Intelligent multi-dimensional air purification cloud intelligent control method and system

Abstract

The present application relates to a cloud-based intelligent control method and system for intelligent multi-dimensional air purification, which solves the problem that the operation of the purifier is adjusted solely based on a single air quality indicator, and cannot fully consider the complexity of the indoor environment, such as factors such as temperature and humidity, concentrations of other pollutants, and different sensitivities of users to air quality. The method includes: parsing the voice control command to extract key information, and parsing the extracted key information, real-time air quality data, and air quality trend information within a preset time range in the future, and inputting them into the cloud-based intelligent decision-making algorithm model, outputting the operating parameters that need to be adjusted for the air purifier, and forming a parameter adjustment decision instruction; sending the formed parameter adjustment decision instruction to the air purifier terminal through the network; the air purifier adjusts its own operating parameters in response to the parameter adjustment decision instruction. The present application has the following effects: improving air quality and meeting the user's needs for the indoor air environment.

Description

Technical Field

[0001] The present invention relates to the technical field of smart homes, and in particular to a cloud-based intelligent control method and system for intelligent multi-dimensional air purification. Background Art

[0002] As people's demands for a higher quality of life continue to rise, optimizing indoor air quality is gaining increasing attention. The SmartConnect multi-dimensional air purification system aims to create a more comfortable and healthy indoor environment for users, using advanced technologies to achieve intelligent control and environmental regulation of the air purifier. In the smart home sector, intelligent air purification technology has gradually become a hot topic in research and application, integrating multiple technologies to meet the diverse needs of users.

[0003] Currently, traditional air purification control methods primarily rely on simple sensor detection and preset operating modes. For example, some air purifiers can only adjust the operating wind speed based on a single detected air quality indicator (such as PM2.5 concentration). When the PM2.5 concentration exceeds the standard, the wind speed is automatically adjusted to a high setting, and then adjusted back to a low setting when the concentration decreases. Some smart air purifiers, while able to receive user voice commands, can only perform simple commands, such as turning the device on and off, switching to fixed modes, etc., and lack comprehensive consideration of complex environmental information and diverse user needs.

[0004] Regarding the above-mentioned related technologies, the inventors found the following defects: simply adjusting the operation of the purifier based on a single air quality indicator cannot fully consider the complexity of the indoor environment, such as temperature and humidity, the concentration of other pollutants, and the different sensitivities of users to air quality. Summary of the Invention

[0005] In order to improve air quality and meet users' needs for indoor air environment, this application provides an intelligent multi-dimensional air purification cloud-based intelligent control method and system.

[0006] In the first aspect, the present application provides a cloud-based intelligent control method for intelligent multi-dimensional air purification, which adopts the following technical solutions:

[0007] A cloud-based intelligent control method for intelligent multi-dimensional air purification, comprising:

[0008] Obtain voice control instructions for the air purifier from users in the room;

[0009] Based on the room's spatial layout information, historical air quality data, and real-time air quality data, a deep learning-based spatiotemporal data analysis algorithm is used to predict and obtain air quality trends within a preset time range in the future.

[0010] The voice control command is parsed to extract key information. The extracted key information, real-time air quality data, and air quality trend information within a preset time range are input into the cloud-based intelligent decision-making algorithm model. The operating parameters that need to be adjusted for the air purifier are output and parameter adjustment decision instructions are formed.

[0011] Send the parameter adjustment decision instruction formed to the air purifier terminal through the network;

[0012] The air purifier adjusts its own operating parameters in response to the parameter adjustment decision instruction.

[0013] By employing this technical solution, the Zhilian Multi-Dimensional Air Purification Cloud Intelligent Control method can understand user needs through voice control commands, predict air quality trends based on spatiotemporal data analysis algorithms, and accurately control the air purifier by outputting operating parameters through an intelligent decision-making algorithm model. This technology has achieved remarkable results, enabling intelligent, precise, and personalized air purification control, improving air quality and meeting user needs for indoor air quality.

[0014] Optionally, the method further includes steps after obtaining the voice control instructions of the user in the room regarding the air purifier, specifically as follows:

[0015] Analyze whether relevant physiological data related to user emotions is obtained;

[0016] If not, maintain the original setting;

[0017] If yes, then based on the obtained physiological data related to the user's emotions, a multimodal emotion fusion and feedback algorithm is used to analyze and obtain emotion-related data, and feature extraction is performed to obtain emotion feature data;

[0018] The pre-processed voice command key information, emotional characteristic data, real-time air quality data and future air quality trend information are input together into an advanced decision-making module based on the integration of fuzzy logic and expert system, and the operating parameters that need to be adjusted for the air purifier are output to form parameter adjustment decision instructions. The parameter adjustment decision instructions formed are sent to the air purifier terminal through the network, and the air purifier adjusts its own operating parameters in response to the parameter adjustment decision instructions.

[0019] By employing these technical solutions, the Zhilian Multi-Dimensional Air Purification Cloud Control method understands user needs through voice commands, predicts air quality trends based on spatiotemporal data analysis algorithms, and outputs operating parameters through intelligent decision-making algorithms to precisely control the air purifier. This technology delivers significant results, enabling intelligent, precise, and personalized air purification control, improving air quality and meeting user needs for an optimal indoor air environment.

[0020] Optionally, the advanced decision-making module based on the fusion of fuzzy logic and expert system is operated as follows:

[0021] The fuzzy logic system performs fuzzification on the input data;

[0022] Reasoning is performed based on a pre-set rule base in the expert system, which contains strategies for air purification and environmental regulation in various situations;

[0023] After fuzzy logic reasoning, the decision module will generate detailed device control instructions, which cover the adjustment of air purifier operating parameters and linkage control of other smart devices.

[0024] By employing the aforementioned technical solution, this operational method combines fuzzy logic with expert systems, first fuzzifying input data and then reasoning based on a rule base. This technology is highly effective, leveraging expert experience to comprehensively consider the diverse range of air purification and environmental conditioning scenarios, accurately generating device control instructions. This not only optimizes air purifier operating parameters but also enables intelligent device linkage, enhancing the precision and intelligence of environmental control, and creating a more comfortable environment for users.

[0025] Optionally, the method further includes steps after obtaining the voice control instructions of the user in the room regarding the air purifier, specifically as follows:

[0026] Analyze whether the acquired voice control commands come from multiple users;

[0027] If not, maintain the original setting;

[0028] If the answer is yes, each user is treated as an agent, and the voice commands are parsed to extract key information. This information, along with the room layout, air quality data for each corner, and air quality trends within a preset timeframe, is then used as input for the multi-agent reinforcement learning algorithm.

[0029] Each agent uses the decision-making model in the multi-agent reinforcement learning algorithm to select matching actions based on its own state information and the acquired environmental information. The scope of actions includes but is not limited to adjusting the operating parameters of the air purifier and controlling the switches or parameters of other smart devices.

[0030] By implementing these technical solutions, the system's adaptability to multi-user scenarios is enhanced. Voice commands are analyzed to maintain stable settings in single-user scenarios. In multi-user scenarios, each user is treated as an agent, and various environmental information is fed into a multi-agent reinforcement learning algorithm. Each agent makes its own decisions, precisely adjusting the air purifier and other equipment to meet the needs of different users, enhancing the personalized and intelligent nature of air purification and environmental conditioning.

[0031] Optionally, the following steps are included: each agent selects a matching action based on its own state information and the acquired environment information using a decision model in a multi-agent reinforcement learning algorithm.

[0032] Various sensors distributed indoors are used to collect real-time data on environmental changes, while user satisfaction feedback on the current environment is regularly collected through user terminal devices.

[0033] Summarize the collected environmental change data and user satisfaction feedback on the current environment to form environmental feedback information;

[0034] Based on the feedback from the environment and the pre-set reward and punishment rules, each agent is rewarded or punished.

[0035] The intelligent agent uses the learning mechanism of the reinforcement learning algorithm to update its own behavior strategy based on the reward or punishment feedback obtained.

[0036] By implementing the above technical solutions, a comprehensive feedback optimization mechanism has been established. Sensors and user terminals collect environmental changes and satisfaction feedback, generating environmental feedback information. Based on reward and punishment rules, agents are rewarded or punished, enabling them to update their behavioral strategies. This allows the system to dynamically adjust based on the actual environment and user experience, continuously optimizing air purification and environmental conditioning, improving user satisfaction, and making intelligent decisions more tailored to actual needs.

[0037] Optionally, the agent uses the learning mechanism of the reinforcement learning algorithm to update its own behavior strategy based on the reward or penalty feedback it receives, including:

[0038] Analyze whether relevant physiological data related to user emotions is obtained;

[0039] If not, the agent uses the learning mechanism of the reinforcement learning algorithm to update its own behavior strategy based on the reward or punishment feedback obtained;

[0040] If yes, then based on the obtained physiological data related to the user's emotions, a multimodal emotion fusion and feedback algorithm is used to analyze and obtain emotion-related data, and feature extraction is performed to obtain emotion feature data;

[0041] We assign weights based on the contribution of different modal data to emotional judgment, fuse the weighted multimodal data, and input it into a deep learning-based sentiment analysis model to determine each user's emotional state.

[0042] Count the number and proportion of agents that want to reduce noise and increase wind speed to deal with pollution, and analyze the preliminary strategies associated with agents whose proportion exceeds the preset threshold;

[0043] Combine the sentiment analysis results with the preliminary strategy derived from the agent proportion analysis.

[0044] By implementing these technical solutions, the agent's behavior strategy update logic is enriched. By analyzing the presence of user physiological data, the multimodal emotion fusion algorithm is determined. By extracting emotional features from physiological data, combining them with deep learning models to determine emotions, and statistically analyzing the agent's strategy, the agent's behavior strategy is then integrated with the emotion and strategy. This allows the agent's behavior strategy to more comprehensively consider user emotions and actual needs, achieving more humane and precise environmental control and improving the user experience.

[0045] Optionally, the process also includes the following steps after integrating the sentiment analysis results with the preliminary strategy derived from the agent proportion analysis:

[0046] Input the current indoor environmental parameters, the existing status of the smart devices, and the control instructions in the fusion strategy to simulate the environmental changes after the strategy is implemented;

[0047] Based on the simulation results and the preset optimization goals, the fusion strategy is further optimized by referring to the success and failure experiences of the intelligent agents in similar historical scenarios.

[0048] Convert the optimized strategy into specific device control instructions.

[0049] By implementing the above technical solutions, this step improves the policy implementation process. Simulating environmental changes after policy implementation allows for early prediction of effectiveness. Optimizing and integrating policies based on pre-set goals and historical experience makes them more scientific and rational. Converting optimized policies into control instructions enables precise control of equipment. This improves the overall effectiveness and accuracy of air purification and environmental conditioning, ensuring that system decisions better meet actual environmental needs and user expectations.

[0050] In a second aspect, the present application provides an intelligent multi-dimensional air purification cloud intelligent control system, which adopts the following technical solutions:

[0051] A smart multi-dimensional air purification cloud intelligent control system includes a memory, a processor, and a program stored in the memory and executable on the processor. When the program is loaded and executed by the processor, the smart multi-dimensional air purification cloud intelligent control method as described in the first aspect is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a flow chart of an intelligent multi-dimensional air purification cloud-based intelligent control method according to an embodiment of the present application.

[0053] Figure 2 This is a flowchart of another embodiment of the present application, which shows the steps after obtaining the voice control instructions of the user in the room regarding the air purifier.

[0054] Figure 3 It is a flowchart of an operation method of an advanced decision-making module based on the fusion of fuzzy logic and expert system according to another embodiment of the present application.

[0055] Figure 4 This is a flowchart of another embodiment of the present application, which shows the steps after obtaining the voice control instructions of the user in the room regarding the air purifier. DETAILED DESCRIPTION

[0056] The present application is further described in detail below with reference to the accompanying drawings.

[0057] Reference Figure 1 , is a cloud-based intelligent control method for intelligent multi-dimensional air purification disclosed in this application, comprising:

[0058] Step S100: Obtain voice control instructions for the air purifier from the user in the room.

[0059] Voice control instructions refer to the operating instructions conveyed by the user to the air purifier in the form of voice, which are intended to enable the air purifier to perform specific tasks to meet the user's needs for air quality and related environmental adjustments.

[0060] Acquisition method: Voice collection devices are placed in appropriate locations in the room, such as the microphone array of a smart speaker or the sound pickup device equipped in the air purifier. When the user issues a voice command, these devices convert the sound signal into an electrical signal, which is then converted into a digital signal through analog-to-digital conversion and transmitted to the system for processing.

[0061] For example: When a user says "adjust the air purifier's wind speed to the maximum" or "turn on the air purifier's sterilization mode", the system obtains such voice commands for subsequent operations.

[0062] In step S200, based on the spatial layout information of the room, historical air quality data, and real-time air quality data, a spatiotemporal data analysis algorithm based on deep learning is used to predict and obtain air quality trend information within a preset time range in the future.

[0063] Room space layout information: describes the room's structure, dimensions, furniture placement, and the location of vents, doors, and windows, which affects the flow of air in the room and the diffusion of pollutants.

[0064] Historical air quality data: Recorded data of various air quality indicators in the room (such as PM2.5, formaldehyde, TVOC concentration, etc.) over the past period of time, reflecting the changing patterns of air quality.

[0065] Real-time air quality data: The air quality index data in the room monitored in real time by the air quality sensor at the current moment.

[0066] Deep learning-based spatiotemporal data analysis algorithm: Utilizes deep learning models, combined with data features in both time and space dimensions, to explore trends in air quality changes over time and its spatial distribution patterns, thereby predicting future air quality.

[0067] The specific process is as follows:

[0068] First, collect the room space layout information, which can be obtained through architectural drawings or measured and modeled using technologies such as laser scanning.

[0069] Historical air quality data is collected from long-term monitoring records and stored in a database. Real-time air quality data is collected in real time by sensors installed at different locations in the room.

[0070] This data is fed into a trained deep learning-based spatiotemporal data analysis algorithm. The model learns the spatiotemporal characteristics of the data, such as how air quality varies across seasons and time periods, and the impact of spatial location on air quality. Using these learned patterns, it predicts air quality trends within a preset timeframe (e.g., the next hour, three hours, etc.), such as whether PM2.5 concentrations will rise, fall, or remain stable.

[0071] In step S300, the voice control command is parsed to extract key information, and the extracted key information, real-time air quality data, and air quality trend information within a preset time range in the future are parsed and input into the cloud-based intelligent decision-making algorithm model, and the operating parameters that need to be adjusted for the air purifier are output to form parameter adjustment decision instructions.

[0072] Voice control command analysis: The voice commands input by the user are converted into text form that can be understood by the computer through natural language processing technology, and the core operation information and target parameters are extracted.

[0073] Key information: Specific operating requirements of the air purifier extracted from voice commands, such as adjusting wind speed, switching modes, setting operating time, and other key content.

[0074] Real-time air quality data: Current air quality index values, such as formaldehyde concentration and PM10 value, are collected and transmitted in real time by various air quality sensors in the room.

[0075] Air quality trends within a preset future timeframe: This forecast uses a deep learning-based spatiotemporal data analysis algorithm to predict air quality trends and indicator estimates for a period of time in the future (e.g., 1-2 hours).

[0076] Intelligent decision-making algorithm model: Based on large amounts of data training, this algorithm model integrates multiple factors for intelligent analysis and decision-making, and can output the optimal operating parameters of the air purifier based on the input information.

[0077] The specific process is as follows:

[0078] With the help of speech recognition and natural language processing tools, voice control commands are parsed. For example, the key information "increase the wind speed of the air purifier" can be extracted from "increase the wind speed".

[0079] Real-time air quality data and future air quality trend information are obtained through sensors and prediction algorithms and transmitted to the cloud.

[0080] These key information, real-time air quality data, and information on future air quality trends are input into the trained intelligent decision-making algorithm model in the cloud. The model comprehensively considers these factors and, after complex calculations, outputs the operating parameters that need to be adjusted for the air purifier, such as fan speed and filter working intensity, to form parameter adjustment decision instructions for precise control of the air purifier.

[0081] Step S400: Send the generated parameter adjustment decision instruction to the air purifier terminal via the network.

[0082] Parameter adjustment decision instructions: Generated by the cloud-based intelligent decision-making algorithm model, it contains information on various operating parameters that need to be adjusted for the air purifier, such as adjusting the wind speed, changing the purification mode, and other specific instructions, which are used to accurately control the operating status of the air purifier.

[0083] Network: covers a variety of network communication methods, such as wireless communication networks such as Wi-Fi, Bluetooth, ZigBee, and wired networks such as Ethernet, responsible for data transmission between the cloud and the air purifier terminal.

[0084] Air purifier terminal: the air purification device actually used by the user, which receives instructions from the cloud and performs corresponding operations to adjust the indoor air quality.

[0085] The specific process is as follows:

[0086] After the cloud completes the generation of the parameter adjustment decision instruction, the instruction is encapsulated in a specific data format, which contains key information such as device identification, instruction type, parameter value, etc., to ensure accurate transmission to the target air purifier.

[0087] The encapsulated instructions are sent over a pre-configured network connection using network protocols such as TCP / IP. In a home environment, this is typically done over a home Wi-Fi network; in commercial locations, this may be done over an enterprise-grade wired or wireless network.

[0088] The air purifier terminal has a built-in corresponding network receiving module, which continuously monitors the network signal. Once it receives an instruction that matches its own device identification, it immediately parses the instruction, extracts the adjustment parameters, and prepares for the next step of executing the instruction.

[0089] In step S500 , the air purifier adjusts its own operating parameters in response to the parameter adjustment decision instruction.

[0090] Parameter adjustment decision instructions: Instructions sent from the cloud carry information on adjusting the operating parameters of the air purifier, including specific parameters such as wind speed, air volume, operating mode, and filter cleaning frequency.

[0091] Operating parameters: various indicators that determine the working status of the air purifier. The adjustment of these parameters directly affects the air purification effect and energy consumption.

[0092] The specific process is as follows:

[0093] The control chip or microprocessor inside the air purifier is responsible for receiving and processing instructions. When receiving a parameter adjustment decision instruction, it first verifies the instruction to ensure that the instruction is complete and error-free.

[0094] After verification, the control chip interprets the command content to determine the operating parameters that need to be adjusted. For example, if the command requires adjusting the wind speed from level 3 to level 5, the control chip will identify the two key pieces of information: "wind speed" and "level 5."

[0095] Based on the analysis results, the control chip sends control signals to the corresponding actuators. For wind speed adjustment, the fan motor speed is changed; for switching operating modes, the control circuit switches to the corresponding operating circuit module, thereby adjusting the air purifier's operating parameters to meet indoor air quality control requirements.

[0096] Reference Figure 2 A cloud-based intelligent control method for intelligent multi-dimensional air purification further includes the following steps after obtaining a voice control instruction of an air purifier from a user in the room:

[0097] Step SA00: Analyze whether relevant physiological data that reflects the user's emotions is obtained. If not, execute step SB00; if yes, execute step SC00.

[0098] Physiological data that reflects user emotions: This refers to physiological indicators that can reflect the user's current emotional state, such as heart rate, blood pressure, galvanic skin response, facial expressions, and muscle movements. Changes in this data often correlate with emotional fluctuations and can be used as a basis for judging user emotions.

[0099] The analysis method is as follows: the system queries connected physiological data collection devices or related data storage areas to confirm whether there is any user physiological data that meets the requirements. For example, it checks whether there are connected smart bracelets (capable of collecting heart rate, galvanic skin response, and other data) or facial expression recognition cameras, and whether these devices are transmitting data to the system normally. If data is transmitted and meets the required format, it is determined that the relevant physiological data reflecting the user's emotions has been obtained; otherwise, it is determined that no data has been obtained.

[0100] Step SB00, maintain the original settings.

[0101] When the system determines that no relevant physiological data to reflect the user's emotions has been obtained, the operating parameters and control strategies of the current air purifier and related smart devices will not be changed, and the system will continue to operate according to the existing settings.

[0102] Step SC00: Based on the acquired physiological data related to the user's emotions, a multimodal emotion fusion and feedback algorithm is used to analyze and acquire emotion-related data, and perform feature extraction to acquire emotion feature data.

[0103] Multimodal Emotion Fusion and Feedback Algorithm: This algorithm comprehensively processes multiple different types of physiological data to more accurately analyze user emotions. These data modalities include but are not limited to heart rate, blood pressure, and facial expressions. By fusing these different modal data, it mines the emotional information contained within and provides corresponding feedback.

[0104] Emotion-related data: Data directly related to emotional expression, parsed from the user's physiological data, such as the amplitude of heart rate changes, muscle movement characteristics of facial expressions, etc. This data can provide a basis for judging emotions.

[0105] Emotional feature data: This data is obtained after feature extraction of emotion-related data. It is a further refinement and abstraction of emotion-related data, is more representative, and is helpful for subsequent emotion analysis and decision-making.

[0106] The specific process is as follows:

[0107] First, the collected multimodal physiological data such as heart rate, blood pressure, and facial expressions are input into the multimodal emotion fusion and feedback algorithm.

[0108] The algorithm preprocesses the data of each modality, such as filtering the heart rate data to remove noise interference; and normalizing the facial expression images to unify the image size.

[0109] Next, the algorithm analyzes and extracts information closely related to emotions from the preprocessed data, such as analyzing the degree of emotional excitement from changes in heart rate, and identifying basic emotion types such as joy, anger, sorrow, and happiness from facial expressions, to obtain emotion-related data.

[0110] Finally, feature extraction techniques, such as principal component analysis (PCA) and linear discriminant analysis (LDA), are used to extract the features that best represent the emotional state from the emotion-related data to form emotion feature data, preparing for subsequent more accurate emotion analysis.

[0111] In step SD00, the pre-processed voice command key information, emotional characteristic data, real-time air quality data, and future air quality trend information are input together into an advanced decision-making module based on the fusion of fuzzy logic and expert system, and the operating parameters that need to be adjusted for the air purifier are output to form parameter adjustment decision instructions. The parameter adjustment decision instructions formed are sent to the air purifier terminal through the network, and the air purifier adjusts its own operating parameters in response to the parameter adjustment decision instructions.

[0112] Key information of pre-processed voice commands: The core operation content, such as adjusting the wind speed and switching modes, is extracted after voice recognition and natural language processing of user voice commands, removing redundant and interfering information.

[0113] Emotional feature data: Feature data extracted from the user's multimodal physiological data that can reflect the user's emotional state and provide a reference for the emotional dimension for decision-making.

[0114] Real-time air quality data: Current indoor air quality index values, such as PM2.5 and formaldehyde concentration, are obtained through real-time monitoring by air quality sensors.

[0115] Future air quality trend information: Use spatiotemporal data analysis algorithms to predict the changing trends and estimated indicators of indoor air quality over a period of time in the future.

[0116] Advanced decision-making module based on the integration of fuzzy logic and expert system: a module that combines the ability of fuzzy logic to process uncertain information with the rich domain knowledge of expert system to make intelligent decisions by integrating multi-source information.

[0117] The specific process is as follows:

[0118] The pre-processed voice command key information, emotional characteristic data, real-time air quality data, and future air quality trend information are organized in a specific format and input into the advanced decision-making module.

[0119] The fuzzy logic system first fuzzifies these input data, converting precise numerical values into fuzzy concepts, such as quantifying fuzzy expressions such as "high concentration of formaldehyde" and "low wind speed requirement".

[0120] The expert system uses a pre-defined rule base and fuzzified information to perform reasoning. The rule base stores air purification and environmental conditioning strategies under various environmental conditions and user needs.

[0121] After fuzzy logic reasoning and expert system rule matching, the decision module outputs the operating parameters that need to be adjusted for the air purifier, such as fan speed, filter replacement prompts, etc., forming parameter adjustment decision instructions.

[0122] The decision instructions are transmitted to the air purifier terminal via Wi-Fi, Bluetooth and other networks.

[0123] The air purifier terminal receives and analyzes the instructions, and adjusts its own operating parameters according to the instructions, such as changing the wind speed and switching the purification mode, to achieve precise control of indoor air quality.

[0124] Reference Figure 3 ,The operation method of the advanced decision making module based on the fusion of fuzzy logic and expert system is as follows:

[0125] In step SD10, the fuzzy logic system performs fuzzification processing on the input data.

[0126] Fuzzy logic system: A mathematical tool that simulates the way human thinking processes uncertainty and fuzzy information. It can convert precise input data into fuzzy concepts that conform to human language habits, thereby solving problems that are difficult to accurately model in complex systems.

[0127] Fuzzy processing: converting clear and precise values or information into elements of fuzzy sets, described by linguistic variables and membership degrees, so that computers can process data in a way similar to human cognition.

[0128] Specific process:

[0129] When the pre-processed voice command key information, emotional characteristic data, real-time air quality data and future air quality trend information are input into the fuzzy logic system, the system will process it according to the pre-defined fuzzy sets and membership functions.

[0130] For example, consider PM2.5 concentrations in real-time air quality data, assuming they are divided into three fuzzy sets: "low," "medium," and "high." For example, if the current PM2.5 concentration is 30 μg / m³, the system calculates its membership in the "low" set to 0.8, its membership in the "medium" set to 0.2, and its membership in the "high" set to 0. This fuzzifies the precise value of 30 μg / m³ into the linguistic variables "low (membership 0.8), medium (membership 0.2)."

[0131] For emotional feature data, such as the "pleasure" value obtained based on facial expression recognition, it will also be mapped to fuzzy sets such as "very happy", "pleasant", "average", "unpleasant", and "very unpleasant" in a similar way, and the corresponding membership will be determined.

[0132] After such fuzzification processing, the originally precise data is converted into fuzzy information that is easy to understand and process, laying the foundation for subsequent reasoning based on fuzzy logic.

[0133] Step SD20 , reasoning is performed based on a rule base preset in the expert system, which contains strategies for air purification and environmental regulation in various situations.

[0134] Expert system: A knowledge-based intelligent system that stores the expertise and experience of domain experts in a specific form and simulates the expert's thinking process to solve problems.

[0135] Rule base: The core component of the expert system, a collection of "if-then" rules. These rules are based on professional knowledge and practical experience in the field of air purification and cover strategies for air purification and environmental regulation in various situations.

[0136] Reasoning: The expert system searches for matching rules in the rule base based on the input fuzzy data and draws conclusions according to a certain reasoning mechanism.

[0137] The specific process is as follows:

[0138] When the fuzzy logic system completes the fuzzification of the input data, the fuzzy information is passed to the expert system.

[0139] The expert system uses the fuzzified key information of the voice command, emotional signature data, real-time air quality data, and information on future air quality trends as its basis, and searches the rule base for matching rules one by one. For example, consider a rule such as "IF (real-time PM2.5 concentration is high) AND (user is irritated) THEN (increase the air purifier's speed and enable the deodorization function)."

[0140] The system determines whether the current input data meets the prerequisites for the rule. If the real-time PM2.5 concentration is indeed "high" after fuzzy processing, and the user's emotion is "irritated" through emotional feature data analysis, then the rule is triggered.

[0141] According to the preset reasoning mechanism, which may be based on forward reasoning (inferring conclusions from conditions), reverse reasoning (finding conditions from goals) or mixed reasoning, the system draws corresponding conclusions based on the triggered rules, that is, the air purifier needs to increase the wind speed and turn on the deodorization function, providing a basis for the subsequent generation of detailed equipment control instructions.

[0142] In step SD30, after fuzzy logic reasoning, the decision module generates detailed device control instructions, which cover the adjustment of operating parameters of the air purifier and the linkage control of other smart devices.

[0143] Fuzzy logic reasoning: The process of calculating and inferring fuzzified input data based on fuzzy logic rules to reach reasonable conclusions. In this scenario, a fuzzy logic system and an expert system rule base are used to comprehensively analyze various types of information related to air purification.

[0144] Decision-making module: This is a functional unit that integrates fuzzy logic reasoning results and generates specific operational instructions based on the overall system objectives. It is the core decision-making component of the entire intelligent multi-dimensional air purification system.

[0145] Device control instructions: A set of commands that clearly instructs a device to perform specific operations, detailing the parameters that the device needs to adjust or the actions to be performed.

[0146] Operating parameter adjustment: For air purifiers, change the parameter settings such as wind speed, air volume, working mode, filter replacement cycle, etc. that determine its operating status and purification effect.

[0147] Smart device linkage control: enables the air purifier to work together with other related smart devices (such as smart windows, fresh air fans, air conditioners, etc.), uniformly coordinate equipment operation according to the environment and user needs, and achieve more optimized indoor environment regulation.

[0148] The specific process is as follows:

[0149] After the fuzzy logic reasoning is complete, the conclusion is passed to the decision module. For example, if the reasoning results indicate that the current air quality is poor and the user requires a higher wind speed, the decision module will integrate this information, taking into account the indoor temperature and humidity.

[0150] For air purifiers, the decision module generates specific operating parameter adjustment instructions. For example, it adjusts the wind speed from the current medium to high range, or changes the filter's working intensity to improve purification efficiency.

[0151] Considering collaboration with other smart devices, the decision module might generate instructions to activate a ventilation fan to increase fresh air intake if indoor air quality is poor and carbon dioxide concentrations are high. Alternatively, if outdoor air quality is good, it might activate smart windows to facilitate natural ventilation. These instructions are aggregated into detailed device control instructions, enabling precise control of each device to optimize the indoor environment and meet user needs.

[0152] Reference Figure 4A cloud-based intelligent control method for intelligent multi-dimensional air purification further includes the following steps after obtaining a voice control instruction of an air purifier from a user in the room:

[0153] In step Sa00, it is analyzed whether the acquired voice control command comes from multiple users. If not, step Sb00 is executed; if yes, step Sc00 is executed.

[0154] Voice control commands: Users give voice commands to the air purification system to control the operation of the air purifier, such as adjusting the wind speed, turning on and off devices, etc.

[0155] Multiple users: Two or more users who issue control commands to the air purifier in the same space.

[0156] The analysis method is as follows: The system uses voice recognition technology and related identity recognition functions to make judgments. For example, through voice and voiceprint recognition technology, the received voice command is compared with the voiceprint characteristics of multiple users registered in the system. If multiple different voiceprint characteristics are matched, it indicates that the voice control command originated from multiple users. If only one voiceprint characteristic is matched or no registered voiceprint is matched (it can be considered a single unknown user), the voice control command is determined not to have originated from multiple users.

[0157] Step Sb00, maintain the original settings.

[0158] When the system determines that the voice control command obtained does not come from multiple users, the air purifier and related smart devices maintain the current operating status and parameter settings without making additional adjustments.

[0159] In step Sc00, each user is regarded as an intelligent agent, and the voice commands of each user are parsed to extract key information. The parsed key information is then used together with the room space layout, air quality data in each corner, and air quality trend information within a preset time range in the future as the input of the multi-agent reinforcement learning algorithm.

[0160] Agents: In a multi-agent system, agents can be considered individuals with the ability to make autonomous decisions, perceive the environment, and take actions to achieve their goals. Each user is considered an agent because they have different needs and decisions regarding the air environment.

[0161] Voice command analysis: Using speech recognition and natural language processing technologies, the user's voice is converted into text that the computer can understand, and key operational intentions are extracted, such as adjusting the wind speed, changing the purification mode, etc.

[0162] Key information: The core operational requirements extracted from the voice command are used to clarify the user's specific expectations for the air purifier or other smart device.

[0163] Multi-agent reinforcement learning algorithm: A machine learning algorithm in which multiple agents interact in a shared environment and optimize their decision-making strategies to achieve common or individual goals by continuously trying and learning from environmental feedback.

[0164] The specific process is as follows:

[0165] After determining that the voice control instructions come from multiple users, the system abstracts each user into an intelligent agent.

[0166] The system analyzes each user's voice commands. For example, if a user says "increase the speed of the air purifier and turn on the humidifier", the system extracts key information such as "increase the speed of the air purifier" and "turn on the humidifier" through voice recognition and natural language processing technology.

[0167] At the same time, the system obtains room space layout information, such as the shape, size, door and window positions of the room, which affects air flow and purification effect.

[0168] Collect air quality data from every corner to fully understand the distribution of indoor air quality.

[0169] Combined with the air quality trend information within a preset future time range predicted by the spatiotemporal data analysis algorithm.

[0170] The key information analyzed above, the room space layout, air quality data in each corner, and information on future air quality trends are integrated as input to the multi-agent reinforcement learning algorithm to provide comprehensive information for subsequent agent decision-making.

[0171] In step Sd00, each intelligent agent uses the decision-making model in the multi-agent reinforcement learning algorithm to select matching actions based on its own state information and the acquired environmental information. The scope of actions includes but is not limited to adjusting the operating parameters of the air purifier and controlling the switches or parameters of other intelligent devices.

[0172] Agent state information: refers to the relevant data representing the specific needs and preferences of each agent (i.e., each user), such as the user's sensitivity to air quality and the customary air purifier operation mode. This information determines the agent's decision-making tendency in a specific environment.

[0173] Environmental information: covers external environmental conditions such as the room layout, air quality data in each corner, and the trend of air quality within a preset time range in the future. The intelligent agent uses this information to judge the current environment and make decisions.

[0174] Decision-making model in multi-agent reinforcement learning algorithm: A model based on the multi-agent reinforcement learning framework, which can calculate the expected reward of each possible action based on the agent's state information and environment information by learning from past experience, thereby guiding the agent to make the best decision.

[0175] Action: The actions taken by the agent to optimize the environment to meet its own needs. In this scenario, this mainly involves controlling the air purifier and other smart devices.

[0176] The specific process is as follows:

[0177] Each agent first collects its own status information. For example, a user is accustomed to setting the air purifier to low-speed silent mode when the air quality reaches a certain standard. This preference is part of the agent's status information.

[0178] At the same time, the intelligent agent obtains shared environmental information, such as the high PM2.5 concentration in a corner of the room, and predicts that the air quality will deteriorate in the future.

[0179] The agent then inputs its state and environment information into the decision-making model within the multi-agent reinforcement learning algorithm. Based on the algorithm's learning mechanism, the model calculates the expected reward for each possible action in the current situation based on the rewards obtained from taking different actions in similar situations in the past (for example, if increasing the air purifier's speed previously resulted in increased user satisfaction, this would be a positive reward).

[0180] Based on the expected reward, the agent chooses the action with the highest expected reward. For example, an agent might choose to increase the speed of an air purifier, close a smart window to prevent outdoor pollutants from entering, or control a fresh air blower to increase the amount of fresh air introduced, thereby improving the environment and meeting its own needs.

[0181] A cloud-based intelligent control method for intelligent multi-dimensional air purification also includes the following steps: each intelligent agent uses a decision model in a multi-agent reinforcement learning algorithm to select a matching action based on its own state information and acquired environmental information.

[0182] In step Se00, various sensors distributed indoors are used to collect environmental change data in real time, and user satisfaction feedback on the current environment is collected regularly through user terminal devices.

[0183] Indoor sensors: These sensors are installed at various locations indoors to monitor environmental parameters. For example, air quality sensors monitor air pollutant concentrations, temperature and humidity sensors detect indoor temperature and humidity, and light sensors measure indoor light intensity. These sensors can sense environmental changes in real time and transmit data to the system.

[0184] Environmental change data: This data, collected by various sensors, reflects real-time changes in the indoor environment, such as fluctuations in air quality indicators, changes in temperature and humidity, and changes in light intensity. This data helps the system understand environmental dynamics and provides a basis for subsequent decision-making.

[0185] User terminal devices: Electronic devices used by users in daily life, such as smartphones and tablets, which interact with the air purification system by installing specific applications.

[0186] User satisfaction feedback on the current environment: Users provide subjective evaluations of the current indoor environment (including air quality, comfort, etc.) through terminal devices. Feedback can be provided by scoring (e.g., 1-5 points), text description, or selecting preset options (e.g., satisfied, average, dissatisfied).

[0187] The specific process is as follows:

[0188] Sensors at various locations indoors continuously operate, capturing real-time environmental information. For example, air quality sensors detect and record concentrations of pollutants such as PM2.5 and formaldehyde at regular intervals (e.g., every minute). Temperature and humidity sensors simultaneously monitor and transmit temperature and humidity values. This data is transmitted in real time to the system's data collection module via wired or wireless means.

[0189] The system also regularly pushes satisfaction surveys to users through an app on their devices. For example, every once in a while (e.g., daily), a simple questionnaire pops up when the user opens the app, asking about their satisfaction with current indoor air quality, temperature, humidity, and other environmental factors. Users respond based on their experience and submit their responses. The system collects this feedback and integrates it with environmental change data collected by sensors to provide comprehensive information for subsequent analysis.

[0190] In step Sf00 , the collected environmental change data and user satisfaction feedback on the current environment are aggregated to form environmental feedback information.

[0191] Environmental change data: Information about the indoor environmental status collected in real time by various indoor sensors, covering dynamic data on air quality (such as PM2.5 and formaldehyde concentration), temperature and humidity, light intensity, etc.

[0192] User satisfaction feedback on the current environment: The user's subjective feelings about the overall condition of the current indoor environment expressed through the terminal device, presented in the form of ratings, text descriptions, or option selections.

[0193] Environmental feedback information: A comprehensive information collection formed by integrating environmental change data and user satisfaction feedback, used to reflect the actual state of the indoor environment and the user's acceptance of this state, providing a key basis for the system's subsequent decision-making.

[0194] The specific process is as follows:

[0195] The system first organizes the environmental change data collected by various sensors. For example, it arranges air quality sensor data, temperature and humidity sensor data, and light sensor data from different time points and locations in an orderly manner according to time series and spatial location to form an environmental change data set.

[0196] At the same time, satisfaction feedback from user terminals is collected. If there are multiple user feedbacks, these feedbacks are aggregated. For example, user A's rating, user B's text description, and user C's selected satisfaction options are all collected.

[0197] The collected environmental change data is then integrated with user satisfaction feedback. Using time as a cue, environmental change data within the same time period is linked to corresponding user satisfaction feedback. For example, environmental data such as air quality, temperature, and humidity at a specific moment are combined with user satisfaction feedback submitted around that time. This ultimately creates environmental feedback information that comprehensively reflects the actual indoor environment and user experience, providing rich data support for subsequent reward and punishment decisions and strategy optimization for the agent.

[0198] Step Sg00: Based on the environmental feedback information and the pre-set reward and punishment rules, each intelligent agent is rewarded or punished.

[0199] Environmental feedback information: Integrates indoor environmental change data (such as air quality, temperature and humidity, etc.) and comprehensive information on user feedback on current environmental satisfaction, fully reflecting the actual environmental conditions and user experience.

[0200] Reward and Penalty Rules: A set of pre-defined criteria used to judge the impact of the agent's actions on the environment and user satisfaction. Actions that meet user expectations and improve the environment are rewarded, while actions that do not are penalized, guiding the agent to optimize its decisions.

[0201] Agent: Here, each user is regarded as an agent, and each agent affects the operation of the air purification and environmental conditioning system through its own decision-making.

[0202] The specific process is as follows:

[0203] After receiving environmental feedback, the system compares each piece of data against the conditions in the reward and penalty rules. For example, if the rule stipulates that rewards are given when air quality improves and the user satisfaction score reaches 4 points or above, the system will check whether the air quality data in the environmental feedback shows an upward trend and also check the user satisfaction score.

[0204] If air quality data shows a decrease in PM2.5 concentrations and harmful gas levels, and a user satisfaction score of 4 or higher, the corresponding agent will receive a reward. This reward may take the form of increasing the agent's weight in system decision-making or providing the user of the agent with some virtual reward, such as points.

[0205] Conversely, if air quality deteriorates and user satisfaction scores are low, such as below 3, the agent will be penalized according to the rules. This penalty may take the form of reducing its influence in system decision-making or reducing its access to certain functions. This reward-penalty mechanism incentivizes agents to continuously adjust their decisions to achieve optimal environmental regulation and user satisfaction.

[0206] In step Sh00, the intelligent agent uses the learning mechanism of the reinforcement learning algorithm to update its own behavior strategy based on the reward or punishment feedback obtained.

[0207] Reinforcement learning: A machine learning algorithm that uses an agent to interact with an environment and learn optimal behavior strategies from reward or penalty signals provided by the environment. The agent takes actions within the environment, and the environment rewards or penalizes the agent based on the results of the actions. The agent's goal is to maximize the long-term cumulative reward through repeated attempts.

[0208] Learning mechanism: The method and rules used in reinforcement learning algorithms to update the agent's behavior strategy. It adjusts the probability or method of the agent choosing different actions in different states based on the reward or penalty feedback received by the agent.

[0209] Behavioral strategy: The rules or methods that an agent uses to choose actions in a specific state, which determines how the agent makes decisions when faced with various environmental situations.

[0210] The specific process is as follows:

[0211] When the agent receives reward or penalty feedback, the reinforcement learning algorithm's learning mechanism kicks in. For example, if the agent is rewarded for increasing the air purifier's speed, which improves air quality and generates higher user satisfaction, the learning mechanism will reinforce the agent's tendency to increase the speed in similar environmental conditions.

[0212] This is typically achieved by adjusting the parameters of the behavioral policy. For example, in a policy gradient-based reinforcement learning algorithm, the algorithm calculates the policy gradient based on the reward signal and updates the policy parameters in a direction that maximizes future rewards. This makes the agent more likely to choose rewarding actions when encountering similar environmental states in the future.

[0213] On the contrary, if the agent's actions lead to environmental degradation or reduced user satisfaction and are punished, the learning mechanism will reduce the probability of the agent taking such actions in similar situations, prompting the agent to explore other actions that may improve the environment and enhance user satisfaction, and gradually optimize its own behavioral strategy to better adapt to the environment and meet user needs.

[0214] The agent uses the learning mechanism of the reinforcement learning algorithm to update its own behavior strategy based on the reward or penalty feedback it receives, including:

[0215] Step Sh10: analyzing whether relevant physiological data that feeds back the user's emotions is obtained.

[0216] Step Sh20: If the answer is no, the agent uses the learning mechanism of the reinforcement learning algorithm to update its own behavior strategy based on the reward or punishment feedback obtained.

[0217] Step Sh30: If yes, based on the acquired physiological data related to the user's emotions, a multimodal emotion fusion and feedback algorithm is used to analyze and acquire emotion-related data, and feature extraction is performed to acquire emotion feature data.

[0218] Overview of the Multimodal Emotion Fusion and Feedback Algorithm: This algorithm aims to comprehensively process physiological data from multiple modalities to achieve more accurate emotion analysis. It not only considers the emotional information contained in each data modality individually, but also focuses on the interrelationships and complementarities between data from different modalities. By integrating this information, it achieves a more comprehensive understanding of user emotions and provides appropriate feedback based on the analysis results to guide the agent's subsequent actions.

[0219] Data preprocessing is as follows: 1. Normalize different types of physiological data. For example, heart rate data may range from tens to over a hundred, while galvanic skin response data may range from a few microsieverts to tens of microsieverts. Normalization brings these data into a similar range, facilitating subsequent analysis. 2. Remove noise and outliers. Physiological data may be subject to various interferences during the acquisition process, such as motion artifacts affecting heart rate monitoring. Using filtering, statistical tests, and other methods, we identify and remove these noise and outliers, improving data quality.

[0220] Emotion-related data is analyzed and obtained as follows: 1. Based on heart rate data: The algorithm analyzes heart rate patterns. For example, a sudden increase in heart rate may be associated with emotions such as tension and excitement, while a steady and low heart rate may indicate relaxation. Heart rate variability metrics, such as the standard deviation of the interbeat interval (SDNN), are calculated. This metric reflects the dynamic changes in the autonomic nervous system's cardiac regulation and, in turn, is associated with emotional state. 2. Based on galvanic skin response data: Galvanic skin response is closely related to the level of emotional arousal. As emotional arousal increases, skin conductance generally increases. The algorithm analyzes changes in the amplitude and frequency of galvanic skin response to determine the level of emotional arousal. 3. Based on facial expression data (if available): The algorithm identifies the movement of key facial muscles, such as the contraction and relaxation of the corrugator supercilii, orbicularis oculi, and zygomaticus major. Different muscle combinations correspond to different basic emotions. For example, frowning and squinting may indicate dissatisfaction or worry, while a raised corner of the mouth may indicate joy. By analyzing the intensity and duration of these muscle movements, emotionally relevant facial expression information is obtained.

[0221] Feature extraction to obtain emotional feature data is as follows: 1. Time domain feature extraction: For data that changes over time, such as heart rate and skin electrical response, extract time domain features such as mean, variance, peak, etc. The mean reflects the average level of physiological indicators over a period of time, the variance reflects the degree of fluctuation of the data, and the peak may be related to the occurrence of specific emotional events. 2. Frequency domain feature extraction: Convert time domain data to the frequency domain through methods such as Fourier transform, and extract features such as the power spectrum density of different frequency components. For example, the power ratio of heart rate variability in different frequency bands (such as low frequency band and high frequency band) is related to the balance of sympathetic and parasympathetic nerve activity in the autonomic nervous system, and can be used as part of the emotional characteristics. 3.

[0222] Machine learning-based feature selection: Machine learning algorithms (such as principal component analysis (PCA) and linear discriminant analysis (LDA)) are used to filter and reduce the dimensionality of the numerous extracted features. PCA can convert high-dimensional feature data into a set of linearly uncorrelated principal components, preserving key information while reducing data dimensionality. LDA selects the most discriminative features based on the differences between different emotion categories, making the extracted emotion feature data more representative and discriminative, facilitating subsequent emotion analysis and intelligent agent decision-making.

[0223] Step Sh40, assigning weights according to the contribution of different modal data to emotion judgment, fusing the weighted multimodal data, and inputting the data into a deep learning-based emotion analysis model to judge the emotional state of each user.

[0224] Weights can be determined using a data-driven approach, specifically as follows: Analyze a large amount of multimodal physiological data with labeled emotional states. For example, collect thousands of samples containing heart rate, galvanic skin response, facial expression data, and other data, with known corresponding emotion labels (such as happiness, sadness, anger, etc.). Using machine learning algorithms such as regression analysis or decision tree algorithms, calculate the correlation between each modality and the final emotion label. The higher the correlation, the greater the contribution of that modality to the emotion judgment, and the higher the weight assigned accordingly. For example, analysis found that facial expression data has a correlation of 0.8 with the actual emotion label when judging "happiness" and "anger," while the correlation for heart rate data is 0.6. Therefore, the weight for facial expression data can be set to 0.5, and the weight for heart rate data to 0.3 (the total weight is 1, and the remaining weights are assigned to the other modal data in the same way).

[0225] Weighted fusion of multimodal data: For each modality of emotion-related data, multiply each eigenvalue by its corresponding weight. For example, suppose facial expression data, after feature extraction, has an eigenvalue of 0.8, with a weight of 0.5; and heart rate data has an eigenvalue of 0.6, with a weight of 0.3. The weighted value of the facial expression data is 0.4, and the weighted value of the heart rate data is 0.18.

[0226] The weighted eigenvalues of all modal data are combined to form a comprehensive feature vector. This vector integrates the information of multiple modal data and takes into account the relative importance of each modal data for emotion judgment.

[0227] Input to the deep learning sentiment analysis model:

[0228] Deep learning model selection: Convolutional neural networks (CNNs), recurrent neural networks (RNNs), or their variants, such as long short-term memory (LSTMs) and gated recurrent units (GRUs), are typically used. For example, when processing spatially structured facial expression image data, CNNs can effectively extract local features within the image; while when processing time-series data such as heart rate and galvanic skin response, RNNs and their variants can better capture temporal variations in the data.

[0229] Model Training: Before using weighted fusion data, the model must be trained on a large amount of multimodal data with labeled emotions. The model parameters (such as the weights and biases of the neural network) are continuously adjusted through the backpropagation algorithm to minimize the error between the model's predicted emotional state and the actual labeled emotional label. For example, the cross-entropy loss function is used to measure the difference between the predicted value and the true value. Model parameters are updated through optimization algorithms such as stochastic gradient descent. After multiple iterations of training, the model accurately learns the mapping between emotional features and emotional states from the input data.

[0230] Emotional judgment: The weighted fused feature vector is input into the trained deep learning model. The model performs calculations and inferences based on the learned patterns, and outputs each user's possible emotional state, such as "happy", "sad", "calm" and other categories, or gives the probability value of each emotional state, thereby providing a quantitative basis for the intelligent agent to understand the user's emotional state, so that the intelligent agent can more accurately adjust its behavioral strategy to meet user needs.

[0231] Step Sh50 counts the number and proportion of agents that want to reduce noise and increase wind speed to deal with pollution, and analyzes the preliminary strategies associated with agents whose proportion exceeds a preset threshold.

[0232] In a multi-agent system, different agents (representing different users) have varying needs for air purification and environmental conditioning. Counting the number and proportion of agents with representative needs—requiring noise reduction and increased wind speed to combat pollution—can help understand user groups' preferences in these two key areas. Analyzing the preliminary strategies associated with agents whose proportion exceeds a preset threshold helps identify the environmental conditioning methods preferred by the majority of users, providing guidance for subsequent policy optimization and ensuring that the final adjustment strategy better aligns with the interests of the majority.

[0233] The number and proportion of statistical agents are as follows:

[0234] Determining the criteria for determining demand: The system first determines whether the agent wishes to reduce noise or increase the fan speed to combat pollution. For example, this can be determined by analyzing the agent's previous actions (e.g., repeatedly reducing the air purifier's fan speed may indicate a desire to reduce noise; frequent requests for fan speed increases coupled with high air quality data may indicate a request to increase fan speed to combat pollution) or by directly analyzing the agent's voice commands (e.g., "Turn up the fan speed, the air is too dirty" or "Turn down the volume").

[0235] Counting and calculating the proportion: Traverse all intelligent agents and count the intelligent agents that meet the two requirements of reducing noise and increasing wind speed to deal with pollution.

[0236] The preliminary strategies associated with agents whose proportion exceeds the preset threshold are as follows:

[0237] Set a preset threshold: Based on actual application scenarios and experience, pre-set a percentage threshold (e.g., 30%). This threshold is used to determine whether the number of agents with a certain type of demand is sufficient to influence overall strategy formulation.

[0238] If the proportion of agents exceeds 30%, meaning the percentage of agents seeking to reduce noise exceeds a preset threshold, the system will analyze these agents' previous actions, related instructions, and environmental feedback to summarize their associated preliminary strategies. For example, it may be found that when air quality is acceptable, these agents tend to set their air purifiers to low-speed silent mode, and may also activate auxiliary purification devices (such as small, low-noise air purifiers) to maintain air quality.

[0239] Similarly, if the proportion of agents is less than 30%, analyzing the behavior patterns of agents that request increased wind speed to combat pollution may reveal that, upon detecting elevated pollution levels, these agents will immediately request to increase the air purifier's wind speed to maximum, and may also simultaneously request to activate the fresh air system to enhance air circulation.

[0240] Step Sh60: Fusion the sentiment analysis results with the preliminary strategy derived from the agent proportion analysis.

[0241] The fusion method is as follows:

[0242] Strategy Adjustment Based on Emotional Type: Positive Emotion: If sentiment analysis results indicate that most users are in a positive emotional state and a high percentage of agents desire noise reduction (as determined by agent proportion analysis), the initial strategy will further strengthen noise-reducing measures. For example, when selecting air purifier operating modes, prioritize low-noise modes that maintain current air quality. For other potentially noisy devices, such as air conditioners, their operating power or frequency can be appropriately reduced to reduce noise. Negative Emotion: If some users are in a negative emotional state and a high percentage of agents request higher fan speeds to combat pollution, the initial strategy will be optimized. For example, while increasing the air purifier's speed, based on the user's specific environmental needs as indicated by sentiment analysis (e.g., a user experiencing irritability may prefer a cooler, faster air flow), the angle and speed of the indoor fan can be adjusted to enhance air circulation. This, in turn, coordinates with adjusting the indoor temperature to create a more comfortable environment.

[0243] Policy weighting based on sentiment intensity: Sentiment analysis not only identifies sentiment type but also its intensity. For situations with higher sentiment intensity, policy adjustments are more forceful. For example, if user sentiment analysis in a particular area reveals strong dissatisfaction, and a high percentage of agents in that area request increased wind speeds to combat pollution, the fusion strategy will increase wind speeds more significantly than usual. This may also increase the ventilation frequency of the fresh air system, improving air quality more quickly and alleviating negative user sentiment.

[0244] 3. Strategy Complementation and Synergy: The non-conflicting parts of the initial strategy derived from sentiment analysis and agent share analysis are complemented and synergized. For example, sentiment analysis reveals that users are unhappy about air quality, while agent share analysis shows that many users are requesting increased wind speeds to combat pollution. In this case, in addition to increasing wind speeds, the fusion strategy can also incorporate other demands that may be suggested by sentiment analysis (such as a higher demand for air freshness) to increase the negative ion release of the air purifier, further improving air quality and meeting potential user needs.

[0245] Forming a Fusion Strategy: Through the various fusion methods described above, the sentiment analysis results are integrated with the preliminary strategy derived from the agent proportion analysis to form a fusion strategy that comprehensively considers user emotions and group needs. This fusion strategy will serve as the basis for subsequent simulation of environmental changes, optimization strategies, and generation of device control commands, ensuring that the final environmental adjustment measures maximize the actual needs of users and improve their satisfaction with the indoor environment.

[0246] A method for intelligent multi-dimensional air purification cloud-based intelligent control further includes the following steps after integrating the sentiment analysis results with the preliminary strategy derived from the agent proportion analysis:

[0247] Step Sh70 , inputting current indoor environmental parameters, the existing status of the smart devices and the control instructions in the fusion strategy, and simulating the environmental changes after the strategy is implemented.

[0248] Current indoor environmental parameters: These parameters encompass a wide range of environmental information, including air quality parameters (PM2.5, PM10, formaldehyde, TVOC, and other pollutant concentrations), temperature and humidity, carbon dioxide concentration, and light intensity. These parameters reflect the current indoor environment and serve as the foundation for simulating environmental changes. For example, a high PM2.5 concentration indicates poor air quality, and the simulation process uses this as a starting point to consider the impact of control instructions in the fusion strategy.

[0249] The current state of smart devices refers to the current operating status of smart devices related to indoor environmental control, such as air purifiers, air conditioners, fresh air fans, and smart windows. This includes the device's on / off status, operating mode (e.g., automatic mode, sleep mode, or high power mode for air purifiers), and set parameters (e.g., set temperature and fan speed for air conditioners). For example, if an air purifier is currently in automatic mode with a medium fan speed, this information will affect the initial conditions for the device's environmental control during the simulation.

[0250] Control instructions within the fusion strategy: These are operational instructions for smart devices within the strategy derived from the fusion of sentiment analysis and agent proportion analysis. For example, instructions might require setting the air purifier's fan speed to high, turning on the fresh air fan at maximum ventilation, or closing south-facing smart windows. These instructions serve as the driving force behind the simulated environmental changes.

[0251] The simulation process is as follows:

[0252] Establishing an environmental model: Utilize mathematical models and algorithms to describe the interrelationships between various factors in the indoor environment and the impact of smart devices on the environment. For example, the law of conservation of mass and diffusion equations can be used to simulate the spread and purification of pollutants in the air; heat transfer principles can be used to simulate changes in temperature and humidity; and the effects of ventilation and air exchange on carbon dioxide concentration can be considered.

[0253] Device response simulation: This model simulates the responses of smart devices based on control commands from the fusion strategy. For example, if a command requests an air purifier to increase its speed to a high setting, the model calculates the impact on indoor air quality based on the air purifier's performance parameters (such as purification efficiency and air volume at different speeds). It also considers the impact of fresh air introduced by the fan on indoor air quality, CO2 concentration, and other parameters, as well as the impact of closed smart windows on indoor and outdoor air exchange and temperature.

[0254] Environmental parameter updates: During the device response simulation, indoor environmental parameters are updated in real time based on the environmental model. For example, as an air purifier removes pollutants, PM2.5 concentrations gradually decrease; fresh air introduced by a fresh air blower reduces carbon dioxide concentrations; and air conditioning operation changes indoor temperature and humidity. Through continuous iterative calculations, the dynamic changes in indoor environmental parameters over a given period of time as the fusion strategy is implemented are simulated.

[0255] Simulation Results Presentation: Simulation results are presented intuitively, such as graphs showing how environmental parameters (including air quality indicators, temperature, humidity, and carbon dioxide concentration) change over time. This graph demonstrates how the indoor environment will evolve after the fusion strategy is implemented. These results provide a clear basis for optimizing the fusion strategy based on pre-set goals and historical experience.

[0256] Step Sh80, based on the simulation results and the preset optimization goals, and referring to the success and failure experiences of the intelligent agent in similar historical scenarios, the fusion strategy is further optimized.

[0257] The evaluation strategy based on the simulation results is as follows:

[0258] Comparison to Preset Optimization Targets: A detailed comparison is performed between the simulated environmental change results and the preset optimization targets. Preset optimization targets may include air quality indicators (e.g., PM2.5 concentration below a certain value, formaldehyde content within a safe range), temperature and humidity comfort zones (e.g., temperature between 22°C and 26°C, humidity between 40% and 60%), etc. For example, if the simulation results show that PM2.5 concentration decreases after implementing the fusion strategy, but remains above the preset target, this indicates that the strategy may not be effective in purifying air quality.

[0259] Analyze trends in key indicators: Focus not only on whether the final environmental indicators meet targets, but also on analyzing trends in key indicators during the simulation. For example, observe whether air quality indicators approach targets quickly or change slowly. Slow changes may indicate a need for improved strategy response. Alternatively, determine whether temperature and humidity fluctuate excessively during the adjustment process. If so, adjustments may be necessary to the control logic of related smart devices.

[0260] Reference to historical similar scenarios experience is as follows:

[0261] Retrieving Historical Data: The system searches historical data records for cases similar to the current simulation scenario. Similar scenarios are determined based on factors such as initial indoor environmental parameters, smart device status, and the main environmental issues being faced. For example, if the current scenario is characterized by high indoor PM2.5 concentrations and slightly elevated temperatures, the system will search for historical records of similar PM2.5 pollution and elevated temperatures.

[0262] Extracting successful and failed experiences: From historical records of similar scenarios, extract the actions taken by the agent and the resulting results. A successful experience might be a specific combination of smart device operations or control parameter settings that effectively improved environmental conditions; a failed experience might be an action that caused environmental deterioration or failed to achieve the expected improvement. For example, in similar past scenarios with high PM2.5 concentrations, simultaneously turning on the air purifier's high-power mode and the fresh air blower, and appropriately adjusting the opening angle of the smart windows, quickly improved air quality. This is a successful experience. On the other hand, there have been cases where over-reliance on air conditioning for cooling and improper ventilation adjustments led to deterioration of indoor air quality. This is a failed experience.

[0263] The optimized fusion strategy is as follows:

[0264] Adjusting Control Instructions: Based on the difference between simulation results and preset targets and historical experience, the control instructions in the fusion strategy are adjusted. For example, if the simulation shows slow improvement in air quality, based on historical experience, the air purifier's operating power can be increased or its operating time can be extended, while the fresh air fan's operating mode can be optimized to more effectively purify the air.

[0265] Optimizing control logic: In addition to adjusting the control instructions for specific devices, it's also possible to optimize the control logic of the entire strategy. For example, if the temperature and humidity adjustments during a simulation are found to be unstable, the coordinated control logic between smart devices can be improved to ensure that air conditioners, humidifiers, dehumidifiers, and other devices work together better when adjusting temperature and humidity, avoiding over-adjustment or delayed adjustment.

[0266] Comprehensively balance multiple objectives: During the optimization process, it's necessary to comprehensively consider the balance between multiple optimization objectives. For example, improving air purification efficiency may increase device noise, affecting user tolerance for noise. Therefore, when optimizing the strategy, it's important to balance air quality improvement with noise control, ensuring that the primary optimization objective is met without significantly impacting other aspects of the user experience. Through these steps, the fusion strategy is comprehensively and meticulously optimized to better meet actual needs, laying the foundation for its eventual translation into specific device control instructions and effective improvement of the indoor environment.

[0267] Step Sh90: converting the optimized strategy into specific device control instructions.

[0268] Based on the same inventive concept, the embodiment of the present invention provides a smart multi-dimensional air purification cloud intelligent control system, including a memory and a processor, the memory stores data that can be executed on the processor to implement the following Figures 1 to 4 Procedure for either method.

[0269] The embodiments of this specific implementation method are all preferred embodiments of the present application and are not intended to limit the scope of protection of the present application. Therefore, any equivalent changes made based on the structure, shape, and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A cloud-based intelligent control method for intelligent multi-dimensional air purification, characterized in that: include: Obtain voice control instructions for the air purifier from users in the room; Based on the room's spatial layout information, historical air quality data, and real-time air quality data, a deep learning-based spatiotemporal data analysis algorithm is used to predict and obtain air quality trends within a preset time range in the future. The voice control command is parsed to extract key information. The extracted key information, real-time air quality data, and air quality trend information within a preset time range are input into the cloud-based intelligent decision-making algorithm model. The operating parameters that need to be adjusted for the air purifier are output and parameter adjustment decision instructions are formed. Send the parameter adjustment decision instruction formed to the air purifier terminal through the network; The air purifier adjusts its own operating parameters in response to the parameter adjustment decision instruction.

2. The cloud-based intelligent control method for intelligent multi-dimensional air purification according to claim 1, characterized in that: The steps after obtaining the voice control instructions of the user in the room regarding the air purifier are also included, which are as follows: Analyze whether relevant physiological data related to user emotions is obtained; If not, maintain the original setting; If yes, then based on the obtained physiological data related to the user's emotions, a multimodal emotion fusion and feedback algorithm is used to analyze and obtain emotion-related data, and feature extraction is performed to obtain emotion feature data; The pre-processed voice command key information, emotional characteristic data, real-time air quality data and future air quality trend information are input together into an advanced decision-making module based on the integration of fuzzy logic and expert system, and the operating parameters that need to be adjusted for the air purifier are output to form parameter adjustment decision instructions. The parameter adjustment decision instructions formed are sent to the air purifier terminal through the network, and the air purifier adjusts its own operating parameters in response to the parameter adjustment decision instructions.

3. The cloud-based intelligent control method for intelligent multi-dimensional air purification according to claim 2 is characterized in that: The operation method of the advanced decision-making module based on the fusion of fuzzy logic and expert system is as follows: The fuzzy logic system performs fuzzification on the input data; Reasoning is performed based on a pre-set rule base in the expert system, which contains strategies for air purification and environmental regulation in various situations; After fuzzy logic reasoning, the decision module will generate detailed device control instructions, which cover the adjustment of air purifier operating parameters and linkage control of other smart devices.

4. The intelligent multi-dimensional air purification cloud intelligent control method according to any one of claims 1 to 3, characterized in that: The steps after obtaining the voice control instructions of the user in the room regarding the air purifier are also included, which are as follows: Analyze whether the acquired voice control commands come from multiple users; If not, maintain the original setting; If the answer is yes, each user is treated as an agent, and the voice commands are parsed to extract key information. This information, along with the room layout, air quality data for each corner, and air quality trends within a preset timeframe, is then used as input for the multi-agent reinforcement learning algorithm. Each agent uses the decision-making model in the multi-agent reinforcement learning algorithm to select matching actions based on its own state information and the acquired environmental information. The scope of actions includes but is not limited to adjusting the operating parameters of the air purifier and controlling the switches or parameters of other smart devices.

5. The cloud-based intelligent control method for intelligent multi-dimensional air purification according to claim 4 is characterized in that: It also includes the steps after each agent selects a matching action based on its own state information and the acquired environment information using the decision model in the multi-agent reinforcement learning algorithm. The specific steps are as follows: Various sensors distributed indoors are used to collect real-time data on environmental changes, while user satisfaction feedback on the current environment is regularly collected through user terminal devices. Summarize the collected environmental change data and user satisfaction feedback on the current environment to form environmental feedback information; Based on the feedback from the environment and the pre-set reward and punishment rules, each agent is rewarded or punished. The intelligent agent uses the learning mechanism of the reinforcement learning algorithm to update its own behavior strategy based on the reward or punishment feedback obtained.

6. The cloud-based intelligent control method for intelligent multi-dimensional air purification according to claim 5, characterized in that: The agent uses the learning mechanism of the reinforcement learning algorithm to update its own behavior strategy based on the reward or penalty feedback it receives, including: Analyze whether relevant physiological data related to user emotions is obtained; If not, the agent uses the learning mechanism of the reinforcement learning algorithm to update its own behavior strategy based on the reward or punishment feedback obtained; If yes, then based on the obtained physiological data related to the user's emotions, a multimodal emotion fusion and feedback algorithm is used to analyze and obtain emotion-related data, and feature extraction is performed to obtain emotion feature data; We assign weights based on the contribution of different modal data to emotional judgment, fuse the weighted multimodal data, and input it into a deep learning-based sentiment analysis model to determine each user's emotional state. Count the number and proportion of agents that want to reduce noise and increase wind speed to deal with pollution, and analyze the preliminary strategies associated with agents whose proportion exceeds the preset threshold; Combine the sentiment analysis results with the preliminary strategy derived from the agent proportion analysis.

7. The cloud-based intelligent control method for intelligent multi-dimensional air purification according to claim 6, characterized in that: It also includes the following steps after fusing the sentiment analysis results with the preliminary strategy derived from the agent share analysis: Input the current indoor environmental parameters, the existing status of the smart devices, and the control instructions in the fusion strategy to simulate the environmental changes after the strategy is implemented; Based on the simulation results and the preset optimization goals, the fusion strategy is further optimized by referring to the success and failure experiences of the intelligent agents in similar historical scenarios. Convert the optimized strategy into specific device control instructions.

8. A smart multi-dimensional air purification cloud intelligent control system, characterized by: It includes a memory, a processor, and a program stored in the memory and executable on the processor, wherein the program can be loaded and executed by the processor to implement an intelligent multi-dimensional air purification cloud intelligent control method as described in any one of claims 1 to 7.

Images