Air purifier air speed control method and device based on indoor crowd density change

Abstract

The embodiment of the invention discloses an air purifier air speed control method and device based on indoor crowd density changes, and belongs to the technical field of smart home. The method comprises the following steps: collecting crowd perception data and environment data including infrared and wireless signals and visual images; calculating a current crowd density index through a weighted fusion algorithm; in combination with the current air quality index, the future air quality is estimated through the prediction model; according to the crowd density change rate and the time information, a gathering, sleeping or conventional operation mode is identified, and mode switching needs to meet a lag condition; and finally, the target wind speed is determined by integrating the operation mode, the crowd density index and the predicted air quality, and the wind speed is dynamically adjusted by adopting a closed-loop control algorithm. According to the method, crowd density perception and air quality prediction are fused, so that advanced response to the pollution trend and scene-adaptive wind speed optimization control are realized, the purification hysteresis is effectively improved, and the purification effect, energy consumption and noise are collaboratively optimized.

Description

Technical Field

[0001] This application relates to the field of smart home technology, and in particular to a method and device for controlling the fan speed of an air purifier based on changes in indoor population density. Background Technology

[0002] With the improvement of living standards and the increasing demand for health management, air purifiers have been widely used in various indoor environments. Currently, most air purification devices on the market still use a feedback control mode based on fixed air quality sensors, adjusting fan speed according to real-time detection of PM2.5, CO2, VOCs, and other indicators. While this method is effective when environmental parameters change relatively slowly, its control logic is essentially a lagging "pollution occurrence-detection-response" model. In real-world scenarios with high population mobility and dynamic changes in gathering levels, this control method reveals significant shortcomings. On the one hand, when people gather rapidly and the pollutant release rate increases sharply (such as at the beginning of family gatherings or meetings), the purifier cannot predict this in advance and only increases the fan speed after the air quality deteriorates, resulting in a significant lag in purification response and indoor air quality remaining at a low level for an extended period. On the other hand, during periods of low population density or stillness (such as nighttime sleep), because the air quality sensor may not detect significant pollution, the device may still maintain a high fan speed, not only generating unnecessary energy consumption but also causing continuous noise interference, affecting users' rest. Summary of the Invention

[0003] This specification provides an air purifier fan speed control method and device based on changes in indoor crowd density, in order to solve at least one of the technical problems mentioned above.

[0004] To solve the above-mentioned technical problems, the embodiments in this specification are implemented as follows: According to a first aspect of the embodiments of this specification, an air purifier fan speed control method based on changes in indoor crowd density is provided, comprising: S1: Collect multi-source sensor data, which includes: crowd perception data for estimating indoor crowd density, environmental data collected by an air quality sensor, time information, and temperature and humidity information; wherein, the crowd perception data includes at least infrared heat source data, WiFi or Bluetooth signal data, and visual image data. S2: Based on the crowd perception data, the current time is calculated using a weighted fusion algorithm. Population density index ; S3: Based on the aforementioned population density index and the current air quality index The air quality index for future moments is predicted using a predictive model. S4: Based on the aforementioned population density index The rate of change and the time information are used to identify the current operating mode of the target air purifier; the operating mode includes at least party mode, sleep mode and normal mode; wherein, switching from one mode to another must meet a preset lag condition; S5: Based on the identified operating mode and the population density index Based on the predicted air quality index, the target wind speed is determined. Based on the target wind speed Based on the deviation from the current wind speed, a closed-loop control algorithm is used to dynamically adjust the wind speed of the target air purifier.

[0005] In some optional implementations, in step S2, the calculation formula of the weighted fusion algorithm is: in, The infrared heat source count is obtained based on the infrared heat source data. The number of connected devices is obtained based on the WiFi or Bluetooth signal data. This is a visual activity intensity index obtained based on the aforementioned visual image data. , , These are weighting coefficients determined based on historical data.

[0006] In some optional implementations, in step S3, the prediction model is: in, Population density index rate of change, , , , These are the model coefficients. The error term is defined as follows: the predicted air quality index for the future time is... .

[0007] In some optional implementations, step S4, identifying the current operating mode of the target air purifier, specifically includes: If the rate of change of the population density index Greater than the first threshold If so, it is identified as a party mode; If the rate of change of the population density index Less than the second threshold If the time information falls within a preset nighttime period, it is identified as a sleep mode; Otherwise, it will be identified as the normal mode.

[0008] In some alternative implementations, in step S5, the target wind speed is determined. The process employs a mode-dependent optimization objective function; in party mode, the objective function prioritizes maximizing the air exchange rate; in sleep mode, the objective function prioritizes minimizing noise and energy consumption; and in normal mode, the objective function balances air quality and energy consumption.

[0009] In some optional implementations, in step S5, the closed-loop control algorithm is a fuzzy PID control algorithm, and its wind speed update formula is: Among them, symbols Indicates air quality deviation. ,symbol , , These are the proportional, integral, and derivative control parameters that are dynamically adjusted according to the operating mode, with the following symbols: Represents the wind speed at the current moment. This represents the target wind speed to be set at the next moment.

[0010] In some optional implementations, between steps S4 and S5, a step of fine-tuning the wind speed based on fuzzy logic is also included, including: Population density change rate Current Air Quality Index and the current time period As fuzzy input variables, each fuzzy input variable is fuzzified, and its corresponding precise value is converted into a fuzzy linguistic value. Based on a pre-set fuzzy rule base, and combined with the fuzzy linguistic values ​​of each fuzzy input variable, inference is performed to output the wind speed adjustment amount. The fuzzy value; The weighted average method is used for defuzzification to obtain the accurate wind speed adjustment. Used to measure the target wind speed Make corrections.

[0011] In some optional implementations, on the cloud server, based on a reinforcement learning framework, the control model is optimized using uploaded data, and the model parameters are updated, specifically including: Define state space The state space Including the current air quality index The population density index The aforementioned operating mode, historical wind speed data, and energy consumption and noise ; Define action space It is the wind speed adjustment amount. The set of all possible values; Define reward function The calculation formula is as follows: Among them, symbols Indicates the rate of improvement in air quality. , , These are the weighting coefficients; The agent is trained using a deep deterministic gradient algorithm to update the model parameters. , , , , To optimize control strategies The update formula for the depth deterministic policy gradient algorithm is as follows: Among them, symbols The action value function output by the Critic network, with the symbol... The action value function output by the target Critic network is represented by the symbol. The policy function output by the Actor network, symbol The symbol represents the policy function output by the target Actor network. Indicates time The system state, symbol For a moment The action performed, symbol Indicates time The system state, symbol Indicates time Execute action The instant reward obtained afterward, symbol Discount factor, symbol The learning rate of the Critic network is represented by the symbol. The learning rate of the Actor network is represented by the symbol. The gradient operator is represented by the symbol. Indicates the number of samples in a batch for training, symbol Indicates the first The target of each sample Value, symbol The parameter set of the Critic network is represented by the symbol. This represents the set of parameters for the Actor network.

[0012] In some optional implementations, the method further includes: The multi-source sensor data and the crowd density index The operating mode and the target wind speed The system uploads actual wind speed and energy consumption data to the cloud server and receives optimized model parameters from the cloud server to update the local control strategy.

[0013] According to a second aspect of the embodiments of this specification, an air purifier fan speed control device based on changes in indoor crowd density is provided, comprising: The data acquisition module is used to acquire multi-source sensor data, which includes: crowd perception data for estimating indoor crowd density, environmental data acquired by an air quality sensor, time information, and temperature and humidity information; wherein, the crowd perception data includes at least infrared heat source data, WiFi or Bluetooth signal data, and visual image data. The crowd density index calculation module is used to calculate the current time based on the crowd perception data using a weighted fusion algorithm. Population density index ; An air quality prediction module is used to predict air quality based on the population density index. and the current air quality index The air quality index for future moments is predicted using a predictive model. The wind speed control module is used to control the wind speed according to the crowd density index. The rate of change and the time information are used to identify the current operating mode of the target air purifier; the operating mode includes at least party mode, sleep mode and normal mode; wherein, switching from one mode to another must meet a preset lag condition; The wind speed dynamic adjustment module is used to adjust the wind speed according to the identified operating mode and the population density index. Based on the predicted air quality index, the target wind speed is determined. Based on the target wind speed Based on the deviation from the current wind speed, a closed-loop control algorithm is used to dynamically adjust the wind speed of the target air purifier.

[0014] One embodiment of this specification can achieve at least the following beneficial effects: This application's technical solution estimates indoor crowd density in real time by integrating multi-source sensing data such as infrared, visual, and wireless signals, and constructs a coupled predictive model of crowd dynamics and pollution trends by combining air quality sensor data. This solution overcomes the lag limitation of traditional air purifiers that rely solely on reactive control based on current air quality parameters. By identifying typical operating modes such as gatherings, sleep, and routines based on crowd density change rates and time information, and using differentiated objective functions to determine wind speed accordingly, it achieves a shift in control strategy from "post-pollution response" to "pre-pollution prevention." This predictive wind speed control mechanism based on crowd density perception and pattern recognition can proactively increase purification intensity before pollutant levels rise rapidly due to crowd gatherings, and actively reduce operating noise and energy consumption during quiet periods. Thus, in dynamically changing usage scenarios, it synergistically optimizes air quality assurance, user comfort, and equipment energy efficiency. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of an air purifier fan speed control method based on changes in indoor crowd density, provided in the embodiments of this specification. Figure 2 Corresponding to Figure 1 A schematic diagram of the structure of an air purifier fan speed control device based on changes in indoor crowd density. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of one or more embodiments of this specification clearer, the technical solutions of one or more embodiments of this specification will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of them. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of one or more embodiments of this specification.

[0018] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another.

[0019] This application provides a method for controlling the fan speed of an air purifier based on changes in indoor population density, such as... Figure 1 As shown, the method may include the following steps: Step S1: Collect multi-source sensor data, which includes: crowd perception data for estimating indoor crowd density, environmental data collected by an air quality sensor, time information, and temperature and humidity information; wherein, the crowd perception data includes at least infrared heat source data, WiFi or Bluetooth signal data, and visual image data.

[0020] In this solution, crowd perception data provides raw signals for indirectly but in real-time estimation of the number, distribution, and activity status of people indoors. This type of data can be inferred indirectly through at least three different sensing technologies, including infrared thermal source data (using infrared sensors to detect the distribution of heat radiated by living organisms), WiFi or Bluetooth signal data (monitoring the number of active devices connected to wireless access points or changes in signal strength), and visual image data (capturing environmental images through image sensors, which can be used for subsequent analysis). Environmental data can be directly collected by dedicated air quality sensors and can include concentration values ​​of pollutants such as PM2.5, carbon dioxide, and volatile organic compounds. This data can directly characterize the current air cleanliness level. Time information can be used to identify different times of day (e.g., daytime, nighttime). Temperature and humidity information can be collected by temperature and humidity sensors, and this data reflects basic physical environmental parameters.

[0021] Step S2: Based on the crowd perception data, calculate the current time using a weighted fusion algorithm. Population density index .

[0022] In this application, considering the limitations and uncertainties that may exist in estimating the number or density of people using a single type of sensor—for example, infrared sensors may not be able to distinguish stationary individuals, wireless signals may be interfered with by factors not carried by people, and visual data may fail in insufficient lighting or privacy conditions—this weighted fusion algorithm, to comprehensively process such multi-source heterogeneous data, first preprocesses and extracts features from the raw crowd perception data from each source (e.g., infrared heat source distribution, wireless device connection status, visual image sequences), transforming them into standardized or normalized intermediate indicators that reflect the presence or activity intensity of people. Subsequently, a specific weight coefficient is assigned to each of these intermediate indicators, reflecting the reliability and contribution of the data source to the estimation of real crowd density under the current environment and algorithm model. Finally, the algorithm performs linear or nonlinear combination operations on these weighted intermediate indicators, and the direct output is the crowd density index. .

[0023] Step S3: Based on the population density index and the current air quality index The air quality index is predicted for future moments using a predictive model.

[0024] This plan takes into account that the deterioration of indoor air quality is often not instantaneous, but rather a cumulative and diffusion process influenced by both the intensity of pollution sources (positively correlated with the number of people and their activities) and air conditions. To characterize this process, the prediction model uses the population density index... This is considered as a leading variable because the presence and activity of people are one of the main sources of indoor pollutants such as carbon dioxide, particulate matter, and volatile organic compounds. Additionally, the current air quality index is taken into account. This represents the initial state of the system, and the model can learn from historical data. , and subsequent The dynamic relationship between these factors allows for the capture of the lagged effects of population gathering on air quality and the inertia of pollutant concentration changes. Therefore, the output of this predictive model—the forecast of future air quality—no longer relies solely on current, deteriorated air quality readings, but rather anticipates potential pollution trends caused by current population density. This enables the system to adjust its operational strategies in advance based on the forecast results, before a perceptible decline in air quality actually occurs.

[0025] Step S4: Based on the population density index The rate of change and the time information are used to identify the current operating mode of the target air purifier; the operating mode includes at least party mode, sleep mode and normal mode; wherein, switching from one mode to another requires meeting a preset lag condition.

[0026] This solution intelligently classifies the operating scenarios of air purifiers based on the dynamic characteristics and temporal patterns of indoor crowd conditions. Specifically, it includes, for example, the crowd density index. When the rate of change shows a significant positive increase, it indicates that people are rapidly entering or activities are intensifying. This usually corresponds to scenarios such as social gatherings and meetings, and can be identified as "gathering mode". When the rate of change shows a negative or extremely low level, especially when combined with time information to determine that it is during the user's preset nighttime rest period, it indicates that the environment is becoming quiet, and can be identified as "sleep mode". For general cases that do not meet the above specific trend and time combination, they can be classified as "normal mode".

[0027] It should be noted that, to avoid frequent and unnecessary switching of operating modes within a short period due to brief personnel entry and exit or normal fluctuations in sensor data, which could cause repeated and sudden changes in wind speed affecting comfort and equipment lifespan, this solution incorporates a "preset hysteresis condition." This means that switching from one mode to another requires not only meeting the immediate criteria for triggering the new mode but also additional constraints such as the new mode state needing to persist for a certain period or the rate of change needing to stably exceed a threshold for a certain time. This design ensures the stability and reliability of operating mode recognition, enabling subsequent wind speed control decisions to be based on a relatively stable and reliable scenario, thus executing a coherent and reasonable control strategy.

[0028] S5: Based on the identified operating mode and the population density index Based on the predicted air quality index, the target wind speed is determined. Based on the target wind speed Based on the deviation from the current wind speed, a closed-loop control algorithm is used to dynamically adjust the wind speed of the target air purifier.

[0029] In this scheme, by integrating the high-level semantic information obtained from previous steps—namely, the operational mode representing the current scenario, the index quantifying population density, and the predicted future air quality index—an ideal target wind speed setpoint is calculated through an embedded decision-making mechanism. Then, the deviation between the target wind speed and the actual operating wind speed of the air purifier fan is calculated in real time. Based on the deviation, a control quantity is dynamically generated to adjust the drive signal of the fan motor. This closed-loop adjustment process continues, which can make the actual wind speed smoothly, stably and gradually approach the target value, and effectively suppress fluctuations caused by external interference or the system's own inertia.

[0030] This application's technical solution estimates indoor crowd density in real time by integrating multi-source sensing data such as infrared, visual, and wireless signals, and constructs a coupled predictive model of crowd dynamics and pollution trends by combining air quality sensor data. This solution overcomes the lag limitation of traditional air purifiers that rely solely on reactive control based on current air quality parameters. By identifying typical operating modes such as gatherings, sleep, and routines based on crowd density change rates and time information, and using differentiated objective functions to determine wind speed accordingly, it achieves a shift in control strategy from "post-pollution response" to "pre-pollution prevention." This predictive wind speed control mechanism based on crowd density perception and pattern recognition can proactively increase purification intensity before pollutant levels rise rapidly due to crowd gatherings, and actively reduce operating noise and energy consumption during quiet periods. Thus, in dynamically changing usage scenarios, it synergistically optimizes air quality assurance, user comfort, and equipment energy efficiency.

[0031] Based on the technical solutions described above, this specification also provides some specific implementation schemes, which are described below.

[0032] In an optional embodiment, in step S2, the calculation formula of the weighted fusion algorithm is: in, The infrared heat source count is obtained based on the infrared heat source data. The number of connected devices is obtained based on the WiFi or Bluetooth signal data. This is a visual activity intensity index obtained based on the aforementioned visual image data. , , These are weighting coefficients determined based on historical data.

[0033] In an optional embodiment, in step S3, the prediction model is: in, Population density index rate of change, , , , These are the model coefficients. The error term is defined as follows: the predicted air quality index for the future time is... .

[0034] In the technical solution of this embodiment It can directly reflect the number of human or animal targets with thermal radiation characteristics; This indicates the number of actively connected devices detected based on Wi-Fi or Bluetooth signals, serving as a supplementary indicator to indirectly infer the possible number of people. This refers to the visual activity intensity index calculated based on visual image data (such as camera footage) using image analysis algorithms (such as motion detection and target recognition). This index not only reflects the number of targets but also, to some extent, characterizes the intensity of human activity. , , These weighting coefficients correspond to the contributions of the three data sources to the final density estimation result. These weighting coefficients can be corrected and determined based on historical data accumulated during system operation, for example, through regression analysis or optimization algorithms, to ensure that the fused result... It best matches the actual population density, thereby improving the accuracy and robustness of the estimation.

[0035] Subsequently, in step S3, the population density index calculated in step S2 is used. Based on the air quality index and its changing trends, a linear regression model for air quality prediction was constructed, the expression of which is shown above. This model uses the air quality index at the previous time point as the basis for prediction. Based on this, the current population density index is introduced. As an explanatory variable, it quantifies the immediate impact of human presence on pollution load. The model also incorporates the rate of change of the population density index. This model is used to capture the leading influence of dynamic trends in population numbers, whether increasing or decreasing, on future changes in air quality. For constant terms, , , These are model coefficients, which can be obtained by training with historical air quality data and corresponding population density data. This represents the random error term that the model failed to explain.

[0036] In an optional embodiment, step S4, identifying the current operating mode of the target air purifier, may specifically include: If the rate of change of the population density index Greater than the first threshold If so, it is identified as a party mode; If the rate of change of the population density index Less than the second threshold If the time information falls within a preset nighttime period, it is identified as a sleep mode; Otherwise, it will be identified as the normal mode.

[0037] In the technical solution of this embodiment, the crowd density index is calculated in real time. rate of change over time And, combined with current time information, perform pattern discrimination based on preset thresholds and time period conditions. If the population density index change rate... Greater than the first threshold If the population density index changes significantly, it indicates a substantial increase in the number of people indoors within a short period, which is classified as "gathering mode." This mode typically corresponds to scenarios with large gatherings and frequent activities, where air quality may deteriorate rapidly. Therefore, air purifiers are required to operate at higher fan speeds to enhance air exchange and pollutant removal capabilities. If the population density index changes... Less than the second threshold If the current time falls within a preset nighttime period (e.g., 10:00 PM to 7:00 AM the next day), it is identified as "Sleep Mode." This mode is suitable for environments where people are at rest and sensitive to noise. The control objective focuses on reducing fan operating noise and energy consumption while maintaining a basic level of air quality. If neither of the above two conditions is met, it is identified as "Normal Mode," which is suitable for general daily activities. In this mode, control can balance air quality maintenance and energy efficiency.

[0038] In an optional embodiment, step S5 involves determining the target wind speed. The process employs a mode-dependent optimization objective function. In party mode, the objective function prioritizes maximizing the air exchange rate; in sleep mode, the objective function prioritizes minimizing noise and energy consumption; and in normal mode, the objective function balances air quality and energy consumption.

[0039] In this embodiment, when the indoor environment is identified as "gathering mode," since this mode corresponds to a scenario where people gather, activities are frequent, and the pollutant generation rate may increase sharply, the optimization objective function will prioritize maximizing the air exchange rate. This means that, under the guidance of this objective function, when calculating the target wind speed, priority will be given to how to quickly dilute and remove pollutants such as carbon dioxide, particulate matter, and volatile organic compounds generated by the increased number of people, in order to rapidly curb the trend of air quality deterioration. The tolerance for increased energy consumption and noise caused by high wind speed operation is relatively high at this time. When identified as "sleep mode," corresponding to an environment where people are resting at night, are sensitive to noise, and pollutant release is significantly reduced, the optimization objective function will prioritize minimizing noise and energy consumption. Under the guidance of this function, the primary constraint in determining the target wind speed is to ensure that the sound pressure level generated by the fan operation is below an acceptable range to avoid disturbing sleep, while minimizing power consumption. The air quality maintenance target is set at a relatively relaxed baseline level suitable for sleep, rather than pursuing the fastest purification speed. In the "normal mode", it is suitable for daily situations with a stable number of people and moderate activity. Its optimization objective function takes into account both air quality and energy consumption. That is, it does not simply pursue the maximization of air exchange rate or the minimization of energy consumption and noise, but seeks a dynamic balance between the two that is adapted to the real-time population density and air quality prediction values, in order to achieve the optimization of energy efficiency while maintaining acceptable indoor air quality and avoiding unnecessary energy waste.

[0040] In an optional embodiment, in step S5, the closed-loop control algorithm is a fuzzy PID control algorithm, and its wind speed update formula is: Among them, symbols Indicates air quality deviation. ,symbol , , These are the proportional, integral, and derivative control parameters that are dynamically adjusted according to the operating mode, with the following symbols: Represents the wind speed at the current moment. This represents the target wind speed to be set at the next moment.

[0041] In the technical solution of this embodiment, the proportional parameter The integral parameter directly affects the controller's immediate response speed to the current deviation. Cumulative value acting on historical bias This is used to eliminate persistent static errors, while the differential parameters Then based on the rate of change of the current deviation from the previous time step Adjustments are made to predict air quality trends and suppress system overshoot. Meanwhile, these three control parameters... , , It is not fixed, but dynamically adjusted according to the operating mode (party mode, sleep mode, normal mode) identified in step S4. For example, in "party mode" which requires a fast response, a larger set of parameters may be automatically used. and Parameters are used to improve the agility of wind speed adjustment and the predictability of pollution trends, while in the "sleep mode" that emphasizes stability and quietness, a smaller set of parameters may be used. and Parameters, and may be adjusted. The parameters are designed to achieve smooth changes in wind speed, avoiding noise interference caused by sudden acceleration or deceleration.

[0042] In an optional embodiment, between steps S4 and S5, a step of fine-tuning the wind speed based on fuzzy logic is further included, which may include: Population density change rate Current Air Quality Index and the current time period As fuzzy input variables, each fuzzy input variable is fuzzified, and its corresponding precise value is converted into a fuzzy linguistic value. Based on a pre-set fuzzy rule base, and combined with the fuzzy linguistic values ​​of each fuzzy input variable, inference is performed to output the wind speed adjustment amount. The fuzzy value; The weighted average method is used for defuzzification to obtain the accurate wind speed adjustment. Used to measure the target wind speed Make corrections.

[0043] In an optional embodiment, on the cloud server, based on a reinforcement learning framework, the control model is optimized and its parameters are updated using uploaded data. Specifically, this may include: Define state space The state space Including the current air quality index The population density index The aforementioned operating mode, historical wind speed data, and energy consumption and noise ; Define action space It is the wind speed adjustment amount. The set of all possible values; Define reward function The calculation formula is as follows: Among them, symbols Indicates the rate of improvement in air quality. , , These are the weighting coefficients; The agent is trained using a deep deterministic gradient algorithm to update the model parameters. , , , , To optimize control strategies The update formula for the depth deterministic policy gradient algorithm is as follows: Among them, symbols The action value function output by the Critic network, with the symbol... The action value function output by the target Critic network is represented by the symbol. The policy function output by the Actor network, symbol The symbol represents the policy function output by the target Actor network. Indicates time The system state, symbol For a moment The action performed, symbol Indicates time The system state, symbol Indicates time Execute action The instant reward obtained afterward, symbol Discount factor, symbol The learning rate of the Critic network is represented by the symbol. The learning rate of the Actor network is represented by the symbol. The gradient operator is represented by the symbol. Indicates the number of samples in a batch for training, symbol Indicates the first The target of each sample Value, symbol The parameter set of the Critic network is represented by the symbol. This represents the set of parameters for the Actor network.

[0044] In this embodiment, between steps S4 and S5, an optional step is provided for fine-tuning wind speed based on fuzzy logic. This step utilizes expert knowledge to refine the initially determined target wind speed. This step first selects three key fuzzy input variables: the real-time rate of change of the crowd density index. Current Air Quality Index and the specific time period at present. The precise numerical values ​​of each input variable (e.g., a change rate of +0.5 people / minute, an AQI of 85, and a time of 14:30) are mapped to a fuzzy set defined by natural language terms such as "low," "medium," "high," "daytime," and "nighttime" through a process called "fuzzification," and assigned corresponding membership degrees. This transforms the precise numerical input into fuzzy linguistic values ​​that the system can process. Subsequently, a pre-defined fuzzy rule base derived from expert experience or historical data (e.g., "if the population density increases rapidly and the air quality is poor, then significantly increase the wind speed") is invoked to combine the fuzzy linguistic values ​​of the above input variables. Through a fuzzy inference mechanism (such as the Mamdani inference method), the output variable, i.e., the wind speed adjustment amount, is calculated. The corresponding fuzzy linguistic values ​​and their membership degrees. Simultaneously, defuzzification methods such as the weighted average method can be used to convert the output fuzzy wind speed adjustment into a definite, precise numerical value. This fine-tuning amount will be used to adjust the target wind speed initially determined in step S5. The settings are then adjusted to produce the final wind speed setting command. Meanwhile, to further enhance the adaptive and optimization capabilities of this solution in long-term operation, this method introduces a reinforcement learning-based model parameter optimization framework at the cloud server level. This framework first defines several elements required for the agent to make decisions, namely the state space. Action space and reward function Among them, the state space It is a multi-dimensional vector that comprehensively describes the system's operating status at any given moment, specifically including the current air quality index. Population density index The identified operating mode, recent historical wind speed data, and current real-time energy consumption. and noise level Action space It is then defined as all possible wind speed adjustments. The set of possible values ​​represents the control operations that the agent can perform. Reward function. It is the criterion that guides the learning direction of the intelligent agent, and its calculation formula is: This function directly quantifies the overall effect of a single-step control action, meaning the positive reward comes from the rate of improvement in air quality. The negative penalty is related to the energy consumption of the action. With noise Cost, weighting factor Used to balance the priorities of different optimization objectives.

[0045] Within this framework, a deep deterministic policy gradient algorithm can be used to train the cloud agent to continuously optimize the control policy. And update key parameters involved in the local control model, such as weighting coefficients. and control parameters The DDPG algorithm, as a deep reinforcement learning algorithm with an actor-critic architecture, utilizes two deep neural networks: the Actor network (policy network). ) and Critic Network (Value Network) This is achieved through co-training. The algorithm's update process involves three formulas, the first of which... Used to calculate the target Q value, which represents the execution of the current action. The instant reward obtained afterwards The cumulative reward expected for the next state after the discount serves as the supervision signal for the Critic network's learning. (Second formula) Describes the parameters of the Critic network. The update method is to minimize the difference between the current Q-value estimate and the target Q-value. The mean square error between them is used to optimize the network, where The learning rate. The third formula. Describes the Actor network parameters The update method aims to improve the action value evaluated by the Critic network. The direction is adjusted by policy gradient. This results in outputting better actions. This corresponds to the learning rate. The formula contains... and These are the target Critic network and the target Actor network, used to stabilize the training process. Discount factor; This represents the batch sample size. Through this iterative training process, the cloud-based model can automatically discover and learn better control parameters and strategies, and then distribute these updated parameters to local devices, thereby achieving personalized performance improvements driven by global experience.

[0046] In an optional embodiment, the method may further include: processing the multi-source sensor data and the crowd density index. The operating mode and the target wind speed The system uploads actual wind speed and energy consumption data to the cloud server and receives optimized model parameters from the cloud server to update the local control strategy.

[0047] In this embodiment, during the data uplink process, the local control unit (i.e., the air purifier or its associated smart terminal) periodically or based on triggering conditions uploads multi-dimensional data generated during operation to the cloud server. This data can include not only raw multi-source sensor data (such as raw or pre-processed data from infrared, wireless signals, and vision, as well as air quality, temperature, and humidity readings), but also intermediate results and decision-making data calculated in real-time by local algorithms, such as crowd density indices. Automatically identified operating mode and target wind speed determined based on optimization objective function. This includes data on actual wind speed and energy consumption, reflecting the effectiveness of the control. Uploading this complete data chain allows the cloud to obtain end-to-end information from environmental perception and intelligent analysis to control execution and feedback. Secondly, on the cloud server side, it utilizes the anonymized or desensitized aggregated data received from numerous devices to run more complex, computationally intensive data analysis and machine learning algorithms (such as deep mining and reinforcement learning training based on historical data) to analyze key parameters involved in the control model (such as weight coefficients in weighted fusion algorithms). Coefficients in air quality prediction models The cloud server performs global optimization calculations on parameters such as gain parameters in the closed-loop control algorithm to generate a set of optimized model parameters. Finally, in the parameter downlink phase, the cloud server sends this optimized set of parameters to the corresponding local device. After receiving the new parameters, the local device replaces or updates the original parameters, thereby achieving iterative optimization of the local control strategy without changing the core algorithm framework.

[0048] Based on the foregoing technical solutions, the present invention also provides an air purifier fan speed control device based on changes in indoor crowd density. From a macroscopic perspective, this device may include the following modules: The data acquisition module 202 is used to acquire multi-source sensor data, which includes: crowd perception data for estimating indoor crowd density, environmental data acquired by an air quality sensor, time information, and temperature and humidity information; wherein, the crowd perception data includes at least infrared heat source data, WiFi or Bluetooth signal data, and visual image data. The crowd density index calculation module 204 is used to calculate the current time based on the crowd perception data using a weighted fusion algorithm. Population density index ; Air quality prediction module 206, used to predict air quality based on the population density index and the current air quality index The air quality index for future moments is predicted using a predictive model. Wind speed control module 208 is used to control the wind speed according to the crowd density index. The rate of change and the time information are used to identify the current operating mode of the target air purifier; the operating mode includes at least party mode, sleep mode and normal mode; wherein, switching from one mode to another must meet a preset lag condition; The wind speed dynamic adjustment module 210 is used to adjust the wind speed according to the identified operating mode and the population density index. Based on the predicted air quality index, the target wind speed is determined. Based on the target wind speed Based on the deviation from the current wind speed, a closed-loop control algorithm is used to dynamically adjust the wind speed of the target air purifier.

[0049] Those skilled in the art will understand that the modules in the apparatus of the embodiments can be distributed in the apparatus of the embodiments as described in the embodiments, or they can be located in one or more devices different from this embodiment with corresponding changes. The modules of the above embodiments can be combined into one module, or they can be further divided into multiple sub-modules.

[0050] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for controlling the fan speed of an air purifier based on changes in indoor crowd density, characterized in that, Includes the following steps: S1: Collect multi-source sensor data, which includes: crowd perception data for estimating indoor crowd density, environmental data collected by an air quality sensor, time information, and temperature and humidity information; wherein, the crowd perception data includes at least infrared heat source data, WiFi or Bluetooth signal data, and visual image data. S2: Based on the crowd perception data, the current time is calculated using a weighted fusion algorithm. Population density index ; S3: Based on the aforementioned population density index and the current air quality index The air quality index for future moments is predicted using a predictive model. S4: Based on the aforementioned population density index The rate of change and the time information are used to identify the current operating mode of the target air purifier; the operating mode includes at least party mode, sleep mode and normal mode; wherein, switching from one mode to another must meet a preset lag condition; S5: Based on the identified operating mode and the population density index Based on the predicted air quality index, the target wind speed is determined. Based on the target wind speed Based on the deviation from the current wind speed, a closed-loop control algorithm is used to dynamically adjust the wind speed of the target air purifier.

2. The air purifier fan speed control method based on indoor population density changes according to claim 1, characterized in that, In step S2, the calculation formula for the weighted fusion algorithm is as follows: in, The infrared heat source count is obtained based on the infrared heat source data. The number of connected devices is obtained based on the WiFi or Bluetooth signal data. This is a visual activity intensity index obtained based on the aforementioned visual image data. , , These are weighting coefficients determined based on historical data.

3. The air purifier fan speed control method based on indoor population density changes according to claim 1, characterized in that, In step S3, the prediction model is: in, Population density index rate of change, , , , These are the model coefficients. The error term is defined as follows: the predicted air quality index for the future time is... .

4. The air purifier fan speed control method based on indoor population density changes according to claim 1, characterized in that, In step S4, identifying the current operating mode of the target air purifier specifically includes: If the rate of change of the population density index Greater than the first threshold If so, it is identified as a party mode; If the rate of change of the population density index Less than the second threshold If the time information falls within a preset nighttime period, it is identified as a sleep mode; Otherwise, it will be identified as the normal mode.

5. The air purifier fan speed control method based on indoor population density changes according to claim 1, characterized in that, In step S5, the target wind speed is determined. The process employs a mode-dependent optimization objective function; in party mode, the objective function prioritizes maximizing the air exchange rate; in sleep mode, the objective function prioritizes minimizing noise and energy consumption; and in normal mode, the objective function balances air quality and energy consumption.

6. The method for controlling the fan speed of an air purifier based on changes in indoor crowd density according to claim 1, characterized in that, In step S5, the closed-loop control algorithm is a fuzzy PID control algorithm, and its wind speed update formula is: Among them, symbols Indicates air quality deviation. ,symbol , , These are the proportional, integral, and derivative control parameters that are dynamically adjusted according to the operating mode, with the following symbols: Represents the wind speed at the current moment. This represents the target wind speed to be set at the next moment.

7. The air purifier fan speed control method based on indoor population density changes according to claim 1, characterized in that, Between steps S4 and S5, there is also a step of fine-tuning the wind speed based on fuzzy logic, including: Population density change rate Current Air Quality Index and the current time period As fuzzy input variables, each fuzzy input variable is fuzzified, and its corresponding precise value is converted into a fuzzy linguistic value. Based on a pre-set fuzzy rule base, and combined with the fuzzy linguistic values ​​of each fuzzy input variable, inference is performed to output the wind speed adjustment amount. The fuzzy value; The weighted average method is used for defuzzification to obtain the accurate wind speed adjustment. Used to measure the target wind speed Make corrections.

8. The air purifier fan speed control method based on indoor population density changes according to claim 1, characterized in that, In the cloud server, based on a reinforcement learning framework, the uploaded data is used to optimize the control model and update the model parameters, specifically including: Define state space The state space Including the current air quality index The population density index The aforementioned operating mode, historical wind speed data, and energy consumption and noise ; Define action space It is the wind speed adjustment amount. The set of all possible values; Define reward function The calculation formula is as follows: Among them, symbols Indicates the rate of improvement in air quality. , , These are the weighting coefficients; The agent is trained using a deep deterministic gradient algorithm to update the model parameters. , , , , To optimize control strategies The update formula for the depth deterministic policy gradient algorithm is as follows: Among them, symbols The action value function output by the Critic network, with the symbol... The action value function output by the target Critic network is represented by the symbol. The policy function output by the Actor network, symbol The symbol represents the policy function output by the target Actor network. Indicates time The system state, symbol For a moment The action performed, symbol Indicates time The system state, symbol Indicates time Execute action The instant reward obtained afterward, symbol Discount factor, symbol The learning rate of the Critic network is represented by the symbol. The learning rate of the Actor network is represented by the symbol. The gradient operator is represented by the symbol. Indicates the number of samples in a batch for training, symbol Indicates the first The target of each sample Value, symbol The parameter set of the Critic network is represented by the symbol. This represents the set of parameters for the Actor network.

9. The air purifier fan speed control method based on indoor population density changes according to claim 1, characterized in that, The method further includes: The multi-source sensor data and the crowd density index The operating mode and the target wind speed The system uploads actual wind speed and energy consumption data to the cloud server and receives optimized model parameters from the cloud server to update the local control strategy.

10. An air purifier fan speed control device based on changes in indoor crowd density, characterized in that, The device includes: The data acquisition module is used to acquire multi-source sensor data, which includes: crowd perception data for estimating indoor crowd density, environmental data acquired by an air quality sensor, time information, and temperature and humidity information; wherein, the crowd perception data includes at least infrared heat source data, WiFi or Bluetooth signal data, and visual image data. The crowd density index calculation module is used to calculate the current time based on the crowd perception data using a weighted fusion algorithm. Population density index ; An air quality prediction module is used to predict air quality based on the population density index. and the current air quality index The air quality index for future moments is predicted using a predictive model. The wind speed control module is used to control the wind speed according to the crowd density index. The rate of change and the time information are used to identify the current operating mode of the target air purifier; the operating mode includes at least party mode, sleep mode and normal mode; wherein, switching from one mode to another must meet a preset lag condition; The wind speed dynamic adjustment module is used to adjust the wind speed according to the identified operating mode and the population density index. Based on the predicted air quality index, the target wind speed is determined. Based on the target wind speed Based on the deviation from the current wind speed, a closed-loop control algorithm is used to dynamically adjust the wind speed of the target air purifier.

Images