An air conditioner control method and system based on area prediction, an air conditioner, and a storage medium

Abstract

The application relates to an air conditioner control method and system based on area prediction, an air conditioner and a storage medium. The air conditioner control method generates a control instruction according to a predicted area parameter and an instantaneous heat load, so that the air conditioner can adjust the air supply and the fan speed in real time according to the actual heat demand, overcooling or overheating under the traditional fixed set value control mode is avoided, unnecessary cooling or heating output is maximally reduced under the premise of ensuring the cooling or heating effect, carbon emission is reduced while energy consumption is saved, energy saving and comfort are synergistically optimized, and good energy saving and environmental protection benefits are achieved.

Description

Technical Field

[0001] This application relates to the field of air conditioning control technology, and more specifically, to an air conditioning control method, system, air conditioner, and storage medium based on area prediction. Background Technology

[0002] In traditional building environments, the design and operation of air conditioning systems typically rely on architectural drawings or manual measurements to determine the space area. However, this approach has many limitations, leading to long-term low-energy-efficiency operation of the air conditioning system, resulting in significant energy waste and carbon emissions.

[0003] First, the data sources are singular and prone to error. Space area information mostly comes from design drawings or manual input. However, in actual projects, drawing versions may not be updated in a timely manner to reflect site changes, and manual measurements may be inaccurate due to improper operation or record-keeping negligence. These factors can all lead to distorted area data, affecting the accuracy of air conditioning load calculations and causing air conditioning selection or control strategies to deviate from actual conditions. The direct consequence of distorted area data is that air conditioning systems operate under conditions deviating from actual needs for extended periods, resulting in the continuous accumulation of ineffective energy consumption and increased carbon emissions.

[0004] Secondly, traditional air conditioning systems lack the ability to sense spatial characteristics. After installation and operation, their control strategies are usually based on initially set area parameters or fixed empirical values, making it difficult to perceive changes in the internal environment. When users add large furniture, change the layout of interior partitions, or alter the function of the space, the effective heat capacity, airflow path, and thermal response characteristics of the space all change. However, traditional air conditioning systems lack the means to detect these changes and continue to use the original control logic, resulting in a mismatch between the air supply parameters and actual heat demand, causing unnecessary energy consumption or a decrease in indoor thermal comfort. Changes in spatial thermal characteristics are often overlooked for a long time, and the continuous output of cooling and heating that does not match actual demand exacerbates energy waste and carbon emissions.

[0005] Furthermore, existing air conditioning systems typically employ fixed setpoints or empirical parameters for control, failing to dynamically optimize based on actual heat load changes. In practice, outdoor weather conditions, indoor occupants, and equipment heat dissipation are constantly fluctuating. Fixed-parameter control methods struggle to adapt to these dynamic fluctuations, often resulting in over-cooling or over-heating, leading to energy waste, or lagging adjustments causing excessive indoor temperature fluctuations, impacting user comfort. This crude control approach results in persistently high air conditioning energy consumption, contradicting the national advocacy for green and low-carbon development, and causing significant waste of electrical resources, generating unnecessary carbon emissions.

[0006] At the technical implementation level, there is a significant disconnect between thermodynamic models and intelligent algorithms in existing research. Thermodynamic models based on physical principles are mostly used for theoretical analysis and offline simulation, while intelligent algorithms such as machine learning primarily rely on large amounts of historical data for pattern recognition and prediction. These two approaches have developed independently and have not yet formed an effective collaborative mechanism. This results in existing air conditioning systems either relying on precisely modeled physical models that are difficult to adapt to complex and variable operating conditions, or on data-driven, purely intelligent algorithms that lack physical interpretability. It is difficult to balance prediction accuracy and control robustness, thus limiting the overall energy efficiency improvement of the system and preventing the full realization of its energy-saving potential.

[0007] Therefore, there is an urgent need for an air conditioning control solution that can overcome the above-mentioned defects, realize automatic sensing and dynamic updating of the space area, and ensure that the air conditioning cooling or heating supply always matches the actual heat demand, thereby effectively reducing the energy consumption and carbon emissions of air conditioning operation and achieving synergistic optimization of energy saving and comfort. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this application provides an air conditioning control method, system, air conditioner, and storage medium based on area prediction. The air conditioning control method adjusts the space where the air conditioner operates to a thermal equilibrium state, calculates the heat exchange based on environmental physical parameters and inherent parameters of the heat exchanger, and ensures that the heat exchange is equal to the instantaneous heat load in the thermal equilibrium state. The environmental physical parameters and the instantaneous heat load are jointly input into the area prediction model, realizing automatic identification of the space area. This allows the air conditioner to adaptively adjust its control strategy according to changes in the space's thermal characteristics, ensuring that the air conditioner output accurately matches the actual heat demand. This effectively avoids overcooling or heating caused by inaccurate area data and fixed setpoint control, significantly reducing air conditioning energy consumption and carbon emissions while ensuring cooling or heating effects, achieving synergistic optimization of energy saving and comfort.

[0009] This solves the problem that traditional air conditioning systems rely on manual input to obtain the space area, lack the ability to automatically sense the thermal characteristics of the space, and thus the control strategy cannot be dynamically optimized according to the actual heat load changes, making it difficult to balance energy efficiency and comfort.

[0010] To achieve the above objectives, this application provides the following technical solution: an air conditioning control method based on area prediction, comprising: Adjust the space where the air conditioner operates to a state of thermal equilibrium; Obtain the environmental physical parameters of the space; Based on the environmental physical parameters and the inherent parameters of the heat exchanger, the heat exchange between the air conditioning system and the space is calculated, and the heat exchange is equal to the instantaneous heat load of the space under the thermal equilibrium state. The environmental physical parameters and the instantaneous heat load are input into the area prediction model to obtain the predicted spatial area; Based on the predicted area of ​​the space and the instantaneous heat load, control commands are generated to adjust the operating parameters of the air conditioning system so that the air conditioning output matches the actual heat demand of the space.

[0011] This application also provides an air conditioning control system based on area prediction, including a heat balance adjustment module, a data acquisition module, a heat exchange calculation module, an area prediction module, and a control module; The heat balance adjustment module is used to adjust the air-conditioned space to a heat balance state. The data acquisition module is used to acquire the environmental physical parameters of the space; The heat exchange calculation module is used to calculate the heat exchange between the air conditioning system and the space based on the environmental physical parameters and the inherent parameters of the heat exchanger. The heat exchange is equal to the instantaneous heat load of the space under the thermal equilibrium state. The area prediction module is used to input the environmental physical parameters and the instantaneous heat load into the area prediction model to obtain the predicted spatial area; The control module is used to generate control commands based on the predicted area of ​​the space and the instantaneous heat load to adjust the operating parameters of the air conditioning system so that the air conditioning output matches the actual heat demand of the space.

[0012] This application also provides an air conditioner, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described above.

[0013] This application also provides a storage medium storing a computer program that, when executed by a processor, implements the method described above.

[0014] Compared with the prior art, this application provides an air conditioning control method, system, air conditioner, and storage medium based on area prediction, which has the following beneficial effects: The air conditioning control method based on area prediction provided in this application adjusts the space under air conditioning to a thermal equilibrium state, calculates the heat exchange based on environmental physical parameters and the inherent parameters of the heat exchanger, and ensures that the heat exchange is equal to the instantaneous heat load in the thermal equilibrium state. The environmental physical parameters and the instantaneous heat load are jointly input into the area prediction model, realizing automatic identification of the space area without relying on manual input or drawing references. When the internal environment of the space changes, the environmental physical parameters change accordingly and are reflected in the prediction results, enabling the air conditioner to automatically sense changes in the thermal characteristics of the space, ensuring that the air conditioning output accurately matches the actual heat demand, effectively avoiding energy waste caused by inaccurate area data, and significantly reducing the energy consumption and carbon emissions of air conditioning operation.

[0015] Based on this, by generating control commands according to the predicted area parameters and instantaneous heat load, the air conditioner can adjust the air volume and fan speed in real time according to the actual heat demand. This avoids overcooling or heating under the traditional fixed setpoint control method. Under the premise of ensuring the cooling or heating effect, unnecessary cold or heat output is reduced to the maximum extent, saving energy and reducing carbon emissions. This achieves synergistic optimization of energy saving and comfort, and has good energy saving and environmental protection benefits. Attached Figure Description

[0016] The present application will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0017] Figure 1 A flowchart of the air conditioning control method provided in this application; Figure 2 A schematic diagram of the deep neural network model structure provided in this application; Figure 3 A schematic diagram of the long short-term memory network model provided in this application; Figure 4 A structural block diagram of the system provided in this application.

[0018] In the picture: 1. Thermal balance adjustment module; 2. Data acquisition module; 3. Heat exchange calculation module; 4. Area prediction module; 5. Control module. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0020] It should be understood that "multiple" as mentioned herein refers to two or more. In the description of the embodiments of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B; "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Furthermore, to facilitate a clear description of the technical solutions of the embodiments of this application, the terms "first," "second," etc., are used in the embodiments of this application to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first," "second," etc., do not limit the quantity or execution order, and the terms "first," "second," etc., do not necessarily imply differentness. Furthermore, the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, such that a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product, or apparatus.

[0021] Traditional air conditioning systems rely on manual input or drawing blueprints to obtain the space area, lacking automatic sensing capabilities. Furthermore, the thermodynamic model and machine learning algorithm are disconnected, resulting in a mismatch between control strategies and actual heat demand, making it difficult to balance energy efficiency and comfort.

[0022] To address the above technical issues, embodiments of this application provide an air conditioning control method, system, air conditioner, and storage medium based on area prediction.

[0023] Please refer to Figures 1-3 As shown, this embodiment provides an air conditioning control method based on area prediction, which can be applied to household split air conditioners, central air conditioning systems or other types of air conditioning equipment.

[0024] S1: Adjust the space where the air conditioner operates to a thermal equilibrium state.

[0025] First, the space under the air conditioner is adjusted to a state of thermal equilibrium. The system adjusts the output power of the air conditioner to achieve thermal equilibrium, meaning that the heat lost from the space to the outside is equal to the heat supplied by the air conditioner. In a state of thermal equilibrium, the temperature of the space remains stable and does not change significantly over time.

[0026] S2: Obtain the environmental physical parameters of the space.

[0027] Next, the system acquires the environmental physical parameters of the space. These parameters, including temperature, humidity, airflow velocity, and air pressure, are collected in real time by a sensor network deployed within the space. The sensors continuously collect data at a preset sampling frequency, forming a time-series data sequence.

[0028] S3: Based on the environmental physical parameters and the inherent parameters of the heat exchanger, the heat exchange between the air conditioning system and the space is calculated. The heat exchange is equal to the instantaneous heat load of the space under thermal equilibrium.

[0029] Next, based on environmental physical parameters and the inherent parameters of the heat exchanger, the heat exchange between the air conditioning system and the space is calculated. The system substitutes the collected environmental physical parameters into the thermodynamic calculation model and combines them with the inherent parameters of the heat exchanger to calculate the heat exchange. The inherent parameters of the heat exchanger refer to the fixed physical parameters of the air conditioning heat exchanger itself, including the overall heat transfer coefficient K and the area of ​​the heat exchanger coil. Under thermal equilibrium conditions, this heat exchange is numerically equal to the instantaneous heat load of the space.

[0030] S4: Input environmental physical parameters and instantaneous heat load into the area prediction model to obtain the predicted spatial area.

[0031] Subsequently, the environmental physical parameters and instantaneous heat load are input into the area prediction model to obtain the predicted spatial area. The predicted spatial area refers to the horizontal floor area of ​​the space served by the air conditioner, that is, the projected floor area of ​​the area served by the air conditioner. The area prediction model is pre-trained using a large amount of historical measured data and can output the predicted area value of the space based on the input environmental physical parameters and instantaneous heat load.

[0032] S5: Based on the predicted area of ​​the space and the instantaneous heat load, generate control commands to adjust the operating parameters of the air conditioning system so that the air conditioning output matches the actual heat demand of the space.

[0033] Finally, based on the predicted area and instantaneous heat load, control commands are generated to adjust the operating parameters of the air conditioning system. The system calculates the required cooling or heating capacity based on the predicted area and instantaneous heat load, converts the calculation results into control commands, and sends them to the air conditioning actuators to achieve dynamic adjustment of the air conditioning operating status.

[0034] In one implementation, the method starts automatically after the air conditioning system is turned on, without requiring any area-related parameter settings or manual input from the user, thus achieving fully automatic intelligent operation.

[0035] This application adjusts the space where the air conditioner operates to a thermal equilibrium state, so that the heat exchange is equal to the instantaneous heat load value under thermal equilibrium state, providing a physically consistent input feature for subsequent area prediction; the area prediction model automatically identifies the space area without relying on manual input or drawing queries, thereby improving the intelligence level of air conditioning control and energy utilization efficiency.

[0036] In one implementation, the heat exchange between the air conditioning system and the space is calculated based on environmental physical parameters and the inherent parameters of the heat exchanger, specifically including two sub-steps: The first step is to calculate the logarithmic mean temperature difference based on environmental physical parameters. The system uses the collected temperature data of the fluids on both sides of the heat exchanger to calculate the logarithmic mean temperature difference. This temperature difference reflects the driving force of the average temperature difference across the entire heat exchange surface and is a key intermediate variable for calculating the heat transfer capacity.

[0037] The second step involves calculating the heat exchange between the air conditioning system and the space based on the logarithmic mean temperature difference and the inherent parameters of the heat exchanger. The system then substitutes the calculated logarithmic mean temperature difference and the inherent parameters of the heat exchanger into the heat exchange calculation formula to obtain the heat exchange amount.

[0038] The heat transfer rate can be calculated by using the logarithmic mean temperature difference method, which can accurately reflect the driving force of the average temperature difference between the fluids on both sides of the heat exchanger, and provide accurate thermodynamic input for subsequent area prediction.

[0039] In one implementation, the system calculates the logarithmic mean temperature difference based on the collected temperature data of the fluids on both sides of the heat exchanger using the following formula: ; in, The logarithmic mean temperature difference The inlet temperature of the hot fluid. , The inlet temperature of the cold fluid This refers to the outlet temperature of the cold fluid.

[0040] and Temperatures were measured by temperature sensors located at the inlet and outlet of the heat exchanger's hot fluid passage. and The temperature readings are measured by temperature sensors located at the inlet and outlet of the cold fluid channel in the heat exchanger. In air conditioning evaporator mode, the cold fluid is indoor air, and the hot fluid is refrigerant; the opposite is true in heating mode. Those skilled in the art can clearly determine which side contains the hot fluid and which side contains the cold fluid based on the actual operating mode of the air conditioner (cooling or heating), and substitute the temperature sensor readings from the corresponding locations into the above formula.

[0041] In one implementation, for an air conditioning system that uses direct refrigerant evaporation or condensation, the temperature of the refrigerant remains essentially constant during the phase change within the heat exchanger (i.e., ≈ or ≈ The above general formula can be simplified accordingly, but the simplified formula is still a variation that can be easily derived by those skilled in the art based on the above general formula, and all of them are within the protection scope of this invention.

[0042] In one implementation, the system calculates the logarithmic mean temperature difference. Based on the inherent parameters of the heat exchanger, the heat exchange capacity Q is calculated using the following formula:

[0043] Where Q is the heat exchange capacity, representing the amount of heat transferred between the air conditioning system and the space per unit time; K is the overall heat transfer coefficient, representing the comprehensive capacity of heat transfer through the heat exchanger, with the unit being W / (m²·K), which is obtained from the performance test calibration of the heat exchanger before it leaves the factory. The area of ​​the heat exchanger coil is an inherent geometric parameter of the heat exchanger coil, measured in m², and is determined by the design dimensions of the heat exchanger.

[0044] Reference Figure 2 and Figure 3 As shown, in one implementation, the area prediction model is a deep neural network model and / or a long short-term memory network model.

[0045] In one implementation, the area prediction model is a deep neural network model, comprising an input layer, three hidden layers, and an output layer. The input layer contains seven nodes, corresponding to the following input features: indoor temperature, outdoor temperature, air humidity, air supply speed, heat source power, time-series information, and instantaneous heat load Q. The time-series information uses hourly encoding, taking the hourly value (0 to 23) at the sampling time as one of the input features, enabling the model to learn the thermal response patterns at different times. The three hidden layers have 128, 128, and 128 neurons respectively, and each hidden layer uses the ReLU activation function to extract the nonlinear relationships of the input features layer by layer and perform feature fusion. The output layer contains one node, uses the Sigmoid activation function, and outputs the predicted spatial area value, which is normalized to the [0,1] interval. The actual area value can be obtained through inverse normalization.

[0046] In one implementation, the area prediction model is a Long Short-Term Memory (LSTM) network model, which includes an input layer, an LSTM layer, a fully connected layer, and an output layer. The input layer contains 7 nodes, corresponding to the input features: indoor temperature, outdoor temperature, air humidity, air supply speed, heat source power, time-series information, and instantaneous heat load Q. The LSTM layer contains 128 LSTM units, each containing an input gate, a forget gate, and an output gate, used to process time-series data and capture the time dependence of the thermal response; the returned sequence parameter is set to True. The fully connected layer contains 64 neurons, employing the ReLU activation function, used to process the features extracted by the LSTM layer. The output layer contains 1 node, employing the Sigmoid activation function, and outputs the predicted spatial area value.

[0047] By using deep neural network models and / or long short-term memory network models, the model can flexibly select a suitable network architecture according to the characteristics of actual data, taking into account both static feature fitting and dynamic temporal modeling needs. By clarifying that the model learns the nonlinear mapping relationship between area and thermal response time constant, the model training objective has a clear physical meaning, improving the model's interpretability and prediction accuracy.

[0048] In one implementation, a deep neural network model and / or a long short-term memory network model are used to learn the nonlinear mapping relationship between spatial area and thermal response time constant.

[0049] In one implementation, the training method for a deep neural network model includes the following steps: During the data preparation phase, historical measured datasets were collected, covering 50 spaces of different sizes (ranging from 20 to 80 square meters). Each space contained approximately 240 samples, totaling about 12,000 samples. Each sample contained 7 input features and 1 label. The input features included indoor temperature Tin, outdoor temperature Tout, air humidity RH, air supply velocity v, heat source power Pheat, time series information, and instantaneous heat load Q. The label was the space area A.

[0050] In the data preprocessing stage, the collected data is cleaned to remove invalid data caused by sensor failure or communication anomalies; missing values ​​are processed by linear interpolation; each input feature is standardized so that the mean is 0 and the standard deviation is 1; the dataset is divided into training set, validation set and test set with a ratio of 70%, 15% and 15% respectively.

[0051] In the model building phase, a deep neural network was constructed. The input layer contained 7 nodes, and the three hidden layers each contained 128 neurons. Each hidden layer used the ReLU activation function. The output layer contained 1 node and used the Sigmoid activation function. The mean squared error (MSE) loss function was used, and the Adam optimizer was used. The initial learning rate was set to 0.001.

[0052] During model training, the batch size was set to 32, the training epochs to 500, and an early stopping strategy was employed: training stopped when the loss function on the validation set no longer decreased for five consecutive training epochs. To prevent overfitting, L2 regularization was added with a regularization coefficient of 0.001. During training, the network parameters were updated using backpropagation and a gradient descent optimizer to minimize the loss function between the predicted and true labels.

[0053] For example, the deep neural network was trained using the above training method, with the training set, validation set, and test set divided into 70%, 15%, and 15% proportions, respectively. The model achieved its best validation set performance in the 387th training epoch, triggering an early stopping strategy and halting training. After training, the model's mean squared error (MSE) on the test set was approximately 0.021, the coefficient of determination (R²) was approximately 0.856, and the mean absolute error was approximately 0.85 square meters. 95% of the predicted values ​​were within ±2 square meters of the actual area. Taking a room with an actual area of ​​35 square meters as an example, the model predicted 34.7 square meters, a deviation of 0.3 square meters from the actual area, validating the model's prediction accuracy under typical working conditions. Compared to traditional methods based on empirical formulas, the prediction accuracy is significantly improved.

[0054] This application, through the training method of the aforementioned deep neural network model, enables the model to effectively learn the nonlinear mapping relationship between the area and the thermal response time constant based on historical measured data, thus ensuring the feasibility of model training and prediction accuracy.

[0055] In one implementation, the training method for a long short-term memory network model includes the following steps: During the data preparation phase, historical measured datasets containing time series were collected. Each sample contains feature data from 24 consecutive time points, with a time series length of 24, representing the feature changes at 24 sampling times within a day. Each time point contains 7 features: indoor temperature, outdoor temperature, air humidity, air supply speed, heat source power, time series information, and instantaneous heat load Q.

[0056] In the data preprocessing stage, the time series of each feature is standardized so that the mean is 0 and the standard deviation is 1. The data is converted into a format suitable for LSTM input using a sliding window method with a window size of 24 time steps. The dataset is divided into training set, validation set and test set with a ratio of 70%, 15% and 15%, respectively.

[0057] During the model building phase, a Long Short-Term Memory (LSTM) network was constructed. The input layer received input data multiplied by 7 features at 24 time steps. The LSTM layer contained 128 LSTM units, with the return sequence parameter set to True. The fully connected layer contained 64 neurons and used the ReLU activation function. The output layer contained 1 node and used the Sigmoid activation function. The loss function was the mean squared error (MSE), and the optimizer was the Adam optimizer. The initial learning rate was set to 0.001.

[0058] During model training, the batch size was set to 64, the training epochs to 300, and an early stopping strategy was adopted: training was stopped when the loss function on the validation set no longer decreased for 5 consecutive training epochs. To prevent overfitting, a Dropout layer was added before the fully connected layer, with a Dropout rate of 0.2.

[0059] For example, the test results of the Long Short-Term Memory (LSTM) network are similar to those of the Deep Neural Network (DNN), and specific test metrics can be obtained by verifying actual data. Taking a room with an actual area of ​​42 square meters as an example, the LSTM model predicted an area of ​​41.6 square meters based on 24 hours of continuous temperature response data, with a deviation of 0.4 square meters from the actual area, verifying the effectiveness of LSTM in processing time series data. Compared to the Deep Neural Network, the gating mechanism of the LSTM network model can effectively handle long-term dependencies in time series data, making it suitable for capturing the time lag characteristics of air conditioning thermal response. It can be flexibly selected according to the characteristics of the actual data. If the collected data is a continuous time series, LSTM can better capture the time lag effect of thermal response; if the collected data is discrete sampling points, the static nonlinear mapping capability of the Deep Neural Network is sufficient to meet the prediction requirements.

[0060] This application utilizes a training method based on a long short-term memory network model to enable the model to effectively handle the thermal inertia features in time-series data. This model is suitable for area prediction scenarios with significant time-series dependencies, improving the prediction accuracy and generalization ability of the model under dynamic conditions.

[0061] Experimental results show that both models have good accuracy and generalization ability in area prediction tasks, and can be flexibly selected according to the characteristics of actual data.

[0062] In one implementation, control commands are generated to adjust the operating parameters of the air conditioning system based on the predicted space area and instantaneous heat load. Specifically, this includes generating control commands to adjust at least one of the air supply volume and fan speed based on the predicted space area and instantaneous heat load.

[0063] The system combines the predicted area with the instantaneous heat load to calculate the current cooling or heating capacity required for the space. Based on the calculation results, the system determines the optimal air volume and fan speed, generates corresponding control commands, and sends them to the air conditioning actuators.

[0064] In one implementation, the system prioritizes adjusting the fan speed to change the air volume. When the fan speed adjustment to the preset range boundary still cannot meet the demand, other parameters (such as the opening degree of the electronic expansion valve or the compressor frequency) are then adjusted.

[0065] In one implementation, the system uses a human thermal comfort model as a constraint to find the lowest energy consumption combination of control parameters while meeting thermal comfort thresholds. The human thermal comfort model can employ the PMV-PPD model, an internationally recognized standard for thermal comfort evaluation. PMV predicts the average temperature perceived by a population (-3 to +3, 0 being moderate), while PPD predicts the percentage of people dissatisfied with the thermal environment. Together, they constitute the specific implementation of the "human thermal comfort model," providing comfort constraints for air conditioning control. A PMV value between -0.5 and +0.5 and a PPD value less than 10% are considered acceptable comfort levels.

[0066] For example, assuming the target temperature is 24°C, the current indoor temperature is 24.5°C, and the relative humidity is 65%, the human thermal comfort model calculates a current PMV value of -0.2 and a PPD value of 5%. If the temperature setpoint is relaxed from 24°C to 24.5°C, the PMV value becomes 0.3 and the PPD value becomes 8%, still within the preset comfort threshold range of -0.5 to +0.5 and PPD less than 10%. The system determines that this adjustment is within an acceptable range and raises the target temperature by 0.5°C. Based on the statistical data of the air conditioning system's operation at different set temperatures, this adjustment can reduce the compressor frequency by an average of about 8%, thereby achieving energy-saving operation.

[0067] In one implementation, the system determines the climate type based on the outdoor ambient temperature and humidity: if it is a hot and humid climate, the dehumidification mode is activated first to reduce the humidity of the air and avoid energy waste caused by excessive cooling; if it is a cold climate, the heat preservation strategy is implemented first to reasonably adjust the heating output and fresh air exchange frequency to prevent heat loss from being too fast.

[0068] In one implementation, the system intelligently controls the opening of the fresh air valve based on the indoor CO2 concentration and outdoor air quality. When the indoor CO2 concentration increases, the opening of the fresh air valve is increased to introduce fresh outdoor air; when the outdoor air quality is poor, the opening of the fresh air valve is decreased or the fresh air valve is closed.

[0069] There is a progressive and synergistic relationship among the above implementation methods: the system first generates basic air volume and fan speed control commands based on the predicted area and instantaneous heat load, and then prioritizes adjusting the fan speed to respond to changes in heat demand; throughout the control process, the human thermal comfort model is used as a global constraint to ensure that all adjustment actions do not exceed the preset comfort threshold; at the same time, the system adjusts the control strategy direction (dehumidification or heat preservation) from a macro level according to outdoor climate conditions, and makes auxiliary adjustments to the fresh air valve according to indoor CO2 concentration and outdoor air quality. The strategies at each level cooperate and work together to achieve the comprehensive goals of precise temperature control, comfort maintenance and energy saving.

[0070] This application adjusts the air supply volume and fan speed according to the predicted area and instantaneous heat load to match the air supply parameters of the air conditioner with the actual heat demand, thus avoiding unnecessary energy consumption. Through the synergistic effect of multiple optimization strategies such as human thermal comfort model constraints, climate adaptability control and intelligent fresh air control, the energy-saving effect is further improved while ensuring user comfort.

[0071] In one implementation, control command generation employs a hierarchical parallel control architecture. This architecture consists of three layers: a bottom layer, a middle layer, and a top layer. Each layer operates independently while also interacting with the others: the middle layer performs online corrections to the control parameters of the bottom layer, and the top layer provides optimized control targets to both the middle and bottom layers.

[0072] The underlying layer employs PID control to perform basic closed-loop regulation. The PID controller uses the deviation between the indoor temperature and the target temperature as input, calculating the control quantity through a weighted combination of proportional, integral, and derivative components, ensuring accurate control and rapid response under stable operating conditions. The PID control has three parameters: proportional gain Kp, integral time Ti, and derivative time Td. These parameters have initial reference values, allowing the PID controller to operate normally from startup. During system operation, the PID parameters serve as the adjustment objects for the intermediate-layer fuzzy control, receiving online corrections from the intermediate layer.

[0073] The intermediate layer employs fuzzy control to correct PID parameters online. The fuzzy controller, based on the current temperature deviation and its rate of change, infers from a pre-defined fuzzy rule base and outputs the correction values ​​for the PID parameters. Specifically, the fuzzy controller uses the temperature deviation *e* and the rate of change *ec* as input variables, dividing the input variables into several fuzzy subsets (e.g., negative large, negative medium, negative small, zero, positive small, positive medium, positive large), and infers from a pre-defined fuzzy rule table to output the correction values ​​ΔKp, ΔKi, and ΔDd for the three PID parameters. In actual control, the PID parameters are obtained by adding the correction values ​​to the reference values. For example, when the temperature deviation is large and the rate of change is rapid, the fuzzy controller increases the proportional gain to speed up the response; when the temperature deviation is small and tends to stabilize, the fuzzy controller reduces the integral action to avoid overshoot.

[0074] The top layer employs reinforcement learning to optimize the control strategy based on long-term cumulative rewards. The reinforcement learning agent continuously learns through interaction with the environment, optimizing the control strategy with the goal of maximizing long-term cumulative rewards. The reward function of reinforcement learning comprehensively considers energy consumption and comfort indicators, enabling the system to find the control strategy with the lowest energy consumption while meeting human thermal comfort requirements. The optimization objective output by reinforcement learning serves as the constraint condition for the intermediate-layer fuzzy control, guiding the inference direction of the fuzzy control.

[0075] In one implementation, the reinforcement learning training process is as follows: A state space is defined, including the current temperature deviation, the rate of change of deviation, the predicted area, and the outdoor ambient temperature; an action space is defined, including the airflow adjustment amount and / or the temperature setpoint adjustment amount; a reward function is designed, which aims to minimize energy consumption and maximize comfort, and is constructed using a weighted summation method; the training algorithm collects samples in simulated environments or historical data, continuously accumulates experience through agent-environment interaction, optimizes the control strategy with the goal of maximizing accumulated rewards, and finally learns the optimal control strategy.

[0076] For example, suppose the air conditioner operates in a living room of approximately 30m², with an outdoor temperature of 35°C and a target temperature of 24°C. Three people are present indoors. When someone opens the door to pick up takeout while the air conditioner is running stably, the influx of hot outdoor air causes the indoor temperature to rise from 24°C to 24.8°C, a temperature deviation of 0.8°C, and the rate of temperature increase is relatively fast. At this point, the bottom-level PID controller, based on the 0.8°C temperature deviation, calculates a basic control quantity (e.g., increasing the compressor frequency from 45Hz to 52Hz) through weighted calculations of the proportional, integral, and derivative components, and begins to increase the cooling output to eliminate the temperature deviation. Simultaneously, the intermediate-level fuzzy controller detects a large rate of change in the deviation, determining it to be a sudden disturbance rather than a gradual change. It immediately infers based on preset fuzzy rules and outputs a correction for the PID parameters: increasing the proportional gain Kp by 20% to accelerate the response speed and decreasing the integral gain Ki by 10% to avoid integral saturation and overshoot. The corrected PID output provides stronger control. The compressor frequency is further increased to 58Hz to quickly respond to temperature fluctuations caused by opening the door. The top-level reinforcement learning agent, based on historical operating data, determines that the outdoor temperature is high and the sunlight is strong during the current period (2:00 PM - 4:00 PM), resulting in a high heat load in the living room. Strictly maintaining the target temperature of 24°C would lead to prolonged high-frequency compressor operation and high energy consumption. After comprehensive judgment, a strategy adjustment signal is output, relaxing the target temperature to 24.3°C (within the acceptable range of the PMV of the human thermal comfort model). It also suggests increasing the cooling capacity 10 minutes before this period for pre-cooling and appropriately reducing the cooling capacity during peak electricity price periods to save on electricity costs. These three levels work together to restore the indoor temperature to 24.3°C within 2 minutes and maintain it stably, reducing the compressor's high-frequency operation time by approximately 15 minutes, achieving energy-saving operation while ensuring user comfort.

[0077] In summary, the aforementioned collaborative process involves the middle layer's fuzzy control dynamically adjusting the control parameters of the lower layer based on real-time operating conditions, while the top layer's reinforcement learning continuously optimizes the inference rules of the middle layer and the control objectives of the lower layer based on long-term operational data. This achieves a three-tiered collaborative control system guided by the top-level strategy, adapted by the middle layer's parameters, and precisely executed by the lower layer. These three layers operate in parallel, with the lower and middle layers responsible for real-time control response, and the top layer responsible for strategy optimization and parameter seeking. Their collaborative work forms a complete control loop.

[0078] The beneficial effects of this implementation are: by using a hierarchical parallel control architecture, the control system can simultaneously possess the fast response capability of PID, the nonlinear processing capability of fuzzy control, and the long-term optimization capability of reinforcement learning, thus taking into account control accuracy, robustness, and energy-saving effect.

[0079] In one implementation, environmental physical parameters are acquired through a distributed sensor network deployed in space. The distributed sensor network consists of multiple sensor nodes deployed at different locations within the space. Each sensor node synchronously collects data according to a preset sampling frequency, and the data is aggregated to a central control unit via a communication network.

[0080] In one implementation, the distributed sensor network uses wireless communication, with each sensor node transmitting data to the central control unit via wireless protocols such as ZigBee, Wi-Fi, or Bluetooth.

[0081] In another implementation, the distributed sensor network uses wired communication, with each sensor node connected to the central control unit via an RS485 bus or Ethernet.

[0082] All sensors are synchronously triggered to acquire data using the same clock source, and the deviation in acquisition time does not exceed a preset threshold to ensure the consistency of features across all dimensions over time. For sensors with different response speeds, the system employs a buffer alignment method, using the sampling period of the temperature sensor as the main period, and synchronously reading the current values ​​of other sensors during each temperature sampling.

[0083] This application acquires environmental physical parameters through a distributed sensor network, achieving high-density, high-frequency sensing of the space environment and providing sufficient real-time data support for subsequent thermodynamic calculations and area prediction.

[0084] In one embodiment, the environmental physical parameters include at least one of indoor temperature, outdoor temperature, air humidity, air supply velocity, air pressure, and heat source power.

[0085] Indoor temperature reflects the thermal state of a space and is a core parameter for calculating heat load and controlling air conditioning output. Outdoor temperature reflects the external thermal environment conditions and is fundamental data for calculating indoor-outdoor temperature difference and heat exchange rate. Air humidity affects human comfort and latent heat load, and is particularly important under cooling conditions. Air supply velocity directly affects the intensity of convective heat transfer and indoor airflow organization. Air pressure is an important factor affecting air density and is valuable in high-altitude areas or in precise thermodynamic calculations. Heat source power reflects the total heat generated by indoor occupants, lighting, and electrical equipment, and is a key source of thermal disturbance within the space.

[0086] In one implementation, the environmental physical parameters include all six parameters simultaneously to provide the most comprehensive information for thermodynamic calculations and area prediction.

[0087] In one implementation, adjusting the air-conditioned space to a thermal equilibrium state specifically includes the following steps: First, we need to capture the first type of heat lost from the space to the outside and the second type of heat supplied to the space by the air conditioner. The first type of heat is... This represents the heat transferred from the space to the outside through the enclosure structure; the second type of heat is... This indicates the amount of heat that the air conditioning system provides to the space.

[0088] In one implementation, the first heat is obtained by measuring heat flux sensors installed on the surfaces of the space enclosure structure, and the summation is performed by multiplying the heat flux density of each surface by its corresponding area. The second heat is obtained by measuring the air enthalpy difference. Temperature and humidity sensors are placed at the air conditioner's return air vent and supply air vent respectively. The enthalpy difference is calculated and then combined with the supply air volume to obtain the final heat. .

[0089] Then, the difference between the first heat and the second heat is calculated. If the difference is greater than a preset threshold, the air conditioner's output power is adjusted, and the first and second heat are re-acquired until the difference is less than or equal to the preset threshold. The specific adjustment method is as follows: If... > This indicates that the space is cooling down; the system should increase its heating capacity or decrease its cooling capacity. < This indicates that the space is heating up, and the system should increase the cooling capacity or decrease the heating capacity.

[0090] Finally, when the difference is less than or equal to a preset threshold, the space is determined to be in thermal equilibrium. At this point, the space temperature remains stable, and the air conditioning output and the space's heat demand are in balance.

[0091] For example, assuming the air conditioner's rated cooling capacity is 3.5kW, and the preset threshold is 5% of the rated cooling capacity, i.e., 0.175kW. The measured value after system startup... =2.8kW =2.0kW, the difference of 0.8kW is greater than 0.175kW, so it is determined to be unbalanced. > The space is heating up, so the system increases its cooling capacity. The measurement was repeated after adjustment. =3.1kW =3.0kW, the difference of 0.1kW is less than 0.175kW, it is determined that the thermal equilibrium state has been reached, the adjustment is stopped, and the collection of environmental physical parameters for area prediction begins.

[0092] In one implementation, the system automatically performs thermal balance judgment and adjustment steps upon initial startup, and re-executes thermal balance judgment at preset time intervals during subsequent operation to ensure that the space is in thermal balance each time data for area prediction is collected.

[0093] In one implementation, the criterion for determining whether the air conditioning system has entered a thermal equilibrium state is that the change in indoor temperature within a preset time (e.g., 5 minutes) does not exceed a preset threshold (e.g., 0.5°C). This method indirectly reflects the thermal equilibrium state through temperature stability.

[0094] This application ensures that the data used for subsequent thermodynamic calculations and area predictions are collected under steady-state conditions by performing thermal balance judgment and adjustment before data acquisition. This avoids the impact of data fluctuations under transient conditions on prediction accuracy and improves the accuracy of area prediction and the reliability of control decisions.

[0095] like Figure 4 As shown, this application also provides an air conditioning control system based on area prediction. This system is used to perform the above-described method.

[0096] The system includes a heat balance adjustment module 1, a data acquisition module 2, a heat exchange calculation module 3, an area prediction module 4, and a control module 5.

[0097] The thermal balance adjustment module 1 is used to adjust the space where the air conditioner operates to a thermal balance state. The module's inputs are the air conditioner's operating parameters and the space's temperature data, and its output is a thermal balance confirmation signal or a power adjustment command. When the space has not reached thermal balance, the thermal balance adjustment module 1 sends a power adjustment command to the air conditioner until the space reaches a thermal balance state.

[0098] Data acquisition module 2 is used to acquire environmental physical parameters of the space. The input of this module is the acquisition signal from each sensor, and the output is time-aligned and formatted multi-dimensional physical parameter time-series data.

[0099] The heat exchange calculation module 3 is used to calculate the heat exchange between the air conditioning system and the space based on environmental physical parameters and the inherent parameters of the heat exchanger. The input to this module is the environmental physical parameters output by the data acquisition module 2, and the output is the heat exchange rate Q. Under thermal equilibrium conditions, this heat exchange rate Q is numerically equal to the instantaneous heat load of the space.

[0100] Area prediction module 4 is used to input environmental physical parameters and instantaneous heat load into the area prediction model to obtain the predicted spatial area. The input of this module is environmental physical parameters and instantaneous heat load, and the output is the predicted area value.

[0101] Control module 5 is used to generate control commands based on the predicted area and instantaneous heat load of the space to adjust the operating parameters of the air conditioning system so that the air conditioning output matches the actual heat demand of the space. The inputs of this module are the predicted area value and the instantaneous heat load q, and the outputs are control commands such as air supply volume and fan speed.

[0102] The five modules are connected in sequence, with the output of the previous module serving as the input of the next, forming a complete data processing chain. After the system starts, the modules work collaboratively in the above order, executing cyclically to form closed-loop control.

[0103] Through the coordinated operation of five functional modules, this application enables the system to automatically complete the entire process from thermal balance adjustment to control command generation, realizing the automation and intelligence of air-conditioned area prediction and dynamic control.

[0104] In one implementation, the heat exchange calculation module 3 calculates the logarithmic mean temperature difference based on environmental physical parameters, and then calculates the heat exchange based on the logarithmic mean temperature difference and the inherent parameters of the heat exchanger. Specifically, the heat exchange calculation module 3 first obtains the temperature data of the fluids on both sides of the heat exchanger from the data acquisition module 2, calculates the logarithmic mean temperature difference, and then calculates the heat exchange based on the inherent parameters of the heat exchanger stored in the module.

[0105] In one implementation, the heat exchange calculation module 3 calculates the heat exchange using the logarithmic mean temperature difference method, while also employing Newton's law of cooling for auxiliary calculation. Newton's law of cooling, based on the difference between the heat exchanger surface temperature and the ambient temperature, combined with a preset surface heat transfer coefficient, independently estimates the current heat dissipation. The system compares the calculation results from the logarithmic mean temperature difference method with those from Newton's law of cooling. If the deviation is within a preset range, it indicates that the heat exchanger is operating normally and the logarithmic mean temperature difference method calculation result is reliable. If the deviation exceeds the preset range, it indicates that the heat exchanger may have problems such as dust accumulation or scaling, and the system can trigger a cleaning reminder or correct the overall heat transfer coefficient K. This achieves the verification and correction of the logarithmic mean temperature difference method calculation results, further improving the accuracy and reliability of the heat exchange calculation.

[0106] In one implementation, the area prediction module 4 employs a deep neural network model and / or a long short-term memory network model for area prediction. The deep neural network model and / or long short-term memory network model are used to learn the nonlinear mapping relationship between spatial area and thermal response time constant.

[0107] When using a deep neural network model, the model is constructed by stacking fully connected layers. The input features undergo multiple nonlinear transformations before outputting the predicted area value. When using a long short-term memory network model, the model is based on a recurrent neural network. It processes time-series data through memory units and gating mechanisms, extracts thermal inertia features, and then performs area prediction.

[0108] In one implementation, the area prediction module 4 uses a combination of a deep neural network model and a long short-term memory network model. First, the long short-term memory network model is used to extract time-series features, and then the time-series features are input into the deep neural network model for regression prediction.

[0109] In one implementation, the area prediction module 4 is pre-trained using supervised learning on a historical measured dataset, which contains temperature and humidity response curves for spaces of different areas under different climatic conditions. Specifically, in the data preparation stage, historical measured datasets are collected, with each sample containing input features and a label. The input features include indoor temperature Tin, outdoor temperature Tout, air humidity RH, air supply speed v, heat source power Pheat, and time series information, while the label is the space area A. In the data preprocessing stage, the collected data is cleaned to remove invalid data caused by sensor malfunctions or communication anomalies, missing values ​​are interpolated, and outliers are removed or corrected. In the model building stage, a deep neural network model or a long short-term memory network model is used to build the area prediction model. The input layer receives multi-dimensional features, and the output layer predicts the space area of ​​continuous values. In the model training stage, the dataset is divided into a training set, a validation set, and a test set. The backpropagation algorithm and gradient descent optimizer are used to update the network parameters by minimizing the loss function between the predicted value and the true label. During training, training stops when the loss function on the validation set no longer decreases for several consecutive training cycles.

[0110] This application, by defining the model type and learning objectives of the area prediction module and disclosing the specific training process, enables the system to handle both static and temporal features, ensuring the feasibility of model training and prediction accuracy.

[0111] In one implementation, the control module 5 generates a control command to adjust at least one of the air supply volume and fan speed based on the predicted area of ​​the space and the instantaneous heat load. Specifically, the control module 5 combines the predicted area with the instantaneous heat load to calculate the current cooling or heating capacity required by the space, determines the optimal air supply volume and fan speed based on the calculation results, and generates corresponding control commands to send to the air conditioning actuator.

[0112] In one implementation, the control module 5 uses a human thermal comfort model as a constraint to find the combination of control parameters with the lowest energy consumption while meeting the thermal comfort threshold.

[0113] In one embodiment, the data acquisition module 2 includes a distributed sensor network deployed in space. The distributed sensor network consists of multiple sensor nodes deployed at different locations in space. Each sensor node synchronously acquires data according to a preset sampling frequency and aggregates the data to the central control unit through a communication network.

[0114] The environmental physical parameters collected by the data acquisition module 2 include at least one of the following: indoor temperature, outdoor temperature, air humidity, air supply speed, air pressure, and heat source power.

[0115] In one implementation, the data acquisition module 2 simultaneously acquires all six parameters, providing the most comprehensive data support for thermodynamic calculations and area prediction.

[0116] In one implementation, the thermal balance adjustment module 1 acquires a first heat loss from the space to the outside and a second heat supply from the air conditioner to the space, and calculates the difference between the first heat loss and the second heat supply. If the difference is greater than a preset threshold, the output power of the air conditioner is adjusted, and the first heat loss and the second heat supply are acquired again until the difference is less than or equal to the preset threshold. When the difference is less than or equal to the preset threshold, the space is determined to be in a thermal balance state.

[0117] In one implementation, the thermal balance adjustment module 1 first performs a thermal balance judgment after the system starts. The thermal balance adjustment module 1 obtains the following data: and The two values ​​are compared and judged. If the difference exceeds a preset threshold, the thermal balance adjustment module 1 sends a power adjustment command to the air conditioning actuator to gradually approach the thermal balance state.

[0118] In one implementation, the thermal balance adjustment module 1 performs a thermal balance judgment before each data acquisition to ensure that the data used for each area prediction is acquired under steady-state conditions.

[0119] In one implementation, the thermal balance adjustment module 1 is executed only once after the system is started, or re-executed after a large fluctuation in the space temperature (such as a sudden temperature change caused by opening doors and windows).

[0120] This application also provides an air conditioner. The air conditioner includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-described air conditioning control method based on area prediction.

[0121] Specifically, the air conditioner includes an indoor unit and an outdoor unit. The indoor unit houses a control motherboard, which integrates a processor and a memory. The memory stores a control program containing computer-executable instructions for performing the methods described above. The processor executes this program to control the thermal balance adjustment module to adjust the space to a thermal equilibrium state, to control the data acquisition module to acquire environmental physical parameters, to call the heat exchange calculation module to calculate the heat exchange, to call the area prediction module to predict the area, and finally, the control module generates control commands and outputs them to the air conditioning actuators via a drive circuit.

[0122] In one embodiment, the air conditioner further includes a sensor assembly for measuring parameters such as indoor temperature, outdoor temperature, air humidity, air supply speed, and air pressure, as well as a remote control receiver module for receiving user commands and a display panel for displaying operating status.

[0123] This application also provides a storage medium. The storage medium stores a computer program, which, when executed by a processor, implements the aforementioned air conditioning control method based on area prediction.

[0124] The storage medium can be a non-volatile computer-readable storage medium such as read-only memory (ROM), random access memory (RAM), flash memory (Flash), hard disk drive (HDD), or solid-state drive (SSD). The storage medium can be installed on the air conditioner's control board or exist independently of the air conditioner, used as a software upgrade package or program update carrier.

[0125] In one embodiment, the storage medium is a removable storage medium such as a USB flash drive or SD card. Users can load the program into the air conditioner's memory by inserting the storage medium into the corresponding interface of the air conditioner, thereby installing or upgrading the air conditioner control program.

[0126] In one implementation, the storage medium is cloud storage space, and the computer program is downloaded and installed into the air conditioner via the network.

[0127] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0128] Although embodiments of this application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An air conditioning control method based on area prediction, characterized in that, include: Adjust the space where the air conditioner operates to a state of thermal equilibrium; Obtain the environmental physical parameters of the space; Based on the environmental physical parameters and the inherent parameters of the heat exchanger, the heat exchange between the air conditioning system and the space is calculated, and the heat exchange is equal to the instantaneous heat load of the space under the thermal equilibrium state. The environmental physical parameters and the instantaneous heat load are input into the area prediction model to obtain the predicted spatial area; Based on the predicted area of ​​the space and the instantaneous heat load, control commands are generated to adjust the operating parameters of the air conditioning system so that the air conditioning output matches the actual heat demand of the space.

2. The method according to claim 1, characterized in that, The step of calculating the heat exchange between the air conditioning system and the space based on the environmental physical parameters and the inherent parameters of the heat exchanger includes: The logarithmic mean temperature difference was calculated based on the aforementioned environmental physical parameters; The heat exchange between the air conditioning system and the space is calculated based on the logarithmic mean temperature difference and the inherent parameters of the heat exchanger.

3. The method according to claim 2, characterized in that, The formula for calculating the logarithmic mean temperature difference based on the environmental physical parameters is as follows: ； in, The logarithmic mean temperature difference The inlet temperature of the hot fluid. , The inlet temperature of the cold fluid This refers to the outlet temperature of the cold fluid.

4. The method according to claim 3, characterized in that, The formula for calculating the heat exchange between the air conditioning system and the space based on the logarithmic mean temperature difference and the inherent parameters of the heat exchanger is as follows: in, To exchange heat, The overall heat transfer coefficient is... This represents the area of ​​the heat exchanger coil.

5. The method according to claim 1, characterized in that, The area prediction model is a deep neural network model and / or a long short-term memory network model.

6. The method according to claim 5, characterized in that, The deep neural network model and / or the long short-term memory network model are used to learn the nonlinear mapping relationship between spatial area and thermal response time constant.

7. The method according to claim 1, characterized in that, The step of generating control commands to adjust the operating parameters of the air conditioning system based on the predicted spatial area and the instantaneous heat load includes: Based on the predicted spatial area and the instantaneous heat load, a control command is generated to adjust at least one of the air supply volume and the fan speed.

8. The method according to claim 7, characterized in that, The generation of the control commands adopts a hierarchical parallel control architecture, which includes: At the lowest level, basic closed-loop regulation is performed through PID control; The intermediate layer uses fuzzy control to correct the PID parameters online. At the top level, the control strategy is optimized based on long-term cumulative returns through reinforcement learning.

9. The method according to claim 1, characterized in that, The environmental physical parameters are acquired through a distributed sensor network deployed within the space.

10. The method according to claim 1 or 9, characterized in that, The environmental physical parameters include at least one of indoor temperature, outdoor temperature, air humidity, air supply speed, air pressure, and heat source power.

11. The method according to claim 1, characterized in that, Adjusting the air-conditioned space to a thermal equilibrium state includes: The first heat loss from the space to the outside and the second heat provided by the air conditioner to the space are obtained; Calculate the difference between the first heat and the second heat; If the difference is greater than a preset threshold, the output power of the air conditioner is adjusted, and the first heat and the second heat are reacquired until the difference is less than or equal to the preset threshold. When the difference is less than or equal to the preset threshold, the space is determined to be in the thermal equilibrium state.

12. An air conditioning control system based on area prediction, characterized in that the system... include: The thermal balance adjustment module (1) is used to adjust the air-conditioned space to a thermal balance state. The data acquisition module (2) is used to acquire the environmental physical parameters of the space; The heat exchange calculation module (3) is used to calculate the heat exchange between the air conditioning system and the space based on the environmental physical parameters and the inherent parameters of the heat exchanger. The heat exchange is equal to the instantaneous heat load of the space under the thermal equilibrium state. The area prediction module (4) is used to input the environmental physical parameters and the instantaneous heat load into the area prediction model to obtain the predicted spatial area; The control module (5) is used to generate control commands to adjust the operating parameters of the air conditioning system based on the predicted area of ​​the space and the instantaneous heat load, so that the air conditioning output matches the actual heat demand of the space.

13. The system according to claim 12, characterized in that, The heat exchange calculation module (3) calculates the logarithmic mean temperature difference based on the environmental physical parameters, and calculates the heat exchange based on the logarithmic mean temperature difference and the inherent parameters of the heat exchanger.

14. The system according to claim 12, characterized in that, The area prediction model is a deep neural network model and / or a long short-term memory network model; The deep neural network model and / or the long short-term memory network model are used to learn the nonlinear mapping relationship between spatial area and thermal response time constant.

15. The system according to claim 12, characterized in that, The control module (5) generates a control command to adjust at least one of the air supply volume and the fan speed based on the predicted spatial area and the instantaneous heat load.

16. The system according to claim 12, characterized in that, The data acquisition module (2) includes a distributed sensor network deployed in the space, and the environmental physical parameters include at least one of indoor temperature, outdoor temperature, air humidity, air supply speed, air pressure and heat source power.

17. The system according to claim 12, characterized in that, The heat balance adjustment module (1) acquires the first heat lost from the space to the outside and the second heat provided by the air conditioner to the space, and calculates the difference between the first heat and the second heat. If the difference is greater than a preset threshold, the output power of the air conditioner is adjusted, and the first heat and the second heat are reacquired until the difference is less than or equal to the preset threshold. When the difference is less than or equal to the preset threshold, the space is determined to be in the thermal equilibrium state.

18. An air conditioner, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as described in any one of claims 1 to 11.

19. A storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 11.

Images