Automatic control method and system of variable frequency range hood

Abstract

This invention discloses an automatic control method and system for a variable frequency range hood, belonging to the field of intelligent control technology for home appliances. The method includes: using non-contact multimodal sensors to collect real-time data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment; constructing a multivariate fusion model to generate a comprehensive environmental state vector through feature extraction and weight allocation; constructing an LSTM prediction model to abstract the current oil fume distribution state and surrounding environmental characteristics, and combining an airflow-energy consumption-noise multi-objective optimization model and a genetic algorithm to generate an optimal fan speed control strategy; and adjusting the fan speed in real-time according to the optimal fan speed control strategy. This invention uses non-contact multimodal sensors to collect real-time data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment, and combines an edge computing chip to run a multivariate fusion algorithm and a time-series prediction model to achieve efficient, energy-saving, and low-noise airflow control.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for home appliances, and in particular to an automatic control method and system for a variable frequency range hood. Background Technology

[0002] With the application of variable frequency technology in the field of range hoods, it has become possible to achieve adaptive airflow control by adjusting the motor speed. However, existing technologies still have shortcomings in terms of multi-parameter collaborative control, dynamic response speed, and level of intelligence in complex kitchen environments, leading to a decline in user experience, increased energy consumption, and shortened lifespan.

[0003] A search revealed Chinese invention patent CN113983518B, which discloses a fuzzy logic intelligent control method and system for a variable frequency range hood. This patent acquires user input information (including expected command value and command input frequency), fuzzifies the input information, and combines it with the current state quantity to determine the control quantity, thereby achieving speed control of the variable frequency range hood. However, this technical solution relies primarily on user input information for control and lacks real-time sensing capabilities for environmental parameters such as oil fume concentration and temperature during actual cooking, resulting in limited dynamic response capabilities. Furthermore, the system's fuzzy rule base needs to be pre-set, making it difficult to adapt to complex and changing actual working conditions, potentially leading to insufficient control accuracy.

[0004] A search revealed Chinese invention patent CN109163364B, which discloses an intelligent range hood and its control method. This patent detects smoke signals using a smoke sensor and monitors changes in image clarity using an imaging device, thereby adjusting the drive speed of a variable frequency motor. This technical solution achieves intelligent control based on changes in image clarity, improving the automation level of the range hood. However, this solution relies on the calibration area and reference image of the imaging device, making it susceptible to factors such as lighting conditions and lens contamination, leading to a decrease in the accuracy of the monitoring results. Furthermore, this patent does not fully consider the dynamic relationship between smoke concentration and exhaust volume, potentially failing to adjust the fan speed in a timely manner under high smoke concentration conditions, affecting purification efficiency and user experience.

[0005] The aforementioned problems indicate that existing variable frequency range hood control technologies still have shortcomings in multimodal environmental perception, dynamic response speed, and multi-parameter collaborative optimization. Therefore, this invention provides an automatic control method and system for variable frequency range hoods. It aims to collect real-time data on fume concentration, temperature, and airflow disturbance using non-contact multimodal sensors, combine this data with an edge computing chip to run a multi-variable fusion algorithm, construct a multi-objective optimization model for airflow, energy consumption, and noise, and introduce a time-series prediction module to predict fume change trends. This significantly improves the system's dynamic response capability and intelligence level, meeting the modern kitchen's demand for efficient, energy-saving, and low-noise range hoods. Summary of the Invention

[0006] To address the technical problems existing in the prior art, this invention provides an automatic control method and system for a variable frequency range hood. Its core lies in the real-time acquisition of oil fume concentration data, temperature data, and airflow disturbance data in the kitchen environment through non-contact multimodal sensors, and the combination of multivariate fusion algorithms and time series prediction models to achieve efficient, energy-saving, and low-noise airflow control.

[0007] In a first aspect, the present invention provides an automatic control method for a variable frequency range hood, comprising the following steps:

[0008] S1. Multimodal data acquisition: Non-contact multimodal sensors are used to collect data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment in real time, and the collected data is transmitted to an edge computing chip for processing.

[0009] S2. Multivariate fusion analysis: Based on the collected oil fume concentration data, temperature data, and airflow disturbance data, a multivariate fusion model is constructed, and a comprehensive environmental state vector is generated through feature extraction and weight allocation.

[0010] S3. Time-series prediction and dynamic optimization: Based on the comprehensive environmental state vector, an LSTM prediction model is constructed to abstract the current oil fume distribution state and surrounding environmental characteristics. The optimal fan speed control strategy is generated by combining the air volume-energy consumption-noise multi-objective optimization model and the genetic algorithm.

[0011] S4. Fan control execution: The optimal fan speed control strategy is sent to the variable frequency motor drive module to adjust the fan speed in real time.

[0012] S5. System monitoring and feedback optimization: Establish a remote communication mechanism to monitor the wind turbine's operating status and environmental parameter changes in real time, and continuously optimize system performance through a closed-loop feedback mechanism.

[0013] Preferably, the multivariate fusion analysis includes the following steps:

[0014] S21. Data Acquisition: Collect historical data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment as a public dataset.

[0015] S22. Data preprocessing: Standardize the public dataset to remove outliers and noise interference, and obtain standardized data.

[0016] S23. Feature engineering: Extracting standard data features from standardized data and assigning weights to them;

[0017] S24. Divide the dataset. Randomly select 70% of the standardized data as the training set for model training and 30% as the test set for model testing.

[0018] S25. Model building: Construct a multivariate fusion model and use an adaptive weighting algorithm to dynamically adjust the weights of each feature;

[0019] S26. Model training: Use the training set to train the multivariate fusion model, and adjust the weight vector of the standardized data through multiple iterations until the model converges.

[0020] S27. Model validation: Input the test set into the multivariate fusion model and use mean squared error for validation.

[0021] S28. Run, and generate a comprehensive environmental state vector in real time through a multivariate fusion model.

[0022] Preferably, the step of extracting relevant features from standardized data and assigning weights, wherein the standardized data features include oil fume concentration features, temperature features, and frequency features, includes the following steps:

[0023] S231. Map the oil fume concentration data to the [0,1] interval through linear transformation to obtain the oil fume concentration features, which are used as the main feature input;

[0024] S232. Transform the temperature data into temperature features through differential calculation and use them as auxiliary feature input;

[0025] S233. Decompose the airflow disturbance data into frequency features through Fourier transform and use them as supplementary feature inputs;

[0026] The features are weighted and fused to form a comprehensive environmental state vector, which includes oil fume concentration, temperature change rate, and main frequencies.

[0027] Preferably, the step of dynamically adjusting the weights of each standard data feature using an adaptive weighting algorithm includes the following steps:

[0028] S251. Initialize the weight vector. Set the initial weight vector to a uniform distribution. The expression is: In the formula, n represents the number of standard data features, w 0 Represents the initial weight vector;

[0029] S252, Weight update: dynamically adjust the weights based on the contribution of each standard data feature to the comprehensive environmental state vector;

[0030] S253. Normalization process: Normalize the updated weight vector into a probability distribution.

[0031] The formula for calculating the weight update is as follows:

[0032]

[0033] The calculation formula for the normalization process is as follows:

[0034]

[0035] In the formula, This represents the weight of the i-th standard data feature at time t. Let α represent the weight of the i-th standard data feature at time t+1, and let ΔE represent the learning rate. i This represents the incremental contribution of the i-th standard data feature to the comprehensive environmental state vector.

[0036] Preferably, the time-series prediction and dynamic optimization includes the following steps:

[0037] S31. Initialize the LSTM model: Construct an LSTM model to predict the current oil fume distribution state and initialize the initial oil fume distribution state.

[0038] S32. Data input: The comprehensive environmental state vector is used as the input sequence of the LSTM model.

[0039] S33, LSTM unit calculation, each LSTM unit contains a forget gate, input gate and output gate; the current oil fume distribution state and surrounding environmental characteristics are obtained through LSTM units;

[0040] S34. Multi-objective optimization: Based on the prediction results of the LSTM model, construct a multi-objective optimization model for air volume, energy consumption, and noise.

[0041] S35. Optimization and solution: The genetic algorithm is used to solve the multi-objective optimization model of air volume-energy consumption-noise.

[0042] S36. Optimal strategy generation: When the multi-objective function of air volume-energy consumption-noise converges to a local optimum, the optimal fan speed control strategy is generated.

[0043] Preferably, the LSTM unit calculation includes the following steps:

[0044] S331. Forget Gate: Calculate the activation value of the forget gate to dynamically filter historical data and retain key historical data. The calculation formula is as follows:

[0045] f t =σ(W f ·[h t-1 ,x t ]+b f );

[0046] S332, Input Gate: Calculates the activation value and candidate oil fume distribution state of the input gate to filter and prepare key new data for the current input, providing candidate values ​​for state updates. The calculation formula is as follows:

[0047] i t =σ(W i ·[h t-1 ,x t ]+b i );

[0048]

[0049] S333. Cell state update: Calculate the new oil fume distribution state to fuse key historical data and key new data, generating the current oil fume distribution state. The calculation formula is as follows:

[0050]

[0051] S334, Output Gate: Calculate the activation value and oil fume distribution state of the output gate, control the output of the oil fume distribution state, and generate a hidden state that focuses on key information. The calculation formula is as follows:

[0052] o t =σ(W o ·[h t-1 ,x t ]+b o );

[0053] h t =o t ·tanh(C t );

[0054] In the formula, f t W represents the activation value of the forget gate. f h represents the weight matrix of the forget gate. t-1 x represents the hidden layer state at the previous time step. t b represents the comprehensive environment state vector at the current time step. f The bias vector representing the forget gate; i t This represents the activation value of the input gate. W represents the distribution state of candidate cooking fumes. i and W C The weight matrices b represent the input gate and the candidate oil fume distribution states, respectively. i and b C C represents the bias vectors of the input gate and the candidate oil fume distribution state, respectively; t C represents the distribution state of oil fume at the current time step. t-1 This indicates the distribution state of oil fume at the previous time step; o t h represents the activation value of the output gate.t W represents the hidden state at the current time step. o and b o σ and σ(·) represent the weight matrix and bias vector of the output gate, respectively; σ(·) represents the Sigmoid activation function; the hidden layer state is a dynamic abstract representation of the current oil fume distribution state and surrounding environmental characteristics by the LSTM model.

[0055] Preferably, the core calculation formula of the air volume-energy consumption-noise multi-objective optimization model is:

[0056] minF(x)=ω1·E(x)+ω2·N(x)+ω3·Q(x);

[0057] In the formula, F(x) represents the multi-objective function of air volume-energy consumption-noise, E(x) represents energy consumption, N(x) represents noise level, Q(x) represents oil fume purification efficiency, and ω1, ω2 and ω3 represent the weighting coefficients of energy consumption, noise and purification efficiency, respectively.

[0058] Preferably, the core calculation formula of the genetic algorithm includes:

[0059]

[0060] In the formula, x k Let x represent the solution of the k-th iteration. k+1 Let represent the solution in the (k+1)th iteration, and η represent the learning rate. This represents the gradient of the objective function.

[0061] Secondly, the present invention also provides an automatic control system for a variable frequency range hood, which applies an automatic control method for a variable frequency range hood as described above. The system includes a multimodal data acquisition module, a multivariable fusion analysis module, a time series prediction and dynamic optimization module, a fan control execution module, and a system monitoring and feedback optimization module.

[0062] The multimodal data acquisition module collects real-time data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment using a non-contact multimodal sensor, and transmits the collected data to the multivariate fusion analysis module.

[0063] The multivariate fusion analysis module is used to receive data sent by the multimodal data acquisition module, generate a comprehensive environmental state vector through feature extraction and weight allocation, and send the comprehensive environmental state vector to the time series prediction and dynamic optimization module.

[0064] The time-series prediction and dynamic optimization module is used to receive the comprehensive environmental state vector sent by the multivariate fusion analysis module, predict and abstract the current oil fume distribution state and surrounding environmental characteristics through the LSTM model, and generate the optimal fan speed control strategy by combining the air volume-energy consumption-noise multi-objective optimization model. The optimal fan speed control strategy is then sent to the fan control execution module and the system monitoring and feedback optimization module.

[0065] The wind turbine control execution module is used to receive the control strategy sent by the timing prediction and dynamic optimization module, and adjust the wind turbine speed in real time;

[0066] The system monitoring and feedback optimization module is used to receive information on the wind turbine operating status and environmental parameter changes sent by the wind turbine control execution module, and to continuously optimize system performance through a closed-loop feedback mechanism.

[0067] Preferably, the multimodal data acquisition module consists of an infrared sensor, a thermocouple sensor, and an ultrasonic sensor, used to acquire real-time data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment.

[0068] The fan control execution module adjusts the fan speed in real time by driving the variable frequency motor. The variable frequency motor drive module generates a corresponding PWM signal according to the target speed value in the control strategy.

[0069] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0070] This invention achieves intelligent, dynamic, and precise control of the kitchen environment through a multi-module collaborative architecture. The multi-modal data acquisition module relies on non-contact sensors to capture real-time multi-dimensional environmental information such as oil fume concentration, temperature, and airflow disturbance data, providing comprehensive data support for subsequent analysis. The multivariate fusion analysis module transforms heterogeneous data into a comprehensive environmental state vector through feature extraction and adaptive weight allocation, accurately representing the dynamic characteristics of the kitchen environment. The time-series prediction module predicts and abstracts the current oil fume distribution state and surrounding environmental characteristics based on an LSTM model, and generates an optimal fan speed control strategy by combining a multi-objective optimization model of airflow, energy consumption, and noise, balancing purification efficiency, energy consumption, and noise control. The fan control execution module adjusts the speed in real-time through frequency conversion drive to ensure efficient strategy implementation. The system monitoring and feedback optimization module continuously tracks the operating status and environmental changes through a closed-loop mechanism, driving iterative upgrades of system performance. This solution effectively improves the adaptability of the range hood to complex kitchen environments, achieving synergistic optimization of efficient purification, energy saving and noise reduction, and a comfortable experience, providing users with a smarter and healthier cooking environment. Attached Figure Description

[0071] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0072] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0073] Figure 1 This is a schematic diagram of the first process of an automatic control method for a variable frequency range hood according to the present invention;

[0074] Figure 2 This is a second flowchart illustrating an automatic control method for a variable frequency range hood according to the present invention.

[0075] Figure 3 This is a flowchart illustrating the structure of an automatic control system for a variable frequency range hood according to the present invention. Detailed Implementation

[0076] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0077] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0078] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" and "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0079] Example 1

[0080] See Figures 1-3As shown, this invention provides an automatic control method for a variable frequency range hood. The method is applied to an automatic control system for the variable frequency range hood. The system includes a multimodal data acquisition module, a multivariable fusion analysis module, a time-series prediction and dynamic optimization module, a fan control execution module, and a system monitoring and feedback optimization module. The method includes the following steps:

[0081] S1. Multimodal data acquisition: Non-contact multimodal sensors are used to collect data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment in real time, and the collected data is transmitted to an edge computing chip for processing.

[0082] S2. Multivariate fusion analysis: Based on the collected oil fume concentration data, temperature data, and airflow disturbance data, a multivariate fusion model is constructed, and a comprehensive environmental state vector is generated through feature extraction and weight allocation.

[0083] S3. Time-series prediction and dynamic optimization: Based on the comprehensive environmental state vector, an LSTM prediction model is constructed to abstract the current oil fume distribution state and surrounding environmental characteristics. The optimal fan speed control strategy is generated by combining the air volume-energy consumption-noise multi-objective optimization model and the genetic algorithm.

[0084] S4. Fan control execution: The optimal fan speed control strategy is sent to the variable frequency motor drive module to adjust the fan speed in real time.

[0085] S5. System monitoring and feedback optimization: Establish a remote communication mechanism to monitor the wind turbine's operating status and environmental parameter changes in real time, and continuously optimize system performance through a closed-loop feedback mechanism.

[0086] Specifically, the multivariate fusion analysis includes the following steps:

[0087] S21. Data Acquisition: Historical data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment will be collected and used as a public dataset.

[0088] S22. Data preprocessing: Standardize the public dataset to remove outliers and noise interference, and obtain standardized data. The standardization process is used to unify data of different dimensions into the same numerical range so as to facilitate subsequent feature extraction and weight allocation operations.

[0089] S23. Feature engineering: Extracting standard data features from standardized data and assigning weights to them. The standard data features include oil fume concentration features, temperature features, and frequency features. This includes the following steps:

[0090] S231. Map the oil fume concentration data to the [0,1] interval through linear transformation to generate oil fume concentration features, which are used as the main feature input;

[0091] S232. Transform the temperature data into temperature features through differential calculation and use them as auxiliary feature input;

[0092] S233. Decompose the airflow disturbance data into frequency features through Fourier transform and use them as supplementary feature inputs;

[0093] The standard data features are weighted and fused to form a comprehensive environmental state vector, which includes oil fume concentration value, temperature change rate and main frequency, and can be expressed as [oil fume concentration value, temperature change rate, main frequency];

[0094] S24. Divide the dataset. Randomly select 70% of the standardized data as the training set for model training and 30% as the test set for model testing.

[0095] S25. Model Construction: Construct a multivariate fusion model and dynamically adjust the weights of each standard data feature using an adaptive weighting algorithm; including the following steps:

[0096] S251. Initialize the weight vector. Set the initial weight vector to a uniform distribution. The expression is: In the formula, n represents the number of features, w 0 Represents the initial weight vector;

[0097] S252, Weight update: dynamically adjust the weights based on the contribution of each standard data feature to the comprehensive environmental state vector;

[0098] S253. Normalization process: Normalize the updated weight vector into a probability distribution.

[0099] The formula for calculating the weight update is as follows:

[0100]

[0101] The calculation formula for the normalization process is as follows:

[0102]

[0103] In the formula, This represents the weight of the i-th standard data feature at time t. Let α represent the weight of the i-th standard data feature at time t+1, and let ΔE represent the learning rate. i This represents the incremental contribution of the i-th standard data feature to the comprehensive environmental state vector;

[0104] S26. Model Training: The multivariate fusion model is trained using the training set. The weight vector is adjusted through multiple iterations until the multivariate fusion model converges. This enables the multivariate fusion model to effectively capture the contribution relationship between the features of each standard data to the environmental state, thereby generating a comprehensive environmental state vector.

[0105] S27. Model validation: Input the test set into the multivariate fusion model and use mean squared error for validation.

[0106] S28. Run, and generate a comprehensive environmental state vector in real time through a multivariate fusion model.

[0107] Specifically, the time series prediction and dynamic optimization include the following steps:

[0108] S31. Initialize the LSTM model and construct an LSTM model to predict the trend of oil fume concentration change. Initialize the initial hidden layer state and the initial oil fume distribution state.

[0109] S32. Data input: The comprehensive environmental state vector is used as the input sequence of the LSTM model. The input at each time step is represented as [oil fume concentration value, temperature change rate, main frequency];

[0110] S33, LSTM unit calculation: The current distribution state of oil fume and surrounding environmental characteristics are obtained through LSTM units. Each LSTM unit contains a forget gate, an input gate, and an output gate; including the following steps:

[0111] S331. Forget Gate: Calculate the activation value of the forget gate to dynamically filter historical data and retain key historical data. The calculation formula is as follows:

[0112] f t =σ(W f ·[h t-1 ,x t ]+b f );

[0113] S332, Input Gate: Calculates the activation value and candidate oil fume distribution state of the input gate to filter and prepare key new data for the current input, providing candidate values ​​for state updates. The calculation formula is as follows:

[0114] i t =σ(W i ·[h t-1 ,x t ]+b i );

[0115]

[0116] S333. Cell state update: Calculate the new cell state to fuse key historical data and key new data, generating the current oil fume distribution state. The calculation formula is as follows:

[0117]

[0118] S334, Output Gate: Calculates the activation value and hidden layer state of the output gate to control the output of the oil fume distribution state and generates the hidden state of the current time step. The calculation formula is as follows:

[0119] o t =σ(W o ·[h t-1 ,x t ]+b o );

[0120] h t =o t ·tanh(C t );

[0121] In the formula, f t W represents the activation value of the forget gate. f h represents the weight matrix of the forget gate. t-1 x represents the hidden layer state at the previous time step. t b represents the comprehensive environment state vector at the current time step. f The bias vector representing the forget gate; i t This represents the activation value of the input gate. W represents the distribution state of candidate cooking fumes. i and W C The weight matrices b represent the input gate and the candidate oil fume distribution states, respectively. i and b C C represents the bias vectors of the input gate and the candidate oil fume distribution state, respectively; t C represents the distribution state of oil fume at the current time step. t-1 This indicates the distribution state of oil fume at the previous time step; o t h represents the activation value of the output gate. t W represents the hidden state at the current time step. o and b o Let represent the weight matrix and bias vector of the output gate, respectively; σ(·) represents the Sigmoid activation function;

[0122] S34. Multi-objective optimization: Based on the prediction results of the LSTM model, construct a multi-objective optimization model for air volume, energy consumption, and noise.

[0123] The core calculation formula of the multi-objective optimization model for air volume, energy consumption, and noise is:

[0124] minF(x)=ω1·E(x)+ω2·N(x)+ω3·Q(x);

[0125] In the formula, F(x) represents the multi-objective function of air volume-energy consumption-noise, E(x) represents energy consumption, N(x) represents noise level, Q(x) represents oil fume purification efficiency, and ω1, ω2 and ω3 represent the weighting coefficients of energy consumption, noise and purification efficiency, respectively.

[0126] S35. Optimization and solution: The genetic algorithm is used to solve the multi-objective optimization model of air volume-energy consumption-noise.

[0127] The core calculation formula of the genetic algorithm includes:

[0128]

[0129] In the formula, x k Let x represent the solution of the k-th iteration. k+1 Let represent the solution in the (k+1)th iteration, and η represent the learning rate. This represents the gradient of the objective function;

[0130] S36. Optimal strategy generation: When the multi-objective function of air volume-energy consumption-noise converges to a local optimum, the optimal fan speed control strategy is generated.

[0131] In the specific implementation process, the multimodal data acquisition module is first activated, using non-contact sensors to collect real-time data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment. These sensors can be infrared sensors, thermocouple sensors, and ultrasonic sensors to ensure the accuracy and real-time nature of data acquisition. The acquired data is transmitted to the multivariate fusion analysis module through a communication interface and enters the data preprocessing stage. The data preprocessing steps include standardization and outlier removal. The purpose of standardization is to unify data of different dimensions into the same numerical range to facilitate subsequent feature extraction and weight allocation operations. After data preprocessing, the acquired data generates standardized data. Then, standard data features are extracted from the standardized data and weighted. For example, oil fume concentration data is linearly transformed and mapped to the [0,1] interval to generate oil fume concentration features, which serve as the main feature input. Temperature data is converted into temperature features through differential calculation, which serve as the auxiliary feature input. Airflow disturbance data is decomposed into frequency features through Fourier transform and serves as the supplementary feature input. The standard data features are then weighted and fused to form a comprehensive environmental state vector, which can be represented as [oil fume concentration value, temperature change rate, main frequency];

[0132] The core of multivariate fusion analysis lies in constructing an adaptive weighting algorithm to dynamically adjust the weights of each standard data feature. After receiving the comprehensive environmental state vector, the weights of each standard data feature are uniformly initialized. Then, the weights of each standard data feature are dynamically adjusted according to their contribution to the comprehensive environmental state vector. To ensure that the weight vector always satisfies the probability distribution characteristics, the weights of the updated standard data features need to be processed according to the normalization calculation formula. After multiple iterations of training, the model can effectively capture the contribution relationship of each standard data feature to the environmental state, thereby generating the comprehensive environmental state vector.

[0133] Next, the integrated environmental state vector is passed to the time series prediction and dynamic optimization module, entering the prediction process of the LSTM model. The LSTM unit comprises four key parts: a forget gate, an input gate, an oil fume distribution state update gate, and an output gate.

[0134] Calculation of the activation value of the forget gate: First, based on the hidden layer state of the previous time step and the comprehensive environmental state vector of the current time step, the activation value of the forget gate is generated by the linear combination of the weight matrix and the bias vector and then processed by the Sigmoid activation function. The activation value of the forget gate is used to quantify the retention ratio of the oil fume distribution state of the previous time step. Since some historical oil fume distribution states may be meaningless for the current prediction, the interference of these irrelevant historical information can be suppressed by the forget gate, so that the model can focus on the recent key state, retain key historical data, and avoid the prediction deviation caused by the model being overloaded with redundant information, such as temperature data that has not changed for a long time and dissipated airflow disturbances.

[0135] The activation value of the input gate is based on the hidden layer state of the previous time step and the comprehensive environmental state vector of the current time step. Linear transformations are performed using independent weight matrices and bias vectors, and the activation value of the input gate is obtained according to the Sigmoid function. Candidate oil fume distribution states are generated according to the Tanh function. The oil fume distribution state is driven by the comprehensive environmental state vector of the current input, but not all inputs are equally important. The input gate filters key new data and generates candidate oil fume distribution data, enabling the LSTM model to update the current state in a targeted manner and avoiding incorrect state corrections from invalid inputs.

[0136] Updating the distribution state of cooking fumes: The activation value of the forget gate is multiplied by the distribution state of cooking fumes from the previous time step to preserve historical features. Then, the product of the activation value of the input gate and the candidate distribution state of cooking fumes is superimposed to fuse new features. The sum of the two is the distribution state of cooking fumes at the current time step. Updating the distribution state of cooking fumes can avoid the one-sidedness of prediction caused by relying solely on the current input.

[0137] Output gate activation value calculation: Based on the hidden layer state of the previous time step and the comprehensive environmental state vector of the current time step, the output gate activation value is obtained through a linear combination of the weight matrix and bias vector and the Sigmoid function. The output gate activation value is multiplied by the Tanh value of the current oil fume distribution state to generate the hidden state of the current time step. The hidden state of the current time step is a dynamic abstraction of the current oil fume distribution state and surrounding environmental features by the LSTM model.

[0138] Based on the hidden state at the current time step, a multi-objective optimization model of air volume, energy consumption, and noise is constructed. This model is used to derive the objective function of the optimal fume control strategy at the current time step. The objective function of the optimal fume control strategy at the current time step is then used as the objective function of the genetic algorithm for iteration. When the objective function converges to a local optimum, the optimal fan speed control strategy is generated.

[0139] After receiving the optimal fan speed control strategy, the fan control execution module adjusts the fan speed in real time by driving the variable frequency motor. The variable frequency motor drive module generates a corresponding PWM signal based on the target speed value in the optimal fan speed control strategy, thereby precisely controlling the fan's operating state. In practical applications, the fan speed adjustment range is typically set between 500 and 2000 revolutions per minute to meet the needs of different cooking scenarios. Simultaneously, the fan control execution module also has overload protection; when abnormal motor operation is detected, it automatically reduces the speed or stops operation to ensure the safety and stability of the equipment.

[0140] The system monitoring and feedback optimization module continuously monitors the fan's operating status and environmental parameter changes through a closed-loop feedback mechanism. It acquires real-time information on fan operating status and environmental parameter changes via remote communication and feeds it back to the multivariate fusion analysis module and the time-series prediction and dynamic optimization module, thereby achieving continuous optimization of system performance. For example, when a sudden increase in oil fume concentration is detected, the system quickly adjusts the fan speed to improve purification efficiency; when a decrease in ambient temperature is detected, the system appropriately reduces the fan speed to reduce energy consumption and noise. Furthermore, the system monitoring and feedback optimization module also allows users to remotely view the device's operating status via a mobile app or smart home platform and manually adjust the fan speed according to their individual needs.

[0141] In summary, this invention utilizes non-contact multimodal sensors to collect real-time data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment. Combined with an edge computing chip running a multivariate fusion algorithm and a time-series prediction model, it significantly improves the system's dynamic response speed and intelligence level. Simultaneously, by introducing a multi-objective optimization model of airflow, energy consumption, and noise, it achieves efficient, energy-saving, and low-noise airflow control. In practical applications, this invention can effectively address various challenges in complex kitchen environments, providing users with a more comfortable and healthier cooking experience.

[0142] Example 2

[0143] Please see Figures 1-3 The present invention also provides an automatic control system for a variable frequency range hood, which applies the automatic control method for a variable frequency range hood as described above. The system includes a multimodal data acquisition module, a multivariable fusion analysis module, a time series prediction and dynamic optimization module, a fan control execution module, and a system monitoring and feedback optimization module.

[0144] The multimodal data acquisition module consists of multiple non-contact sensors used to acquire real-time data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment. These sensors transmit the acquired data to the multivariate fusion analysis module for processing via wireless or wired means.

[0145] The multivariate fusion analysis module is used to receive data sent by the multimodal data acquisition module, and generate a comprehensive environmental state vector by extracting features and assigning weights to the acquired data, providing support for subsequent time series prediction and dynamic optimization.

[0146] The time-series prediction and dynamic optimization module is used to receive the comprehensive environmental state vector sent by the multivariate fusion analysis module, predict and abstract the current oil fume distribution state and surrounding environmental characteristics through the LSTM model, and generate the optimal fan speed control strategy by combining the air volume-energy consumption-noise multi-objective optimization model, and send the control strategy to the fan control execution module and the system monitoring and feedback optimization module.

[0147] The wind turbine control execution module is used to receive the control strategy sent by the timing prediction and dynamic optimization module, and adjust the wind turbine speed in real time by driving the variable frequency motor;

[0148] The system monitoring and feedback optimization module continuously monitors the wind turbine's operating status and environmental parameter changes through a closed-loop feedback mechanism, thereby further optimizing system performance.

[0149] These modules work together to complete the entire process from data acquisition to fan speed adjustment.

[0150] Specifically, the multimodal data acquisition module consists of an infrared sensor, a thermocouple sensor, and an ultrasonic sensor, and is used to acquire real-time data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment.

[0151] The fan control execution module adjusts the fan speed in real time by driving the variable frequency motor. The variable frequency motor drive module generates a corresponding PWM signal according to the target speed value in the control strategy.

[0152] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. An automatic control method for a variable frequency range hood, characterized in that, Including the following steps: S1. Multimodal data acquisition: Non-contact multimodal sensors are used to collect data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment in real time, and the collected data is transmitted to an edge computing chip for processing. S2. Multivariate fusion analysis: Based on the collected oil fume concentration data, temperature data, and airflow disturbance data, a multivariate fusion model is constructed, and a comprehensive environmental state vector is generated through feature extraction and weight allocation. S3. Time-series prediction and dynamic optimization: Based on the comprehensive environmental state vector, an LSTM prediction model is constructed to abstract the current oil fume distribution state and surrounding environmental characteristics. Combined with the air volume-energy consumption-noise multi-objective optimization model and genetic algorithm, the optimal fan speed control strategy for the current time step is generated. S4. Fan control execution: The optimal fan speed control strategy for the current time step is sent to the variable frequency motor drive module to adjust the fan speed in real time. S5. System monitoring and feedback optimization: Establish a remote communication mechanism to monitor the wind turbine's operating status and environmental parameter changes in real time, and continuously optimize system performance through a closed-loop feedback mechanism. The multivariate fusion analysis includes the following steps: S21. Data Acquisition: Collect historical data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment as a public dataset. S22. Data preprocessing: Standardize the public dataset to remove outliers and noise interference, and obtain standardized data. S23. Feature engineering: Extracting standard data features from standardized data and assigning weights to them; S24. Divide the dataset. Randomly select 70% of the standardized data as the training set for model training and 30% as the test set for model testing. S25. Model building: Construct a multivariate fusion model and use an adaptive weighting algorithm to dynamically adjust the weights of each standard data feature; S26. Model training: Use the training set to train the multivariate fusion model, and adjust the weight vector of the standardized data through multiple iterations until the model converges. S27. Model validation: Input the test set into the multivariate fusion model and use mean squared error for validation. S28. Run, and generate a comprehensive environmental state vector in real time through a multivariate fusion model; The time series prediction and dynamic optimization include the following steps: S31. Initialize the LSTM model: Construct an LSTM model to predict the current oil fume distribution state and initialize the initial oil fume distribution state. S32. Data input: The comprehensive environmental state vector is used as the input sequence of the LSTM model. S33, LSTM unit calculation, each LSTM unit contains a forget gate, input gate and output gate; the current oil fume distribution state and surrounding environmental characteristics are obtained through LSTM units; S34. Multi-objective optimization: Based on the prediction results of the LSTM model, construct a multi-objective optimization model for air volume, energy consumption, and noise. S35. Optimization and solution: The genetic algorithm is used to solve the multi-objective optimization model of air volume-energy consumption-noise. S36. Optimal strategy generation: When the multi-objective function of air volume-energy consumption-noise converges to a local optimum, the optimal fan speed control strategy is generated.

2. The automatic control method for a variable frequency range hood according to claim 1, characterized in that, The step of extracting standard data features from standardized data and assigning weights, wherein the standard data features include oil fume concentration features, temperature features, and frequency features, includes the following steps: S231. Map the oil fume concentration data to the [0,1] interval through linear transformation to obtain the oil fume concentration features, and use them as the main feature input; S232. Transform the temperature data into temperature features through differential calculation and use them as auxiliary feature input; S233. Decompose the airflow disturbance data into frequency features through Fourier transform and use them as supplementary feature inputs; The standard data features are weighted and fused to form a comprehensive environmental state vector, which includes oil fume concentration, temperature change rate, and main frequencies.

3. The automatic control method for a variable frequency range hood according to claim 1, characterized in that, The method of dynamically adjusting the weights of each standard data feature using an adaptive weighting algorithm includes the following steps: S251. Initialize the weight vector. Set the initial weight vector to a uniform distribution. The expression is: In the formula, n represents the number of standard data features. Represents the initial weight vector; S252, Weight update: dynamically adjust the weights based on the contribution of each standard data feature to the comprehensive environmental state vector; S253. Normalization process: Normalize the updated weight vector into a probability distribution. The formula for calculating the weight update is as follows: The calculation formula for the normalization process is as follows: In the formula, This represents the weight of the i-th standard data feature at time t. Let Ei represent the weight of the i-th standard data feature at time t+1, α represent the learning rate, and ΔEi represent the incremental contribution of the i-th standard data feature to the comprehensive environmental state vector.

4. The automatic control method for a variable frequency range hood according to claim 1, characterized in that, The LSTM unit calculation includes the following steps: S331. Forget Gate: Calculate the activation value of the forget gate to dynamically filter historical data and retain key historical data. The calculation formula is as follows: S332, Input Gate: Calculates the activation value and candidate oil fume distribution state of the input gate to filter and prepare key new data for the current input, providing candidate values ​​for state updates. The calculation formula is as follows: S333. Cell state update: Calculate the new oil fume distribution state to fuse key historical information with key new data, generating the current oil fume distribution state. The calculation formula is as follows: S334, Output Gate: Calculate the activation value and oil fume distribution state of the output gate, control the output of the oil fume distribution state, and generate a hidden state that focuses on key information. The calculation formula is as follows: In the formula, This represents the activation value of the forget gate. The weight matrix represents the forget gate. This represents the hidden layer state at the previous time step. This represents the comprehensive environmental state vector at the current time step. The bias vector representing the forget gate; This represents the activation value of the input gate. Indicates the distribution status of candidate cooking fumes. and These represent the weight matrices for the input gate and the candidate oil fume distribution states, respectively. and These represent the bias vectors for the input gate and the candidate oil fume distribution state, respectively; This indicates the distribution status of oil fume at the current time step. This shows the distribution of oil fume at the previous time step; This indicates the activation value of the output gate. This indicates the hidden layer state at the current time step. and σ and σ(·) represent the weight matrix and bias vector of the output gate, respectively; σ(·) represents the Sigmoid activation function; the hidden layer state is a dynamic abstract representation of the current oil fume distribution state and surrounding environmental characteristics by the LSTM model.

5. The automatic control method for a variable frequency range hood according to claim 1, characterized in that, The core calculation formula of the multi-objective optimization model for air volume, energy consumption, and noise is: In the formula, F(x) represents the multi-objective function of air volume-energy consumption-noise, E(x) represents energy consumption, N(x) represents noise level, Q(x) represents oil fume purification efficiency, and ω1, ω2 and ω3 represent the weighting coefficients of energy consumption, noise and purification efficiency, respectively.

6. The automatic control method for a variable frequency range hood according to claim 1, characterized in that, The core calculation formula of the genetic algorithm includes: In the formula, This represents the solution for the k-th iteration. Let represent the solution of the (k+1)th iteration, η represent the learning rate, and represent the gradient of the objective function.

7. An automatic control system for a variable frequency range hood, employing the automatic control method for a variable frequency range hood as described in any one of claims 1 to 6, characterized in that, The system includes a multimodal data acquisition module, a multivariate fusion analysis module, a time series prediction and dynamic optimization module, a wind turbine control execution module, and a system monitoring and feedback optimization module. The multimodal data acquisition module collects real-time data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment using a non-contact multimodal sensor, and transmits the collected data to the multivariate fusion analysis module. The multivariate fusion analysis module is used to receive data sent by the multimodal data acquisition module, generate a comprehensive environmental state vector through feature extraction and weight allocation, and send the comprehensive environmental state vector to the time series prediction and dynamic optimization module. The time-series prediction and dynamic optimization module receives the comprehensive environmental state vector sent by the multivariate fusion analysis module, predicts and abstracts the current oil fume distribution state and surrounding environmental characteristics through an LSTM model, and generates the optimal fan speed control strategy by combining the air volume-energy consumption-noise multi-objective optimization model. The optimal fan speed control strategy is then sent to the fan control execution module and the system monitoring and feedback optimization module. The fan control execution module receives the control strategy sent by the time-series prediction and dynamic optimization module and adjusts the fan speed in real time. The system monitoring and feedback optimization module is used to receive information on the wind turbine operating status and environmental parameter changes sent by the wind turbine control execution module, and to continuously optimize system performance through a closed-loop feedback mechanism.

8. The automatic control system for a variable frequency range hood according to claim 7, characterized in that, The multimodal data acquisition module consists of an infrared sensor, a thermocouple sensor, and an ultrasonic sensor, and is used to acquire real-time data on oil fume concentration, temperature, and airflow disturbance in the kitchen environment. The fan control execution module adjusts the fan speed in real time by driving the variable frequency motor. The variable frequency motor drive module generates a corresponding PWM signal according to the target speed value in the optimal fan speed control strategy.

Images