback optimization mechanism, based on visual and thermal imaging feedback, detects temperature distribution and adjusts control parameters;

[0049] Enhanced user interface supports voice, touch screen and multi-language operation, and integrates thermal imaging charts and energy statistics functions.

[0050] Specifically, the distributed sensor network is deployed in a grid layout, collecting data and transmitting it to a low-power edge computing unit for processing. The environmentally adaptive energy management system adjusts the power supply ratio of the energy storage unit through a DC-DC converter based on the input data from the solar panel. The cross-device collaborative control network enables multi-device data exchange through a wireless communication module. The adaptive learning algorithm calculates and controls the output based on the data processed by the edge computing unit. The multimodal user feedback optimization mechanism generates feedback signals through the visual feedback system and the thermal imaging feedback system, which are input into the adaptive learning algorithm. The enhanced user interaction interface receives user commands through voice control, touch screen operation, and multi-language switching, and displays thermal imaging charts and energy usage statistics.

[0051] The various modules of the system are connected through the serial port protocol UART or ZigBee protocol. After the data stream is processed by the low-power edge computing unit from the distributed sensor network, it is input into the environmental adaptive energy management system, cross-device collaborative control network, adaptive learning algorithm and multimodal user feedback optimization mechanism, and finally presented through the enhanced user interaction interface.

[0052] The adaptive learning algorithm performs standardized preprocessing on temperature, humidity, occupancy density, light intensity, and thermal imaging data, uses the gradient descent method to optimize the weight matrix and bias vector, and updates the weights in real time based on environmental changes. The prediction output formula of the adaptive learning algorithm is:

[0053] ;

[0054] in, For the The control output at each moment, is the input feature vector, is the weight matrix, is the bias vector.

[0055] Specifically, the standardization preprocessing step includes the following operations:

[0056] First, the raw data from the distributed sensor network is collected, and the temperature range is set to , humidity range is to , occupant density is measured in people per square meter, light intensity is measured in lux, thermal imaging data is represented by grayscale values, and data is obtained through: , normalized, where is the normalized eigenvalue, is the original data, are the minimum and maximum values ​​of the corresponding features, respectively.

[0057] The gradient descent optimization process uses batch gradient descent method, and the weight matrix and the bias vector Initialized to zero matrix and zero vector, the optimization goal is to minimize the prediction error, and the error function is defined as in is the mean square error, is the actual control output, is the total number of training samples. The weight and bias vector are updated in each iteration, and the update formula is: and , in is the learning rate, The gradients of the weight matrix and bias vector are calculated by backpropagation based on the chain rule.

[0058] The real-time weight update mechanism is based on environmental change detection. The edge computing unit scans the input feature vector every 5 seconds. When the change of any feature exceeds the preset threshold, the weight re-optimization is triggered. The optimization process is performed under the TensorFlow Lite framework of the low-power edge computing unit, using a recursive neural network model, specifically a long short-term memory network structure, including one hidden layer with 32 nodes, and the input layer receives the normalized , the output layer generates .

[0059] In the algorithm implementation, the data storage uses a circular buffer with a capacity of 1000 records. The excess data will overwrite the earliest data. The weight matrix and the bias vector Stored in the non-volatile memory of the edge computing unit to ensure that the parameters are retained after power failure.

[0060] By standardizing preprocessing multi-dimensional environmental data such as temperature, humidity, and occupancy density, the system eliminates interference caused by scale differences between different data types. Gradient descent is used to optimize the weight matrix and bias vector, continuously reducing prediction error and improving control output accuracy. The edge computing unit scans for feature changes every 5 seconds and triggers real-time weight updates, enabling rapid response to dynamic environmental changes and avoiding control lag. Combined with the recurrent neural network's ability to process time-series data, the system effectively captures the impact of historical data on the current environment, enhancing the rationality of the prediction logic. The efficient data storage of the circular buffer ensures the required historical data volume for learning while adapting to the low-power storage requirements of the edge computing unit. Storing the weights and bias vectors in non-volatile memory ensures parameter loss after power outages, avoiding the need for retraining after system restarts and ensuring continuous and stable operation. This enables the adaptive learning algorithm to accurately and dynamically adjust central air conditioning control parameters in real time, significantly improving the system's adaptability to complex environments and control efficiency.

[0061] The environmental adaptive energy management system dynamically adjusts the proportion of solar power and battery power through the energy allocation formula, which is:

[0062] ;

[0063] in, is the total power, is the solar input power, For battery power supply, is the distribution coefficient.

[0064] Specifically, Collected by solar panels, Provided by the energy storage unit, The allocation coefficient represents the proportion of solar power supply. The solar energy collection panel uses a monocrystalline silicon photovoltaic panel, the energy storage unit is a lithium-ion battery pack, and the charging status is monitored by the battery management system. The DC-DC converter is a buck-boost type to match the power input of the edge computing unit.

[0065] The energy distribution process is performed by a low-power edge computing unit, which collects data every 10 seconds. and electricity demand data , the distribution coefficient The initial value is 0.5, and the formula is adjusted through feedback Dynamic updates, including To adjust the step size, For the The real-time power demand at the moment is calculated by the edge computing unit based on the data of the distributed sensor network and the control output of the adaptive learning algorithm. Increase to give priority to the use of solar energy, when Reduce to increase the battery power ratio. During operation, energy management data is stored in the circular buffer of the edge computing unit with a capacity of 500 records. The excess data will overwrite the earliest data.

[0066] The multimodal user feedback optimization mechanism calculates the average comfort through the comfort evaluation formula, which is:

[0067] ;

[0068] in, For average comfort, For the Thermal imaging value of the area, The total number of regions.

[0069] Specifically, is the average comfort level, indicating the uniformity of the overall temperature distribution in the target area. Collected by thermal imaging feedback system, is the total number of regions, defined as the number of sub-regions the target space is divided into, which is determined by the room size.

[0070] The comfort evaluation process begins with the edge computing unit performing image preprocessing, applying a Gaussian filter to remove noise, and then using The mean clustering algorithm is used to segment the thermal image and calculate the average gray value of each sub-region. By accumulating and divided by The calculation frequency is synchronized with data acquisition, and the temperature adjustment trigger condition is based on the formula in, is the temperature adjustment amount, is the target temperature, is the current average temperature, Conversion (grayscale value 255 corresponds to Corresponding to 0C), is the proportional coefficient, set to 0.5.

[0071] Feedback data transmission uses the I2C protocol. The visual feedback system and thermal imaging feedback system are connected to the edge computing unit through independent I2C buses respectively. The adjustment signal output is sent to the adaptive learning algorithm to trigger weight update. The update cycle is consistent with the acquisition cycle.

[0072] The cross-device collaborative control network coordinates the operation of multiple devices through the regional optimization target formula, which is:

[0073] ;

[0074] in, For regional optimization goals, For the The weight of the device, For the Device No. The control output at each moment, is the total number of devices.

[0075] Specifically, the wireless communication module adopts the ZigBee protocol and is installed on the control panel of each air-conditioning equipment. The network coordinator is located in the low-power edge computing unit and is responsible for device registration and data routing.

[0076] The regional optimization process is performed every 10 seconds. The coordinator collects data from each device. Data, weight The edge computing unit calculates the temperature and personnel density data of the distributed sensor network based on the distribution principle that the greater the temperature deviation, the lower the weight. The formula is: in, is the target temperature, For the The current temperature of the area where the device is located.

[0077] Negotiated updates via formula Implementation, where To negotiate the step size, is the objective function right The partial derivative of is calculated as , the update results are broadcast to each device through the ZigBee protocol.

[0078] The distributed sensor network deploys humidity sensors, light sensors, and thermal imaging sensors in a grid layout to ensure uniformity of data collection; the low-power edge computing unit uses the embedded operating system FreeRTOS, communicates with the distributed sensor network through the serial port protocol UART, and supports data caching.

[0079] Specifically, the distributed sensor network adopts a 5×5 grid layout, with humidity sensors installed at grid nodes, light sensors installed at the top of grid nodes, and thermal imaging sensors installed at the four corners of grid nodes. The sensors are connected to local data acquisition nodes through the I2C bus, and each node is equipped with an 8-bit microcontroller for primary data formatting.

[0080] The low-power edge computing unit communicates with the distributed sensor network through a 4-way UART interface. The data cache function adopts a double buffer and a ping-pong operation mode. When the current buffer is full, it switches to the backup buffer. After the data is written, it is transferred to the SD card via DMA.

[0081] The enhanced user interface integrates the D3.js library to generate thermal imaging charts and energy usage statistics. The grid layout of the distributed sensor network is verified in a constant temperature and humidity laboratory. A thermal imaging camera is used to collect temperature distribution data. Combined with the weight optimization of the adaptive learning algorithm, the uniformity of data collection and the system response speed are verified.

[0082] Specifically, the enhanced user interaction interface is based on the D3.js library, which converts the thermal imaging data collected by the distributed sensor network into dynamic thermal imaging charts, intuitively presenting the indoor temperature distribution. At the same time, it integrates energy usage data to generate multi-dimensional statistical charts, and supports voice control, touch screen operation and multi-language switching to improve user operation convenience and data perception. The performance verification is carried out by building a standardized test platform in a constant temperature and humidity laboratory. The solar simulator can simulate solar energy input scenarios under different light intensities and angles, and the thermal imaging camera captures the indoor temperature field distribution in real time. By simulating diverse environmental conditions, the dynamic adjustment accuracy of the control parameters of the adaptive learning algorithm, the dynamic allocation stability of the environmental adaptive energy management system between solar energy and battery power, the comfort evaluation and parameter adjustment response speed of the multimodal user feedback optimization mechanism based on thermal imaging data, and the timeliness of data interaction between multiple air-conditioning equipment and the achievement of regional optimization goals in the cross-device collaborative control network are verified to ensure the stability and reliability of the collaborative operation of each module.

[0083] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent control device for a central air-conditioning system, characterized in that: include: Distributed sensor network to collect temperature, humidity, population density, light intensity and thermal imaging data in real time; A low-power edge computing unit that connects to the sensor network and initially processes environmental data; An environmentally adaptive energy management system, including solar energy collection panels and energy storage units, connected to an edge computing unit via a converter to monitor solar energy input and power demand; Cross-device collaborative control network, integrated wireless communication module, to achieve multi-device data interaction; Adaptive learning algorithm, using recursive neural network model to process time series data and dynamically adjust control parameters; Multimodal user feedback optimization mechanism, based on visual and thermal imaging feedback, detects temperature distribution and adjusts control parameters; Enhanced user interface supports voice, touch screen and multi-language operation, and integrates thermal imaging charts and energy statistics functions.

2. The intelligent control device for a central air-conditioning system according to claim 1, characterized in that: The adaptive learning algorithm performs standardized preprocessing on temperature, humidity, population density, light intensity and thermal imaging data, uses the gradient descent method to optimize the weight matrix and bias vector, and updates the weights in real time according to environmental changes.

3. The intelligent control device for a central air-conditioning system according to claim 2, characterized in that: The prediction output formula of the adaptive learning algorithm is: ； in, For the The control output at each moment, is the input feature vector, is the weight matrix, is the bias vector.

4. The intelligent control device for a central air-conditioning system according to claim 1, characterized in that: The environmentally adaptive energy management system dynamically adjusts the proportion of solar power and battery power through the energy allocation formula, which is: ； in, is the total power, is the solar input power, For battery power supply, is the distribution coefficient.

5. The intelligent control device for a central air-conditioning system according to claim 1, characterized in that: The multimodal user feedback optimization mechanism calculates the average comfort through the comfort evaluation formula, which is: ； in, For average comfort, For the Thermal imaging value of the area, The total number of regions.

6. The intelligent control device for a central air-conditioning system according to claim 1, characterized in that: The cross-device collaborative control network coordinates the operation of multiple devices through the regional optimization target formula, which is: ； in, For regional optimization goals, For the The weight of the device, For the Device No. The control output at each moment, The total number of devices.

7. The intelligent control device for a central air-conditioning system according to claim 1, characterized in that: The distributed sensor network deploys humidity sensors, light sensors and thermal imaging sensors in a grid layout to ensure uniformity of data collection.

8. The intelligent control device for a central air-conditioning system according to claim 1, characterized in that: The low-power edge computing unit adopts the embedded operating system FreeRTOS, communicates with the distributed sensor network through the serial port protocol UART, and supports data caching function.

9. The intelligent control device for a central air-conditioning system according to claim 1, characterized in that: The enhanced user interaction interface integrates the D3.js library to generate thermal imaging charts and energy usage statistics.

10. An intelligent control device for a central air-conditioning system according to any one of claims 1 to 9, characterized in that: The grid layout of the distributed sensor network was verified by performing performance verification in a constant temperature and humidity laboratory. A thermal imaging camera was used to collect temperature distribution data, and the weight optimization of the adaptive learning algorithm was combined to verify the uniformity of data collection and the system response speed.

Images