Air conditioning system control method and control system for dynamic airflow adjustment

Abstract

The invention discloses an air conditioning system control method and system for dynamic airflow regulation, and the method comprises the steps: collecting environment data in a temperature regulation and control region through a sensor unit, and transmitting the environment data to a central control unit; an intelligent model arranged in the central control unit is used for processing the environment data, outputting a control instruction and transmitting the control instruction to the air conditioning system; the air conditioning system executes an adjustment action of the air conditioning system according to the control instruction, stores the adjustment action into a memory and feeds back the adjustment action to the intelligent model; and the intelligent model stores the environment data and the adjustment action generated by the environment data in a memory as historical reference data, and the intelligent model is adjusted and updated. The air conditioner system control method and control system capable of achieving dynamic airflow adjustment have the advantages that the air outlet direction and the air volume of an air conditioner can be automatically adjusted according to the distribution and activity conditions of people in a room, and intelligent airflow adjustment is achieved.

Description

Technical Field

[0001] The invention relates to an intelligent control technology based on a neural network, and in particular to a control method and a control system for an air conditioning system with dynamic airflow regulation. Background Art

[0002] Air conditioning system refers to a system that uses artificial methods to process the temperature, humidity, cleanliness and air velocity of indoor air. Air conditioning system can make certain places obtain air with a certain temperature, humidity and air quality to meet the requirements of users and production processes and improve labor hygiene and indoor climate conditions. As people's requirements for comfortable life and energy conservation and environmental protection continue to increase, air conditioning system has become one of the necessary infrastructures for homes, offices and public places.

[0003] The air conditioning system adjusts the ambient temperature by adjusting indicators such as the air outlet temperature, air outlet speed and air outlet direction. At present, most existing air conditioning systems use manual or preset modes to adjust the air outlet direction and air volume. This conventional control method often has the following disadvantages: (1) It leads to unsatisfactory ambient temperature distribution. It is not suitable to stay in the position where the air outlet blows directly for a long time, and cold or hot air is not suitable to reach local areas; (2) The indoor temperature is uneven, making some people feel cold and some people feel hot.

[0004] Therefore, the traditional air-conditioning control method in the prior art lacks dynamism and intelligence, and is difficult to adapt to real-time changes in the room environment, resulting in uneven air flow distribution, excessive energy consumption, or poor user comfort. Summary of the invention

[0005] The present invention aims to avoid the deficiencies in the above-mentioned prior art and to provide an air conditioning system control method and control system with dynamic airflow regulation, so as to perform temperature control operations according to the distribution of people in the environment and improve the comfort and energy saving of the air conditioning.

[0006] The present invention adopts the following technical solutions to solve the technical problems.

[0007] The present invention provides a method for controlling an air conditioning system with dynamic airflow regulation, the method comprising the following steps:

[0008] Step 1: Data collection step: Collect environmental data in the temperature control area through the sensor unit and send the environmental data to the central control unit;

[0009] Step 2: Data processing step; the central control unit has a built-in intelligent model, and the environmental data is input into the intelligent model; the intelligent model performs calculations and outputs control instructions, and transmits the control instructions to the air conditioning system;

[0010] Step 3: Control execution step; the air conditioning system executes the adjustment action of the air conditioning system according to the control instruction, stores the adjustment action into the memory and feeds back to the intelligent model;

[0011] Step 4: Feedback update step; the intelligent model stores the environmental data and the adjustment actions generated by the environmental data in a memory as historical reference data to adjust and update the intelligent model.

[0012] The characteristics of the air conditioning system control method of the present invention are also as follows:

[0013] Furthermore, in step 1, the environmental data includes environmental parameters and environmental images; the environmental parameters include the temperature T and humidity H in the temperature control area; the personnel parameters are obtained based on the environmental image; the personnel parameters include the number of personnel n, the position coordinates of the personnel P and the activity intensity A.

[0014] Furthermore, in step 2, the central control unit pre-processes the environmental data before inputting it into the intelligent model.

[0015] Furthermore, in step 2, the intelligent model includes a CNN network and a DQN network.

[0016] Furthermore, in step 2, the CNN network is used to process environmental data.

[0017] Furthermore, in step 2, the DQN network is used to receive the output vector Ocnn of the CNN network processing environment data, and use the output vector Ocnn as the input vector of the DQN network to calculate the control instruction.

[0018] The present invention also discloses a dynamic airflow regulating air conditioning system control system, comprising a sensor unit, a central control unit, a multi-directional air outlet execution unit and an airflow regulating execution unit;

[0019] The sensor unit is used to collect environmental data in the temperature control area and send the environmental data to the central control unit;

[0020] The intelligent model of the central control unit is used to process and calculate the environmental data, obtain control instructions, and send the control instructions to the multi-directional air outlet execution unit and the air flow adjustment execution unit of the air-conditioning system. The multi-directional air outlet execution unit and the air flow adjustment execution unit execute the control instructions to complete the adjustment action, and save the environmental data and the adjustment action generated by the environmental data in the memory as historical reference data to adjust and update the intelligent model.

[0021] The present invention also discloses an electronic device, comprising at least one processor and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the air conditioning system control method with dynamic airflow adjustment.

[0022] The present invention also discloses a non-transient computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the air conditioning system control method for dynamic airflow regulation.

[0023] The present invention also discloses a computer program product, comprising a computer program; when the computer program is executed by a processor, the air conditioning system control method for dynamic airflow regulation is realized.

[0024] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0025] The present invention discloses a control method and control system for an air-conditioning system with dynamic airflow regulation, wherein environmental data in a temperature control area is collected by a sensor unit, and the environmental data is sent to a central control unit; an intelligent model built into the central control unit processes the environmental data, and the intelligent model outputs control instructions, and transmits the control instructions to the air-conditioning system; the air-conditioning system executes adjustment actions of the air-conditioning system according to the control instructions, and stores the adjustment actions in a memory and feeds back to the intelligent model; the intelligent model stores the environmental data and the adjustment actions generated by the environmental data in the memory as historical reference data, and adjusts and updates the intelligent model.

[0026] The air conditioning system control method and control system for dynamic airflow regulation of the present invention have the advantages of being able to automatically adjust the air outlet direction and air volume of the air conditioner according to the distribution and activities of people in the room, and realizing intelligent airflow regulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 The present invention is a flow chart of a method for controlling an air conditioning system with dynamic airflow regulation.

[0028] Figure 2 The block diagram of a control system of an air conditioning system with dynamic airflow regulation according to the present invention.

[0029] The present invention will be further described below through specific implementation modes in conjunction with the accompanying drawings. DETAILED DESCRIPTION

[0030] See also Figure 1-2 The present invention provides a method for controlling an air conditioning system with dynamic airflow regulation, the method comprising the following steps:

[0031] Step 1: Data collection step: Collect environmental data in the temperature control area through the sensor unit and send the environmental data to the central control unit;

[0032] Step 2: Data processing step; the central control unit has a built-in intelligent model, and the environmental data is input into the intelligent model; the intelligent model performs calculations and outputs control instructions, and transmits the control instructions to the air conditioning system;

[0033] Step 3: Control execution step; the air conditioning system executes the adjustment action of the air conditioning system according to the control instruction, stores the adjustment action into the memory and feeds back to the intelligent model;

[0034] Step 4: Feedback update step; the intelligent model stores the environmental data and the adjustment actions generated by the environmental data in a memory as historical reference data to adjust and update the intelligent model.

[0035] The control method of the air conditioning system with dynamic airflow regulation of the present invention comprises a control system including a sensor unit, a central control unit, a multi-directional air outlet structure and an airflow regulation execution unit. The precise guidance of the airflow is achieved through the multi-directional adjustable air outlet structure and the intelligent control algorithm of the central control unit. The multi-directional adjustable air outlet structure comprises a plurality of independently controlled fan blades or air guide plates, the angle and direction of which are adjusted by the servo motor of the multi-directional air outlet execution unit, and the airflow regulation execution unit is dynamically controlled according to the distribution of people in the room to adjust the direction and intensity of the airflow, thereby improving the comfort and energy saving of the air conditioner.

[0036] During specific implementation, in step 1, the environmental data includes environmental parameters and environmental images; the environmental parameters include the temperature T and humidity H in the temperature control area; the personnel parameters are obtained based on the environmental image; the personnel parameters include the number of personnel n, the personnel position coordinates P and the activity intensity A.

[0037] The sensor unit includes multiple sensors. The multiple sensors are used to collect different environmental parameters in the room in real time. The environmental parameters include but are not limited to temperature T, humidity H and CO2 concentration C CO2 The camera is used to collect the environment image in the room, and the environment image is processed to obtain the personnel parameters such as the number of people n, the position coordinates of the people P and the activity intensity A. i Represents the horizontal coordinate of the i-th person, y i represents the ordinate of the i-th person. Therefore, the environmental data can be represented by the environmental vector Ev, which includes T, H, C CO2 ,n,x1,y1,...x n ,y nand A. The environment vector Ev includes all the parameters required for air conditioning system control in an ideal state. However, in specific implementation, some parameters are used for auxiliary control and are not necessarily used as input quantities for the intelligent model. The specific sensor configuration includes: The temperature sensor is used to detect the temperature T in the room. The humidity sensor is used to monitor the humidity H in the air. The camera is used to obtain the environment image in the room. The CO2 sensor is used to monitor the CO2 concentration C in the room. CO2 , as an auxiliary optimization parameter to further optimize ventilation requirements. Sensors are arranged in various locations in the room to ensure that the collected environmental data can reflect the overall condition of the room.

[0038] The activity intensity A uses the frequency of a person's body movements as an indicator of activity intensity. For example, the body movements of a person are captured by a camera, and the number of changes in the body movements of a person (such as obvious movements such as arm swings and body rotations) per unit time (such as per minute) is analyzed using image recognition technology, and this value is used as a quantitative indicator of activity intensity A. Alternatively, the frequency of movement of a person in a certain area detected by an infrared sensor, such as the number of times the person crosses different detection areas in a short period of time (such as 5 minutes), is used to measure the activity intensity. The more times, the greater the activity intensity, and this statistical value is included in the activity intensity parameter in the state vector.

[0039] The sensor unit continuously collects environmental data in the room and transmits the data to the central control unit. The central control unit first uses the CNN network to process the environmental image or other complex sensor data, extracts the location, number and activity characteristics of people in the room, and finally converts it into an environmental vector Ev (T, H, n, x1, y1, ... x n ,y n , A).

[0040] During specific implementation, in step 2, the central control unit pre-processes the environmental data before inputting it into the intelligent model.

[0041] The central control unit preprocesses the collected environmental parameters and environmental images, including data normalization and noise filtering. The environmental images are input into the CNN deep learning model for feature extraction and analysis. The specific process is as follows: Data preprocessing: The collected environmental parameters such as temperature T and humidity H can be normalized using the Min-Max Normalization method.

[0042] Use infrared sensors, ultrasonic sensors or cameras to collect real-time environmental data in the room (such as temperature, humidity, personnel location and activity status, etc.).

[0043] Normalize or extract features from a certain sensor data (such as the temperature value collected by a temperature sensor or the humidity value collected by a humidity sensor), and use the processed data as input features. For example, if the original data value of a certain sensor data is z, the minimum value of the feature where the data is located is z min , the maximum value is z max , the normalized value z n The calculation formula is shown in the following formula (1).

[0044] z n =(z- z min ) / (z max -z min ) (1)

[0045] Through the above formula (1), all environmental parameters obtained by sensors are normalized in this way, and the value range of each environmental parameter is unified to the interval [0, 1]. In the normalization process, a reasonable threshold can be set to remove some obviously abnormal or erroneous data (i.e., unqualified parameters). For example, if the temperature sensor collects data that exceeds the normal indoor temperature range (such as below -20°C or above 50°C, and a reasonable boundary is set according to the actual situation), the data is determined to be abnormal data and is removed and not included in the subsequent processing flow.

[0046] Preprocess the environmental images captured by the camera to ensure that all images have the same format, such as common JPEG, PNG, etc. If the image formats of the environmental images are inconsistent, they need to be converted to a unified format. For environmental images (such as the images captured by the camera), use CNN to extract features from the environmental images to identify the distribution, number, and activity area of ​​people in the room.

[0047] The processing of the environment image includes:

[0048] (1) Resize the environment image;

[0049] (2) Using methods such as cropping and scaling, the environment image is uniformly scaled to the resolution required by the model, such as 224×224 pixels or other specified sizes;

[0050] (3) Normalize the pixel values ​​of the environment image to a specific range. The most common method is to normalize the pixel values ​​of the RGB image from [0, 255] to [0, 1].

[0051] Preprocessing of environmental parameters and environmental images helps the intelligent model converge faster, and to a certain extent reduces the impact of data differences on intelligent model training, and filters out possible noise. The environmental parameters and environmental images are processed through the CNN network, and finally the environmental vector Ev (T, H, n, x1, y1, ... x n ,y n , A).

[0052] During specific implementation, in step 2, the intelligent model includes a CNN network and a DQN network.

[0053] The current state of the environment vector Ev is input into DQN, and the best adjustment action (including adjusting the outlet angle, wind speed, and switch status, etc.) is selected through reinforcement learning. The corresponding airflow adjustment operation is performed according to the selected adjustment action, and the feedback data after adjustment (such as temperature change and user comfort) is used as the reward value to update the Q value function of DQN for optimization strategy.

[0054] During specific implementation, in step 2, the CNN network is used to process environmental data.

[0055] In specific implementation, one of the corner points of the room can be used as the origin O of the coordinate system, and the straight lines where the projection lines of the two adjacent and perpendicular walls of the corner point on the floor of the room are located are used as the X-axis and Y-axis respectively to establish a two-dimensional plane rectangular coordinate system. For the partitioning of personnel position coordinates, a clustering algorithm (such as the K-Means clustering algorithm) can be used. For example, the room is divided into 4 areas (the number of partitions K can be adjusted according to actual needs): Set the number of clusters K = 4, and the position coordinates P of the i-th person i (x i ,y i ) as data points for cluster analysis. Through algorithm calculation, four cluster centers can be obtained, corresponding to different areas, such as the upper left corner area, the upper right corner area, the lower left corner area, and the lower right corner area. The personnel position coordinates in each area belong to the corresponding cluster, and the personnel position information of the area in the state vector is identified in the form of area number or area average coordinates, and then output through the convolution layer, pooling layer and fully connected layer of the CNN network respectively.

[0056] 1. In the convolution layer of the CNN network, it is assumed that the convolution kernel size is 3×3 (the convolution kernel can be adjusted according to factors such as the actual input image resolution), the step size is set to 1, and the padding method uses SAME padding. For the input environment image (the number of channels in the convolution layer is Cj, and the size of the environment image is height H×width W), the size of the feature map output after the convolution operation is calculated using formula (2) to calculate the height Hout of the output feature map of the convolution layer.

[0057]

[0058] The feature map size calculation formula (3) after the convolution operation is used to calculate the output feature map width Wout of the convolution layer.

[0059]

[0060] In formula (2) and formula (3), H is the height of the input image of the input convolution layer, W is the width of the input image of the input convolution layer, k is the convolution kernel size of the convolution layer of the CNN network, p is the padding size of the convolution layer of the CNN network, and S j is the stride of the convolutional layer of the CNN network.

[0061] The number of channels Cj of the output feature map is determined by the number of convolution kernels. If the number of convolution kernels is N, the number of channels of the output feature map is also N. The activation function ReLU is shown in the following formula (4).

[0062] ReLU(f(px)=max(0,px) (4)

[0063] In formula (4), f(px)=max(0,px), where px represents the pixel of the environment image input to .

[0064] The activation function ReLU is used to enhance the nonlinear expression capability of the CNN network, making it easier to extract more complex features such as personnel locations.

[0065] 2. In the pooling layer of the CNN network, maximum pooling can be used. Set the pooling kernel size to 2×2 and the step size to 2. The input feature map size is Hout and Wout (i.e., the output volume Hout and Wout of the convolutional layer), that is. The calculation formula for the output feature map size after the pooling operation is shown in the following formulas (5) and (6).

[0066]

[0067] In formula (5) and formula (6), Hpool is the height of the image output after pooling, Wpool is the width of the image output after pooling, and S c is the stride of the pooling layer of the CNN network.

[0068] The dimensionality reduction operation of the pooling layer can reduce the number of features of the environmental image while retaining important information such as key personnel distribution and activity areas, which can reduce the subsequent calculation amount and improve the robustness of the features.

[0069] 3. The fully connected layer of the CNN network is used for feature flattening operation. Assume that the multidimensional feature map obtained after convolution, pooling and other operations is represented by a three-dimensional matrix X, and the dimension of the three-dimensional matrix X is represented as Cq×Hq×Wq; where Cq represents the number of channels of the fully connected layer, Hq represents the height of the multidimensional feature map input to the fully connected layer after convolution, pooling and other operations, and Wq represents the width of the multidimensional feature map input to the fully connected layer after convolution, pooling and other operations. Through the flattening operation, the three-dimensional matrix X is converted into a one-dimensional vector. Assume that the flattened feature vector is Xp, and the length of the feature vector Xp is M, which is represented by the following formula (7).

[0070] Xp=Flatten(X) (7)

[0071] In formula (7), Flatten(X) represents a flattening function, which transforms the three-dimensional matrix X into a one-dimensional vector Xp and satisfies the condition: M = Cq × Hq × Wq, where M is the number of elements in the one-dimensional vector Xp.

[0072] In the specific calculation, the one-dimensional vector Xp needs to be mapped to the environment vector Ev (T, H, n, x1, y1, ... x n ,y n , A), the dimension of the environment vector Ev is expressed as D; D is the number of elements in the environment vector Ev, then the number of neurons in the fully connected layer is set to D accordingly. In addition, there is a weight matrix Ww, whose dimension is M×D (that is, the above-mentioned eigenvector is the length M of Xt); the dimension of the bias vector b is a D×1 column vector (or a D-dimensional column vector, written in column vector form for the convenience of subsequent matrix operations). The flattened eigenvector Xp has a dimension of 1×M row vector. (Written in row vector form for the convenience of matrix multiplication operations). The flattened eigenvector Xp is linearly transformed by the weight matrix Ww, and the intermediate vector Z is obtained by performing matrix multiplication rules according to the formula (8).

[0073] Z=Xp×Ww (8)

[0074] In formula (8), the intermediate vector Z is an intermediate result vector after linear transformation, and its dimension is 1×D (or a D-dimensional row vector).

[0075] The intermediate vector Z is biased to obtain the output vector Ocnn of the fully connected layer (i.e., the output vector of the CNN network). The intermediate vector Z is added with the bias vector b for bias adjustment. According to the matrix addition rule (the dimensions of the matrices involved in the addition are required to be the same), the bias vector b is added to the intermediate vector Z to finally obtain the output vector Ocnn of the fully connected layer. The calculation formula is shown in formula (9).

[0076] Ocnn=Z+b=Xp×Ww+b (9)

[0077] The output vector Ocnn and the environment vector Ev(T, H, n, x1, y1, ... x n ,y n , A) is consistent with the dimension expected by the environment vector Ev. The output vector Ocnn also includes environmental parameters such as temperature T and humidity H and personnel parameters such as the number of personnel n, personnel position coordinates P and activity intensity A.

[0078] During specific implementation, in step 2, the DQN network is used to receive the output vector Ocnn of the CNN network processing environment data, and calculate the output vector Ocnn as the input vector of the DQN network to obtain the control instruction.

[0079] According to the optimal control action output by the DQN network, the airflow control execution unit adjusts the angle and wind speed of the multi-directional air outlet to achieve precise control of the airflow. If it is detected that a certain area is densely populated, the airflow coverage of the area will be increased; if the number of people in the area decreases, the air volume output of the area will be reduced to achieve energy saving.

[0080] Assume that the input vector of the DQN network is Ocnn(T, H, n, x1, y1, ... x n ,y n , A). In specific implementation, the number of elements in the input vector may be greater than that of the environment vector Ev(T, H, n, x1, y1, ... x n ,y n , A) has fewer elements. The present invention assumes in the calculation process that the output vector Ocnn is the same as the environment vector Ev.

[0081] The DQN network selects the best action based on the current state through the Q value function, which is the strategy for adjusting the air outlet direction and air volume of the air conditioner. The best actions may include: adjusting the outlet angle (horizontal and vertical), changing the wind speed, opening and closing the outlet, etc.

[0082] The control instructions include the control of adjusting the horizontal angle, adjusting the vertical angle, adjusting the wind speed, and opening and closing the air outlet. Therefore, the control instruction is also expressed by a row vector, and the control instruction Cc = (Js, Jc, v, Sw), where Js represents the horizontal angle of the air outlet after adjustment, Jc represents the vertical angle of the air outlet after adjustment, v represents the wind speed after adjustment, and Sw represents the opening and closing state of the air outlet after adjustment.

[0083] In the feedback update step, the reward value is calculated based on the feedback after the system executes the action (such as the temperature change detected by the sensor, the comfort score of the personnel, etc.). The reward function R can be designed as: reward value = comfort improvement value - energy consumption increase value, so that the system can reduce energy consumption as much as possible while ensuring comfort.

[0084] The reward value and the new environment state are used to update the Q value function of the DQN network, so that DQN gradually optimizes the airflow control strategy. This update process is repeated and iterated, and the system continuously optimizes the control strategy of the air conditioner. The optimal action is selected based on the learned Q value function. By continuously optimizing its Q value function, DQN can learn the optimal airflow control strategy under different environmental conditions.

[0085] The reward function R of the DQN network is expressed as follows: The reward function R is used to make the air conditioning system save energy as much as possible while improving user comfort.

[0086] R=α×(Scd- Scd0)-β×Nh (10)

[0087] In formula (10), α is the weight of comfort, Scd0 is the comfort score of the air conditioner in the initial state, Scd is the comfort score of the air conditioner system in the current state given by the user, which is used as the comfort value; β is the weight of the energy consumption index of the air conditioner system in the current state; Nh is the energy consumption value of the air conditioner system in the current state, which can be expressed by the current output power of the air conditioner, etc.

[0088] The comfort score Scd is a comprehensive indicator that indicates the user's satisfaction with the current state of the air conditioning system. Multiple factors should be considered in the specific implementation. For example, the contribution weights of temperature T, humidity H, wind speed v, airflow direction and other parameters to comfort can be set as wT, wH, wv and wD respectively.

[0089] Assuming that the comfort score of the air conditioner in the initial state is Scd0, and the comfort score in the current state after the air conditioner is adjusted is Scd, then the user comfort score changes to the difference between the current comfort score and the comfort score: Scd-Scd0. Scd can be calculated by the following formula (11).

[0090] Scd=wT×Tf+ wH×Hf +wv×vf+wD×Df (11)

[0091] In formula (11), Tf, Hf, vf and Df are the comfort contribution values ​​of temperature T, humidity H, wind speed v and airflow direction after the air conditioner is adjusted.

[0092] Taking temperature T as an example, the temperature comfort contribution value Tf can be calculated by the following formula (12) according to the temperature comfort range of the human body.

[0093] Tf=(TT min ) / (T max -T min ) (12)

[0094] In formula (12), T min and T max The lower and upper limits of the temperature comfort range respectively.

[0095] In addition, similar formula (12) can be used to calculate the comfort contribution values ​​Hf, vf and Df of humidity H, wind speed v and airflow direction.

[0096] According to formula (12), Scd0 = wT×T0f+wH×H0f+wv×v0f+wD×D0f; where T0f, H0f, v0f and D0f are the comfort contribution values ​​of temperature T, humidity H, wind speed v and airflow direction in the initial state, respectively.

[0097] Then the user comfort score change Scd-Scd0 can be expressed as the following formula (13).

[0098] Scd-Scd0=wT×(Tf-T0f)+wH×(Hf-H0f)+wv×(vf-v0f)+wD×(Df-D0f) (12)

[0099] Substituting the above formula (12) into formula (10), the reward value R of the current state can be obtained.

[0100] The reward value R and the new environment vector Ev, control command Cc and other existing data are used to update the Q value function of the DQN network, so that DQN gradually optimizes the airflow regulation strategy. The update rule of Q-learning is shown in the following formula (13).

[0101] Q(S,A)←Q(S,A)+η×[R+γ×max(Q(Ev',Cc'))-Q(Ev,Cc)] (13)

[0102] In formula (13), η is the learning rate, γ is the discount factor, Ev' is the environment vector of the next state, and Cc' is the control instruction of the next action.

[0103] The system of the present invention optimizes the airflow adjustment strategy by continuously monitoring environmental changes and user feedback. The feedback update process is as follows:

[0104] 1. Environmental data feedback: Real-time monitoring of environmental changes after adjustment, such as changes in temperature and humidity, and updating of system status.

[0105] 2. User feedback collection (optional): When the user manually adjusts the air conditioning settings, the system records this behavior and automatically makes similar adjustments in subsequent similar scenarios.

[0106] 3. Reward mechanism: Use environmental data feedback and user feedback as reward signals to update the Q-value function of DQN, thereby improving the accuracy of the adjustment strategy.

[0107] 4. Online learning or offline training: The system can choose online learning (real-time update of model parameters) or strategy optimization through regular offline training data sets to ensure that the system adapts to dynamically changing environments.

[0108] Users can also manually adjust the angle and direction of the air outlet through the mobile phone app or the air conditioner remote control. The system will provide a virtual room layout map, and users only need to select the corresponding area to adjust the airflow direction, which is very intuitive.

[0109] The air conditioning system control method with dynamic airflow regulation of the present invention realizes intelligent and adaptive regulation of airflow by using artificial intelligence algorithm, improves user comfort and reduces energy consumption, and has broad application prospects in homes, offices and large public places.

[0110] The present invention also discloses a dynamic airflow regulating air conditioning system control system, comprising a sensor unit, a central control unit, a multi-directional air outlet execution unit and an airflow regulating execution unit;

[0111] The sensor unit is used to collect environmental data in the temperature control area and send the environmental data to the central control unit;

[0112] The intelligent model of the central control unit is used to process and calculate the environmental data, obtain control instructions, and send the control instructions to the multi-directional air outlet execution unit and the air flow adjustment execution unit of the air-conditioning system. The multi-directional air outlet execution unit and the air flow adjustment execution unit execute the control instructions to complete the adjustment action, and save the environmental data and the adjustment action generated by the environmental data in the memory as historical reference data to adjust and update the intelligent model.

[0113] The present invention also discloses an electronic device, comprising at least one processor and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the air conditioning system control method with dynamic airflow adjustment.

[0114] The present invention also discloses a non-transient computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the air conditioning system control method for dynamic airflow regulation.

[0115] The present invention also discloses a computer program product, comprising a computer program; when the computer program is executed by a processor, the air conditioning system control method for dynamic airflow regulation is realized.

[0116] The central control unit has built-in artificial intelligence algorithms, including Convolutional Neural Networks (CNN) and Deep Q-Network (DQN), which are used to process sensor data and execute airflow adjustment decisions. The multi-directional air outlet structure consists of multiple independently controllable air outlets, each of which can adjust the angle and air volume to achieve precise control of airflow.

[0117] The air conditioning system control method for dynamic airflow regulation of the present invention has the following characteristics.

[0118] 1. Discretization of state and action space: In order to reduce the computational complexity, the environment vector Ev and the control command Cc can be discretized, for example, the air outlet angle can be divided into several fixed angles.

[0119] 2. Balance of reward function R: By adjusting the values ​​of α and β, the weights of comfort and energy saving are balanced to avoid extreme situations (such as excessive energy saving leading to reduced comfort).

[0120] 3. Separation of training and inference: Pre-train the algorithm in an offline environment, and then make small adjustments in a real-time environment to ensure system responsiveness.

[0121] The control method of the air conditioning system with dynamic airflow regulation of the present invention, its control process and algorithm scheme can realize intelligent regulation of the air conditioning system, improve the user's comfort, and reduce energy consumption.

[0122] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0123] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for controlling an air conditioning system with dynamic airflow regulation, characterized in that: The steps include: Step 1: Data collection step: Collect environmental data in the temperature control area through the sensor unit and send the environmental data to the central control unit; Step 2: Data processing step; the central control unit has a built-in intelligent model, and the environmental data is input into the intelligent model; the intelligent model performs calculations and outputs control instructions, and transmits the control instructions to the air conditioning system; Step 3: Control execution step; the air conditioning system executes the adjustment action of the air conditioning system according to the control instruction, stores the adjustment action into the memory and feeds back to the intelligent model; Step 4: Feedback update step; the intelligent model stores the environmental data and the adjustment actions generated by the environmental data in a memory as historical reference data to adjust and update the intelligent model.

2. The method for controlling an air conditioning system with dynamic airflow regulation according to claim 1, characterized in that: In step 1, the environmental data includes environmental parameters and environmental images; the environmental parameters include the temperature T and humidity H in the temperature control area; the personnel parameters are obtained according to the environmental image; the personnel parameters include the number of personnel n, the position coordinates P of the personnel and the activity intensity A.

3. The method for controlling an air conditioning system with dynamic airflow regulation according to claim 1, characterized in that: In step 2, the central control unit pre-processes the environmental data before inputting it into the intelligent model.

4. The method for controlling an air conditioning system with dynamic airflow regulation according to claim 1, characterized in that: In step 2, the intelligent model includes a CNN network and a DQN network.

5. The method for controlling an air conditioning system with dynamic airflow regulation according to claim 1, characterized in that: In step 2, the CNN network is used to process environmental data.

6. A method for controlling an air conditioning system with dynamic airflow regulation according to claim 5, characterized in that: In step 2, the DQN network is used to receive the output vector Ocnn of the CNN network processing environment data, and use the output vector Ocnn as the input vector of the DQN network to calculate the control instruction.

7. An air conditioning system control system with dynamic airflow adjustment according to claim 1, characterized in that: It includes a sensor unit, a central control unit, a multi-directional air outlet execution unit and an airflow adjustment execution unit; The sensor unit is used to collect environmental data in the temperature control area and send the environmental data to the central control unit; The intelligent model of the central control unit is used to process and calculate the environmental data, obtain control instructions, and send the control instructions to the multi-directional air outlet execution unit and the air flow adjustment execution unit of the air-conditioning system. The multi-directional air outlet execution unit and the air flow adjustment execution unit execute the control instructions to complete the adjustment action, and save the environmental data and the adjustment action generated by the environmental data in the memory as historical reference data to adjust and update the intelligent model.

8. An electronic device, characterized in that: It includes at least one processor and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the air conditioning system control method with dynamic airflow regulation described in any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: in, The computer instructions are used to enable the computer to execute the air conditioning system control method with dynamic air flow adjustment according to any one of claims 1-6.

10. A computer program product comprising a computer program; wherein: When the computer program is executed by a processor, the computer program implements the air conditioning system control method with dynamic airflow adjustment according to any one of claims 1 to 6.

Images