Intelligent air conditioner management and control and scenario mode adjustment system

Abstract

The application relates to the technical field of air conditioning systems, and discloses an intelligent air conditioning management and control and scene mode adjustment system, which comprises the following: a comprehensive monitoring module that monitors air quality parameters and user physiological data indoors and outdoors in real time; a strategy generation module that analyzes and processes the collected data to generate an intelligent air conditioning control strategy; a control execution module that adjusts the operation parameters of the air conditioning system in real time according to the generated control strategy; a scene mode identification module that analyzes the air quality parameters and user physiological data to identify the current scene; a scene mode adjustment module that adjusts the operation parameters of the air conditioning system to match the operation state of the current scene mode; and a user interaction interface that provides a user interaction mode and can realize manual switching of the scene mode and adjustment of the operation parameters of the air conditioning system. The application significantly improves the intelligent level of the air conditioning system, optimizes energy consumption management, and improves the comfort and use experience of users.

Description

Technical Field

[0001] This invention relates to the field of air conditioning system technology, and in particular to an intelligent air conditioning control and scene mode adjustment system. Background Technology

[0002] Existing air conditioning systems often fail to dynamically adjust operating parameters based on changes in indoor and outdoor environments and actual user needs, leading to energy waste and reduced user comfort. For example, traditional air conditioning systems cannot simultaneously consider humidity, air quality, and other factors when adjusting temperature, resulting in users enjoying comfortable temperatures while facing poor air quality or excessive energy consumption. Furthermore, existing systems lack precise parameter adjustments when switching between scene modes, which can easily cause user discomfort.

[0003] Existing air conditioning systems suffer from insufficient accuracy in their scene mode switching functions. For example, traditional air conditioning systems often fail to accurately identify the current scene when switching modes, resulting in operating parameters that do not match actual needs. Furthermore, existing systems do not provide smooth parameter adjustments during scene mode switching, which can easily cause user discomfort. For instance, the sudden change in temperature and fan speed when switching from "operating mode" to "sleep mode" may cause discomfort to the user.

[0004] Existing air conditioning systems have certain limitations in terms of user interaction. For example, while traditional voice control and gesture recognition provide convenient interaction methods, their accuracy and response speed still need improvement. Furthermore, the display interfaces of existing systems are relatively simple, unable to display detailed air quality data and scene mode status in real time, making it difficult for users to intuitively understand the system's operating status.

[0005] Existing air conditioning systems have shortcomings in data monitoring and processing. For example, traditional air quality monitoring modules typically only monitor a single parameter, failing to comprehensively reflect indoor environmental quality. Furthermore, existing systems have limited data processing capabilities, making it impossible to analyze and predict air quality changes in real time, and thus difficult to generate precise control strategies. Therefore, this invention proposes an intelligent air conditioning control and scenario mode adjustment system. Summary of the Invention

[0006] The purpose of this invention is to solve the problems of tedious and complex calculation processes and large calculation workload in the prior art, and to propose an intelligent air conditioning control and scenario mode adjustment system.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent air conditioning control and scene mode adjustment system, comprising:

[0008] The integrated monitoring module uses multi-sensor fusion technology to monitor indoor and outdoor air quality parameters and user physiological data in real time.

[0009] The strategy generation module analyzes and processes the collected data using an improved deep learning algorithm to predict changes in air quality and generate intelligent air conditioning control strategies.

[0010] The control execution module adjusts the operating parameters of the air conditioning system in real time according to the generated control strategy through frequency conversion technology and intelligent air direction adjustment technology, so as to control air pollutants and reduce energy consumption.

[0011] The scenario pattern recognition module uses support vector machine (SVM) to analyze air quality parameters and user physiological data to identify the current scenario. It also optimizes the recognition logic through improved machine learning algorithms to improve the accuracy of scenario pattern recognition.

[0012] The scenario mode adjustment module generates a user preference model through the Transformer architecture, predicts user preferences, and gradually adjusts the operating parameters of the air conditioning system through a progressive parameter adjustment algorithm to match the current scenario mode's operating status, thus achieving smooth switching between scenario modes.

[0013] The user interface provides a way for users to interact with the system. It supports voice control and gesture recognition, and allows users to manually switch scene modes and adjust the operating parameters of the air conditioning system. It also displays the current air quality data and scene mode status in real time.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0015] This invention incorporates a comprehensive monitoring module that employs multi-sensor fusion technology to monitor indoor and outdoor air quality parameters and user physiological data in real time. Furthermore, it dynamically adjusts the fusion strategy through an adaptive fusion algorithm, thereby improving the accuracy of data monitoring and enhancing the system's data processing capabilities.

[0016] This invention includes a strategy generation module that generates intelligent air conditioning control strategies. This not only optimizes energy consumption but also adjusts operating parameters according to the user's actual needs, achieving the dual goals of energy consumption optimization and comfort balance.

[0017] This invention includes a scenario mode recognition module and a scenario mode adjustment module, which can accurately identify the current scenario, precisely predict user preferences, and achieve smooth switching between scenario modes, significantly improving the accuracy of scenario mode switching and user experience.

[0018] This invention features a user interface that employs a deep recurrent neural network combined with an attention mechanism to improve the accuracy of speech recognition and semantic parsing. Users can directly adjust air conditioning parameters and switch scene modes through touch operations. This intuitive and convenient interaction method significantly enhances the user experience. Attached Figure Description

[0019] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the system configuration provided in an embodiment of the present invention. Detailed Implementation

[0021] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following detailed description, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the intelligent air conditioning control and scene mode adjustment system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0023] The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0024] The specific solution of the intelligent air conditioning control and scene mode adjustment system provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0025] Example

[0026] An embodiment of the present invention provides an intelligent air conditioning control and scene mode adjustment system, including:

[0027] The integrated monitoring module uses multi-sensor fusion technology to monitor indoor and outdoor air quality parameters and user physiological data in real time.

[0028] The strategy generation module analyzes and processes the collected data using an improved deep learning algorithm to predict changes in air quality and generate intelligent air conditioning control strategies.

[0029] The control execution module adjusts the operating parameters of the air conditioning system in real time according to the generated control strategy through frequency conversion technology and intelligent air direction adjustment technology, so as to control air pollutants and reduce energy consumption.

[0030] The scenario pattern recognition module uses support vector machine (SVM) to analyze air quality parameters and user physiological data to identify the current scenario. It also optimizes the recognition logic through improved machine learning algorithms to improve the accuracy of scenario pattern recognition.

[0031] The scenario mode adjustment module generates a user preference model through the Transformer architecture, predicts user preferences, and gradually adjusts the operating parameters of the air conditioning system through a progressive parameter adjustment algorithm to match the current scenario mode's operating status, thus achieving smooth switching between scenario modes.

[0032] The user interface provides a way for users to interact with the system. It supports voice control and gesture recognition, and allows users to manually switch scene modes and adjust the operating parameters of the air conditioning system. It also displays the current air quality data and scene mode status in real time.

[0033] Please refer to Figure 1 This is a schematic diagram of the system configuration provided in an embodiment of the present invention.

[0034] I. Comprehensive Monitoring Module

[0035] The integrated monitoring module includes air quality sensors, physiological sensors, and a data fusion unit;

[0036] Air quality sensors, including temperature sensors, humidity sensors, PM2.5 sensors, VOCs sensors, and carbon dioxide sensors, are used to collect indoor and outdoor air quality parameters, including temperature, humidity, PM2.5, VOCs, and carbon dioxide.

[0037] Physiological sensors, including heart rate sensors and respiratory rate sensors, are used to collect users' physiological data, including heart rate and respiratory rate.

[0038] The data fusion unit uses multi-sensor fusion technology to fuse sensor data and generate a comprehensive data signal. The multi-sensor fusion technology employs an adaptive fusion algorithm and adjusts the fusion strategy through an online learning algorithm.

[0039] It should be noted that the temperature sensor is used to measure the ambient temperature and provide temperature data.

[0040] Humidity sensor: Used to measure ambient humidity and provide humidity data.

[0041] PM2.5 sensor: Used to measure the concentration of fine particulate matter (PM2.5) in the air and provide air quality data.

[0042] VOCs sensors: used to measure the concentration of volatile organic compounds (VOCs) and provide air quality data.

[0043] Carbon dioxide sensor: Used to measure the concentration of carbon dioxide (CO2) and provide air quality data.

[0044] Physiological sensors are used to collect users' physiological data to reflect their comfort and health status. By collecting users' physiological data, the system can better understand users' actual needs, thereby achieving more precise comfort control.

[0045] Heart rate sensor: Used to measure the user's heart rate and provide physiological data.

[0046] Respiratory rate sensor: Used to measure the user's respiratory rate and provide physiological data.

[0047] Multi-sensor fusion technology is a technique that integrates data from multiple sensors to improve the system's perception capabilities and decision-making accuracy. In intelligent air conditioning control and scenario mode adjustment systems, multi-sensor fusion technology can effectively integrate multiple data sources to generate more comprehensive and accurate integrated data signals, thereby improving the accuracy and reliability of the data.

[0048] Adaptive fusion algorithm is an algorithm that dynamically adjusts the data fusion strategy. It can automatically optimize the fusion process based on real-time data and environmental changes. In intelligent air conditioning systems, adaptive fusion algorithm can automatically optimize the data fusion process according to different environments and user states, ensuring that the generated comprehensive data signal can truly reflect the current environment and user state.

[0049] Online learning algorithms are algorithms that can update model parameters in real time during system operation. In intelligent air conditioning systems, online learning algorithms enable the system to continuously learn and adapt to new data patterns and automatically adjust fusion strategies to cope with changes in the environment and user behavior.

[0050] II. Strategy Generation Module

[0051] The strategy generation module includes an image data acquisition unit, a 3D reconstruction unit, a surgical path planning unit, and a 3D printed positioning guide unit;

[0052] The data preprocessing unit performs validity checks on the collected air quality parameters and user physiological data, removes invalid and duplicate values, and uses the Min-Max normalization method to normalize all data to the [0,1] interval, eliminating the differences between data of different dimensions.

[0053] The deep learning model unit uses an improved deep learning algorithm to extract and analyze features from preprocessed data and predict air quality change trends. The improved deep learning algorithm is a combination of Long Short-Term Memory Network (LSTM) and Convolutional Neural Network (CNN).

[0054] The strategy generation unit constructs a multi-objective optimization function based on the prediction results of the deep learning model. Combining the preset comfort and energy consumption targets, it uses the particle swarm optimization algorithm to optimize the operating parameters of the air conditioning system and generate an intelligent air conditioning control strategy. The intelligent air conditioning control strategy includes temperature setting, fan speed adjustment, air direction adjustment, and purification mode switching.

[0055] Furthermore, in the policy generation module, the improved deep learning algorithm is a combination of a Long Short-Term Memory (LSTM) network and a Convolutional Neural Network (CNN), wherein:

[0056] Improved deep learning algorithms include convolutional layers, recurrent layers, feature fusion layers, fully connected layers, optimization algorithms, loss functions, and regularization units;

[0057] Convolutional layers are used to extract spatial features of data. Air quality parameters and user physiological data are treated as multi-channel inputs, and local features are extracted through convolution operations.

[0058] The recurrent layer uses a long short-term memory network, taking the features extracted by the convolutional layer as input to capture the changing trends of time series data;

[0059] The feature fusion layer dynamically adjusts the weights of the output features of the convolutional and LSTM layers through an attention mechanism to achieve the fusion of spatial and temporal features;

[0060] The fully connected layer processes the fused features to generate prediction results, which include air quality change trends and air conditioning control strategies.

[0061] The optimization algorithm employs an improved Adam optimizer, combined with a learning rate decay strategy, to improve the training efficiency and convergence speed of the model.

[0062] The loss function uses a weighted mean squared error loss function, which assigns different weights based on the importance and reliability of the data to improve the model's prediction accuracy.

[0063] The regularization unit introduces Dropout technology to prevent model overfitting and improve the model's generalization ability.

[0064] It should be noted that validity checking is an important step in data preprocessing, aiming to ensure the accuracy and reliability of the data. In intelligent air conditioning systems, validity checking includes invalid value detection: by setting a threshold range, values ​​outside the range are considered invalid; and duplicate value detection: by comparing adjacent data points or using a hash table to detect duplicate data and remove duplicate values.

[0065] Invalid values: These refer to unusable data collected by the sensor, abnormally high or low values ​​caused by sensor malfunction, or default values ​​when the sensor was not properly initialized. These values ​​need to be identified and removed to avoid misleading subsequent analysis.

[0066] Duplicate values: These are data points that appear repeatedly in a dataset. This may be due to repeated sampling by the sensor or errors in data transmission. Duplicate values ​​need to be detected and removed to reduce data redundancy and improve processing efficiency.

[0067] Min-Max normalization is a commonly used data normalization method used to scale data to the [0,1] interval. Normalization can eliminate the differences between data of different dimensions, making the data comparable. Normalized data helps improve the training efficiency and prediction accuracy of machine learning models.

[0068] Air quality change trends refer to the changes in air quality parameters (temperature, humidity, PM2.5 concentration, VOCs concentration, carbon dioxide) over time. By analyzing these trends, future air quality changes can be predicted, providing a basis for the control strategies of intelligent air conditioning systems.

[0069] Long Short-Term Memory (LSTM) networks are a special type of recurrent neural network (RNN) specifically designed for processing time series data. By introducing a gating mechanism, LSTM can effectively capture long-term dependencies and avoid the gradient vanishing problem of traditional RNNs.

[0070] Convolutional Neural Networks (CNNs) are a type of deep learning model that excels at processing data with a grid structure, including images and multidimensional time series data. CNNs use a structure of convolutional layers, pooling layers, and fully connected layers to automatically extract local features from the data.

[0071] Multi-objective optimization functions are optimization methods that consider multiple objectives simultaneously. In intelligent air conditioning systems, multi-objective optimization functions need to comprehensively consider both comfort and energy consumption objectives.

[0072] Particle Swarm Optimization (PSO) is a swarm intelligence-based optimization algorithm that finds the optimal solution by simulating the foraging behavior of bird flocks. In intelligent air conditioning systems, PSO can optimize the operating parameters of the air conditioning system.

[0073] Intelligent air conditioning control strategy refers to dynamically adjusting the operating parameters of the air conditioning system based on predicted air quality change trends and optimization results in order to achieve a balance between comfort and energy consumption.

[0074] Temperature setting: The air conditioner's set temperature is dynamically adjusted based on predicted air quality trends and user preferences.

[0075] Fan speed adjustment: The air conditioner's fan speed is dynamically adjusted according to the indoor and outdoor environment and user needs.

[0076] Airflow adjustment: The air conditioner dynamically adjusts the airflow direction based on the user's location and the surrounding environment to avoid blowing directly on the human body.

[0077] Purification mode switching: Automatically switches air purification modes (HEPA filtration, activated carbon adsorption, negative ion generation) based on air quality parameters.

[0078] Attention mechanisms are mechanisms that simulate human attention, dynamically adjusting the model's focus on different input features. In deep learning, attention mechanisms improve model performance by calculating the importance weights of input features and concentrating more attention on important features.

[0079] The Adam optimizer is a gradient descent-based optimization algorithm that combines the advantages of momentum and RMSProp. The improved Adam optimizer introduces a learning rate decay strategy to dynamically adjust the learning rate, thereby improving the training efficiency and convergence speed of the model.

[0080] The learning rate decay strategy dynamically adjusts the learning rate to avoid training instability caused by an excessively high learning rate, while reducing the learning rate in the later stages of training to improve the model's convergence accuracy.

[0081] The weighted mean squared error loss function is an improved mean squared error loss function that increases the model's attention to important data points by assigning different weights to different data points.

[0082] Dropout is a regularization method that prevents overfitting and improves the model's generalization ability by randomly dropping a portion of neurons during training.

[0083] The following is the code for the improved deep learning algorithm:

[0084] import tensorflow as tf

[0085] from tensorflow.keras.layers import Conv1D, LSTM, Dense, Dropout, TimeDistributed, Flatten

[0086] from tensorflow.keras.models import Model

[0087] from tensorflow.keras.optimizers import Adam

[0088] from tensorflow.keras.callbacks import ReduceLROnPlateau

[0089] # The shape of the input data

[0090] time_steps = 100 # Time step size

[0091] features = 10 # Number of features

[0092] input_shape = (time_steps, features)

[0093] # Input layer

[0094] inputs = tf.keras.Input(shape=input_shape)

[0095] # Convolutional layer

[0096] conv1d = Conv1D(filters=64, kernel_size=3, activation='relu', padding='same')(inputs)

[0097] # LSTM layer

[0098] lstm = LSTM(units=64, return_sequences=True)(conv1d)

[0099] # Attention Mechanism

[0100] def attention_layer(inputs, units):

[0101] input_dim = int(inputs.shape[2])

[0102] attention_score = Dense(input_dim, activation='tanh')(inputs)

[0103] attention_score = Dense(1)(attention_score)

[0104] attention_score = tf.nn.softmax(attention_score, axis=1)

[0105] context_vector = tf.reduce_sum(inputs * attention_score, axis=1)

[0106] return context_vector

[0107] # Feature Fusion

[0108] attention = attention_layer(lstm, 64)

[0109] # Fully Connected Layer

[0110] dense = Dense(128, activation='relu')(attention)

[0111] dropout = Dropout(0.5)(dense)

[0112] # Predicting air quality trends

[0113] predictions = Dense(1, activation='linear')(dropout)

[0114] # Generate air conditioning control strategy

[0115] control_strategy = Dense(4, activation='softmax')(dropout)

[0116] # Define the optimizer

[0117] optimizer = Adam(learning_rate=0.001)

[0118] # Learning rate decay

[0119] reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.5,patience=5, min_lr=0.00001)

[0120] # Custom weighted mean squared error loss function

[0121] def weighted_mse(y_true, y_pred):

[0122] weights = tf.constant([1.0, 2.0, 1.5, 1.0]) # Assume weights

[0123] return tf.reduce_mean(tf.square(y_true - y_pred) * weights)

[0124] # Compilation Model

[0125] model = Model(inputs=inputs, outputs=[predictions, control_strategy])

[0126] model.compile(optimizer=optimizer, loss=[weighted_mse, 'categorical_crossentropy'], loss_weights=[1.0, 0.5])

[0127] # Model Summary

[0128] model.summary()

[0129] III. Control Execution Module

[0130] The control and execution module includes a frequency converter control unit, an intelligent airflow adjustment unit, and an air purification unit;

[0131] The variable frequency control unit adjusts the operating frequency of the air conditioner compressor in real time according to the generated control strategy. The PID controller adjusts the compressor frequency according to the deviation between the indoor temperature and the set temperature to achieve temperature control and energy consumption optimization.

[0132] The intelligent airflow adjustment unit uses multi-dimensional airflow adjustment technology. Through multiple micro motors and sensors installed at the air conditioner outlet, it automatically adjusts the airflow direction of the air conditioner according to the indoor space layout and the user's position to avoid blowing directly on the human body and improve comfort.

[0133] The air purification unit automatically switches between air purification modes based on air quality parameters to remove pollutants from the air. The air purification modes include HEPA filtration, activated carbon adsorption, and negative ion generation.

[0134] It should be noted that the PID controller is a common feedback controller that adjusts the control quantity through three parameters: proportional (P), integral (I), and derivative (D). In intelligent air conditioning systems, the PID controller dynamically adjusts the compressor's operating frequency based on the deviation between the indoor temperature and the set temperature.

[0135] Proportional (P): Directly adjusts the control quantity based on the current deviation, quickly responding to changes in deviation.

[0136] Integral (I): Adjust the control quantity based on the cumulative value of the deviation to eliminate steady-state error.

[0137] Derivative (D): Adjust the control quantity according to the rate of change of the deviation, predict the trend of deviation change, adjust the control quantity in advance, and reduce overshoot.

[0138] Adjusting the compressor frequency: By adjusting the compressor frequency, the inverter control unit can reduce energy consumption while meeting temperature control requirements. When the indoor temperature is close to the set temperature, the compressor frequency is reduced to reduce energy consumption; when the indoor temperature deviates from the set temperature, the compressor frequency is increased to quickly adjust the temperature.

[0139] Multi-dimensional airflow adjustment technology: Through multiple micro motors and sensors, the intelligent airflow adjustment unit can achieve multi-dimensional airflow adjustment. By adjusting the airflow in both horizontal and vertical directions, it ensures that the airflow direction can be optimized according to the indoor space layout and the user's location.

[0140] Avoid direct airflow to the human body: The intelligent airflow adjustment unit can automatically adjust the airflow direction according to the user's position and activity status to avoid direct airflow to the human body and reduce user discomfort.

[0141] HEPA filtration: HEPA (High-Efficiency Particulate Filter) can effectively filter fine particulate matter in the air, including PM2.5, pollen, and dust, improving air quality.

[0142] Activated carbon adsorption: Activated carbon has a strong adsorption capacity and can adsorb harmful gases and odors in the air, including volatile organic compounds (VOCs) such as formaldehyde and benzene.

[0143] Negative ion generation: Negative ion generators can produce negative ions, which can combine with pollutants in the air and cause them to settle, thereby purifying the air. Negative ions also have certain health benefits, including improving sleep and reducing fatigue.

[0144] Automatic mode switching: The air purification unit can automatically switch between different purification modes based on real-time monitored air quality parameters. When a high PM2.5 concentration is detected, the system automatically switches to HEPA filtration mode; when a high VOCs concentration is detected, the system automatically switches to activated carbon adsorption mode.

[0145] IV. Scene Pattern Recognition Module

[0146] The contextual pattern recognition module includes a data analysis unit, a deep learning classifier unit, and a logic optimization unit;

[0147] The data analysis unit performs real-time analysis on the collected air quality parameters and user physiological data. Through statistical analysis and feature extraction algorithms, it extracts feature vectors, which include temperature, humidity, PM2.5 concentration, heart rate, and respiratory rate.

[0148] The deep learning classifier unit uses a support vector machine (SVM) to classify feature vectors, identify the current scenario mode, classify real-time data using a trained classification model, and output the current scenario mode. Scenario modes include sleep mode, work mode, leisure mode, sports mode, party mode, and energy-saving mode.

[0149] The logic optimization unit optimizes the classifier's recognition logic through improved machine learning algorithms, adjusting the classifier's parameters and thresholds to improve the accuracy of scene pattern recognition.

[0150] Furthermore, in the context pattern recognition module, the logic optimization unit optimizes the classifier's recognition logic using an improved machine learning algorithm, adjusting the classifier's parameters and thresholds to improve the accuracy of context pattern recognition. Specifically:

[0151] The improved machine learning algorithm combines reinforcement learning algorithms, Bayesian optimization methods, online learning algorithms, cross-validation mechanisms, and ensemble learning methods.

[0152] Reinforcement learning algorithms dynamically adjust the parameters of the classifier through reward and punishment mechanisms to maximize recognition accuracy;

[0153] Bayesian optimization methods dynamically adjust the threshold of the classifier through a Bayesian probability model, thereby improving the robustness of recognition.

[0154] Online learning algorithms continuously update the classifier's model parameters based on real-time feedback data, adapting to changes in the environment and user behavior.

[0155] Cross-validation evaluates the classifier's performance through multiple rounds of validation, avoiding overfitting and underfitting.

[0156] Ensemble learning methods combine the outputs of multiple classifiers to improve recognition accuracy.

[0157] It should be noted that statistical analysis includes descriptive statistics: calculating the mean, median, and standard deviation of the data to understand the distribution of the data; and correlation analysis: calculating the correlation coefficients between different features to identify features that have an important impact on scene pattern recognition.

[0158] Feature extraction algorithms include Principal Component Analysis (PCA): reducing feature dimensionality through dimensionality reduction techniques while retaining the main information of the data; Wavelet Transform: used to extract time-frequency features of signals, suitable for processing time series data; Discrete Cosine Transform (DCT): used to extract frequency domain features of signals; Adaptive Feature Extraction: dynamically adjusting the feature extraction method according to the characteristics of the data to adapt to different scenario modes.

[0159] Temperature: Indoor temperature, reflecting the thermal comfort of the environment.

[0160] Humidity: Indoor humidity affects human comfort.

[0161] PM2.5 concentration: The concentration of fine particulate matter in the air, reflecting air quality.

[0162] Heart rate: The user's heart rate reflects the user's physiological state.

[0163] Breathing rate: The user's breathing rate reflects the user's activity status.

[0164] Support Vector Machine (SVM) is a commonly used classification algorithm that separates data of different categories by finding the optimal hyperplane. It can effectively distinguish data of different categories and improve the accuracy of scene pattern recognition. Through kernel tricks, SVM can handle non-linearly separable data and has good robustness.

[0165] Sleep mode: An environment setting suitable for users while they sleep, optimizing temperature, humidity, and fan speed to provide a comfortable sleep environment;

[0166] Working mode: Suitable for the user's working environment settings, optimizing air quality (such as reducing PM2.5 concentration) and temperature to improve work efficiency;

[0167] Leisure Mode: An environment setting suitable for users during leisure time, optimizing temperature, humidity, and wind direction to provide a comfortable leisure environment;

[0168] Sports Mode: Environmental settings suitable for users exercising indoors, optimizing temperature, humidity and wind speed to accommodate higher activity intensity;

[0169] Party Mode: Suitable for setting up environments for large gatherings, optimizing air quality, temperature, and wind direction to suit the needs of large group activities;

[0170] Energy-saving mode: Optimizes air conditioning operating parameters to reduce energy consumption while ensuring comfort.

[0171] Reinforcement learning algorithms dynamically adjust the parameters of the classifier through reward and punishment mechanisms to maximize recognition accuracy. The core of reinforcement learning is that the agent takes actions in the environment and adjusts the policy based on the feedback from the environment.

[0172] Reward mechanism: When the classifier correctly identifies the context pattern, a positive reward is given; when the classifier misidentifies, a negative reward is given.

[0173] Bayesian optimization methods dynamically adjust the threshold of the classifier through a Bayesian probability model to improve the robustness of recognition. The core of Bayesian optimization is to use prior knowledge and observation data to update the posterior distribution through Bayesian inference, thereby optimizing the parameters.

[0174] Bayesian probabilistic model: By constructing a Bayesian probabilistic model, the threshold of the classifier can be dynamically adjusted.

[0175] Online learning algorithms continuously update the classifier's model parameters based on real-time feedback data, adapting to changes in the environment and user behavior. The core of online learning is to gradually update the model parameters through incremental learning methods, avoiding the need to retrain the entire model.

[0176] Cross-validation evaluates the performance of a classifier through multiple rounds of validation, avoiding overfitting and underfitting. The core of cross-validation is to divide the dataset into multiple subsets and perform multiple rounds of validation to evaluate the model's performance.

[0177] Ensemble learning methods combine the outputs of multiple classifiers to improve recognition accuracy. The core of ensemble learning is to improve the robustness of the system by combining the outputs of multiple classifiers through voting or weighted averaging.

[0178] The following is the code for the improved machine learning algorithm:

[0179] import numpy as np

[0180] import gym

[0181] from stable_baselines3 import PPO

[0182] from skopt import gp_minimize

[0183] from skopt.space import Real, Integer

[0184] from skopt.utils import use_named_args

[0185] from sklearn.linear_model import SGDClassifier

[0186] from sklearn.model_selection import cross_val_score

[0187] from sklearn.ensemble import RandomForestClassifier

[0188] from sklearn.ensemble import VotingClassifier

[0189] from sklearn.linear_model import LogisticRegression

[0190] from sklearn.svm import SVC

[0191] # Define the environment

[0192] class ScenarioRecognitionEnv(gym.Env):

[0193] def __init__(self):

[0194] super(ScenarioRecognitionEnv, self).__init__()

[0195] self.action_space = gym.spaces.Discrete(2) # 0: Adjust parameters, 1: Keep unchanged

[0196] self.observation_space = gym.spaces.Box(low=-1, high=1, shape=(n_features,), dtype=np.float32)

[0197] self.state = np.zeros(n_features)

[0198] self.reward = 0

[0199] def step(self, action):

[0200] if action == 0:

[0201] # Adjust parameters

[0202] self.state += np.random.normal(0, 0.1, size=n_features)

[0203] else:

[0204] # Remain unchanged

[0205] pass

[0206] self.reward = self.calculate_reward()

[0207] done = self.is_done()

[0208] return self.state, self.reward, done, {}

[0209] def reset(self):

[0210] self.state = np.zeros(n_features)

[0211] self.reward = 0

[0212] return self.state

[0213] def calculate_reward(self):

[0214] # Assume the reward function is the recognition accuracy

[0215] accuracy = self.evaluate_accuracy(self.state)

[0216] return accuracy

[0217] def is_done(self):

[0218] # Assume the process ends when the accuracy reaches a certain threshold.

[0219] return self.calculate_reward()>0.9

[0220] def evaluate_accuracy(self, state):

[0221] # Accuracy evaluation function of the hypothesis

[0222] return np.random.uniform(0.7, 0.9)

[0223] # Initialize environment and model

[0224] env = ScenarioRecognitionEnv()

[0225] model = PPO("MlpPolicy", env, verbose=1)

[0226] # Training Model

[0227] model.learn(total_timesteps=10000)

[0228] # Save Model

[0229] model.save("scenario_recognition_model")

[0230] # Bayesian optimization

[0231] @use_named_args(dimensions=[Real(0.1, 1.0, name='learning_rate'),

[0232] Integer(10, 50, name='num_leaves'),

[0233] Integer(10, 50, name='max_depth')])

[0234] def objective(**params):

[0235] # Hypothetical optimization objective function

[0236] return -np.random.uniform(0.7, 0.9)

[0237] dimensions = [Real(0.1, 1.0, name='learning_rate'),

[0238] Integer(10, 50, name='num_leaves'),

[0239] Integer(10, 50, name='max_depth')]

[0240] result = gp_minimize(objective, dimensions, n_calls=10, random_state=0, verbose=True)

[0241] print("Best parameters:", result.x)

[0242] # Online learning

[0243] model_online = SGDClassifier(max_iter=1, tol=None)

[0244] for t in range(100):

[0245] X_t = np.random.rand(1, n_features) # Features at the current time step

[0246] y_t = np.random.randint(0, 2) # Label of the current time step

[0247] model_online.partial_fit(X_t, [y_t], classes=np.unique([y_t]))

[0248] # Cross-validation

[0249] X = np.random.rand(100, n_features)

[0250] y = np.random.randint(0, 2, 100)

[0251] clf = RandomForestClassifier()

[0252] scores = cross_val_score(clf, X, y, cv=5)

[0253] print("Cross-validation scores:", scores)

[0254] print("Mean cross-validation score:", np.mean(scores))

[0255] # Integrated Learning

[0256] clf1 = LogisticRegression()

[0257] clf2 = SVC(probability=True)

[0258] clf3 = RandomForestClassifier()

[0259] eclf = VotingClassifier(estimators=[('lr', clf1), ('svc', clf2), ('rf', clf3)], voting='soft')

[0260] eclf.fit(X, y)

[0261] print("Predictions:", eclf.predict(X))

[0262] V. Scenario Mode Adjustment Module

[0263] The scenario mode adjustment module includes a user preference prediction unit, a parameter adjustment unit, and a mode switching unit;

[0264] The user preference prediction unit collects users' historical preference data, combines it with real-time environmental data and the current scenario mode, and generates a user preference model through the Transformer architecture to predict users' preferences under different scenario modes. Historical preference data includes historical temperature settings, wind speed preferences, wind direction preferences, and the frequency of use of purification mode.

[0265] The parameter adjustment unit adjusts the operating parameters of the air conditioning system according to the predicted user preferences through a preset parameter mapping table and dynamic adjustment algorithm, matching the operating status of the current scenario mode. The operating parameters include temperature, wind speed, wind direction and purification intensity.

[0266] The mode switching unit is used to achieve smooth switching between scenario modes, avoiding user discomfort caused by sudden parameter changes. Through a progressive parameter adjustment algorithm, the operating parameters of the air conditioning system are gradually adjusted when switching scenario modes to achieve a continuous user experience.

[0267] It should be noted that the historical temperature setting refers to the temperature value set by the user in different scenario modes in the past. In sleep mode, the user may prefer a lower temperature (24°C), while in work mode, the user may prefer a slightly higher temperature (26°C).

[0268] Wind speed preference: The wind speed level set by the user in different scenario modes in the past. In leisure mode, the user may prefer low wind speed, while in sports mode, the user may prefer high wind speed.

[0269] Wind direction preference: The wind direction set by the user in different scenario modes in the past. In sleep mode, the user may prefer the wind to avoid the head, while in work mode, the user may prefer the wind to blow directly towards the work area.

[0270] Purification mode usage frequency: How often users use different purification modes in different scenarios in the past. In environments with poor air quality, users may use the HEPA filter mode more frequently; in environments with good air quality, users may use the negative ion generation mode more frequently.

[0271] Real-time environmental data: including current indoor and outdoor temperature, humidity, and air quality, which reflect the current environmental conditions.

[0272] The Transformer architecture is a deep learning model based on the attention mechanism, widely used in natural language processing and sequence data processing. In the context mode adjustment module, the Transformer architecture is used to generate user preference models and predict user preferences in different context modes.

[0273] The preset parameter mapping table is a predefined table used to map scenario modes to specific air conditioning operating parameters, including temperature, fan speed, airflow direction, and purification intensity. The preset parameter mapping table defines the operating parameters for different scenario modes:

[0274] Sleep mode: Temperature 24°C, low fan speed, airflow away from the head, purification mode generates negative ions.

[0275] Operating mode: Temperature 26°C, fan speed medium, airflow directly towards the work area, purification mode is HEPA filtration.

[0276] Leisure mode: Temperature 25°C, low wind speed, evenly distributed airflow, purification mode is activated carbon adsorption.

[0277] Exercise mode: Temperature 23°C, high wind speed, wind direction directly blows towards the exercise area, purification mode generates negative ions.

[0278] Party mode: Temperature 27°C, fan speed medium, airflow evenly distributed, purification mode HEPA filtration.

[0279] Energy-saving mode: Temperature 28°C, low wind speed, evenly distributed airflow, purification mode is activated carbon adsorption.

[0280] The dynamic adjustment algorithm adjusts the operating parameters of the air conditioning system in real time based on the output of the user preference prediction unit to ensure that the system's operating status is consistent with the user's preferences.

[0281] The progressive parameter adjustment algorithm is used to achieve smooth switching between scenario modes and avoid user discomfort caused by sudden parameter changes. When switching scenario modes, the system does not immediately adjust to the target parameters. The progressive parameter adjustment algorithm dynamically adjusts the parameters according to real-time data to ensure the smoothness of the switching process.

[0282] VI. User Interface

[0283] The user interface includes a voice control unit, a gesture recognition unit, a display unit, and a manual operation unit;

[0284] The voice control unit uses voice recognition technology to collect user voice commands through a built-in multi-microphone array. It uses a deep recurrent neural network (RNN) combined with an attention mechanism for voice recognition and semantic parsing, enabling users to adjust the operating parameters of the air conditioning system and switch scene modes via voice commands.

[0285] The gesture recognition unit captures the user's three-dimensional gestures through a camera, and combines them with a 3D convolutional neural network to perform gesture recognition and action analysis, thereby realizing non-contact interactive control. It also supports custom gesture settings, allowing users to define specific gestures to adjust the operating parameters of the air conditioning system and switch scene modes.

[0286] The display unit uses a touch screen to display the current air quality data, scene mode status, and air conditioning operating parameters in real time through a graphical interface. It also supports users to directly adjust the operating parameters of the air conditioning system and switch scene modes through touch operation.

[0287] The manual operation unit provides both physical buttons and a touchscreen interface. Through multi-touch technology and physical button feedback mechanism, users can manually adjust the operating parameters of the air conditioning system and switch scene modes.

[0288] It should be noted that the speech recognition technology uses a deep recurrent neural network (RNN) combined with an attention mechanism for speech recognition and semantic parsing, which can effectively process speech signals and improve the accuracy and response speed of recognition.

[0289] Multi-microphone array: The built-in multi-microphone array is used to collect user voice commands, improve the quality and stability of voice signals, and reduce background noise through beamforming technology to improve the accuracy of voice recognition.

[0290] Deep Recurrent Neural Networks (RNNs): RNNs can process time-series data and are suitable for processing speech signals. Combined with attention mechanisms, the model can dynamically focus on key parts of the speech signal, improving recognition performance.

[0291] Semantic parsing: Through semantic parsing technology, the system can understand the user's voice commands and convert them into corresponding control commands. The user can say "set the temperature to 24°C" or "switch to sleep mode", and the system can accurately parse and execute these commands.

[0292] 3D Convolutional Neural Networks: Used for gesture recognition and motion analysis, 3D convolutional neural networks can process three-dimensional data, capture the shape and motion trajectory of gestures, and improve the accuracy of recognition.

[0293] Custom gesture settings: Users can define specific gestures to achieve specific functions, including adjusting temperature and switching scene modes. The system supports user-defined gestures, increasing the flexibility of interaction.

[0294] Non-contact interactive control: By capturing hand gestures through a camera, users can control the device without touching it, improving convenience and hygiene.

[0295] Touchscreen display: It adopts a high-resolution touchscreen display, providing an intuitive graphical interface. Users can directly adjust the operating parameters of the air conditioning system and switch scene modes by touching the screen.

[0296] Physical buttons: Provides a physical button operation interface, allowing users to directly adjust the air conditioning system's operating parameters and switch scene modes via buttons. The physical buttons provide clear operation feedback, improving the reliability of use.

[0297] Multi-touch technology: Supports multi-touch operation, allowing users to perform various operations via the touchscreen, including swiping to adjust the temperature and tapping to switch scene modes.

[0298] Feedback mechanism: The physical buttons and touch screen interface are combined with a feedback mechanism, which allows users to receive clear feedback during operation and improves the user experience.

[0299] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. An intelligent air conditioning control and scene mode adjustment system, characterized in that, include: The integrated monitoring module uses multi-sensor fusion technology to monitor indoor and outdoor air quality parameters and user physiological data in real time. The strategy generation module analyzes and processes the collected data using an improved deep learning algorithm to predict changes in air quality and generate intelligent air conditioning control strategies. The control execution module adjusts the operating parameters of the air conditioning system in real time according to the generated control strategy through frequency conversion technology and intelligent air direction adjustment technology, so as to control air pollutants and reduce energy consumption. The scenario pattern recognition module uses support vector machine (SVM) to analyze air quality parameters and user physiological data to identify the current scenario. It also optimizes the recognition logic through improved machine learning algorithms to improve the accuracy of scenario pattern recognition. The scenario pattern recognition module includes: The data analysis unit performs real-time analysis on the collected air quality parameters and user physiological data. Through statistical analysis and feature extraction algorithms, it extracts feature vectors, which include temperature, humidity, PM2.5 concentration, heart rate, and respiratory rate. The deep learning classifier unit uses a support vector machine (SVM) to classify feature vectors, identify the current scenario mode, classify real-time data using a trained classification model, and output the current scenario mode, which includes sleep mode, work mode, leisure mode, sports mode, party mode, and energy-saving mode. The logic optimization unit optimizes the classifier's recognition logic using an improved machine learning algorithm, adjusting the classifier's parameters and thresholds to improve the accuracy of scene pattern recognition. Specifically: The improved machine learning algorithm combines reinforcement learning algorithms, Bayesian optimization methods, online learning algorithms, cross-validation mechanisms, and ensemble learning methods. The reinforcement learning algorithm dynamically adjusts the classifier parameters through reward and punishment mechanisms to maximize recognition accuracy. The Bayesian optimization method dynamically adjusts the classifier threshold through a Bayesian probability model, thereby improving the robustness of recognition. The online learning algorithm continuously updates the classifier's model parameters based on real-time feedback data to adapt to changes in the environment and user behavior. The cross-validation mechanism evaluates the classifier's performance through multiple rounds of validation, avoiding overfitting and underfitting. The ensemble learning method combines the outputs of multiple classifiers to improve recognition accuracy. The scenario mode adjustment module generates a user preference model using the Transformer architecture, predicts user preferences, and gradually adjusts the operating parameters of the air conditioning system using a progressive parameter adjustment algorithm to match the current scenario mode's operating state, achieving smooth switching between scenario modes. The scenario mode adjustment module includes: The user preference prediction unit collects users' historical preference data, combines it with real-time environmental data and the current scenario mode, and generates a user preference model through the Transformer architecture to predict users' preferences under different scenario modes. The historical preference data includes historical temperature settings, wind speed preferences, wind direction preferences, and the frequency of use of purification mode. The parameter adjustment unit adjusts the operating parameters of the air conditioning system according to the predicted user preferences through a preset parameter mapping table and dynamic adjustment algorithm, matching the operating status of the current scenario mode. The operating parameters include temperature, wind speed, wind direction and purification intensity. The mode switching unit is used to achieve smooth switching between scenario modes, avoiding user discomfort caused by sudden parameter changes. Through a progressive parameter adjustment algorithm, the operating parameters of the air conditioning system are gradually adjusted when switching scenario modes to achieve a continuous user experience. The user interface provides a way for users to interact with the system. It supports voice control and gesture recognition, and allows users to manually switch scene modes and adjust the operating parameters of the air conditioning system. It also displays the current air quality data and scene mode status in real time.

2. The intelligent air conditioning control and scene mode adjustment system according to claim 1, characterized in that, The integrated monitoring module includes: An air quality sensor, including a temperature sensor, a humidity sensor, a PM2.5 sensor, a VOCs sensor, and a carbon dioxide sensor, is used to collect indoor and outdoor air quality parameters, including temperature, humidity, PM2.5, VOCs, and carbon dioxide. Physiological sensors, including heart rate sensors and respiratory rate sensors, are used to collect physiological data from users, including heart rate and respiratory rate. The data fusion unit uses multi-sensor fusion technology to fuse sensor data and generate a comprehensive data signal. The multi-sensor fusion technology employs an adaptive fusion algorithm and adjusts the fusion strategy through an online learning algorithm.

3. The intelligent air conditioning control and scene mode adjustment system according to claim 1, characterized in that, The strategy generation module includes: The data preprocessing unit performs validity checks on the collected air quality parameters and user physiological data, removes invalid and duplicate values, and uses the Min-Max normalization method to normalize all data to the [0,1] interval, eliminating the differences between data of different dimensions. The deep learning model unit uses an improved deep learning algorithm to extract and analyze features from preprocessed data and predict air quality change trends. The improved deep learning algorithm is a combination of Long Short-Term Memory Network (LSTM) and Convolutional Neural Network (CNN). The strategy generation unit constructs a multi-objective optimization function based on the prediction results of the deep learning model. Combining the preset comfort and energy consumption targets, it uses a particle swarm optimization algorithm to optimize the operating parameters of the air conditioning system and generate an intelligent air conditioning control strategy. The intelligent air conditioning control strategy includes temperature setting, fan speed adjustment, fan direction adjustment, and purification mode switching.

4. The intelligent air conditioning control and scene mode adjustment system according to claim 3, characterized in that, The improved deep learning algorithm is a combination of a Long Short-Term Memory (LSTM) network and a Convolutional Neural Network (CNN), wherein: The improved deep learning algorithm includes convolutional layers, recurrent layers, feature fusion layers, fully connected layers, optimization algorithms, loss functions, and regularization units; The convolutional layer is used to extract spatial features of the data, treating air quality parameters and user physiological data as multi-channel inputs, and extracting local features through convolution operations. The recurrent layer uses a long short-term memory network, taking the features extracted by the convolutional layer as input to capture the changing trends of time series data; The feature fusion layer dynamically adjusts the weights of the output features of the convolutional layer and LSTM layer through an attention mechanism to achieve the fusion of spatial and temporal features; The fully connected layer processes the fused features to generate prediction results, which include air quality change trends and air conditioning control strategies. The optimization algorithm employs an improved Adam optimizer, combined with a learning rate decay strategy, to improve the training efficiency and convergence speed of the model. The loss function uses a weighted mean squared error loss function, which assigns different weights according to the importance and reliability of the data to improve the prediction accuracy of the model. The regularization unit introduces Dropout technology to prevent model overfitting and improve the model's generalization ability.

5. The intelligent air conditioning control and scene mode adjustment system according to claim 1, characterized in that, The control execution module includes: The variable frequency control unit adjusts the operating frequency of the air conditioner compressor in real time according to the generated control strategy. The PID controller adjusts the compressor frequency according to the deviation between the indoor temperature and the set temperature to achieve temperature control and energy consumption optimization. The intelligent airflow adjustment unit uses multi-dimensional airflow adjustment technology. Through multiple micro motors and sensors installed at the air conditioner outlet, it automatically adjusts the airflow direction of the air conditioner according to the indoor space layout and the user's position to avoid blowing directly on the human body and improve comfort. The air purification unit automatically switches air purification modes based on air quality parameters to remove pollutants from the air. The air purification modes include HEPA filtration, activated carbon adsorption, and negative ion generation.

6. The intelligent air conditioning control and scene mode adjustment system according to claim 1, characterized in that, The user interface includes: The voice control unit uses voice recognition technology to collect user voice commands through a built-in multi-microphone array. It uses a deep recurrent neural network (RNN) combined with an attention mechanism for voice recognition and semantic parsing, enabling users to adjust the operating parameters of the air conditioning system and switch scene modes via voice commands. The gesture recognition unit captures the user's three-dimensional gestures through a camera, and combines them with a 3D convolutional neural network to perform gesture recognition and action analysis, thereby realizing non-contact interactive control. It also supports custom gesture settings, allowing users to define specific gestures to adjust the operating parameters of the air conditioning system and switch scene modes. The display unit uses a touch screen to display the current air quality data, scene mode status, and air conditioning operating parameters in real time through a graphical interface. It also supports users to directly adjust the operating parameters of the air conditioning system and switch scene modes through touch operation. The manual operation unit provides both physical buttons and a touchscreen interface. Through multi-touch technology and physical button feedback mechanism, users can manually adjust the operating parameters of the air conditioning system and switch scene modes.

Images