Method, system, medium and product for indoor air purification by independent double-duct range hood

Abstract

A method, system, medium, and product for indoor air purification using an independent dual-duct range hood, relating to the smart home field, are disclosed. The method includes: based on the current cooking mode and predicted oil fume concentration, calculating and setting a dynamic ideal air pressure field distribution map in real time within a three-dimensional dynamic airflow field calculation model, with a preset high-temperature smoke core area of ​​the cooktop as the protection target; using the dynamic ideal air pressure field distribution map as the target, calculating the airflow curves and airflow direction angles of the exhaust duct and the make-up air duct required to establish or maintain the dynamic clean air barrier zone through the corrected three-dimensional dynamic airflow field calculation model, and driving the fan and airflow direction adjustment mechanism to execute these parameters. Implementing the above technical solution accurately establishes and maintains the dynamic clean air barrier zone, effectively improving smoke extraction efficiency.

Description

Technical Field

[0001] This application relates to the field of smart homes, and in particular to a method, system, medium, and product for purifying indoor air with an independent dual-duct system for a range hood. Background Technology

[0002] As people's living standards improve, the demand for kitchen air purification is also increasing. As a cooking area, the kitchen generates a large amount of oil fumes and odors. If these are not effectively and promptly removed, they will not only affect the air quality within the kitchen but may also threaten the health of the residents. Furthermore, the airflow within the kitchen is complex and influenced by various factors, such as the location of the stove and the opening and closing of doors and windows, making kitchen air purification a challenging problem. Therefore, how to efficiently purify kitchen air and create a healthy and comfortable cooking environment has become a hot research topic in the home furnishing industry.

[0003] Traditional kitchen air purification primarily relies on ordinary range hoods for smoke extraction. Ordinary range hoods typically use a simple fan to draw cooking fumes from near the stove into the exhaust duct and then vent them outdoors. This method merely removes the fumes without considering the overall airflow and distribution within the kitchen. Furthermore, ordinary range hoods cannot adjust to different cooking modes and fume concentrations, resulting in relatively fixed efficiency and difficulty adapting to the complex and ever-changing kitchen environment. Additionally, some ordinary range hoods are not adequately designed for air intake, leading to poor air circulation and easy diffusion of fumes throughout the kitchen. Existing ordinary range hood exhaust methods do not consider the positional relationship between the range hood and the stove, the layout of other objects in the room, or the opening and closing of doors and windows, and cannot adjust in real-time according to cooking modes and fume concentrations. This makes them ineffective in coping with the complex and ever-changing kitchen environment, resulting in easy escape of fumes and poor purification performance. Summary of the Invention

[0004] This application provides a method, system, medium, and product for indoor air purification of a range hood with independent dual air ducts, which accurately establishes and maintains a dynamic clean air barrier zone, effectively improving smoke exhaust efficiency.

[0005] In a first aspect, this application provides a method for purifying indoor air in a range hood with independent dual ducts, the method comprising: Based on the fixed spatial positions of the range hood and cooktop, as well as the real-time indoor object layout and door and window opening and closing status, a three-dimensional dynamic airflow field calculation model is established, including the inlet and outlet of the exhaust duct, the inlet and outlet of the make-up air duct, and the cooktop area. Based on the current cooking mode and the predicted oil fume concentration, the three-dimensional dynamic air flow field calculation model calculates and sets a dynamic ideal air pressure field distribution map in real time with the preset high-temperature smoke core area of ​​the stove as the protection target. The dynamic ideal air pressure field distribution map includes a low-pressure capture zone formed above the oil fume generation area, and a dynamic clean air barrier zone with an air pressure gradient between the low-pressure capture zone and the make-up air duct. By deploying a sensor array in the stove area, the exhaust duct, and the make-up air duct, real-time multi-physics field data of the physical space is collected. The multi-physics field data includes air pressure, airflow velocity, and temperature data. The multi-physics field data is compared with the prediction data of the three-dimensional dynamic airflow field calculation model. Based on the multi-dimensional model prediction error signal set, the three-dimensional dynamic airflow field calculation model is corrected in real time through an adaptive filtering algorithm. Using the dynamic ideal air pressure field distribution map as the target, the air volume curves and air direction angles of the exhaust duct and the make-up air duct required to establish or maintain the dynamic clean air barrier area are calculated by the corrected three-dimensional dynamic air flow field calculation model, and the fan and air direction adjustment mechanism are driven to perform the calculation.

[0006] By employing the above technical solution, a low-pressure zone is dynamically established and maintained above the cooktop, enabling precise and efficient capture of the source of cooking fumes and effectively preventing their escape. A dynamic clean air barrier with a pressure gradient is formed between the fume capture zone and the supplementary air duct, which not only blocks the spread of fumes but also guides fresh air in an orderly manner, improving overall ventilation and purification efficiency. Through real-time feedback from a sensor array and model self-correction, the system can dynamically adapt to environmental changes (such as door and window opening / closing, object movement) and cooking conditions, ensuring that the control strategy remains precise and effective. With the goal of establishing and maintaining an ideal airflow field, the system calculates and adjusts the optimal airflow and direction of the dual air ducts in real time, avoiding the blind high-volume operation of traditional range hoods, thus achieving energy-saving operation at high efficiency.

[0007] In some embodiments, the step of calculating and setting a dynamic ideal air pressure field distribution map in real time in the three-dimensional dynamic airflow field calculation model based on the current cooking mode and predicted oil fume concentration, with a preset high-temperature smoke core area of ​​the stove as the protection target, specifically includes: Based on the stove status parameters and the recognition of oil fume patterns by image sensors, the current cooking mode is determined in real time, and the oil fume concentration and diffusion trend within the preset time window are predicted. Based on the oil fume concentration and diffusion trend, the dynamic spatial attributes of the low-pressure capture zone are dynamically calculated. The dynamic spatial attributes include the target pressure value, the three-dimensional spatial range, and the relative distance between the low-pressure capture zone and the make-up air duct. Based on the dynamic spatial properties, calculate the pressure gradient curve decreasing from the make-up air side to the fume side within the dynamic clean air barrier zone between the low-pressure capture zone and the inlet of the make-up air duct. The target pressure value of the low-pressure capture zone, the pressure gradient curve of the dynamic clean air barrier zone, and the ambient air pressure value of the kitchen are fused together to generate the dynamic ideal air pressure field distribution map.

[0008] By employing the aforementioned technical solution, based on cooking mode recognition and oil fume diffusion prediction, the system can pre-plan the flow field before a large amount of oil fume is generated, transforming passive suction and exhaust into active guidance and interception, thus improving the timeliness and effectiveness of control. The system dynamically calculates the pressure value and spatial range of the low-pressure capture zone based on oil fume prediction, ensuring that the suction strength and the area of ​​action are precisely matched to real-time needs, guaranteeing capture effectiveness while avoiding energy waste. By accurately calculating and setting the continuous air pressure gradient from the air inlet to the capture zone, a controlled airflow path is essentially planned, forming an invisible dynamic "air curtain" in key areas that effectively isolates oil fumes and guides fresh air. The air pressure settings of local key areas (capture zone, barrier zone) are integrated with the overall ambient air pressure to generate a complete dynamic air pressure field distribution map, providing a unified and quantifiable high-order control target for the subsequent coordinated adjustment of exhaust and air supply ducts.

[0009] In some embodiments, the real-time correction of the three-dimensional dynamic airflow field calculation model using an adaptive filtering algorithm based on the multi-dimensional model prediction error signal set includes: Based on the spatial locations of the dynamic clean air barrier zone and the low-pressure capture zone, the inlet and outlet locations of the smoke exhaust duct and the make-up air duct, and the physical layout of the sensor array, a set of spatial key points for model calibration are determined. In each control cycle, pressure scalar error, airflow velocity vector error, and temperature scalar error are extracted from the actual multiphysics data and the prediction data of the three-dimensional dynamic airflow field calculation model at the corresponding spatial key points, respectively, and together they constitute the multidimensional model prediction error signal set. Based on the amplitude and frequency characteristics of each error component in the multi-dimensional model prediction error signal set, the optimal smooth estimate of the model bias is obtained through an adaptive Kalman filter. By using a pre-calibrated mapping relationship, the optimal smoothing estimate is inversely analyzed into a quantitative correction amount for specific key parameters in the three-dimensional dynamic airflow field calculation model. The three-dimensional dynamic airflow field calculation model is then corrected based on the quantitative correction amount. The specific key parameters include turbulent viscosity coefficient, wall drag coefficient, or local momentum source term coefficient.

[0010] By employing the above technical solution, and through pre-setting spatial key points and extracting multi-dimensional physical quantity errors, the system can accurately locate and quantify complex real-world flow field deviations into structured data, providing clear input for correction. Using an adaptive Kalman filter to process the error signal effectively filters out sensor noise and transient interference, obtaining the optimal smooth estimate of the model deviation and avoiding oscillations or instability caused by data fluctuations during the correction process. Through pre-calibrated mapping relationships, surface observation errors are inversely analyzed into quantitative corrections to key intrinsic parameters such as turbulent viscosity and wall drag, directly improving the accuracy of the physical mechanism of the computational model, rather than simply compensating for results.

[0011] In some embodiments, the step of using the dynamic ideal air pressure field distribution map as a target, calculating the air volume curves and air direction angles of the exhaust duct and the make-up air duct required to establish or maintain the dynamic clean air barrier zone through the corrected three-dimensional dynamic airflow field calculation model, and driving the fan and air direction adjustment mechanism to execute, specifically includes: Using the corrected three-dimensional dynamic airflow field calculation model as the predictive controller and the dynamic ideal air pressure field distribution map as the tracking target, rolling time domain optimization calculation is performed. In the rolling time-domain optimization, the optimization objective is to minimize the system tracking error and the overall energy consumption. The future control sequence that satisfies the performance constraints of the dynamic clean air barrier zone is solved. The future control sequence defines the optimal air volume curve and air direction angle of the exhaust duct and the make-up air duct in a preset time period from the current moment. The first control quantity of the future control sequence is sent to the fan and the wind direction adjustment mechanism, and the actual execution feedback of the fan and the wind direction adjustment mechanism is collected. Adjustments are made based on the comparison result between the actual execution feedback and the first control quantity.

[0012] By employing the aforementioned technical solution, a predictive controller based on a calibration model performs rolling time-domain optimization. This enables the system to not only accurately track the current pressure field target but also proactively plan the optimal control actions for a future period, improving control accuracy and smoothness. The optimization objective directly integrates minimizing tracking error and minimizing overall energy consumption. When solving for control commands, it simultaneously weighs "effectiveness" and "efficiency," thus actively seeking and executing the most energy-efficient operating strategy under the constraint of dynamically maintaining air barrier effectiveness. By comparing and adjusting the theoretical control quantity (the first control quantity) with the actual execution feedback, a closed loop from command to execution is formed, effectively compensating for actuator errors or external disturbances, ensuring that control intentions are reliably translated into actual actions. This process combines the dynamic ideal target (pressure field) formed in the preceding steps with the calibrated cognitive model (flow field model). Through optimized calculation output, it can directly drive the optimal control sequence of the hardware, ultimately completing the entire chain of intelligent decision-making.

[0013] In some embodiments, the step of using the corrected three-dimensional dynamic airflow field calculation model as a predictive controller and the dynamic ideal air pressure field distribution map as the tracking target to perform rolling time-domain optimization calculations specifically includes: The corrected three-dimensional dynamic airflow field calculation model is discretized into a state space prediction model suitable for real-time control. The state space prediction model takes the control input of the smoke exhaust fan corresponding to the smoke exhaust duct and the make-up air fan corresponding to the make-up air duct as variables, and the air pressure and airflow velocity at the key points in the dynamic ideal air pressure field distribution map as state outputs. In each control cycle, the current physical state fed back by the sensor array is used as the initial condition, and the state space prediction model is used to perform forward simulation prediction of the dynamic evolution of the system state in the future preset time domain. Within the future preset time domain, key performance indicators of the dynamic clean air barrier zone are defined as path constraints that must be met in the optimization process. The key performance indicators include the barrier zone pressure gradient retention rate and the oil fume escape concentration threshold. Based on the results of forward simulation prediction and the path constraints, an optimization problem is constructed with the objective of minimizing the deviation between the predicted trajectory and the ideal target trajectory.

[0014] By employing the above technical solution, the complex three-dimensional flow field calculation model is discretized into a state-space prediction model with fan control input as the variable and the flow field state at key points as the output, providing a precise and efficient computational framework for online real-time optimization calculation. In each control cycle, forward simulation of the future system dynamics is performed starting from the current measured state, ensuring that control decisions are no longer merely responses to current deviations, but rather proactive planning based on predictions of future states, significantly improving the anticipation and smoothness of control. The efficiency indicators of the dynamic clean air barrier zone (such as the pressure gradient retention rate) are explicitly defined as path constraints that optimization must adhere to, fundamentally ensuring that the core purification and isolation functions of the system are not sacrificed regardless of energy-saving optimizations, achieving a reliable balance between performance and energy efficiency. Finally, minimizing the deviation between the predicted trajectory and the ideal target trajectory is defined as the optimization objective, giving the entire rolling optimization process a clear mathematical direction: driving the actual flow field to infinitely approximate the dynamically calculated ideal flow field, forming a direct closed loop between the theoretical objective and real-time control.

[0015] In some embodiments, solving for the future control sequence that satisfies the dynamic clean air barrier zone performance constraints includes: In the optimization problem, the sequence of changes in air volume and air direction angle of the exhaust duct and the make-up air duct in the future time domain is defined as the control variables to be optimized. A comprehensive cost function is established, wherein the first term of the comprehensive cost function is the tracking error between the predicted system state and the target value of the dynamic ideal pressure field distribution map, and the second term of the comprehensive cost function is the comprehensive energy consumption estimate that is positively correlated with the cubic speed of the wind turbine. The minimum value of the comprehensive cost function is obtained by satisfying all path constraints and actuator physical constraints. The smoothness and feasibility of the optimal control sequence are verified to obtain the future control sequence.

[0016] By employing the aforementioned technical solution, the future time-domain airflow and direction change sequences of the dual-duct system are jointly defined as optimization variables, enabling coordinated forward planning of smoke extraction and air supply actions in the time dimension, ensuring the continuity and overall optimality of system actions. By constructing a comprehensive cost function that integrates tracking error (effectiveness) and fan cubic power consumption (efficiency), the core multi-objective optimization problem of the system is transformed into a computable single-objective mathematical problem, providing precise quantitative basis for intelligent trade-offs. When solving for the minimum value, the effectiveness path constraints of the air barrier and the physical limit constraints of the actuator are considered simultaneously, ensuring that the obtained optimal control sequence is not only mathematically optimal but also a feasible solution that can be safely and reliably executed in a real-world system. After obtaining the theoretically optimal solution, additional smoothness and feasibility verification is performed, filtering out commands that may cause mechanical oscillations or exceed the response capabilities of the actuator, ultimately outputting a smoother and safer actual control sequence, improving the system's stability and lifespan.

[0017] In some embodiments, the method further includes: Before the first control quantity is issued, the transient impact of the flow field that may be caused by the execution of the control command is predicted based on the state space prediction model, and a dynamic feedforward compensation signal is generated and superimposed on the first control quantity. After the first control variable is executed, the actual performance index of the dynamic clean air barrier zone is calculated based on the real-time data of the sensor array, and compared with the performance index predicted by the model to generate a real-time performance micro-deviation signal. Based on the historical trend of the performance micro-deviation signal, the weight ratio of the system tracking error term and the comprehensive energy consumption term in the comprehensive cost function is dynamically adjusted by an adaptive law in the rolling time-domain optimization calculation. The dynamic feedforward compensation signal and the adjusted weight ratio are applied to the rolling time-domain optimization calculation of the next control cycle.

[0018] By employing the above technical solution, transient flow field impacts that may be triggered by model-predictive control commands are actively offset through feedforward compensation signals, significantly improving the system's response speed and stability, and avoiding overshoot or oscillations during the adjustment process. By comparing actual and predicted barrier zone performance indicators and generating micro-deviation signals, a millimeter-level closed-loop feedback on the control effect is formed, enabling real-time correction of subtle deviations caused by model predictions or environmental disturbances, ensuring high reliability of core functions. Based on the historical trend of performance deviation, the weight ratio of tracking effect and comprehensive energy consumption in the optimization objective is adaptively adjusted, allowing the system to intelligently adjust its strategy focus under different operating conditions, achieving globally adaptive optimization. Applying feedforward compensation experience and weight adjustment strategies to the next control cycle allows the entire control logic to continuously iterate and improve based on historical execution results, resulting in continuous improvement in system performance over time.

[0019] In a second aspect, embodiments of this application provide a computer system including a memory, a processor, and a computer program stored in the memory; the processor executes the computer program to implement the steps of the method described in any possible implementation of the first aspect.

[0020] Thirdly, embodiments of this application provide a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implement the steps of the method described in any possible implementation of the first aspect.

[0021] Fourthly, embodiments of this application provide a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the method described in any possible implementation of the first aspect.

[0022] It is understood that the computer system provided in the second aspect, the storage medium provided in the third aspect, and the computer program product provided in the fourth aspect are all used to execute the method provided in this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0023] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. Traditional range hoods rely on creating a large negative pressure zone beneath the hood to draw in diffused cooking fumes. The core of this method is to pre-calculate and proactively construct an ideal aerodynamic environment. It precisely creates a low-pressure capture zone at the source of the fumes (above the stove) for accurate fume extraction; simultaneously, it constructs a dynamic clean air barrier zone with a pressure gradient around this zone. This is equivalent to setting up a "smart wall" of airflow along the path of fume diffusion, not only absorbing the fumes but also proactively guiding fresh air and limiting the spread of the fumes, thereby significantly improving purification efficiency. 2. The system can predict the trend of oil fume generation based on the cooking mode and calculate the dynamic ideal air pressure field distribution map required to achieve the best capture and isolation effect. By comparing the model prediction with actual sensor data, the system can automatically correct the model in real time to ensure consistency between the digital world and the physical world. Then, based on the corrected accurate model, it calculates the optimal actions (airflow curve and airflow angle) that the actuators (fan, airflow adjustment mechanism) need to perform and drives them to execute. 3. Due to the establishment of a real-time dynamic model and correction mechanism, the system can automatically adapt to various environmental changes and disturbances. For example, when kitchen doors and windows are opened, people walk around, or new items are placed, the original airflow field will change. Traditional range hoods are powerless to address this, while this application can detect these changes through sensors and quickly adjust the model and control strategy to re-establish an effective low-pressure capture zone and air barrier, thereby ensuring the stability of the purification effect in various complex real-world environments; 4. Traditional range hoods often deal with cooking fumes simply by increasing the airflow (higher speed), resulting in high energy consumption. This method, through precise modeling and calculation, can clearly determine the minimum and optimal energy required to maintain a dynamic clean air barrier zone. By dynamically adjusting the airflow and direction angle of the dual air ducts, it achieves the purification goal in the most precise and coordinated way, avoiding energy waste and achieving a balance between high efficiency and low energy consumption. Attached Figure Description

[0024] Figure 1 This is a schematic flowchart of a method for purifying indoor air with an independent dual-duct system for a range hood, as described in an embodiment of this application. Figure 2 This is a schematic diagram of the smoke exhaust duct and the air supply duct in the embodiments of this application; Figure 3 This is a schematic diagram of an exemplary hardware structure of a computer system in an embodiment of this application. Detailed Implementation

[0025] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0026] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0027] The following is combined Figure 1 The method of the embodiments of this application will be described below.

[0028] Figure 1 This is a flowchart illustrating a method for indoor air purification using a range hood with independent dual ducts, as described in an embodiment of this application. Figure 1 As shown, the method includes the following steps: S101. Based on the fixed spatial position of the range hood and the stove, and the real-time acquisition of the indoor object layout and the opening and closing status of doors and windows, establish a three-dimensional dynamic air flow field calculation model including the inlet and outlet of the exhaust duct, the inlet and outlet of the make-up air duct, and the stove area. S102. Based on the current cooking mode and the predicted oil fume concentration, calculate and set a dynamic ideal air pressure field distribution map in real time in the three-dimensional dynamic air flow field calculation model with the preset high-temperature smoke core area of ​​the stove as the protection target. The dynamic ideal air pressure field distribution map includes a low-pressure capture zone formed above the oil fume generation area, and a dynamic clean air barrier zone with an air pressure gradient between the low-pressure capture zone and the make-up air duct. S103. Using a sensor array arranged in the stove area, the exhaust duct, and the make-up air duct, real-time multi-physics field data of the physical space is collected. The multi-physics field data includes air pressure, airflow velocity, and temperature data. The multi-physics field data is compared with the prediction data of the three-dimensional dynamic airflow field calculation model. Based on the multi-dimensional model prediction error signal set, the three-dimensional dynamic airflow field calculation model is corrected in real time through an adaptive filtering algorithm. S104. Using the dynamic ideal air pressure field distribution map as the target, the air volume curves and air direction angles of the exhaust duct and the make-up air duct required to establish or maintain the dynamic clean air barrier area are calculated by the corrected three-dimensional dynamic air flow field calculation model, and the fan and air direction adjustment mechanism are driven to perform the calculation.

[0029] The fixed 3D coordinates of the range hood, stove, exhaust duct outlet, and make-up air duct inlet are input, forming the model's fixed boundaries. Approximate positions and dimensions of large objects (such as cabinets, refrigerators, and dining tables) and the initial states of doors and windows (open / closed and degree of opening) are obtained using visual sensors within the kitchen (such as depth cameras) or user-preset information, serving as the model's initial internal boundary conditions. Using computational fluid dynamics, the physical space is discretized into a 3D mesh. The mesh is refined in key areas (such as above the stove and near the duct inlets / outlets) to improve computational accuracy. Door and window states are acquired in real-time via magnetic sensors or visual recognition. When a state changes, the boundary condition is correspondingly changed from a wall to an opening with environmental pressure, or vice versa, in the computational model. Movement of large objects (such as someone entering) can be detected by sensors (such as ToF sensors), simplifying them in the model as an obstacle with specific roughness. To achieve the required real-time control, a reduced-order model or a surrogate model trained with extensive pre-computation can be used. For example, Method 1 (physics-based simplification): A set of transfer functions is established, taking fan speed as input and flow velocity and pressure at key points (such as the center of the capture zone) as outputs, and coupled through simplified forms of flow control equations (such as Bernoulli's equation and continuity equation). Method 2 (data-driven): Massive amounts of flow field data are generated in advance through high-fidelity simulations or experiments for different layouts, window / door conditions, and fan operating conditions, and a neural network model is trained. During online runtime, the current state is input into the network to quickly predict the distribution of the entire flow field. The core is to strike a balance between computational accuracy and speed, ensuring the model can complete a flow field update within tens to hundreds of milliseconds. Typically, a fast solver based on linearized flow assumptions or a lightweight neural network is used.

[0030] The cooking mode (e.g., simmering over low heat or stir-frying over high heat) is determined by a combination of the cooktop's power sensor (firepower level), temperature sensor (pot bottom temperature), and image sensor (observing the state of matter inside the pot, such as the shape and amount of fumes produced during frying, boiling, etc.). Based on the identified mode and historical data, the trend of fume production in the near future (the next 30 seconds) is predicted (e.g., for stir-fry mode, it is predicted that a period of high fume production will begin immediately and continue for a period of time). The parameters of the low-pressure capture zone are dynamically adjusted according to the predicted fume concentration and diffusion trend. For example, for stir-fry mode: a low-pressure zone with a lower pressure value and a larger spatial range is set to generate strong capture power. For low-heat mode: a low-pressure zone with a slightly higher pressure value and a smaller spatial range is set, sufficient to capture the fumes. The specific target pressure value and spatial shape (which can be simplified to an ellipsoidal range) of the low-pressure capture zone are obtained by querying a preset experience database, which is linked to "cooking mode - fume prediction - ideal capture parameters". A dynamic clean air barrier zone is defined between the low-pressure capture zone and the inlet of the make-up air duct. Within this dynamic clean air barrier zone, a linear pressure gradient is established: it decreases uniformly from the air intake side (pressure close to or slightly higher than ambient pressure) towards the low-pressure capture zone side (lowest pressure). This gradient causes the incoming fresh air to naturally flow from the high-pressure side to the low-pressure side, both replenishing the air exhausted by the range hood and forming a flow barrier to prevent the lateral diffusion of cooking fumes. Integrating the pressure settings of the low-pressure capture zone and the dynamic clean air barrier zone with the rest of the kitchen (usually set to standard ambient pressure) generates a three-dimensional pressure target map covering the entire computational domain, which is the dynamic ideal pressure field distribution map.

[0031] Miniature barometers, hot-wire anemometers, and temperature sensors are installed at different heights directly above the stove, at the inlet grille of the exhaust duct, and at the outlet grille of the make-up air duct. These sensors synchronously collect data at a fixed frequency (e.g., 10Hz). In each control cycle, the predicted air pressure, wind speed, and temperature values ​​at grid points corresponding exactly to the physical locations of the sensors are read from the three-dimensional dynamic airflow field calculation model. The measured values ​​from the sensors are subtracted from the model predictions point by point to obtain a set of error signals (air pressure error, velocity vector error, and temperature error). These error signals contain measurement noise and actual model bias. Using algorithms such as adaptive Kalman filters, the true and smooth trend of model bias can be estimated. For example, if multiple sensors continuously show that the actual wind speed in a certain area is lower than the predicted value, the filter will output a stable negative bias estimate. The system internally maintains the model's key physical parameters (such as turbulence intensity coefficient and local drag coefficient of the duct). Based on the deviation estimate obtained after filtering, these internal parameters are fine-tuned through a pre-calibrated inverse mapping relationship (e.g., if the measured wind speed is too low, it may be because the duct drag coefficient in the model is set too low, and this coefficient should be increased). After adjustment, the predicted values ​​of the three-dimensional dynamic airflow field calculation model will be closer to the sensor's measured values ​​in the next calculation, thereby improving the model's realism and reliability.

[0032] The corrected 3D dynamic airflow field calculation model is used as the predictor, and the dynamic ideal air pressure field distribution map is used as the tracking target. In each control cycle (e.g., per second), the system performs an optimization calculation: assuming that in the future (e.g., the next 10 seconds), the exhaust and makeup air fans operate with different combinations of airflow and direction, how will the flow field predicted by the model evolve? The task of the optimization algorithm (e.g., model predictive control algorithm) is to find an optimal sequence from countless possible fan action combinations, such that: the predicted flow field approximates the ideal air pressure field distribution map as closely as possible (i.e., the tracking error is minimized), and the air pressure gradient of the dynamic clean air barrier zone remains effective (performance constraint); under the premise of meeting the above requirements, the total energy consumption of the fans (approximately proportional to the cube of the rotational speed) is minimized. Solving this optimization problem yields a set of optimal airflow and direction angle curves for the exhaust and makeup air fans in the future time domain. The system immediately extracts the first instruction from this optimal sequence (i.e., the fan speed and direction angle that should be executed immediately at the current moment) and issues it to the variable frequency fan and the servo-driven direction adjustment mechanism for execution. Simultaneously, the system monitors the execution effect: comparing the actual barrier zone effectiveness reported by the sensors with the effectiveness predicted by the model. If a slight deviation exists, a compensation signal is generated and superimposed on the optimization calculation of the next control cycle for dynamic fine-tuning. In the next control cycle, the system repeats the entire process: correcting the model based on the latest sensor data, updating the ideal target according to the latest cooking status, performing rolling optimization again, and executing the first new instruction. This cycle repeats continuously, forming a fully closed-loop adaptive intelligent control system of perception-modeling-planning-execution-correction.

[0033] Figure 2 This is a schematic diagram of the smoke exhaust duct and the make-up air duct in the embodiments of this application, as shown below. Figure 2 As shown, the smoke exhaust duct and the air supply duct are independent of each other. The smoke exhaust duct is used in conjunction with the smoke exhaust fan to exhaust the oil fumes generated indoors to the outside, while the air supply duct is used in conjunction with the air supply fan to bring filtered fresh air from the outside into the room.

[0034] In some embodiments, the step of calculating and setting a dynamic ideal air pressure field distribution map in real time in the three-dimensional dynamic airflow field calculation model based on the current cooking mode and predicted oil fume concentration, with a preset high-temperature smoke core area of ​​the stove as the protection target, specifically includes: Based on the stove status parameters and the recognition of oil fume patterns by image sensors, the current cooking mode is determined in real time, and the oil fume concentration and diffusion trend within the preset time window are predicted. Based on the oil fume concentration and diffusion trend, the dynamic spatial attributes of the low-pressure capture zone are dynamically calculated. The dynamic spatial attributes include the target pressure value, the three-dimensional spatial range, and the relative distance between the low-pressure capture zone and the make-up air duct. Based on the dynamic spatial properties, calculate the pressure gradient curve decreasing from the make-up air side to the fume side within the dynamic clean air barrier zone between the low-pressure capture zone and the inlet of the make-up air duct. The target pressure value of the low-pressure capture zone, the pressure gradient curve of the dynamic clean air barrier zone, and the ambient air pressure value of the kitchen are fused together to generate the dynamic ideal air pressure field distribution map.

[0035] Cooktop status parameters: Directly connect to the cooktop's power level electrical signal (e.g., 0-10V analog or digital setting) to obtain real-time heat load data. Simultaneously, install an infrared temperature sensor at or near the bottom of the pot to monitor the pot's temperature and its heating rate. Image visual recognition: Deploy a wide-angle high-definition camera on the range hood, covering the cooktop area. Analyze the video stream in real-time using computer vision algorithms (e.g., image classification and object detection models based on convolutional neural networks (CNNs) to identify the density (transparency), rising speed, intensity of churning, and color of the fumes (e.g., light blue smoke and thick white smoke). For example, rapidly churning thick white smoke usually corresponds to stir-frying with high moisture and oil content. Assist in identifying cooking stages, such as "heating oil," "adding ingredients and stir-frying," and "adding water and simmering." Process the above signals (power, temperature, visual features) as a time series to capture their changing trends. Input the extracted feature vectors (e.g., power value, temperature rise rate, and visual feature vectors of fumes) into a pre-trained classifier (e.g., support vector machine (SVM) or lightweight neural network). The classifier outputs several preset "cooking mode" labels and their confidence scores, such as: "high-heat stir-fry mode," "medium-heat frying mode," "low-heat steaming mode," and "static preheating mode." The system associates each "cooking mode" with an empirical prediction model. This model can be a simple predefined curve or a time-series model (such as ARIMA) learned from historical data. For example, when the system determines it to be "high-heat stir-fry mode," it immediately calls the corresponding prediction model, which outputs the predicted fume source intensity (amount generated per unit time) and the main diffusion direction (primarily driven upwards by thermal buoyancy, but possibly affected by lateral airflow) within a preset time window (e.g., the next 60 seconds). The prediction results can be discrete levels (e.g., "low, medium, high, extremely high") or continuous estimates. The system internally maintains a dynamic parameter mapping table, which uses "predicted fume concentration level" and "predicted diffusion trend" (e.g., whether the main direction is stable) as key indices. Target pressure value: Based on the index, a negative pressure target value (relative to ambient air pressure) is retrieved from the mapping table. For example, "extremely high" concentration corresponds to an extremely low negative pressure value (e.g., -15 Pa) to generate strong suction, while "low" concentration corresponds to a weaker negative pressure value (e.g., -5 Pa). Three-dimensional spatial range: Similarly, parameters of a three-dimensional control volume are determined according to the mapping table. This can typically be simplified to an inverted elliptical cone or cuboid, whose base covers the cookware area and is slightly larger than the cookware. Its height is adjusted based on the predicted oil fume concentration—the higher the concentration, the higher the height required to prevent escape. The range parameters are defined by the center point coordinates and the half-axis length / side length. Relative distance: This attribute mainly depends on the fixed physical layout of the kitchen (the distance between the installation location of the make-up air duct inlet and the stove) and is usually a fixed value. However, in some designs, the make-up air direction is adjustable; in this case, the vector distance between the effective inlet and the capture area needs to be dynamically calculated based on the current make-up air direction.Define key points and interpolation: The starting point is the center point of the supply air duct outlet plane or the core point of effective air supply. Its target air pressure value P_s is usually set slightly higher than the ambient air pressure (e.g., +2 Pa) to provide driving force; the ending point is the edge point of the low-pressure capture zone near the supply air side, and its target air pressure value P_c has been determined by the above steps (a negative value, e.g., -10 Pa). Construct the gradient curve: Between the starting and ending points, set a series of intermediate interpolation points along the path connecting the two points (or the path adjusted according to spatial obstacles). Assign air pressure values ​​to these intermediate points using linear interpolation or smooth curve interpolation (e.g., cubic spline) methods to ensure that the air pressure continuously and monotonically decreases from P_s to P_c. This air pressure value change curve along the spatial path is the air pressure gradient curve. It defines the "pressure slope" of air flowing naturally and smoothly from the supply air inlet to the capture zone. Map the three-dimensional spatial range of the low-pressure capture zone onto a grid, and directly set the "target air pressure value" of all grid points within this range as P_c. The target pressure value is calculated and assigned to grid points within a certain width of the pressure gradient curve path and its vicinity, based on their distance from the make-up air inlet. The target pressure value for most grid points in the kitchen not covered by the above two areas is set as the standard environmental reference pressure (usually 0 Pa, as a relative pressure benchmark). To avoid abrupt changes in target pressure at the boundary between areas (which would lead to control difficulties), a narrow-band smoothing transition (e.g., using Gaussian filtering or linear attenuation) is required between the low-pressure capture / barrier zone and the environmental zone, allowing the pressure value to transition smoothly to the environmental value. After the above assignment and smoothing, each three-dimensional grid point has a clearly defined target pressure value. This three-dimensional scalar field data, covering the entire computational domain and containing fine structure (low-pressure zone, gradient band), is the dynamic ideal pressure field distribution map. This distribution map is regenerated every control cycle (e.g., every second) based on the latest identification and prediction results, serving as the dynamically changing ultimate target that the subsequent model predictive controller tracks in real time.

[0036] In some embodiments, the real-time correction of the three-dimensional dynamic airflow field calculation model using an adaptive filtering algorithm based on the multi-dimensional model prediction error signal set includes: Based on the spatial locations of the dynamic clean air barrier zone and the low-pressure capture zone, the inlet and outlet locations of the smoke exhaust duct and the make-up air duct, and the physical layout of the sensor array, a set of spatial key points for model calibration are determined. In each control cycle, pressure scalar error, airflow velocity vector error, and temperature scalar error are extracted from the actual multiphysics data and the prediction data of the three-dimensional dynamic airflow field calculation model at the corresponding spatial key points, respectively, and together they constitute the multidimensional model prediction error signal set. Based on the amplitude and frequency characteristics of each error component in the multi-dimensional model prediction error signal set, the optimal smooth estimate of the model bias is obtained through an adaptive Kalman filter. By using a pre-calibrated mapping relationship, the optimal smoothing estimate is inversely analyzed into a quantitative correction amount for specific key parameters in the three-dimensional dynamic airflow field calculation model. The three-dimensional dynamic airflow field calculation model is then corrected based on the quantitative correction amount. The specific key parameters include turbulent viscosity coefficient, wall drag coefficient, or local momentum source term coefficient.

[0037] Key point selection strategy: The dynamic clean air barrier zone and low-pressure capture zone are the areas that directly reflect the system's function, and a sufficient number of points must be selected within them (e.g., one point each at the beginning, middle, and end of the barrier zone, and points at the center and edge of the capture zone); the flow field state near the exhaust duct inlet and the makeup air duct outlet directly affects the control effect and must be considered as key points; the physical locations of all sensor arrays must have corresponding grid nodes in the three-dimensional computational grid, and these points are the only window for obtaining real-world data. By merging the above three types of locations (functional areas, duct openings, and sensor points), removing duplicates, a suitable set of spatial key points (e.g., 15-25) is finally determined. At the start of each control cycle, two operations are performed simultaneously: First, acquire the latest multiphysics data from all sensors deployed at key points, including air pressure (scalar), airflow velocity (vector, including magnitude and direction), and temperature (scalar). Second, perform a rapid forward calculation on the current model state (based on the correction results of the previous cycle and the current fan control commands), outputting the model's predicted air pressure, airflow velocity, and temperature values ​​at each of the aforementioned key points under the current operating conditions. Then, compare the multiphysics data with the model predictions point-by-point and quantity-by-quantity. Air pressure scalar error = measured air pressure - predicted air pressure (unit: Pascal); airflow velocity vector error = measured velocity vector - predicted velocity vector, which is typically decomposed into velocity magnitude error and direction angle error; temperature scalar error = measured temperature - predicted temperature (unit: degrees Celsius). Arrange all these error values ​​at key points sequentially into an array or vector; this set is the multidimensional model prediction error signal set. The Kalman filter is an optimal estimation algorithm that dynamically provides the optimal estimate of the state based on model predictions and measured data, naturally filtering out high-frequency noise. A standard Kalman filter requires known statistical characteristics (covariance matrices) of the system process noise and measurement noise. In a kitchen environment, these noise characteristics can change (e.g., increased disturbance due to door and window opening and closing). An adaptive Kalman filter can estimate or adjust these noise covariance matrices online in real time, keeping the filter in optimal operating condition and thus more robust to environmental changes. Internally, the filter maintains an estimate of the model bias state. It combines the bias estimate from the previous period (prior), the current error observation, and the adaptive noise estimate, using a set of recursive formulas to calculate the optimal smoothed estimate of the model bias for the current period. This optimal smoothed estimate is a vector with the same dimension as the error signal set, but it is no longer a coarse instantaneous error; instead, it is a filtered and smoothed, more stable signal that better reflects the systematic model bias. For example, it might determine that "there is a persistent systematic underestimation of -0.5 Pa in the air pressure forecast for the eastern region." During the system development phase, it is necessary to establish a mapping relationship between model parameters and flow field characteristics through a large number of simulation experiments or physical bench experiments.In each control cycle, the optimal smoothed estimate of the model bias (i.e., the systematic prediction bias pattern) is input into the pre-calibrated mapping relationship described above. This mapping relationship (such as a trained neural network) is analyzed and outputs a set of quantitative corrections. The system immediately applies these quantitative corrections to the internal parameters of the three-dimensional dynamic airflow field calculation model. The model uses these updated, more accurate parameters when performing prediction calculations in the next control cycle.

[0038] In some embodiments, the step of using the dynamic ideal air pressure field distribution map as a target, calculating the air volume curves and air direction angles of the exhaust duct and the make-up air duct required to establish or maintain the dynamic clean air barrier zone through the corrected three-dimensional dynamic airflow field calculation model, and driving the fan and air direction adjustment mechanism to execute, specifically includes: Using the corrected three-dimensional dynamic airflow field calculation model as the predictive controller and the dynamic ideal air pressure field distribution map as the tracking target, rolling time domain optimization calculation is performed. In the rolling time-domain optimization, the optimization objective is to minimize the system tracking error and the overall energy consumption. The future control sequence that satisfies the performance constraints of the dynamic clean air barrier zone is solved. The future control sequence defines the optimal air volume curve and air direction angle of the exhaust duct and the make-up air duct in a preset time period from the current moment. The first control quantity of the future control sequence is sent to the fan and the wind direction adjustment mechanism, and the actual execution feedback of the fan and the wind direction adjustment mechanism is collected. Adjustments are made based on the comparison result between the actual execution feedback and the first control quantity.

[0039] The corrected three-dimensional dynamic airflow field calculation model serves as the predictive model within the Model Predictive Control (MPC) framework. MPC operates within a continuously scrolling time window, divided into two parts: the prediction time domain and the control time domain. In each control cycle (e.g., at t=k), MPC performs the following operations: using the current sensor-measured actual state as initial conditions, within the prediction time domain (e.g., in the future T=10 seconds), it simulates the system's response under different future control sequences (i.e., a series of assumptions about how the wind turbine should operate from k to k+T). Through an optimization algorithm, it finds an optimal sequence from countless possible control sequences, ensuring the predicted response best matches the target. The first control action of this optimal control sequence (i.e., the action that should be executed immediately at t=k) is then actually issued to the wind turbine. In the next cycle (t=k+1), the system repeats the entire process based on the new measured state, and the window scrolls forward one step, hence the term "scrolling time domain optimization." The optimization variable is the future control sequence, which arranges the values ​​of exhaust fan speed, make-up air fan speed, exhaust air direction angle, and make-up air direction angle into a long vector at each discrete time step (e.g., every 0.5 seconds) in the future control time domain. The optimization algorithm works to find the optimal value of this vector. The cost function is used to evaluate the quality of any hypothetical control sequence. It contains two core parts: (1) Tracking error term: calculates the sum of squares of the differences between the air pressure / velocity predicted by the model at the spatial key points and the corresponding target values ​​in the dynamic ideal air pressure field distribution map in the prediction time domain. Minimizing this term means forcing the actual flow field to approximate the ideal flow field. (2) Comprehensive energy consumption term: estimates the energy consumed in executing the control sequence. The fan power consumption is approximately proportional to the cube of the speed. This term is usually the weighted sum of the cubes of the exhaust and make-up air fan speeds in the prediction time domain. Minimizing this term means pursuing energy saving. The two terms are weighted and summed by a weight coefficient (λ) to form a comprehensive cost function: total cost = tracking error + λ * comprehensive energy consumption. Adjusting λ can change the system's emphasis on "purification effect" and "energy saving". Optimization must be performed under the following hard constraints: Performance path constraints: This is the most important performance constraint. It requires that key performance indicators of the dynamic clean air barrier zone (such as "barrier zone pressure gradient retention rate" greater than 90%, or "predicted oil fume escape concentration" lower than a certain threshold) must always be met throughout the entire prediction time domain. This ensures that any optimized control strategy will not sacrifice core purification functions. Actuator physical constraints: Fans have minimum and maximum speed limits; airflow adjustment mechanisms have angular rotation range and angular velocity limits. These are applied directly to the optimization variables as boundary constraints. The optimization variables, comprehensive cost function, and constraints defined above can be expressed as a standard constrained optimization mathematical problem (usually a nonlinear programming problem).An efficient real-time optimization solver (such as Sequential Quadratic Programming (SQP), interior-point method, or an efficient solver for convex problems) is used for online solving. The solver output is the optimal control sequence that satisfies all constraints and minimizes the comprehensive cost function. The first control variable (i.e., the control command for the current time k) is extracted from the optimal control sequence and prepared to be issued to the fan and wind direction adjustment mechanism. Before issuance, the system collects actual execution feedback (such as actual speed and actual angle) through the feedback signals from the frequency converter (actual speed and current) and the encoder feedback from the wind direction mechanism. The actual execution feedback is compared with the initially issued "first control variable" command. If there is a steady-state error (e.g., the command is 3000 rpm, but the actual speed is only 2950 rpm) or dynamic lag, the system records this error and generates a fast PID-type feedback correction variable, which is immediately compensated for in the output of the next control cycle. The characteristic error of this actuator is also fed back to the model correction stage to fine-tune the response characteristic parameters of the fan / damper in the model (such as the flow-speed curve coefficient), making the model prediction more accurate in the next cycle.

[0040] In some embodiments, the step of using the corrected three-dimensional dynamic airflow field calculation model as a predictive controller and the dynamic ideal air pressure field distribution map as the tracking target to perform rolling time-domain optimization calculations specifically includes: The corrected three-dimensional dynamic airflow field calculation model is discretized into a state space prediction model suitable for real-time control. The state space prediction model takes the control input of the smoke exhaust fan corresponding to the smoke exhaust duct and the make-up air fan corresponding to the make-up air duct as variables, and the air pressure and airflow velocity at the key points in the dynamic ideal air pressure field distribution map as state outputs. In each control cycle, the current physical state fed back by the sensor array is used as the initial condition, and the state space prediction model is used to perform forward simulation prediction of the dynamic evolution of the system state in the future preset time domain. Within the future preset time domain, key performance indicators of the dynamic clean air barrier zone are defined as path constraints that must be met in the optimization process. The key performance indicators include the barrier zone pressure gradient retention rate and the oil fume escape concentration threshold. Based on the results of forward simulation prediction and the path constraints, an optimization problem is constructed with the objective of minimizing the deviation between the predicted trajectory and the ideal target trajectory.

[0041] Even with simplification, the original three-dimensional dynamic airflow field calculation model is overly complex and computationally time-consuming due to the relationship between its inputs (boundary conditions) and outputs (physical quantities of the entire three-dimensional field), making it unsuitable for direct use in controllers requiring optimization problems to be solved multiple times per second. State-space models, the standard form in control theory, use a set of state variables, input variables, and output variables to describe the dynamic behavior of a system through matrix equations, offering extremely high computational efficiency. First, determine the control input variables (u): these are the quantities that the controller can directly manipulate. In this system, these typically include the exhaust fan's speed command (n_exhaust), the makeup air fan's speed command (n_supply), and their respective airflow adjustment angles (θ_exhaust, θ_supply). These four (or more) variables constitute the control input vector u. Next, determine the state output variables (y): these are the quantities that need to be controlled and tracked. These quantities directly originate from the core focus of the dynamic ideal pressure field distribution diagram. Instead of tracking tens of thousands of grid points across the entire field, the system selects a representative set of key spatial points. These points must cover: the core and boundaries of the low-pressure capture zone; the beginning, middle, and end of the dynamic clean air barrier zone; and other sensitive or easily disturbed areas. Each key point needs to track its pressure value (p) and the component (v) of its airflow velocity vector in the key direction. For example, selecting 20 key points, each outputting one pressure and one dominant directional velocity, constitutes a 40-dimensional output vector y. State-space equations are established: through system identification or model reduction methods, a discrete-time state-space model of the following form is derived from the corrected flow field model: x(k+1)=A*x(k)+B*u(k); y(k) = C*x(k) + D*u(k); Here, x(k) is the internal state vector at time k (which may not have direct physical meaning, but can encompass the dynamic characteristics of the flow field, such as vorticity and kinetic energy), u(k) is the control input vector at time k (fan speed, angle), y(k) is the observed output vector at time k (air pressure and velocity at key points), and A, B, C, and D are constant matrices obtained through data fitting or physical derivation. This model is a state-space prediction model. Given the current state x(k) and a series of future control inputs u(k), u(k+1)..., it can predict a series of future outputs y(k+1), y(k+2)... extremely quickly (in microseconds).

[0042] At the start of each control cycle, the sensor array provides the actual air pressure and velocity measurements y_measured at those key points in the physical space. The controller uses a state observer (such as a Kalman filter) to calculate the optimal estimate x(k) of the current system state, combining the current measurement y_measured with the state estimate x(k-1) from the previous cycle. This x(k) serves as the initial condition for this prediction. The controller performs an internal loop calculation. It assumes that a series of control input sequences to be optimized [u(k), u(k+1), ..., u(k+N-1)] will be applied within a preset time domain in the future (e.g., N=20 control cycles in the future, corresponding to 10 seconds in the future). Substituting the initial state x(k) and the assumed control sequence into the state-space prediction model, it recursively calculates a series of predicted output sequences [ŷ(k+1), ŷ(k+2), ..., ŷ(k+N)]. This sequence describes how the system will evolve in the future if the assumed control actions are executed, specifically how the air pressure and velocity at the key points will change. If only tracking the target and energy saving are pursued, the optimization algorithm may give an "energy-saving" solution with extremely low airflow, but this obviously cannot purify the fumes. Therefore, hard constraints must be imposed. At each future moment in the prediction time domain, the air pressure gradient value on the specified path within the barrier zone is calculated based on the predicted output ŷ. This gradient value must always be greater than or equal to a preset minimum effective threshold (e.g., not less than 85% of the ideal gradient value). This ensures that the aerodynamic structure of the barrier is effective throughout the entire prediction future. Fume escape concentration threshold: A simplified fume transport sub-model can be integrated into the prediction model. Based on the predicted airflow velocity field ŷ, the diffusion path of fume particles is simulated. It is required that the fume concentration at the boundary of the kitchen personnel activity area or non-capture area is always below a health- or sensory-permissible maximum value (e.g., 1 mg / m³) within the prediction time domain. This effectively prevents fume escape. These indicators are quantified as inequalities with respect to the predicted state ŷ and / or control input u, which must be satisfied at every step of the optimization solution, and are therefore called path constraints. Optimization variables: the future control sequence to be solved [u(k), u(k+1), ..., u(k+N-1)]. Objective function: a scalar function is constructed, the core of which is to minimize the deviation between the predicted trajectory and the ideal target trajectory. Ideal target trajectory: derived from the "dynamic ideal pressure field distribution map", which provides the ideal pressure and velocity target values ​​y_ref(k+1), ..., y_ref(k+N) for each spatial key point at each future time (corresponding to the prediction time domain). Deviation calculation: at each time point in the prediction time domain, the difference between the predicted output ŷ and the ideal target y_ref is calculated. Typically, the sum of the squares of the pressure and velocity deviations at each key point is taken to emphasize large deviations.The objective function is typically in the form: J = Σ[(ŷ(i)-y_ref(i))^T*Q*(ŷ(i)-y_ref(i))], where the summation covers the entire prediction time domain, and Q is a weight matrix used to adjust the degree of emphasis on different key points and different physical quantities. Minimizing J is to drive the predicted flow field to infinitely approximate the ideal flow field.

[0043] In some embodiments, solving for the future control sequence that satisfies the dynamic clean air barrier zone performance constraints includes: In the optimization problem, the sequence of changes in air volume and air direction angle of the exhaust duct and the make-up air duct in the future time domain is defined as the control variables to be optimized. A comprehensive cost function is established, wherein the first term of the comprehensive cost function is the tracking error between the predicted system state and the target value of the dynamic ideal pressure field distribution map, and the second term of the comprehensive cost function is the comprehensive energy consumption estimate that is positively correlated with the cubic speed of the wind turbine. The minimum value of the comprehensive cost function is obtained by satisfying all path constraints and actuator physical constraints. The smoothness and feasibility of the optimal control sequence are verified to obtain the future control sequence.

[0044] The controller divides time forward into several small steps (e.g., 0.5 seconds per step for the next 10 seconds). For each future time step, it needs to plan an airflow rate (corresponding to fan speed) and a wind direction angle for both the exhaust fan and the makeup air fan. Arranging these planned speed and angle values ​​in chronological order creates a long list of tasks, i.e., a sequence of control variables. The optimization algorithm's task is to find the best one from this list of countless possibilities. This best one is determined by a comprehensive cost function. The comprehensive cost function has two core objectives: tracking error and comprehensive energy consumption estimation. Tracking error (performance priority): This measure is the difference between the system's predicted future flow field (pressure and wind speed at key points) and the target set in the dynamic ideal pressure field distribution map if a certain hypothetical control scheme is implemented. Calculation method: Compare the differences between the predicted and target values ​​at each key point and at each future time, and sum the squares of all differences. The smaller the difference and the smaller the sum of squares, the closer the control effect is to the ideal blueprint. Minimizing this term ensures that the purification system forms an effective low-pressure capture zone and air barrier. Comprehensive Energy Consumption Estimation (Efficiency Priority): This estimates the electrical energy required to execute the control scheme. Fan power consumption is physically highly correlated with the cube of rotational speed (i.e., a slight increase in speed leads to a significant increase in power consumption). Calculation Method: The rotational speeds of the exhaust and makeup air fans at each moment within the prediction time domain are cubed and weighted summed. The smaller this value, the lower the estimated energy consumption of the scheme. Minimizing this value aims to pursue system economy while ensuring effectiveness, avoiding unnecessary energy waste. The final score is the weighted sum of these two items. By adjusting the weighting coefficients between the two items, it can be determined whether the control system prioritizes effectiveness or energy saving. The smaller the value of this comprehensive cost function, the better the overall performance of the control scheme in terms of effectiveness and energy efficiency. Path Constraints (Efficiency Red Line): This is a mandatory requirement. Throughout the entire future period of optimization, key performance indicators of the dynamic clean air barrier zone (such as the degree of pressure gradient maintenance and the predicted concentration of oil fume escape) must always meet the preset safety or performance standards. Any solution that could lead to barrier failure or excessive oil fume emissions, no matter how energy-efficient, will be immediately rejected. Physical constraints (execution limits): The fan has maximum and minimum speed limits; the airflow adjustment mechanism has rotation angle range and speed limits. The optimized solution cannot require the actuator to perform actions it cannot. The controller calls the built-in optimization solver, which explores various possible combinations of control sequences. It evaluates whether each solution violates the above constraints and calculates its comprehensive cost. Finally, it outputs the solution with the lowest comprehensive cost score while satisfying all constraints, i.e., the optimal control sequence. The optimal sequence is checked to ensure that changes in fan speed and airflow angle are gradual. Sudden, large jumps in commands may cause fan overload, mechanical vibration or noise, and may also impact the flow field, disrupting the established stable barrier.If the sequence is found to be insufficiently smooth, the instruction sequence will be filtered or smoothed. A more refined secondary check will also be performed to ensure that each instruction falls within the precise controllable range of the actuator. For example, considering more nuanced characteristics such as motor response delay and mechanism dead zones, the instructions will be fine-tuned to ensure accurate and reliable execution. After verification and necessary corrections, the resulting future control sequence is truly ready to be issued to the hardware for execution. The controller will immediately retrieve the first instruction of this sequence (i.e., the action that should be executed immediately at the current moment) and send it to the fan and regulating mechanism.

[0045] In some embodiments, the method further includes: Before the first control quantity is issued, the transient impact of the flow field that may be caused by the execution of the control command is predicted based on the state space prediction model, and a dynamic feedforward compensation signal is generated and superimposed on the first control quantity. After the first control variable is executed, the actual performance index of the dynamic clean air barrier zone is calculated based on the real-time data of the sensor array, and compared with the performance index predicted by the model to generate a real-time performance micro-deviation signal. Based on the historical trend of the performance micro-deviation signal, the weight ratio of the system tracking error term and the comprehensive energy consumption term in the comprehensive cost function is dynamically adjusted by an adaptive law in the rolling time-domain optimization calculation. The dynamic feedforward compensation signal and the adjusted weight ratio are applied to the rolling time-domain optimization calculation of the next control cycle.

[0046] When the controller decides to significantly increase the fan speed or rapidly change the airflow direction, this step command forces a sudden change in the flow field. Before the airflow stabilizes again, brief vortices, pressure fluctuations, or even suction voids may occur, which can instantly disrupt the established dynamic clean air barrier. Before officially issuing the first control variable (u(k)) to the hardware, the system inputs it into the state-space prediction model for an ultra-short-term transient simulation (e.g., predicting dynamic details within the next 0.5 seconds). The model quickly predicts how the airflow velocity and pressure at key points within the dynamic clean air barrier will change instantaneously under the action of the command u(k). The system pays particular attention to whether these predicted values ​​will exhibit drastic reverse fluctuations or fall below the threshold for maintaining effectiveness. If a harmful transient shock is predicted (e.g., a sudden 30% drop in wind speed within the barrier), the control algorithm immediately generates a dynamic feedforward compensation signal. This signal is typically a pulse or small waveform with a rapidly decaying amplitude, opposite in phase to the main control command u(k). The compensation signal is superimposed on the original command u(k) to form a corrected final execution command. For example, if the original command is "increase the speed from 1000 to 2000 RPM," the compensation signal might be "briefly increase to 2200 RPM and then drop back to 2000 RPM," allowing the actual flow field to transition to the new state more smoothly. After a slight delay following command execution (waiting for the flow field to initially stabilize, such as 100-200 milliseconds), the system performs calculations based on the latest real-time data from the sensor array. For example, multiple sensors along the barrier path are selected, and the current actual pressure gradient is calculated based on their pressure values; or, combined with oil fume observations from image sensors, the current oil fume escape concentration is assessed to determine the perception level, which is the actual performance index. Simultaneously, the performance index values ​​(i.e., predicted barrier gradient and predicted escape concentration) predicted by the model during the previous optimization cycle are retrieved from within the MPC. The actual values ​​are compared with the model predictions, and their relative or absolute differences are calculated to generate a real-time performance micro-deviation signal. For example, the actual pressure gradient is 5% lower than the model prediction. Instead of reacting drastically to a single minor deviation, the system continuously records the performance deviation signal and analyzes its historical trend (e.g., whether the deviation has been consistently positive, consistently negative, or fluctuating randomly around zero over the past 30 seconds). If the performance deviation remains negative (i.e., the actual performance is consistently slightly worse than the model prediction), it indicates that the current optimization strategy may be overly focused on energy saving (the weight of the overall energy consumption term is relatively too high), resulting in a slightly conservative control action and insufficient performance margin. In this case, an adaptive law (a set of pre-defined heuristic rules or a lightweight learning algorithm) dynamically increases the weight of the tracking error term in the overall cost function while correspondingly decreasing the weight of the overall energy consumption term. If the performance deviation remains positive and stable (the actual performance is consistently better than the prediction), the system can cautiously and slightly increase the weight of the energy consumption term to explore more energy-efficient possibilities.Through this slow, long-term feedback-based adjustment, the system can find a dynamically optimal balance point that matches environmental conditions, ensuring sufficient performance margin while pursuing ultimate energy efficiency. Empirical parameters for generating the compensation signal (such as the compensation amplitude and time constant) are recorded. If similar control commands recur, the system can generate compensation signals faster and more accurately. The new weight ratios, adjusted by the adaptive law, will be directly applied to the rolling time-domain optimization calculation in the next control cycle. In the next cycle, MPC will use the new weights to calculate a new optimal control sequence, with its decision strategy fine-tuned based on the performance feedback from the previous cycle. Simultaneously, new control commands will benefit from more accurate feedforward compensation before being issued, resulting in smoother execution. After execution, new performance feedback will be generated, further driving the fine-tuning of weights and the model.

[0047] The above describes a method for purifying indoor air with an independent dual-duct system for a range hood. The computer system in the embodiments of this application will be described in detail below in conjunction with the above method for purifying indoor air with an independent dual-duct system for a range hood.

[0048] Please see Figure 3 This is a schematic diagram of an exemplary hardware structure of a computer system in an embodiment of this application.

[0049] In some embodiments, the computer system 300 includes a computer device, which may be a terminal device. The computer device includes a processor 301, a memory 302, a sensor module 303, a communication module 304, an input device 305, and an output device 306 connected via a system bus. The processor 301 of the computer device provides computing and control capabilities. The memory 302 of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database is used to store data.

[0050] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0051] In some embodiments of this application, a computer-readable storage medium is provided, including instructions that, when executed on a computer system 300, cause the computer system 300 to perform a method for purifying indoor air in a range hood with independent dual ducts according to an embodiment of this application.

[0052] In some embodiments of this application, a computer program product is also provided, which, when run on a computer system 300, causes the computer system 300 to execute a method for purifying indoor air in a range hood with independent dual ducts according to an embodiment of this application.

[0053] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0054] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0055] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for purifying indoor air with an independent dual-duct system in a range hood, characterized in that, include: Based on the fixed spatial positions of the range hood and cooktop, as well as the real-time indoor object layout and door and window opening and closing status, a three-dimensional dynamic airflow field calculation model is established, including the inlet and outlet of the exhaust duct, the inlet and outlet of the make-up air duct, and the cooktop area. Based on the current cooking mode and the predicted oil fume concentration, the three-dimensional dynamic air flow field calculation model calculates and sets a dynamic ideal air pressure field distribution map in real time with the preset high-temperature smoke core area of ​​the stove as the protection target. The dynamic ideal air pressure field distribution map includes a low-pressure capture zone formed above the oil fume generation area, and a dynamic clean air barrier zone with an air pressure gradient between the low-pressure capture zone and the make-up air duct. By deploying a sensor array in the stove area, the exhaust duct, and the make-up air duct, real-time multi-physics field data of the physical space is collected. The multi-physics field data includes air pressure, airflow velocity, and temperature data. The multi-physics field data is compared with the prediction data of the three-dimensional dynamic airflow field calculation model. Based on the multi-dimensional model prediction error signal set, the three-dimensional dynamic airflow field calculation model is corrected in real time through an adaptive filtering algorithm. Using the dynamic ideal air pressure field distribution map as the target, the air volume curves and air direction angles of the exhaust duct and the make-up air duct required to establish or maintain the dynamic clean air barrier area are calculated by the corrected three-dimensional dynamic air flow field calculation model, and the fan and air direction adjustment mechanism are driven to perform the calculation.

2. The method of claim 1, wherein, The process of calculating and setting a dynamic ideal air pressure field distribution map in real time within the three-dimensional dynamic airflow field calculation model based on the current cooking mode and predicted oil fume concentration, with the preset high-temperature smoke core area of ​​the stove as the protection target, specifically includes: Based on the stove status parameters and the recognition of oil fume patterns by image sensors, the current cooking mode is determined in real time, and the oil fume concentration and diffusion trend within the preset time window are predicted. Based on the oil fume concentration and diffusion trend, the dynamic spatial attributes of the low-pressure capture zone are dynamically calculated. The dynamic spatial attributes include the target pressure value, the three-dimensional spatial range, and the relative distance between the low-pressure capture zone and the make-up air duct. Based on the dynamic spatial properties, calculate the pressure gradient curve decreasing from the make-up air side to the fume side within the dynamic clean air barrier zone between the low-pressure capture zone and the inlet of the make-up air duct. The target pressure value of the low-pressure capture zone, the pressure gradient curve of the dynamic clean air barrier zone, and the ambient air pressure value of the kitchen are fused together to generate the dynamic ideal air pressure field distribution map.

3. The method of claim 1, wherein, The multi-dimensional model-based prediction error signal set is used to correct the three-dimensional dynamic airflow field calculation model in real time through an adaptive filtering algorithm, including: Based on the spatial locations of the dynamic clean air barrier zone and the low-pressure capture zone, the inlet and outlet locations of the smoke exhaust duct and the make-up air duct, and the physical layout of the sensor array, a set of spatial key points for model calibration are determined. In each control cycle, pressure scalar error, airflow velocity vector error, and temperature scalar error are extracted from the actual multiphysics data and the prediction data of the three-dimensional dynamic airflow field calculation model at the corresponding spatial key points, respectively, and together they constitute the multidimensional model prediction error signal set. Based on the amplitude and frequency characteristics of each error component in the multi-dimensional model prediction error signal set, the optimal smooth estimate of the model bias is obtained through an adaptive Kalman filter. By using a pre-calibrated mapping relationship, the optimal smoothing estimate is inversely analyzed into a quantitative correction amount for specific key parameters in the three-dimensional dynamic airflow field calculation model. The three-dimensional dynamic airflow field calculation model is then corrected based on the quantitative correction amount. The specific key parameters include turbulent viscosity coefficient, wall drag coefficient, or local momentum source term coefficient.

4. The method of claim 1, wherein, The process involves using the dynamic ideal air pressure field distribution map as a target, and through the corrected three-dimensional dynamic airflow field calculation model, calculating the airflow curves and airflow direction angles of the exhaust duct and the make-up air duct required to establish or maintain the dynamic clean air barrier zone, and driving the fan and airflow direction adjustment mechanism to execute these calculations. Specifically, this includes: Using the corrected three-dimensional dynamic airflow field calculation model as the predictive controller and the dynamic ideal air pressure field distribution map as the tracking target, rolling time domain optimization calculation is performed. In the rolling time-domain optimization, the optimization objective is to minimize the system tracking error and the overall energy consumption. The future control sequence that satisfies the performance constraints of the dynamic clean air barrier zone is solved. The future control sequence defines the optimal air volume curve and air direction angle of the exhaust duct and the make-up air duct in a preset time period from the current moment. The first control quantity of the future control sequence is sent to the fan and the wind direction adjustment mechanism, and the actual execution feedback of the fan and the wind direction adjustment mechanism is collected. Adjustments are made based on the comparison result between the actual execution feedback and the first control quantity.

5. The method of claim 4, wherein, The step of using the corrected three-dimensional dynamic airflow field calculation model as a predictive controller and the dynamic ideal air pressure field distribution map as the tracking target to perform rolling time-domain optimization calculations specifically includes: The corrected three-dimensional dynamic airflow field calculation model is discretized into a state space prediction model suitable for real-time control. The state space prediction model takes the control input of the smoke exhaust fan corresponding to the smoke exhaust duct and the make-up air fan corresponding to the make-up air duct as variables, and the air pressure and airflow velocity at the key points in the dynamic ideal air pressure field distribution map as state outputs. In each control cycle, the current physical state fed back by the sensor array is used as the initial condition, and the state space prediction model is used to perform forward simulation prediction of the dynamic evolution of the system state in the future preset time domain. Within the future preset time domain, key performance indicators of the dynamic clean air barrier zone are defined as path constraints that must be met in the optimization process. The key performance indicators include the barrier zone pressure gradient retention rate and the oil fume escape concentration threshold. Based on the results of forward simulation prediction and the path constraints, an optimization problem is constructed with the objective of minimizing the deviation between the predicted trajectory and the ideal target trajectory.

6. The method according to claim 5, characterized in that, The solution for the future control sequence that satisfies the dynamic clean air barrier zone effectiveness constraint includes: In the optimization problem, the sequence of changes in air volume and air direction angle of the exhaust duct and the make-up air duct in the future time domain is defined as the control variables to be optimized. A comprehensive cost function is established, wherein the first term of the comprehensive cost function is the tracking error between the predicted system state and the target value of the dynamic ideal pressure field distribution map, and the second term of the comprehensive cost function is the comprehensive energy consumption estimate that is positively correlated with the cubic speed of the wind turbine. The minimum value of the comprehensive cost function is obtained by satisfying all path constraints and actuator physical constraints. The smoothness and feasibility of the optimal control sequence are verified to obtain the future control sequence.

7. The method of claim 6, wherein, The method further includes: Before the first control quantity is issued, the transient impact of the flow field that may be caused by the execution of the control command is predicted based on the state space prediction model, and a dynamic feedforward compensation signal is generated and superimposed on the first control quantity. After the first control variable is executed, the actual performance index of the dynamic clean air barrier zone is calculated based on the real-time data of the sensor array, and compared with the performance index predicted by the model to generate a real-time performance micro-deviation signal. Based on the historical trend of the performance micro-deviation signal, the weight ratio of the system tracking error term and the comprehensive energy consumption term in the comprehensive cost function is dynamically adjusted by an adaptive law in the rolling time-domain optimization calculation. The dynamic feedforward compensation signal and the adjusted weight ratio are applied to the rolling time-domain optimization calculation of the next control cycle.

8. A computer system comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program comprises instructions that, when executed by the processor, cause the processor to perform the method of any one of claims 1 to 7. The processor executes the computer program to implement the steps of the method according to any one of claims 1-7.

9. A computer readable storage medium having stored thereon computer programs / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method described in any one of claims 1-7.

10. A computer program product comprising computer programs / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method described in any one of claims 1-7.

Images