Air purification system based on intelligent perception and self-adaptive regulation and control method thereof

Abstract

The invention belongs to the technical field of intelligent control, and particularly relates to an air purification system based on intelligent perception and a self-adaptive regulation and control method thereof. Comprising a multi-dimensional environment perception unit, an environment data fusion and modeling unit, a self-adaptive regulation and optimization decision unit, a modular air purification execution unit and a man-machine interaction and user preference setting unit. According to the method, multi-element parameters are collected, fusion modeling is carried out, a pollution source is identified and predicted, a multi-target optimization algorithm is applied to generate an optimal purification strategy, and then a purification execution unit is controlled. The indoor air quality can be finely and dynamically managed, the purification efficiency is remarkably improved, energy consumption, operation noise and the service life of a filter material are intelligently balanced, and a personalized healthy breathing environment is provided.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of intelligent control, and in particular to an air purification system based on intelligent sensing and an adaptive regulation method thereof. BACKGROUND

[0002] With the advancement of industrialization and the increase in urban population density, indoor air quality problems have attracted increasing attention. As the core place for people's daily work and life, the air quality of indoor space directly affects human health and living comfort. Air purification systems, as key equipment for improving indoor air quality, have been widely used in residential, commercial buildings, medical institutions and industrial clean spaces. Such systems remove particulate matter, harmful gases, microorganisms and odors from the air to provide a healthy and clean breathing environment for users. The development of this technology is of great significance to improving public health and quality of life.

[0003] Current mainstream air purification systems mainly rely on physical and chemical principles to achieve purification functions, including high-efficiency air filters to trap fine particulate matter, activated carbon to adsorb volatile organic compounds and odors, ultraviolet light to inactivate microorganisms, and ion generators to settle charged particles. Meanwhile, fan units are integrated to drive air flow through various treatment modules for purification. In terms of control strategy, existing systems mostly use timed on-off, manual adjustment of air speed, or threshold comparison based on a single sensor (such as PM2.5 and carbon dioxide sensors) to trigger a preset working mode. Such solutions are simple in structure and easy to implement, and have played a certain role in solving basic indoor air pollution problems.

[0004] However, existing technologies are difficult to adapt to complex indoor environments and refined application requirements, and limitations are gradually emerging. On the one hand, existing systems mostly rely on passive response control of single-point or a small number of sensors, which cannot comprehensively sense the dynamic changes of indoor environment. Indoor pollution sources are diverse and have spatial and temporal heterogeneity, making it difficult for a single sensor to cover multiple pollution parameters, let alone identify pollution source types and diffusion patterns. This can lead to missed detection or delayed response of some pollutants, and fixed purification modes triggered by a single parameter can ignore the balance between purification efficiency and energy consumption and noise, resulting in resource waste. On the other hand, existing systems lack multi-dimensional and multi-regional intelligent analysis capabilities for indoor air quality, making it difficult to construct an air quality "holographic picture" and develop optimized strategies based on spatial layout, personnel activity and user demand. In complex conditions, it is difficult to balance purification efficiency, energy efficiency and equipment life, and it is difficult to achieve truly intelligent and personalized air environment management. Therefore, the present application provides an air purification system based on intelligent sensing and an adaptive regulation method thereof. SUMMARY

[0005] In order to make up for the shortcomings of the prior art and solve the technical problems proposed in the background art.

[0006] The technical scheme adopted by the present application to solve its technical problems is: the air purification system based on intelligent sensing and the adaptive regulation and control method thereof, comprising a multi-dimensional environment sensing unit, an environment data fusion and modeling unit, an adaptive regulation and control and optimization decision unit, a modular air purification execution unit, and a man-machine interaction and user preference setting unit. The multi-dimensional environment sensing unit is used to collect multiple parameters of indoor environment in real time; the environment data fusion and modeling unit is used to receive and process the multiple parameters, construct a dynamic holographic picture of indoor air quality, and identify pollution sources and predict pollution diffusion trend; the adaptive regulation and control and optimization decision unit is used to generate an optimized air purification strategy according to the dynamic holographic picture, pollution source identification result, pollution diffusion prediction, and preset user preference; the modular air purification execution unit is used to execute corresponding purification operation according to the optimized strategy; and the man-machine interaction and user preference setting unit is used to provide user setting preference and receive system state information.

[0007] Further, the multi-dimensional environment sensing unit comprises at least one PM2.5 sensor, at least one carbon dioxide sensor, at least one volatile organic compound sensor, at least one temperature and humidity sensor, at least one light intensity sensor, and at least one personnel presence sensor. The PM2.5 sensor adopts laser scattering principle and is used to detect the concentration of particulate matter with a diameter less than 2.5 microns in the air, with a measurement range of 0 to 999 micrograms per cubic meter and a resolution of 1 microgram per cubic meter. The CO2 sensor adopts non-dispersive infrared principle and is used to detect the volume concentration of carbon dioxide in the air, with a measurement range of 0 to 5000 ppm and a resolution of 1 ppm. The VOCs sensor adopts metal oxide semiconductor technology and is used to detect the total amount of various volatile organic compounds in the air, with an output that can be converted into total volatile organic compound concentration or air quality index, with a measurement range of 0 to 1000 ppb. The temperature and humidity sensor adopts capacitive or resistive sensing principle and is used to detect environmental temperature, with a measurement range of -20℃ to 70℃ and an accuracy of ±0.5℃, and relative humidity, with a measurement range of 0%RH to 100%RH and an accuracy of ±3%RH. The light intensity sensor adopts a photodiode array and is used to detect environmental illumination, with a measurement range of 0 to 10000 lux and a resolution of 1 lux. The personnel presence sensor adopts passive infrared principle or millimeter wave radar principle and is used to detect whether there is personnel activity in the specified area, with a detection distance of 5 meters to 10 meters and a detection angle of 120 degrees.

[0008] As a preferred embodiment of the present application, the sensors of the multi-dimensional environment perception unit are arranged in a distributed manner to cover multiple areas or spaces in the indoor environment. Specifically, the system can be configured to include multiple sub-perception nodes, each of which integrates some or all of the above-mentioned types of sensors and transmits data to the central controller through a wireless communication module. The sub-perception nodes are embedded with microcontroller units for collecting, calibrating, and preliminarily filtering raw sensor data, and packaging and sending the processed data to the environment data fusion and modeling unit through the wireless communication module. The central controller can also integrate some sensors as a master perception node, forming a hierarchical perception network. This distributed arrangement aims to obtain heterogeneous data on the spatial dimension of indoor air quality, thereby enabling accurate identification of local pollution events.

[0009] Further, the environment data fusion and modeling unit includes a data collection and preprocessing module, a multi-source data fusion module, and an environment modeling and pollution source identification module. The data collection and preprocessing module receives raw sensor data from the multi-dimensional environment perception unit and performs timestamp labeling, data format unification, missing value interpolation, outlier removal, and noise filtering operations. Specifically, the noise filtering can use Kalman filtering or moving average filtering algorithms to improve data stability and accuracy. The multi-source data fusion module integrates processed data from different types of sensors and different perception nodes. The fusion method can use weighted averaging, Bayesian networks, or deep learning techniques to comprehensively evaluate the current indoor environment state. The environment modeling and pollution source identification module constructs a dynamic model of indoor air quality based on the fused multi-dimensional environment data. The dynamic model can use machine learning-based methods such as recurrent neural networks or long short-term memory networks to capture the time series variation of air pollutant concentrations. The pollution source identification analyzes the correlation between different pollutant concentration changes and their correlation with human activity, external environment, and other factors to infer the possible source of pollution. When VOCs concentration suddenly increases and PM2.5 concentration also increases, combined with external environment data, it may indicate cooking activities or outdoor pollution penetration. The environment modeling and pollution source identification module is also responsible for pollution diffusion trend prediction, which can predict the concentration changes of pollutants in each area in the future based on historical data, current airflow models, and machine learning prediction models, providing prospective basis for adaptive regulation.

[0010] As an embodiment of the present application, the adaptive control and optimization decision unit is the core intelligent part of the system, which includes: an intelligent decision engine, an optimization algorithm module, and a purification strategy generation module. The intelligent decision engine receives the environmental model, pollution source identification results, and pollutant dispersion prediction data from the environmental data fusion and modeling unit, while obtaining the user preference settings provided by the human-computer interaction and user preference setting unit. The user preference settings can include but are not limited to: priority settings, specific pollutant thresholds, work and rest schedules, and preferences for specific purification modes.

[0011] Further, the optimization algorithm module embeds a multi-objective optimization algorithm, which aims to balance multiple conflicting performance indicators such as energy consumption, operating noise level, filter life, purification efficiency, and user comfort, while meeting indoor air quality standards. The optimization algorithm can use a heuristic search-based method, such as genetic algorithm or particle swarm optimization, or a reinforcement learning-based method to learn the optimal strategy through continuous interaction with the environment. The optimization algorithm module defines a multi-dimensional objective function, whose input parameters include: the current indoor air quality state, the predicted future air quality trend, the operating costs of the purification system, the purification efficiency parameters of different purification modules, and the user preference factors. The objective of the objective function is to minimize the total cost while ensuring that the air quality indicators remain within the safety range set by the user or national standards. For example, when the indoor PM2.5 concentration is higher than the preset threshold and there is no human activity, the system tends to choose a high-purification-efficiency but low-noise and low-energy-consumption operating mode to slowly purify over a longer period of time; when there is human presence and the VOCs concentration is rapidly rising, the system tends to quickly start a high-efficiency, possibly slightly noisy purification module to rapidly remove pollutants, and quickly switch back to a silent and energy-saving mode after the pollutant concentration decreases.

[0012] Specifically, the optimization algorithm module generates a series of candidate purification system operating parameter combinations, such as the activation / deactivation state of each purification module, the fan operating speed, and the operating intensity of each purification module, through iterative calculation. The purification strategy generation module selects the optimal combination from the candidate operating parameter combinations generated by the optimization algorithm module as the current air purification strategy. The strategy generation module will also consider the dynamic response characteristics and delays of the system to ensure smooth switching and effective implementation of the strategy. For example, to avoid equipment wear and tear caused by frequent start-stop, the strategy generation module can introduce minimum operating time or minimum downtime constraints.

[0013] As an embodiment of the present application, the modular air purification execution unit comprises: a fan unit, a primary filter module, a high-efficiency particulate filter module, an activated carbon adsorption module, a UV sterilization module, and a negative ion generation module. The fan unit adopts a centrifugal fan or an axial flow fan driven by a brushless DC motor, and its speed can be stepless or multi-step adjusted by a pulse width modulation signal, so as to realize accurate control of air volume. The primary filter module is used to remove larger particulate matters in the air, such as hair, dust, etc., and its filter material is polyester fiber or nylon net. The high-efficiency particulate filter module adopts a high-efficiency air filter of H13 or H14 grade, which is used to efficiently remove fine particulate matters such as PM2.5 and PM10, and its removal efficiency for particulate matters with a diameter of 0.3 microns can reach more than 99.97%. The activated carbon adsorption module adopts high-iodine-value coconut activated carbon or coal activated carbon, which is used to adsorb harmful gases such as volatile organic compounds, formaldehyde, and odors, and its specific surface area is greater than 1000 square meters per gram. The UV sterilization module is configured with a UVC ultraviolet lamp tube with a wavelength of 253.7 nanometers, which is used to inactivate bacteria, viruses, and molds in the air, and its radiation intensity can be adjusted according to the strategy, such as by controlling the power-on time or driving current of the lamp tube. The negative ion generation module generates negative oxygen ions by high-voltage ionization of air, which is used to adsorb and settle fine particulate matters in the air, and has a certain air freshening effect, and its release concentration and working time can be controlled by the system. Each module has an independent control interface, and can be independently started, stopped, or intensity adjusted according to the instructions issued by the adaptive control and optimization decision unit, so as to realize flexible combination and optimized operation.

[0014] Further, the human-computer interaction and user preference setting unit comprises a display screen, input keys or a touch screen, and an interface for communication with external devices. The display screen is used to display real-time indoor air quality data, system operation mode, air volume level, filter screen life reminder, and historical data charts and other information. The input keys or touch screen are used for user to manually adjust system parameters, switch operation modes, or set personalized preferences. The communication interface can adopt Bluetooth or Wi-Fi technology, allowing users to remotely monitor and control the air purification system through a mobile application, and set more detailed personalized purification strategies. The user preference settings can specifically include: priority weight configuration for different time periods; upper and lower threshold values for specific pollutant concentrations; additional purification requirements for specific allergens such as pollen and dust mites for allergic users; and preference settings for system energy consumption and filter replacement cycle. These preference settings will be input as constraint conditions or weight factors into the optimization algorithm module of the adaptive control and optimization decision unit to affect the final strategy generation.

[0015] As a method of the present application, an adaptive control method of an intelligent sensing-based air purification system comprises the following steps: S1: Real-time collection of multi-element parameters of indoor environment by multi-dimensional environment perception unit, said multi-element parameters including but not limited to PM2.5 concentration, CO2 concentration, VOCs concentration, environmental temperature, relative humidity, light intensity and personnel presence state; S2: Transmission of said multi-element parameters to environment data fusion and modeling unit, pre-processing of data, multi-source data fusion by the unit, and construction of dynamic holographic scene of indoor air quality, said holographic scene containing pollutant distribution status of different areas and different time points in the room; S3: Based on said dynamic holographic scene, identification of indoor pollution source type and its occurrence area by environment data fusion and modeling unit, and prediction of diffusion trend of pollutant concentration in each area in the future period of time; S4: Transmission of said dynamic holographic scene, pollution source identification result, pollutant diffusion prediction and user preset preference to adaptive regulation and optimization decision unit; S5: In said adaptive regulation and optimization decision unit, using multi-objective optimization algorithm, according to said input information, balancing multiple performance indicators such as indoor air quality, energy consumption, operating noise and filter life, calculating and generating the current optimal air purification strategy, said strategy including activation state of purification module, fan operating speed and operating intensity of each purification module; S6: Transmission of said optimal air purification strategy to modular air purification execution unit; S7: The modular air purification execution unit controls the fan unit to run at a set wind speed according to said strategy, and selectively starts or adjusts the operating intensity of the primary filter module, high-efficiency particulate matter filter module, activated carbon adsorption module, ultraviolet sterilization module and negative ion generation module, thereby realizing purification of indoor air; S8: Circulating execution of S1 to S7 to realize continuous monitoring of indoor air quality and real-time adaptive regulation of air purification system operating parameters.

[0016] Further, in S3, said pollution source identification can be specifically implemented by: establishing an association model between different pollution sources and different pollutant concentration change patterns, said model can be based on machine learning algorithm, for example, using a classifier to train sensor data to identify typical pollution events such as cooking, personnel gathering, external pollution penetration, cleaning activities, etc. The pollutant diffusion trend prediction can use a time series prediction model, for example, based on recurrent neural network or its variant long short-term memory network, combining historical environmental data, current environmental state and external influencing factors, to predict the concentration change of PM2.5, CO2, VOCs and other key pollutants in different areas in the future several minutes to several hours.

[0017] As a preferred regulation method of the present application, in S5, the optimization objectives of the multi-objective optimization algorithm can include: maintaining the concentrations of PM2.5, CO2, and VOCs within the safe range set by the user or the national standard; minimizing the total energy consumption of the system; minimizing the noise generated during the operation of the system; maximizing the service life of the filter material; and maximizing the user comfort. In each iteration, the algorithm evaluates these objective functions and finds a solution that satisfies all the constraints and optimizes the weighted objective function through the optimization strategy. The weighting factors are provided by the user preference setting unit to reflect the user's priority weight for different performance indicators. For example, when the user sets the "sleep mode", the weight factor of noise will be significantly higher than that of energy consumption and purification speed; when the user sets the "fast purification mode", the weight factor of purification efficiency will be significantly improved.

[0018] The beneficial effects of the present application are as follows: 1. The intelligent sensing-based air purification system and its adaptive regulation method of the present application overcome the inherent limitations of single-point sensing, passive response, and "one-size-fits-all" control strategies in the prior art through its multi-dimensional environmental sensing, intelligent data fusion and prediction, and multi-objective optimization decision-making capabilities. The present application realizes fine and dynamic management of indoor air quality, can accurately identify pollution sources and predict their diffusion trend, and thus takes more accurate and forward-looking purification measures. Such a strategy not only significantly improves the efficiency and effectiveness of air purification, but also intelligently balances the relationship between purification performance and key performance indicators such as energy consumption, operating noise, etc. according to actual needs and user preferences, avoiding unnecessary resource waste, prolonging the service life of equipment and filter material, and ultimately providing users with a more personalized, comfortable, and healthy indoor breathing environment. The modules interact with each other through standardized communication protocols to ensure the scalability and modular update capability of the system. The central controller can transmit data and issue instructions to each sub-sensing node and modular air purification execution unit through wired (such as Ethernet, RS485) or wireless (such as Wi-Fi, Zigbee) methods, forming an efficient and collaborative intelligent network. BRIEF DESCRIPTION OF DRAWINGS

[0019] The present application will be further described below with reference to the accompanying drawings.

[0020] Figure 1 is a structural framework diagram of an intelligent sensing-based air purification system of the present application; Figure 2 is a flowchart of an adaptive regulation method of an intelligent sensing-based air purification system of the present application.

[0021] In the figure: 10, multi-dimensional environment perception unit; 11, PM2.5 sensor; 12, carbon dioxide sensor; 13, volatile organic compound sensor; 14, temperature and humidity sensor; 15, light intensity sensor; 16, personnel presence sensor; 17, sub-perception node; 18, wireless communication module; 19, microcontroller unit; 20, central controller; 30, environmental data fusion and modeling unit; 31, data acquisition and preprocessing module; 32, multi-source data fusion module; 33, environmental modeling and pollution source identification module; 40, adaptive regulation and optimization decision unit; 41, intelligent decision engine; 42, optimization algorithm module; 43, purification strategy generation module; 50, modular air purification execution unit; 51, fan unit; 52, primary filter module; 53, high-efficiency particulate matter filter module; 54, activated carbon adsorption module; 55, ultraviolet sterilization module; 56, negative ion generation module; 60, human-computer interaction and user preference setting unit; 61, display screen; 62, input key; 63, touch screen; 64, communication interface. DETAILED DESCRIPTION

[0022] In order to make the technical means, creative features, purposes and effects realized by the present application easy to understand, the present application is further described below in combination with specific embodiments.

[0023] REFERENCE Figure 1 The disclosed air purification system based on intelligent perception includes a multi-dimensional environment perception unit 10, an environmental data fusion and modeling unit 30, an adaptive regulation and optimization decision unit 40, a modular air purification execution unit 50, and a human-computer interaction and user preference setting unit 60. The multi-dimensional environment perception unit 10 is responsible for real-time acquisition of multiple parameters of indoor environment, and transmits these data to the environmental data fusion and modeling unit 30. The environmental data fusion and modeling unit 30 receives and processes the multiple parameters, constructs a dynamic holographic picture of indoor air quality, further identifies the type of pollution source and its occurrence area, and predicts the diffusion trend of pollutants in indoor space. The adaptive regulation and optimization decision unit 40 generates an optimized air purification strategy according to the dynamic holographic picture, pollution source identification result, pollutant diffusion prediction, and preset user preference. This strategy is then sent to the modular air purification execution unit 50, which performs corresponding purification operations according to the strategy. The human-computer interaction and user preference setting unit 60 provides functions of setting personalized preferences, checking system status information, and manual control for users. The units interact with each other through standardized communication protocols, ensuring the overall collaborative work efficiency and accuracy of information flow of the system. The central controller 20, as the core coordination node of the system, is responsible for integrating part of the perception functions, and managing data transmission and instruction distribution.

[0024] Further, the configuration of the multi-dimensional environment perception unit 10 is particularly crucial, which includes various high-precision sensors to obtain detailed data of the indoor environment. Specifically, this unit integrates at least one PM2.5 sensor 11, at least one carbon dioxide (CO2) sensor 12, at least one volatile organic compounds (VOCs) sensor 13, at least one temperature and humidity sensor 14, at least one light intensity sensor 15, and at least one person presence sensor 16. The PM2.5 sensor 11 adopts the laser scattering principle to accurately detect the concentration of particulate matter with a diameter less than 2.5 microns in the air, with a measurement range designed for 0 to 999 micrograms per cubic meter and a resolution of 1 microgram per cubic meter, ensuring sensitive monitoring of fine particulate matter. The CO2 sensor 12 adopts the non-dispersive infrared (NDIR) principle, which can accurately detect the volume concentration of carbon dioxide in the air, with a measurement range covering 0 to 5000 ppm and a resolution of 1 ppm, meeting the evaluation needs of indoor ventilation conditions. The VOCs sensor 13 adopts the metal oxide semiconductor (MOS) technology to detect the total amount of various volatile organic compounds in the air, and its output signal can be converted into total volatile organic compound (TVOC) concentration or indoor air quality index (IAQ) after internal algorithm processing, with a measurement range of 0 to 1000 ppb, providing basic data for identifying odors and harmful gases. The temperature and humidity sensor 14 adopts capacitive or resistive sensing principles to detect environmental temperature, with a measurement range of -20°C to 70°C and an accuracy of ±0.5°C, and relative humidity, with a measurement range of 0%RH to 100%RH and an accuracy of ±3%RH. These environmental parameters are crucial for assessing human comfort and affecting the diffusion speed of pollutants. The light intensity sensor 15 adopts a photodiode array to detect ambient illumination, with a measurement range of 0 to 10000 lux and a resolution of 1 lux, which can assist in determining indoor activity patterns or sleep status. The person presence sensor 16 adopts passive infrared (PIR) principle or millimeter wave radar principle to detect whether there is personnel activity in the specified area, with a detection distance of 5 to 10 meters and a detection angle of 120 degrees, providing key information for people flow and activity area judgment.

[0025] As a preferred embodiment of the present application, the sensors of the multi-dimensional environment perception unit 10 are arranged in a distributed manner to achieve comprehensive coverage of multiple areas or spaces in the room. The system is configured to include multiple sub-perception nodes 17, each of which integrates some or all of the above-mentioned types of sensors and transmits data to the central controller 20 through a wireless communication module 18. The wireless communication module 18 can be a Wi-Fi module based on the IEEE 802.11b / g / n standard to utilize existing wireless network infrastructure for high-bandwidth data transmission; or a Zigbee module based on the IEEE 802.15.4 standard for low-power, multi-node ad hoc network communication, suitable for battery-powered node deployment. Each sub-perception node 17 is embedded with a microcontroller unit (MCU) 19, which is responsible for collecting, preliminary calibrating, linearizing, and preliminarily filtering raw sensor data, and packaging the processed data in a predetermined communication protocol, such as JSON or Protobuf format, and then sending it to the central controller 20 through the wireless communication module 18.

[0026] The central controller 20 can also integrate some core sensors as a master perception node, forming a hierarchical perception network architecture. The core purpose of this distributed arrangement is to obtain heterogeneous data of indoor air quality in the spatial dimension, such as the air quality difference between the living room and the bedroom, the kitchen and the study, so as to achieve accurate identification and positioning of local pollution events (such as local cooking fume, local CO2 rise caused by personnel gathering), and provide a basis for subsequent precise purification.

[0027] Further, the environmental data fusion and modeling unit 30 is the basis of system intelligence, and its detailed composition includes a data acquisition and preprocessing module 31, a multi-source data fusion module 32, and an environmental modeling and pollution source identification module 33. The data acquisition and preprocessing module 31 receives raw sensor data from the multi-dimensional environment perception unit 10 and performs a series of strict data quality control operations. These operations include accurate timestamp labeling for each piece of data, unifying heterogeneous data from different sensors and nodes into a standard data format, using interpolation algorithms to interpolate missing values that occur intermittently (e.g., using linear interpolation or spline interpolation), removing outliers through statistical methods (such as the 3σ criterion or local outlier factor detection), and filtering random noise in the data using signal processing algorithms (e.g., Kalman filtering or moving average filtering) to significantly improve data stability and accuracy. Kalman filtering performs outstandingly in processing dynamic and uncertain system data, effectively estimating system state, while moving average filtering smooths short-term fluctuations by averaging recent data points.

[0028] The multi-source data fusion module 32 integrates the processed data from different types of sensors (such as PM2.5, CO2, VOCs) and different perception nodes (such as living room, bedroom sub-nodes). The fusion method has high flexibility and configurability. Weighted average method can be used, in which different weights are assigned to the reliability and importance of different sensors; Bayesian network can also be used to establish the causal relationship between the data of each sensor through probability inference to achieve deeper fusion; or deep learning technology such as multi-modal neural network can be used to automatically learn features from complex multi-source data and perform high-dimensional information integration to comprehensively evaluate the current indoor environment state. For example, by fusing temperature and humidity, light, personnel presence and CO2 data, it can accurately determine whether there is personnel activity in the room and its state.

[0029] The environmental modeling and pollution source identification module 33, the core function of which is to construct a dynamic model of indoor air quality based on the fused multi-dimensional environmental data. The dynamic model can use machine learning-based methods such as recurrent neural networks (RNN) or their variants long short-term memory networks (LSTM), which are particularly suitable for processing time series data and can effectively capture the time series variation, periodicity and long-term dependence of air pollutant concentration. The input features of the model include current sensor data, historical air quality data, timestamp information, and external environmental parameters (such as outdoor temperature, relative humidity, AQI index, etc.), and the output of the model predicts the concentration trend of the main indoor pollutants (such as PM2.5, CO2, VOCs) in the future. The pollution source identification function is achieved by analyzing the correlation, rate of change of different pollutant concentrations, and their correlation with personnel activity, external environment, etc. to infer the possible source of the pollutants. For example, when the CO2 concentration rapidly rises in a short time, and the PM2.5 and VOCs concentrations also rise simultaneously, and if the personnel presence sensor data shows an increase in the number of people in the area, the system can intelligently infer that it is caused by personnel gathering. When the VOCs concentration suddenly rises, the PM2.5 concentration also shows an upward trend, but the CO2 concentration changes little, combined with external environmental data (such as wind direction, outdoor AQI) and user historical behavior data, the system may indicate cooking activities, new furniture release or outdoor pollution penetration. The environmental modeling and pollution source identification module 33 also has the function of predicting the diffusion trend of pollutants, which can predict the concentration change path and speed of pollutants in different areas in the future minutes to hours based on historical data, current airflow model (established by a few distributed wind speed sensors or a pre-set indoor space layout model), and machine learning prediction model (such as support vector regression SVR or Gaussian process regression GPR), providing prospective and predictive basis for adaptive regulation.

[0030] As an embodiment of the present application, the adaptive control and optimization decision unit 40 is the core intelligent part of the system, which is composed of an intelligent decision engine 41, an optimization algorithm module 42, and a purification strategy generation module 43. The intelligent decision engine 41 serves as an information aggregation and distribution center, which receives the environmental model, pollution source identification results, and pollutant dispersion prediction data from the environmental data fusion and modeling unit 30, and obtains the user's preset personalized preference settings from the human-computer interaction and user preference setting unit 60. The user preference settings are a set of configurable parameters, which can include but are not limited to: priority settings for different performance indicators (e.g., prioritizing air quality, prioritizing energy consumption reduction, prioritizing noise reduction), upper and lower threshold values for specific pollutant concentrations (e.g., PM2.5 below 25 μg / m³, CO2 below 800 ppm), user's work and rest schedule (e.g., sleep time, work time, and time period away from home), and preference for specific purification modes or module combinations.

[0031] Further, the optimization algorithm module 42 embeds advanced multi-objective optimization algorithms, whose core goal is to balance multiple conflicting performance indicators such as energy consumption, operating noise level, filter life, purification efficiency, and user comfort under the premise of meeting indoor air quality standards. The optimization algorithm can use heuristic search-based methods such as genetic algorithm (Genetic Algorithm, GA) or particle swarm optimization (Particle Swarm Optimization, PSO), which can efficiently find global optimal solutions in complex multi-dimensional search spaces. Genetic algorithm simulates natural selection and genetic mechanisms to iteratively generate and optimize candidate strategies; particle swarm optimization simulates bird hunting behavior by sharing experiences between individuals and groups to converge to the optimal solution. In addition, reinforcement learning (Reinforcement Learning, RL) based methods can also be used to learn optimal strategies through continuous interaction with the environment, which is particularly suitable for continuous decision-making processes in dynamic environments. The system learns autonomously through trial and error and reward mechanisms to maximize long-term performance indicators.

[0032] The optimization algorithm module 42 internally defines a multi-dimensional objective function, whose input parameter set is extensive and dynamic, including: current indoor air quality status (provided by the environmental model, such as real-time concentration and spatial distribution of PM2.5, CO2, VOCs), predicted future air quality trend (pollutant diffusion prediction), various operating costs of the purification system (including power consumption, filter wear and tear, noise level generated during operation, which can be quantified as part of the cost function), inherent purification efficiency parameters of different purification modules (such as HEPA efficiency for PM2.5, activated carbon adsorption capacity for VOCs), and user preference factors from the user preference setting unit (as weights or constraints of the objective function). The objective of the objective function is to minimize the weighted total cost while ensuring that all air quality indicators (such as PM2.5, CO2, VOCs concentration) are maintained within the safety range set by the user or national standards. For example, when the user sets "sleep mode", the weight factor of noise will be significantly higher than energy consumption and purification speed; while when the user sets "fast purification mode", the weight factor of purification efficiency will be significantly increased. In a specific scenario, when the indoor PM2.5 concentration is higher than the preset threshold and there is no human activity, the system tends to choose a high purification efficiency operation mode while ensuring low noise and low energy consumption to purify smoothly over a long period of time; while when there is a person present and the VOCs concentration is rising rapidly, the system tends to quickly start high-efficiency, possibly slightly noisy purification modules (such as high-speed fan, full-power activated carbon module), to quickly remove pollutants, and quickly switch back to a silent and energy-saving mode when the pollutant concentration drops to a safe level, achieving dynamic balance.

[0033] Specifically, the optimization algorithm module 42 generates a series of candidate purification system operation parameter combinations through iterative calculation. These parameter combinations are refined to specific execution levels, such as: the activation / deactivation state (ON / OFF) of each purification module, the fan operating speed (for example, set to low, medium, and high three fixed gears, or through pulse width modulation PWM signal for 0-100% stepless speed control), and the operating intensity of each purification module (for example, the power level of the UV lamp, the release concentration of the ion generator, or the working time). The purification strategy generation module 43 selects and determines the optimal combination as the air purification strategy at the current time from the candidate operation parameter combinations generated by the optimization algorithm module 42 according to the target function evaluation results. The strategy generation module 43 will also fully consider the dynamic response characteristics and potential operation delay of the system to ensure the smoothness, effectiveness and reliability of the strategy switching. For example, to avoid equipment wear and tear and operation jitter caused by frequent start and stop, the strategy generation module can introduce a minimum operating time constraint (for example, once a module is started, it must run for at least 5 minutes) or a minimum downtime constraint (for example, a module must be closed for at least 2 minutes before it can be started again) to prolong the service life of the equipment and improve user experience.

[0034] As an embodiment of the present application, the modular air purification execution unit 50 has high flexibility and maintainability, and its internal structure includes a fan unit 51, a primary filter module 52, a high-efficiency particulate filter module 53, an activated carbon adsorption module 54, a UV sterilization module 55, and a negative ion generation module 56. The fan unit 51 is the core of the system air flow circulation, which uses a centrifugal fan or an axial flow fan driven by a brushless direct current motor (BLDC) to achieve high efficiency, low noise and long service life. The fan speed can be stepless or multi-step precisely adjusted by a pulse width modulation (PWM) signal from 0 to 100%, so as to achieve fine control of indoor circulating air volume from low wind volume to high wind volume mode required for rapid purification. The primary filter module 52 is located at the air inlet and is used to remove larger particles in the air, such as hair, pet dander, large particle dust, etc. The filter material is usually polyester fiber or nylon mesh, which has good air permeability and washability, and can significantly prolong the service life of the subsequent high-efficiency filter. The high-efficiency particulate filter module 53 is the key to removing fine particles, which uses a H13 or H14 grade high-efficiency air filter (HEPA) with a removal efficiency of more than 99.97% for particles with a diameter of 0.3 microns, which can efficiently remove PM2.5, PM10 and other fine particles, bacteria spores and allergens such as pollen. The activated carbon adsorption module 54 is mainly used for the removal of gaseous pollutants, which uses high-iodine-value coconut activated carbon or coal activated carbon to ensure a high specific surface area (usually greater than 1000 square meters per gram), so as to efficiently adsorb volatile organic compounds (VOCs), formaldehyde, benzene series, odor and other harmful gases, and capture them through molecular adsorption principle. The UV sterilization module 55 is configured with a UVC ultraviolet lamp tube with a wavelength of 253.7 nanometers, which has strong sterilization and inactivation ability, and is used to effectively inactivate bacteria, viruses and molds in the air. Its radiation intensity can be adjusted by controlling the power-on time or driving current of the lamp tube to provide additional biological purification function under specific requirements. The negative ion generation module 56 generates negative oxygen ions by ionizing air at high voltage, which can adsorb and settle fine particles in the air, and also has the effect of freshening the air and improving human comfort. Its release concentration and working time can be accurately controlled by the system according to the purification strategy. Each module has an independent electrical control interface, which can be independently started, stopped or running intensity adjusted by the instructions issued by the adaptive control and optimization decision unit 40, for example, the running intensity of the HEPA module is directly related to the fan speed, and the start of the activated carbon module is highly related to the VOCs concentration, so as to realize flexible combination and targeted optimization operation, and avoid unnecessary energy consumption.

[0035] Further, the human-computer interaction and user preference setting unit 60 is the bridge for the user to interact with the system, which is designed to provide intuitive and comprehensive information display and control functions. This unit includes an integrated display screen 61, input buttons 62 or high-sensitivity touch screen 63, and an interface 64 for communication with external devices. The display screen 61 uses high-resolution color liquid crystal display technology to display real-time key data of the current indoor air quality (such as the real-time concentrations of PM2.5, CO2, VOCs, IAQ index, etc.), the current operating mode of the system (such as automatic mode, sleep mode, intensive purification mode), air volume level, remaining percentage of filter life, and historical data charts (such as the pollution concentration trend in the past 24 hours or 7 days), making the information easy to understand. The input buttons 62 or touch screen 63 provide multiple operation modes, allowing the user to manually adjust system parameters (such as air volume, start / stop of specific modules), switch operating modes, or set more personalized purification preferences. The communication interface 64 can use Bluetooth or Wi-Fi technology to allow the user to remotely monitor and control the air purification system through a mobile application (APP) on a smartphone or tablet, and set more detailed personalized purification strategies. The user preference settings can be as detailed as: priority weight configuration for different time periods (such as sleep period, work period, and away-from-home period) (for example, the highest priority for noise during the sleep period, the system will prefer to choose a low-noise mode; the highest priority for air quality during the work period, the system will prioritize efficient purification; the highest priority for energy consumption during the away-from-home period, the system can selectively disable some modules or enter energy-saving standby state); upper and lower thresholds for specific pollutant concentrations (such as the user can set the upper limit of PM2.5 to 25 μg / m³ and the upper limit of CO2 to 800 ppm); for users with allergic constitution, the system can set additional purification requirements for specific allergens such as pollen and dust mites, for example, automatically starting the high-efficiency particulate filter module and increasing air volume during the pollen season; and user preferences for system energy consumption and filter replacement period (such as the user may prefer to reduce energy consumption and accept slightly lower purification speed, or prefer to extend the service life of the filter). These detailed preference settings will serve as constraints or weight factors, transmitted to the optimization algorithm module 42 of the adaptive regulation and optimization decision unit 40 through the communication interface, thereby directly affecting the final strategy generation, achieving truly personalized intelligent purification.

[0036] Reference Figure 2 The adaptive regulation method of the intelligent sensing-based air purification system disclosed in the present application covers a complete closed loop from environmental sensing to purification execution, specifically including the following steps: S1: Real-time collection of multi-parameter of indoor environment by multi-dimensional environment perception unit 10. The multi-parameter is comprehensive, including PM2.5 concentration, CO2 concentration, VOCs concentration, environmental temperature, relative humidity, light intensity, and personnel presence state, etc. These data are continuously collected at high frequency (e.g. every 5 or 10 seconds), and after preliminary processing by the microcontroller unit 19 of the sub-perception node 17, they are transmitted to the central controller 20 through the wireless communication module 18.

[0037] S2: Transmission of the multi-parameter to the environmental data fusion and modeling unit 30. This unit first performs time stamping, format unification, missing value interpolation, outlier removal, and noise filtering on the data by the data acquisition and preprocessing module 31. Subsequently, the multi-source data fusion module 32 integrates data from different types of sensors and different perception nodes, such as through a Bayesian network model to fuse sensor data from different spatial locations, to construct a dynamic holographic picture of indoor air quality. The holographic picture not only contains the real-time concentration of pollutants in different areas of the room at different time points, but also includes their historical change trends and spatial distribution characteristics.

[0038] S3: Based on the dynamic holographic picture, the environmental modeling and pollution source identification module 33 in the environmental data fusion and modeling unit 30 identifies the type of indoor pollution source and its occurrence area, and predicts the diffusion trend of pollutant concentration in each area in the future. For example, when detecting that the PM2.5 concentration in the living room increases from 20 μg / m³ to 80 μg / m³ in a short time, accompanied by an increase in VOCs concentration, combined with outdoor AQI data and wind direction, the system may identify it as outdoor pollution penetration. At the same time, through the LSTM model, the system can predict that without intervention, the PM2.5 concentration will continue to rise to 120 μg / m³ in the next 30 minutes, and may spread to the adjacent dining room area.

[0039] S4: Transmission of the dynamic holographic picture, pollution source identification result, pollutant diffusion prediction, and user's personalized preference set through human-computer interaction and user preference setting unit 60 to adaptive regulation and optimization decision unit 40. These information are transmitted in real time through standardized data interfaces (such as MQTT or RESTful API).

[0040] S5: In the adaptive regulation and optimization decision unit 40, the intelligent decision engine 41 receives all input information and feeds it to the optimization algorithm module 42. The optimization algorithm module 42 uses a multi-objective optimization algorithm (such as a genetic algorithm or reinforcement learning algorithm) to balance multiple competing performance indicators such as indoor air quality targets (such as maintaining PM2.5 below 25 μg / m³), energy consumption (minimizing total system power), operating noise (maintaining within user acceptable range, such as below 30 dB(A) at night), and filter life (extending filter replacement cycle) based on the input information, to calculate the current optimal air purification strategy. The strategy describes in detail the activation state of the purification module (e.g. HEPA ON / OFF), fan operating speed (e.g. PWM duty cycle 60%), and operating intensity of each purification module (e.g. UV lamp power level, negative ion release concentration).

[0041] S6: The optimal air purification strategy is transmitted to the modular air purification execution unit 50 in the form of an instruction set. The instruction set contains specific operating parameters for each controllable module.

[0042] S7: The modular air purification execution unit 50 controls the fan unit 51 to operate at the set wind speed (by adjusting the BLDC motor speed) according to the strategy, and selectively starts or adjusts the operating intensity of the primary efficiency filter module 52, high-efficiency particulate matter filter module 53, activated carbon adsorption module 54, ultraviolet sterilization module 55, and negative ion generation module 56 according to the strategy, thereby achieving dynamic and precise purification of indoor air. For example, when the strategy indicates rapid removal of VOCs, the fan will be raised to medium-high speed, the activated carbon adsorption module will be operated at full power, and the UV module may remain closed.

[0043] S8: The system performs S1 to S7 in a cycle to achieve continuous monitoring of indoor air quality and real-time adaptive regulation of air purification system operating parameters. The entire cycle period can be set to tens of seconds to minutes according to system response requirements, ensuring rapid response of the system to environmental changes and continuous optimization of the strategy.

[0044] Further, in S3, the pollution source identification can be achieved by establishing a correlation model between different pollution sources and different pollution concentration variation patterns. The model can be based on machine learning algorithms, such as training the collected sensor data (concentration, rate of change of PM2.5, CO2, VOCs, presence of people, etc.) using classifiers (such as Support Vector Machine SVM or Random Forest). Through a large amount of historical data, the system can learn and identify the "fingerprint" of typical pollution events such as cooking activities (usually accompanied by rapid rise of PM2.5 and VOCs), gathering of people (rapid rise of CO2), external pollution penetration (simultaneous rise of PM2.5 and CO2, but the CO2 rise rate is not related to people), cleaning activities (short-term rise of VOCs), etc. The pollution diffusion trend prediction can use advanced time series prediction models, such as based on Recurrent Neural Network (RNN) or its variant Long Short-Term Memory Network (LSTM), combined with historical environmental data, current environmental state and external influencing factors (such as outdoor air quality index, wind speed and direction in weather forecast, humidity, etc.), to predict the concentration variation trend of key pollutants such as PM2.5, CO2, VOCs in different areas in the next few minutes to hours, providing data support for the system to take proactive measures.

[0045] As a preferred regulation method of the present application, in S5, the optimization target of the multi-objective optimization algorithm is multi-dimensional, which usually includes: maintaining PM2.5, CO2, VOCs concentration within the safe range set by the user or national standard (for example, PM2.5 below 25 μg / m³, CO2 below 800 ppm, TVOC below 0.3 mg / m³); minimizing the total energy consumption of the system (for example, by optimizing the fan speed and module start-stop); minimizing the noise generated by the system during operation (for example, controlling the noise below 30 dB(A) at night); maximizing the service life of the filter material (avoiding excessive wear and unnecessary full-power operation); and maximizing user comfort (by balancing air quality, noise, temperature and humidity, etc.). The algorithm quantitatively evaluates these objective functions in each iteration, and finds a solution that satisfies all the constraints and makes the weighted objective function optimal through optimization strategies. The weighting factors are provided by the user preference setting unit 60 to reflect the user's priority weight of different performance indicators. For example, when the user sets "sleep mode", the weight factor of noise will be significantly higher than that of energy consumption and purification speed, and the system will prefer to choose a low-noise operation scheme; when the user sets "fast purification mode", the weight factor of purification efficiency will be significantly improved, and the system will prefer to adopt a high-air-volume, multi-module collaborative operation scheme to reach the target concentration as soon as possible.

[0046] The intelligent perception-based air purification system and the adaptive regulation method thereof provided by the present application overcome the inherent limitations of the single-point perception, passive response and "one-size-fits-all" control strategy in the prior art through multi-dimensional environmental perception, intelligent data fusion and prediction, and multi-objective optimization decision-making capabilities. The present application realizes fine and dynamic management of indoor air quality, can accurately identify pollution sources and predict their diffusion trend, so as to take more accurate and forward-looking purification measures. Such a strategy not only significantly improves the efficiency and effectiveness of air purification, but also intelligently balances the relationship between purification performance and key performance indicators such as energy consumption, operating noise, etc. according to actual needs and user preferences, avoiding unnecessary resource waste, prolonging the service life of equipment and filter materials, and ultimately providing users with a more personalized, comfortable and healthy indoor breathing environment. The system modules interact with each other through standardized communication protocols, ensuring the scalability and modular update capability of the system. The central controller 20 can transmit data and issue instructions to each sub-perception node 17 and modular air purification execution unit 50 through wired (such as Ethernet, RS485) or wireless (such as Wi-Fi, Zigbee) methods, forming an efficient and collaborative intelligent network.

[0047] The advantages of the present application are further illustrated by a specific example and comparative examples as follows: Example: Performance of the intelligent perception system in the scenario of night pollution outbreak Suppose a bedroom in a residential area of a city, with an area of 15 square meters, a height of 2.8 meters, and a total volume of 42 cubic meters. The bedroom is equipped with the intelligent perception air purification system of the present application. The user has set the "sleep mode" through the human-computer interaction and user preference setting unit 60: the priority weight is noise > air quality > energy consumption, and the PM2.5 threshold is 25 μg / m³, the CO2 threshold is 800 ppm, and the night noise target is lower than 30 dB(A).

[0048] At 2 a.m. in the night, the system is in a standby or low wind volume running state with low power consumption and low noise. At this time, the PM2.5 sensor 11 in the bedroom reads 15 μg / m³, the CO2 sensor 12 reads 700 ppm, the VOCs sensor 13 reads 80 ppb, and the personnel presence sensor 16 shows that there is personnel activity. Due to urban industrial pollution, the PM2.5 concentration begins to rise rapidly and penetrates into the indoor through the not completely sealed window gap.

[0049] At 2:05, the PM2.5 concentration in the bedroom starts to slowly rise from 15 μg / m³ to 25 μg / m³. The environmental data fusion and modeling unit 30 detects this change and, in combination with external weather data and historical data, identifies it as an "outdoor PM2.5 pollution infiltration" event and predicts that, without intervention, the PM2.5 concentration will reach 50 μg / m³ in the next 30 minutes.

[0050] The adaptive regulation and optimization decision unit 40, based on this prediction and the user's "sleep mode" preference (noise priority highest), immediately initiates a multi-objective optimization algorithm. The algorithm weighs the purification efficiency, noise, and energy consumption. Due to the high noise priority, the system does not immediately adjust the fan to the highest speed. It calculates the optimal strategy as: The fan unit 51 operates at medium-low speed (PWM duty cycle from 20% to 45%), ensuring that the noise increases to an acceptable 35 dB(A).

[0051] The high-efficiency particulate filter module 53 is fully activated, ensuring maximum PM2.5 removal efficiency.

[0052] The primary efficiency filter module 52, activated carbon adsorption module 54, ultraviolet sterilization module 55, and negative ion generation module 56 remain dormant or operate at low power to save energy and reduce unnecessary noise or interference.

[0053] This strategy is expected to reduce the PM2.5 concentration to below 20 μg / m³ within 30 minutes and maintain it within the target threshold thereafter.

[0054] The system executes this strategy. At 2:35, 30 minutes after the strategy is executed, the PM2.5 concentration successfully drops to 20 μg / m³ and is stably maintained in subsequent operation. During this process, the average noise level in the bedroom rises from 28 dB(A) in standby mode to 35 dB(A) for 30 minutes, but does not reach the user's set unacceptable noise upper limit. The total energy consumption increases by about 5 watt-hours in this 30-minute period. The system achieves air quality protection while considering user sleep quality (noise control) and energy efficiency.

[0055] Comparative Example: Performance of a Traditional Fixed-Speed Air Purifier in the Same Scenario Suppose the same bedroom is equipped with a traditional fixed-speed air purifier without intelligent sensing and adaptive regulation functions. This purifier usually provides low, medium, and high speed, and cannot sense specific pollution sources and user preferences.

[0056] Comparative Example 1: Traditional system (fixed medium mode) If the user sets it to fixed medium mode, this mode is normally low noise (about 40 dB(A)) and moderate energy consumption. When PM2.5 rises from 15 μg / m³ to 25 μg / m³, the system runs at fixed medium, and its purification efficiency may not be sufficient to quickly respond to the penetration of external pollution. The PM2.5 concentration may continue to rise to 40 μg / m³ or even higher after 30 minutes, resulting in long-term substandard indoor air quality and affecting user health. It will take longer to eventually reduce PM2.5 to the target value, and during this period the user may have already noticed the decline in air quality.

[0057] Comparative Example 2: Traditional system (user manually adjusts to high mode) If the user is awakened by discomfort (such as difficulty breathing) in their sleep, or discovers that PM2.5 has risen by checking their smartphone, they manually adjust the purifier to high mode. High mode is usually above 55 dB(A) in noise and significantly increases energy consumption. Although the purification speed is fast, the high noise will seriously interfere with sleep, and after the concentration of pollutants decreases, the user may forget to adjust the purifier back to low mode because they fall asleep again, resulting in long-term noise and energy waste, and accelerating the wear and tear of the filter.

[0058] Data Comparison Table The following table quantitatively compares the key performance indicators of the embodiments of the present application and the two comparative examples in the above-mentioned scenario of sudden PM2.5 pollution at night:

[0059] From the comparison of the above examples and comparative examples, it can be seen that the air purification system based on intelligent sensing and its adaptive control method of the present application can significantly improve the adaptability and operating efficiency of the system in complex and variable environments. It avoids the disadvantages of passive response or over-response of traditional systems through precise sensing, intelligent prediction and multi-objective optimization, and achieves a dynamic balance between purification efficiency, energy consumption, noise and user comfort, providing a more intelligent and personalized indoor air management solution for users.

[0060] The application can be further extended in function and application range. For example, a voice recognition module is integrated in the human-computer interaction and user preference setting unit 60, allowing the user to control the system and set preferences through natural language instructions, such as "purifier, night mode, reduce noise". The system can also interconnect with a smart home platform to realize linkage control with other smart devices, such as automatically triggering the smart window to close or opening the whole house fresh air system when the air quality deteriorates. The environment data fusion and modeling unit 30 can also introduce external data sources, such as real-time weather data, city air quality monitoring station data, and pollen forecast, to further improve the accuracy of pollution source identification and pollutant diffusion prediction. The data acquisition and preprocessing module 31 can use edge computing technology to perform preliminary data processing and part of the decision logic at the sub-sensing node 17 or the central controller 20, to reduce the dependence on cloud services, reduce communication delay, and improve the real-time response capability of the system. The various sensors, controllers and actuators used in the application are mature and reliable technical components in the industry. Through innovative system architecture and intelligent control algorithms, they are integrated to achieve performance improvement far beyond the existing technology, providing a solid technical foundation for building a healthy, comfortable and intelligent living space.

[0061] The basic principles, main features and advantages of the application are shown and described above. Those skilled in the art should understand that the application is not limited by the above examples, and the above examples and descriptions in the specification are only to illustrate the principles of the application. Without departing from the spirit and scope of the application, various changes and improvements can be made to the application, and these changes and improvements all fall within the scope of the claimed application. The scope of protection of the application is defined by the appended claims and their equivalents.

Claims

1. An intelligent sensing based air purification system, characterized in that, The application relates to an indoor air purification system, comprising: a multi-dimensional environment perception unit (10) for collecting multi-element parameters of an indoor environment in real time; a central controller (20) for integrating part of the perception function and managing data transmission and instruction distribution; an environment data fusion and modeling unit (30) connected with the multi-dimensional environment perception unit (10) and used for receiving and processing the multi-element parameters, constructing a dynamic holographic picture of indoor air quality, identifying the type of a pollution source and the area where the pollution source occurs, and predicting the diffusion trend of pollutants; an adaptive regulation and optimization decision unit (40) connected with the environment data fusion and modeling unit (30) and used for generating an optimized air purification strategy according to the dynamic holographic picture, the pollution source identification result, the diffusion prediction of pollutants and preset user preferences; a modular air purification execution unit (50) connected with the adaptive regulation and optimization decision unit (40) and used for executing corresponding purification operations according to the optimized strategy; a man-machine interaction and user preference setting unit (60) connected with the adaptive regulation and optimization decision unit (40) and used for providing user setting preferences and receiving system state information.

2. The smart sensing based air purification system as claimed in claim 1, wherein, The multi-dimensional environment perception unit (10) comprises a PM2.5 sensor (11), a carbon dioxide sensor (12), a volatile organic compound sensor (13), a temperature and humidity sensor (14), an illumination intensity sensor (15) and a personnel presence sensor (16); the PM2.5 sensor (11) adopts a laser scattering principle and is used for detecting the concentration of particulate matters with a diameter less than 2.5 microns in the air; the carbon dioxide sensor (12) adopts a non-dispersive infrared principle and is used for detecting the volume concentration of carbon dioxide in the air; the volatile organic compound sensor (13) adopts a metal oxide semiconductor technology and is used for detecting the total amount of various volatile organic compounds in the air and outputting the total volatile organic compound concentration or air quality index; the temperature and humidity sensor (14) adopts a capacitive or resistive sensing principle and is used for detecting the environmental temperature and relative humidity; the illumination intensity sensor (15) adopts a photodiode array and is used for detecting the environmental illumination; the personnel presence sensor (16) adopts a passive infrared principle or a millimeter wave radar principle and is used for detecting whether there is personnel activity in a specified area, the detection distance is 5-10 meters, and the detection angle is 120 degrees.

3. The smart sensing based air purification system as claimed in claim 2, wherein, The sensors of the multi-dimensional environment perception unit (10) are distributedly arranged and comprise a plurality of sub-perception nodes (17), each of the sub-perception nodes (17) integrates part or all types of sensors and transmits data to the central controller (20) through a wireless communication module (18); each of the sub-perception nodes (17) is embedded with a microcontroller unit (19) for collecting, calibrating and preliminarily filtering sensor raw data and packaging the processed data to send to the environment data fusion and modeling unit (30) through the wireless communication module (18).

4. The smart sensing based air purification system as claimed in claim 1, wherein, The environment data fusion and modeling unit (30) comprises: The data acquisition and preprocessing module (31) is configured to receive raw sensor data from the multi-dimensional environment perception unit (10), and perform time stamping, data format unification, missing value interpolation, outlier rejection, and noise filtering operations, wherein the noise filtering adopts a Kalman filtering algorithm or a sliding average filtering algorithm. The multi-source data fusion module (32) is configured to integrate processed data from different types of sensors and different perception nodes. The environment modeling and pollution source identification module (33) is configured to construct a dynamic model of indoor air quality based on the fused multi-dimensional environment data, wherein the dynamic model adopts a machine learning-based method, including a recurrent neural network or a long short-term memory network, to capture the time series variation law of air pollutant concentration; the pollution source identification is performed by analyzing the correlation of different pollutant concentration variations and the correlation between the pollutant concentration variations and factors such as personnel activities and external environment, to infer the source of the pollutants; and the environment modeling and pollution source identification module (33) is also responsible for pollution diffusion trend prediction, which can be based on historical data, a current airflow model, and a machine learning prediction model, including a support vector regression or a Gaussian process regression, to predict the pollutant concentration variation of each region in a future period of time.

5. The smart sensing based air purification system as claimed in claim 1, wherein, The adaptive regulation and optimization decision unit (40) includes: An intelligent decision engine (41) configured to receive the environment model, the pollution source identification result, and the pollution diffusion prediction data from the environment data fusion and modeling unit (30), and obtain user preference settings provided by the human-computer interaction and user preference setting unit (60); An optimization algorithm module (42) embedded with a multi-objective optimization algorithm, which can adopt a heuristic search-based method, including a genetic algorithm or a particle swarm optimization, or a reinforcement learning-based method; the optimization algorithm module (42) defines a multi-dimensional objective function, the input parameters of which include the current indoor air quality state, the predicted future air quality trend, the operation cost of the purification system, the purification efficiency parameters of different purification modules, and the user preference factor, and the objective function aims to minimize the total cost while ensuring that the air quality indicators are maintained within the safety range of the user settings or national standards; A purification strategy generation module (43) configured to select the optimal combination from a series of candidate operation parameter combinations generated by the optimization algorithm module (42) as the air purification strategy at the current time, and the strategy generation module (43) also considers the dynamic response characteristics and delay of the system, and introduces minimum operation time or minimum downtime constraints.

6. The smart sensing based air purification system as claimed in claim 1, wherein, The modular air purification execution unit (50) includes: A fan unit (51) adopting a direct-current brushless motor-driven centrifugal fan or axial flow fan, the rotational speed of which can be steplessly or multi-staged adjusted through a pulse width modulation signal to realize accurate control of air volume; a primary efficiency filtration module (52) configured to remove larger particulate matters in the air, and the filter material is polyester fiber or nylon net; A high-efficiency particulate matter filtration module (53) configured to remove fine particulate matters; An activated carbon adsorption module (54) adopting high-iodine-value coconut shell activated carbon or coal-based activated carbon, and configured to adsorb volatile harmful gases. An ultraviolet sterilization module (55) is configured with ultraviolet light tubes with a wavelength of 253.7 nanometers, which are used to inactivate bacteria, viruses, and molds in the air; A negative ion generation module (56) generates negative oxygen ions by ionizing air at high voltage, which is used to adsorb and settle fine particulate matter in the air; Each module has an independent control interface, which can be independently started, stopped, or intensity-adjusted through the instructions issued by the adaptive control and optimization decision unit (40).

7. The smart sensing based air purification system as claimed in claim 1, wherein, The human-computer interaction and user preference setting unit (60) includes a display screen (61), input keys (62), or a touch screen (63), and a communication interface (64); The display screen (61) is used to display real-time indoor air quality data, system operation mode, air volume level, filter life reminder, and historical data chart information; The input keys (62) or the touch screen (63) are used for user manual adjustment of system parameters, switching of operation modes, or setting of personalized preferences; The communication interface (64) uses Bluetooth technology, allowing users to remotely monitor and control the air purification system through a mobile application; The user preference setting specifically includes priority weight configuration for different time periods, upper and lower threshold values for specific pollutant concentrations, additional purification requirements for specific allergens for users with allergic constitution, and preference settings for system energy consumption and filter replacement cycle.

8. An adaptive regulation method of an intelligent sensing-based air purification system, applicable to the intelligent sensing-based air purification system of any one of claims 1-7, characterized in that, The steps include: S1: Real-time collection of multi-element parameters of indoor environment through a multi-dimensional environment perception unit (10), including but not limited to PM2.5 concentration, carbon dioxide concentration, volatile organic compound concentration, environmental temperature, relative humidity, light intensity, and personnel presence status; S2: Transmission of the multi-element parameters to an environmental data fusion and modeling unit (30), which pre-processes the data, fuses multi-source data, and constructs a dynamic holographic scene of indoor air quality, including pollutant distribution in different areas and at different time points; S3: Based on the dynamic holographic scene, the environmental data fusion and modeling unit (30) identifies the type and location of indoor pollution sources and predicts the diffusion trend of pollutant concentrations in each area; S4: Transmission of the dynamic holographic scene, pollution source identification results, pollutant diffusion prediction, and user preset preferences to an adaptive control and optimization decision unit (40); S5: In the adaptive control and optimization decision unit (40), a multi-objective optimization algorithm is used to balance the performance indicators of indoor air quality, energy consumption, operating noise, and filter life based on the input information, and to calculate and generate the current optimal air purification strategy, which includes the activation state of the purification module, the fan operating speed, and the operating intensity of each purification module; S6: Transmission of the optimal air purification strategy to a modular air purification execution unit (50); S7: The modular air purification execution unit (50) controls the fan unit (51) to operate at a set air speed and selectively activates or adjusts the operation intensity of the primary filter module (52), the high-efficiency particulate matter filter module (53), the activated carbon adsorption module (54), the ultraviolet sterilization module (55), and the negative ion generation module (56) according to the strategy, to achieve purification of indoor air; S8: S1 to S7 are executed cyclically to achieve continuous monitoring of indoor air quality and real-time adaptive control of air purification system operation parameters.

9. The self-adaptive regulation method of the intelligent perception-based air purification system according to claim 8, wherein, In S3, the pollution source identification is achieved by establishing an association model between different pollution sources and different pollution concentration change patterns. The association model can be based on a machine learning algorithm, including training sensor data using a classifier to identify typical pollution events. 10.The self-adaptive regulation method of the intelligent perception based air purification system according to claim 8, wherein, In S5, the optimization objectives of the multi-objective optimization algorithm include: Maintaining PM2.5, carbon dioxide, and volatile organic compound concentrations within the safety range set by the user or national standards; Minimizing total system energy consumption; Minimizing noise generated during system operation; Maximizing filter service life; Maximizing user comfort; The multi-objective optimization algorithm evaluates the objective function in each iteration and finds a solution that satisfies all constraints and optimizes the weighted objective function through an optimization strategy. The weighting factors are provided by the user preference setting unit (60) to reflect the user's priority weight for different performance indicators.

Images