An intelligent control system for kitchen environment based on multi-sensor fusion

Abstract

The present invention discloses a kitchen environment intelligent control system based on multi-sensor fusion, which relates to the field of artificial intelligence technology, including a data acquisition module, a data fusion evaluation module, an intelligent control module and a remote feedback warning module; the present invention adopts a fuzzy reasoning algorithm to fuse multi-source data, calculates a comprehensive kitchen environment quality evaluation value, and comprehensively reflects the real-time status of the kitchen environment. This comprehensive evaluation capability effectively solves the limitations of single sensor monitoring, constructs an intelligent control model based on a deep Q network algorithm through an intelligent control module, automatically adjusts kitchen equipment through a state vector monitored in real time, and motivates the system to take actions to improve environmental quality and reduce energy waste through the training of intelligent bodies and the design of reward functions. This intelligent control not only improves the efficiency of intelligent control and environmental comfort, but also achieves optimal use of energy.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an intelligent control system for a kitchen environment based on multi-sensor fusion. Background Art

[0002] As the main place for cooking in the family, the environmental quality of the kitchen directly affects the cooking experience and the health of family members. Traditional monitoring equipment usually relies on a single sensor for data collection and lacks multi-sensor data fusion technology, which limits the comprehensive assessment of air quality. Due to the lack of comprehensive assessment capabilities, these devices cannot provide indicators that fully reflect the environmental quality, resulting in the inability to accurately capture and respond to complex environmental changes in the kitchen; at the same time, in terms of automation and intelligent control, most kitchen equipment requires users to manually operate based on personal experience and intuition. This method is not only inefficient, but also difficult to achieve the best regulation effect, and will also lead to energy waste and unstable environmental comfort. Summary of the invention

[0003] The purpose of the present invention is to provide an intelligent control system for kitchen environment based on multi-sensor fusion, which realizes comprehensive perception, automatic optimization and safety monitoring of the kitchen environment through multi-sensor data fusion, intelligent control and remote monitoring and early warning, and effectively improves the quality of the kitchen environment.

[0004] The purpose of the present invention can be achieved through the following technical solutions:

[0005] The present application provides a kitchen environment intelligent control system based on multi-sensor fusion, including a data fusion evaluation module and an intelligent control module.

[0006] The data fusion evaluation module fuses the collected kitchen environment data through a fuzzy inference algorithm to calculate the kitchen environment quality evaluation value;

[0007] The intelligent control module constructs an intelligent control model to perform intelligent control. The model combines the kitchen environment quality assessment value and outputs the optimal action to automatically adjust the kitchen equipment.

[0008] Specifically, intelligent regulation is carried out by building an intelligent regulation model, including:

[0009] Preprocess the sensor data of the kitchen environment, including normalization and feature extraction, to obtain the state space;

[0010] The preprocessed data is used as the state vector, including temperature, humidity, odor concentration, and particle concentration;

[0011] According to the state vector, the action performed by the intelligent agent is determined to form a defined action space, including the speed level of the range hood fan, the fire power of the gas stove, the output power and power-on time of the induction cooker, and the opening and closing status of the smart window.

[0012] Furthermore, it also includes a data acquisition module, which is used to monitor the kitchen environment data in real time by installing odor sensors, particle concentration sensors, temperature and humidity sensors, and gas composition sensors at various locations in the kitchen; and transmits the acquired kitchen environment data to the data fusion evaluation module;

[0013] The odor sensor is used to detect volatile organic compounds or other odor molecules in the air; the particle concentration sensor is used to measure the size and number of suspended particles in the air; the temperature and humidity sensor is used to record the temperature and humidity levels of the environment; the gas composition sensor is used to analyze the concentration of specific gases in the air, such as carbon monoxide and carbon dioxide.

[0014] Furthermore, the calculation of the kitchen environment quality assessment value specifically includes:

[0015] The collected raw data is preprocessed, the raw data is normalized, and converted into a fuzzy processing format. The normalization processing formula is expressed as:

[0016]

[0017] in, is the normalized value, and are the minimum and maximum values ​​of the i-th sensor data respectively;

[0018] Calculate the membership of the normalized fuzzy value in each fuzzy set:

[0019] The membership degree is expressed as: ,

[0020] in, is a fuzzy set The center of is the width parameter;

[0021] Using membership functions to determine weight coefficients for environmental data from each sensor;

[0022] The weight coefficient is expressed as:

[0023] in, is the weight coefficient of the i-th sensor, is the importance score of the sensor, is the maximum value among all sensor scores, is the sensitivity parameter;

[0024] The normalized membership degree of each sensor is multiplied by the weight coefficient to obtain a comprehensive fuzzy evaluation;

[0025]

[0026] in, It is a comprehensive fuzzy evaluation. Represents the "AND" operation of fuzzy logic;

[0027] The comprehensive evaluation is fuzzily converted into specific values ​​to obtain a comprehensive kitchen environment quality evaluation value, which is used to reflect the overall condition of the kitchen environment;

[0028] The specific kitchen environment quality evaluation formula is expressed as: ;

[0029] in, is the comprehensive kitchen environment quality evaluation value after defuzzification. is a fuzzy set The associated specific numerical values ​​are: m is the number of fuzzy sets and n is the number of sensors.

[0030] Furthermore, the construction of the intelligent control model also includes: training of the intelligent agent and design of the reward function. The reward function design quantifies the quality of each action through the reward function according to the execution action output by the intelligent control model and the kitchen environment quality evaluation value. It is specifically expressed as:

[0031]

[0032] in, is the reward obtained by executing action a in state s, It is the reward obtained after performing the action, and it is the feedback given by the kitchen environment. is the discount factor (0 ), used to balance the importance of immediate rewards and future rewards, It is the next state The reward for the assessment value is quantified based on the kitchen environment quality assessment value.

[0033] Furthermore, the DQN algorithm is used to train the agent, including:

[0034] Initialize the main network and target network of DQN to estimate the Q value of taking an action in a given state;

[0035] Define the environment in which the agent lives, including the state, action, and reward mechanism, where the environment receives the action of the agent and returns the corresponding state and reward;

[0036] The agent performs actions in the environment to generate a state-action-reward-next-state quadruple and stores it in the experience replay buffer, which is used to store past interactions and perform sample training.

[0037] Among them, samples are randomly drawn from the experience replay buffer and the target network is used to calculate the target value;

[0038] Specifically expressed as: ; where r is the immediate reward, is the discount factor, Target network parameters, is the new state after executing action a, is in state The best action among all actions under

[0039] Use the main network to predict the Q value of the current state-action pair, which is used to evaluate the expected reward of performing an action in a given state. It takes the state vector as input and outputs the predicted Q value of each action.

[0040] Use an optimization algorithm to update the parameters of the main network according to the loss function, regularly copy the parameters of the main network to the target network, and continuously perform iterative training until the agent reaches the predetermined training cycle.

[0041] Further, the updating of the parameters of the main network according to the loss function specifically includes: evaluating the predicted Q value and the target by calculating the loss function The difference between the values, the loss function is expressed as:

[0042]

[0043] Where N is the batch size, which is the number of samples randomly drawn from the experience replay buffer for one training iteration to calculate the loss function and update the parameters of the neural network. are the parameters of the neural network, s represents the current state, and a represents the action chosen by the agent in a given state.

[0044] Furthermore, the trained DQN model is deployed in the actual kitchen environment for real-time status monitoring and intelligent regulation. The main network predicts the Q value of each action under a specific state. The intelligent agent combines the kitchen environment quality assessment value and adjusts the kitchen equipment through the optimal action of predicting the Q value output by the DQN model.

[0045] Furthermore, it also includes a remote feedback warning module, which remotely monitors the real-time status of the kitchen environment through a mobile phone APP, displays the current data and evaluation results of the kitchen environment, and sends warning information to the user remotely when an abnormal situation is detected.

[0046] Furthermore, the remote feedback warning module is specifically used to: display the current environmental data and evaluation results of the kitchen on the display screen of the mobile phone APP; if there is an abnormal situation, send a warning message to the user; the warning information includes information that the smoke concentration is too high or the harmful gas exceeds the standard.

[0047] Furthermore, the manual intervention decision-making process through the mobile phone APP remote control interface includes: confirming the warning information and executing to stop the current cooking task, reduce the equipment workload or increase the ventilation volume.

[0048] The beneficial effects of the present invention are:

[0049] (1) The data acquisition module integrates multiple sensors such as odor, particle concentration, temperature, humidity, and gas composition to achieve comprehensive monitoring of the kitchen environment. The data fusion evaluation module uses fuzzy inference algorithm to fuse multi-source data and calculate a comprehensive kitchen environment quality evaluation value, which fully reflects the real-time status of the kitchen environment. This comprehensive evaluation capability effectively solves the limitations of single sensor monitoring and improves the accuracy and reliability of environmental quality judgment.

[0050] (2) Through the intelligent control module, an intelligent control model is constructed based on the deep Q network algorithm. The kitchen equipment, such as range hoods, gas stoves and induction cookers, is automatically adjusted through the real-time monitoring state vector. Through the training of the intelligent agent and the design of the reward function, the system is encouraged to take actions to improve the environmental quality and reduce energy waste. This intelligent control not only improves the efficiency of intelligent control and environmental comfort, but also optimizes the use of energy and reduces long-term operating costs.

[0051] (3) Users can view kitchen environment data and evaluation results in real time. The system can also proactively send warning information to users when it detects abnormal conditions such as excessive smoke concentration or excessive harmful gases. In addition, users can manually intervene in control decisions through the APP, such as stopping cooking tasks or increasing ventilation, forming a closed-loop feedback with the intelligent system. This remote monitoring and early warning mechanism significantly improves the safety of the kitchen environment and the controllability of users. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] For better understanding and implementation, the technical solution of the present application is described in detail below with reference to the accompanying drawings.

[0053] Figure 1 A schematic diagram of the structure of a kitchen environment intelligent control system based on multi-sensor fusion provided in Example 1 of the present application;

[0054] Figure 2 A schematic diagram of a flow chart of a kitchen environment quality evaluation value calculated by a kitchen environment intelligent control system based on multi-sensor fusion provided in Example 1 of the present application;

[0055] Figure 3 A schematic diagram of a process of constructing an intelligent control model for performing intelligent control of a kitchen environment intelligent control system based on multi-sensor fusion provided in Example 1 of the present application;

[0056] Figure 4 A flowchart of using the DQN algorithm to train an intelligent agent in a kitchen environment intelligent control system based on multi-sensor fusion provided in Example 1 of the present application. DETAILED DESCRIPTION

[0057] In order to further explain the technical means and effects taken by the present invention to achieve the predetermined invention purpose, exemplary embodiments will be described in detail here, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are only examples of methods and systems consistent with some aspects of the present application as detailed in the attached claims.

[0058] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0059] The specific implementation methods, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0060] Example 1

[0061] See also Figure 1-Figure 4 This embodiment provides a kitchen environment intelligent control system based on multi-sensor fusion. Through multi-sensor data fusion, intelligent control and remote monitoring and early warning, it realizes comprehensive perception, automatic optimization and safety monitoring of the kitchen environment, and effectively improves the quality of the kitchen environment.

[0062] The present invention provides a kitchen environment intelligent control system based on multi-sensor fusion, including a data acquisition module, a data fusion evaluation module, an intelligent control module and a remote feedback warning module;

[0063] The data acquisition module acquires the odor, particle concentration, temperature, humidity and specific gas composition in the kitchen environment, and transmits the acquired kitchen environment data to the data fusion evaluation module;

[0064] The data fusion evaluation module fuses the collected kitchen environment data and calculates the kitchen environment quality evaluation value;

[0065] The intelligent control module constructs an intelligent control model for intelligent control, and the model combines the kitchen environment quality evaluation value to output the optimal action for automatic adjustment of kitchen equipment;

[0066] The remote feedback warning module remotely monitors the real-time status of the kitchen environment through a mobile phone APP, displays the current data and evaluation results of the kitchen environment, and sends warning information to the user remotely when an abnormal situation is detected.

[0067] Furthermore, the data acquisition module installs odor sensors, particle concentration sensors, temperature and humidity sensors, and gas composition sensors at various locations in the kitchen to monitor the kitchen's environmental data in real time;

[0068] Among them, odor sensors are used to detect volatile organic compounds (VOCs) or other specific odor molecules in the air; particle concentration sensors are used to measure the size and number of suspended particles in the air; temperature and humidity sensors are used to record the temperature and humidity levels of the environment; gas composition sensors are used to analyze the concentration of specific gases in the air, such as carbon monoxide, carbon dioxide, etc.

[0069] Furthermore, the environmental data of each sensor is obtained, and the environmental data of each sensor is integrated through the fuzzy inference algorithm to calculate the kitchen environment quality evaluation value, which specifically includes:

[0070] S11, pre-processing the collected raw data, normalizing the raw data, and converting it into a fuzzy processing format. The normalization processing formula is expressed as:

[0071]

[0072] in, is the normalized value, and are the minimum and maximum values ​​of the i-th sensor data respectively;

[0073] S12, calculating the membership degree of the normalized fuzzy value in each fuzzy set;

[0074] The membership degree is expressed as:

[0075] in, is a fuzzy set The center of is the width parameter;

[0076] S13, using a membership function to determine a weight coefficient for the environmental data of each sensor;

[0077] The weight coefficient is expressed as:

[0078] in, is the weight coefficient of the i-th sensor, is the importance score of the sensor, is the maximum value among all sensor scores, is the sensitivity parameter;

[0079] S14, multiplying the normalized membership degree of each sensor by the weight coefficient to obtain a comprehensive fuzzy evaluation;

[0080]

[0081] in, It is a comprehensive fuzzy evaluation. Represents the "AND" operation of fuzzy logic;

[0082] S15. The comprehensive evaluation is fuzzily converted into a specific numerical value to obtain a comprehensive kitchen environment quality evaluation value, which is used to reflect the overall condition of the kitchen environment.

[0083] The specific kitchen environment quality evaluation formula is expressed as: ;

[0084] in, is the comprehensive kitchen environment quality evaluation value after defuzzification. is a fuzzy set The associated specific numerical values ​​are: m is the number of fuzzy sets and n is the number of sensors.

[0085] Furthermore, the intelligent control module performs intelligent control by constructing an intelligent control model, including:

[0086] S21, preprocessing the sensor data of the kitchen environment, including normalization and feature extraction, to obtain a state space;

[0087] S22, using the preprocessed data as a state vector, including temperature, humidity, odor concentration, and particle concentration;

[0088] S23. According to the state vector, determine the action performed by the intelligent agent to form a defined action space, including the speed level of the range hood fan, the fire power of the gas stove, the output power and power-on time of the induction cooker, and the opening and closing status of the smart window.

[0089] Furthermore, constructing the intelligent control model also includes: training of the intelligent agent and design of the reward function. The reward function design quantifies the quality of each action through the reward function according to the execution action output by the intelligent control model and the kitchen environment quality evaluation value. It is specifically expressed as:

[0090]

[0091] in, is the reward obtained by executing action a in state s, It is the reward obtained after performing the action, and it is the feedback given by the kitchen environment. is the discount factor (0 ), used to balance the importance of immediate rewards and future rewards, It is the next state The reward for the evaluation value is quantified based on the kitchen environment quality evaluation value. When the kitchen environment quality evaluation value is low, the reward is reduced to encourage the agent to take measures to improve the environment in the kitchen.

[0092] Furthermore, the use of the DQN algorithm to train the agent includes:

[0093] S201, initialize the main network and target network of DQN, the target network is a copy of the main network, the main network is a deep neural network used to estimate the Q value of taking an action in a given state, this network receives the state vector as input;

[0094] S202, defining the environment in which the agent is located, including states, actions, and reward mechanisms, wherein the environment receives the actions of the agent and returns corresponding states and rewards;

[0095] S203, the agent performs actions in the environment to generate a four-tuple of state-action-reward-next state, and stores it in the experience playback buffer, which is used to store past interactions and perform sample training;

[0096] Among them, samples are randomly drawn from the experience replay buffer and the target network is used to calculate the target value;

[0097] Specifically expressed as: ; where r is the immediate reward, is the discount factor, Target network parameters, is the new state after executing action a, is in state The best action among all actions under

[0098] S204, using the main network to predict the Q value of the current state-action pair, which is used to evaluate the expected reward of performing an action in a given state, taking the state vector as input and outputting the predicted Q value of each action;

[0099] S205. Use an optimization algorithm to update the parameters of the main network according to the loss function, and regularly copy the parameters of the main network to the target network to maintain the stability of the target network. Continue to perform iterative training until the agent learns an effective strategy or reaches a predetermined training cycle.

[0100] Among them, the predicted Q value and the target are evaluated by calculating the loss function The difference between the values, expressed as:

[0101]

[0102] Where N is the batch size, which is the number of samples randomly drawn from the experience replay buffer for one training iteration to calculate the loss function and update the parameters of the neural network. are the parameters of the neural network, s represents the current state, and a represents the action chosen by the agent in a given state.

[0103] Furthermore, the trained DQN model is deployed in the actual kitchen environment for real-time state monitoring and intelligent regulation. The main network predicts the Q value of each action under a specific state. The intelligent agent combines the kitchen environment quality assessment value and uses the DQN model to output the predicted Q value to regulate the kitchen equipment. When the DQN algorithm model is selected for action decision-making, the optimization process of each parameter can be performed independently or in parallel.

[0104] Furthermore, the remote feedback warning module is specifically used to: display the current environmental data and evaluation results of the kitchen on the display screen of the mobile phone APP; if there is an abnormal situation, send a warning message to the user; the warning information includes information that the smoke concentration is too high or the harmful gas exceeds the standard.

[0105] Furthermore, the manual intervention decision-making process through the mobile phone APP remote control interface includes: confirming the warning information and executing to stop the current cooking task, reduce the equipment workload or increase the ventilation volume.

[0106] This embodiment provides a comprehensive solution for modern families through highly integrated data acquisition, evaluation, regulation and early warning modules to optimize the kitchen environment and ensure the health and safety of users. The data acquisition module uses odor, particle concentration, temperature, humidity and gas composition sensors distributed in key locations in the kitchen to monitor air quality in real time. The data fusion evaluation module uses fuzzy reasoning algorithm to deeply fuse the data and calculate the comprehensive kitchen environment quality evaluation value, so as to comprehensively reflect the current status of the kitchen environment; the intelligent regulation module receives the evaluation value and combines the intelligent regulation model constructed by the deep Q network (DQN) algorithm to learn the optimal regulation strategy by training the intelligent agent. The input of the model is the preprocessed state vector, and the output is the automatic adjustment action for the kitchen equipment, such as the range hood fan speed, gas stove fire, induction cooker power and window opening and closing status, so as to realize intelligent environmental management.

[0107] In addition, the training process of the intelligent agent involves the design of a reward function to quantify the pros and cons of each action and encourage the intelligent agent to take measures to improve the quality of the environment. The remote feedback and early warning module is implemented through a mobile phone APP. Users can remotely monitor the real-time data and evaluation results of the kitchen environment. Once the system detects an abnormal situation, such as excessive smoke concentration or excessive harmful gases, the module will immediately send an early warning message to the user to ensure timely manual intervention. Users can also manually confirm the early warning through the APP and execute intervention measures, such as stopping the cooking task, reducing equipment load or increasing ventilation, forming a closed-loop feedback with the intelligent control system.

[0108] Through the implementation of this system, the problem of single and inefficient means of monitoring and controlling the traditional kitchen environment has been solved, and real-time monitoring, intelligent control and early warning of the kitchen environment have been realized, which has significantly improved the user's cooking experience and health and safety protection. In addition, the system's self-learning and optimization capabilities ensure the continuous improvement and adaptability of the control strategy over time, demonstrating the great potential of intelligent technology in improving the quality of the living environment.

[0109] Example 2

[0110] The present invention provides a kitchen environment intelligent control system based on multi-sensor fusion. The difference between this embodiment and embodiment 1 is that: data from different sensors are fused through a particle filter to calculate the kitchen environment quality evaluation value. The rest of the process is the same as that of embodiment 1 and will not be repeated here.

[0111] The particle filter is used to fuse the data from different sensors and calculate the kitchen environment quality assessment value, including:

[0112] Define state variables and determine the state variables that describe the quality of the kitchen environment, such as temperature, humidity, odor concentration, particulate matter concentration, and specific gas concentration;

[0113] Initialize particles, randomly generate a group of particles in the state space, each particle represents a possible state, and each particle is initially given the same weight;

[0114] Sensor data preprocessing: preprocessing the data collected by each sensor, including denoising and standardization;

[0115] Observation model establishment, establish the observation model, describe how to generate the probability distribution of observation data based on the current state, and perform the following two main stages at each time step:

[0116] a. Prediction stage: predict the next state of each particle based on the state transition probability model;

[0117] b. Update phase: adjust the weight of each particle according to the new observation data and observation model.

[0118] Calculate weights, using the observation model and the latest sensor data to calculate the weight of each particle. The weight reflects the consistency of the particle with the actual observation data.

[0119] Resampling: resampling is performed according to the weight of the particles to form a new set of particles, ensuring that more particles represent important states;

[0120] State estimation, which is to calculate the state estimation of the kitchen environment quality based on the particle set, such as calculating the weighted average or other statistics;

[0121] Evaluation value calculation, based on the state estimation, calculates the kitchen environment quality evaluation value using a predefined evaluation standard or function.

[0122] Through particle filters, the system is able to process data from different sensors and provide robust state estimation and environmental quality assessment even in the event of sensor failure or data loss, providing accurate information support for intelligent control.

[0123] Example 3

[0124] The present invention provides a kitchen environment intelligent control system based on multi-sensor fusion. The difference between this embodiment and embodiment 1 is that the data from different sensors are fused through a Kalman filter to calculate the kitchen environment quality evaluation value. The rest of the process is the same as that of embodiment 1 and will not be repeated here.

[0125] The Kalman filter is used to fuse the data from different sensors and calculate the kitchen environment quality assessment value, including:

[0126] Define the state model and determine the variables that describe the state of the kitchen environment, such as temperature, humidity, odor concentration, particle concentration, and gas component concentration;

[0127] Initialize the Kalman filter, set the initial state estimate and covariance matrix of the Kalman filter, and the covariance matrix of the process noise and observation noise;

[0128] Sensor data preprocessing: preprocessing the data collected by each sensor, including denoising and standardization;

[0129] Establish state transfer and observation models, define the state transfer matrix, describe the evolution of the state over time, define the observation matrix, and map the state variables to the observation space;

[0130] Prediction step: Use the state transition model to predict the state estimate for the next time step and update the covariance of the state estimate;

[0131] Update step, collect sensor data for the current time step, calculate Kalman gain based on the observation model and predicted state estimate, and use Kalman gain to update the state estimate and covariance matrix;

[0132] Fusion of sensor data. If multiple sensors measure the same variable, the Kalman filter will automatically fuse the data to provide the best estimate.

[0133] Calculate the environmental quality assessment value, and calculate the kitchen environment quality assessment value based on the updated state estimate using a predefined assessment function or standard.

[0134] Through the Kalman filter, the system can accurately estimate the real-time state of the kitchen environment and calculate a comprehensive environmental quality assessment value, providing a scientific basis for intelligent regulation. The Kalman filter is particularly suitable for systems with linear state space and Gaussian noise, and can provide optimal linear unbiased estimates.

[0135] Example 4

[0136] Another embodiment of the present application provides a kitchen environment intelligent control system based on multi-sensor fusion, which is applied to cooking in the kitchen to ensure the kitchen air quality and cooking efficiency and improve the user's cooking experience. The present application can determine multiple parameters that need to be optimized by calculating the kitchen environment quality evaluation value, and optimize these parameters at the same time, select and control the kitchen equipment in combination with the intelligent control model to improve the air quality in the kitchen.

[0137] Specific implementation process: Through the data acquisition module, various sensors are arranged in the kitchen to monitor environmental data, specifically sensors related to oil fume concentration and air quality:

[0138] Odor sensor: detects the concentration of oil smoke and odor; Particle concentration sensor: monitors the concentration of particulate matter in the air; Temperature sensor: records temperature changes during cooking; Humidity sensor: monitors ambient humidity; Gas composition sensor: detects and analyzes the concentration of specific gases in the air, such as carbon monoxide, carbon dioxide, etc.

[0139] The data fusion evaluation module processes and fuses the collected kitchen environment data, including: preprocessing: normalizing the sensor data; fuzzy reasoning: using fuzzy logic reasoning to calculate the membership and weight of each sensor data to obtain a comprehensive kitchen environment quality evaluation value.

[0140] The intelligent control module selects the optimal control action based on the kitchen environment quality assessment value, including:

[0141] Option 1: State space: Contains data such as oil fume concentration, particle concentration, temperature, and humidity collected by sensors. The DQN algorithm is used to train the agent and the selection of the range hood fan speed level is optimized through the reward function.

[0142] Action space: Define the range hood fan speed level, such as low speed, medium speed, and high speed.

[0143] Option 2: Use the DQN algorithm to train the agent and optimize the choice of window opening and closing status through the reward function;

[0144] Action space: defines the window opening and closing states, such as closed, half-open, and fully open.

[0145] Option 3: Use the DQN algorithm to train the agent and optimize the selection of the gas stove fire size through the reward function;

[0146] Action space: defines the fire power of the gas stove, such as low fire, medium fire, high fire,

[0147] Option 4: Use the DQN algorithm to train the agent and optimize the output power and power-on time of the induction cooker through the reward function;

[0148] Action space: defines the output power and power-on time of the induction cooker.

[0149] The specific content is: Based on the environmental quality assessment value, the intelligent control module determines the parameters that currently need to be optimized, including the range hood fan speed, window opening and closing status, gas stove firepower, induction cooker output power and power-on time.

[0150] According to different optimization goals, the corresponding DQN algorithm model is selected for action decision-making. When the fume concentration is high, it is preferred to optimize the range hood fan speed level and adjust it to high speed. When the air quality is poor, it is preferred to optimize the window opening and closing status and adjust it to fully open. The system adjusts the working status of related kitchen equipment according to the optimal action instructions output by the DQN algorithm, which can optimize the air quality in the kitchen in real time, ensure cooking efficiency, and enhance the user's cooking experience.

[0151] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A kitchen environment intelligent control system based on multi-sensor fusion, characterized by: Including data fusion evaluation module and intelligent control module, The data fusion evaluation module fuses the collected kitchen environment data through a fuzzy inference algorithm to calculate the kitchen environment quality evaluation value; The intelligent control module constructs an intelligent control model for intelligent control, and the model combines the kitchen environment quality evaluation value to output the optimal action for automatic adjustment of kitchen equipment; Intelligent regulation through building an intelligent regulation model specifically includes: Preprocess the sensor data of the kitchen environment, including normalization and feature extraction, to obtain the state space; The preprocessed data is used as the state vector, including temperature, humidity, odor concentration, and particle concentration; According to the state vector, the action performed by the agent is determined to form a defined action space, including the speed level of the range hood fan, the fire power of the gas stove, the output power and power-on time of the induction cooker, and the opening and closing status of the smart window; It also includes a data acquisition module, which is used to monitor the kitchen environment data in real time by installing odor sensors, particle concentration sensors, temperature and humidity sensors, and gas composition sensors at various locations in the kitchen, and transmit the acquired kitchen environment data to the data fusion evaluation module; The odor sensor is used to detect volatile organic compounds or other odor molecules in the air, the particle concentration sensor is used to measure the size and number of suspended particles in the air, the temperature and humidity sensor is used to record the temperature and humidity level of the environment; the gas composition sensor is used to analyze the concentration of specific gases in the air; The calculation of the kitchen environment quality assessment value specifically includes: The collected raw data is preprocessed, the raw data is normalized, and converted into a fuzzy processing format. The normalization processing formula is expressed as: ; in, is the normalized value, and are the minimum and maximum values ​​of the i-th sensor data respectively; Calculate the membership of the normalized fuzzy value in each fuzzy set; The membership degree is expressed as: ,in, is a fuzzy set The center of is the width parameter; Using membership functions to determine weight coefficients for environmental data from each sensor; The weight coefficient is expressed as: ; in, is the weight coefficient of the ith sensor, is the importance score of the sensor, is the maximum value among all sensor scores, is the sensitivity parameter; The normalized membership degree of each sensor is multiplied by the weight coefficient to obtain a comprehensive fuzzy evaluation; ;in, It is a comprehensive fuzzy evaluation. Represents the "AND" operation of fuzzy logic; The comprehensive evaluation is fuzzily converted into specific values ​​to obtain a comprehensive kitchen environment quality evaluation value, which is used to reflect the overall condition of the kitchen environment; The specific kitchen environment quality evaluation formula is expressed as: ; in, is the comprehensive kitchen environment quality evaluation value after defuzzification. is a fuzzy set The associated specific values, m is the number of fuzzy sets, n is the number of sensors; The construction of the intelligent control model also includes the training of the intelligent agent and the design of the reward function. The reward function is designed to quantify the quality of each action through the reward function according to the execution action output by the intelligent control model and the kitchen environment quality evaluation value. It is specifically expressed as: ;in, is the reward obtained by executing action a in state s, It is the reward obtained after performing the action, and it is the feedback given by the kitchen environment. is the discount factor , used to balance the importance of immediate rewards and future rewards, It is the next state The reward for the assessment value is quantified based on the kitchen environment quality assessment value; The DQN algorithm is used to train the agent, specifically including: Initialize the main network and target network of DQN to estimate the Q value of taking an action in a given state; Define the environment in which the agent lives, including the state, action, and reward mechanism, where the environment receives the action of the agent and returns the corresponding state and reward; The agent performs actions in the environment to generate a state-action-reward-next-state quadruple and stores it in the experience replay buffer, which is used to store past interactions and perform sample training. Among them, samples are randomly drawn from the experience replay buffer and the target network is used to calculate the target value; Specifically expressed as: ; where r is the immediate reward, is the discount factor, Target network parameters, is the new state after executing action a, is in state The best action among all actions under Use the main network to predict the Q value of the current state-action pair, which is used to evaluate the expected reward of performing an action in a given state. It takes the state vector as input and outputs the predicted Q value of each action. Use an optimization algorithm to update the parameters of the main network according to the loss function, regularly copy the parameters of the main network to the target network, and continuously perform iterative training until the agent reaches the predetermined training cycle; The updating of the parameters of the main network according to the loss function specifically includes: evaluating the predicted Q value and the target by calculating the loss function The difference between the values, the loss function is expressed as: ; Where N is the batch size, which is the number of samples randomly drawn from the experience replay buffer for one training iteration, used to calculate the loss function and update the parameters of the neural network. are the parameters of the neural network, s represents the current state, and a represents the action chosen by the agent in a given state; The trained DQN model is deployed in the actual kitchen environment for real-time status monitoring and intelligent regulation. The main network predicts the Q value of each action under a specific state. The intelligent agent combines the kitchen environment quality assessment value and uses the DQN model to output the optimal action with the predicted Q value to regulate the kitchen equipment. It also includes a remote feedback warning module, which can remotely monitor the real-time status of the kitchen environment through a mobile phone APP, display the current data and evaluation results of the kitchen environment, and send warning information to users remotely when abnormal conditions are detected; The remote feedback warning module is specifically used to: display the current environment data and evaluation results of the kitchen on the display screen of the mobile phone APP; if there is an abnormal situation, send warning information to the user; the warning information includes information that the smoke concentration is too high or the harmful gas exceeds the standard; The manual intervention decision-making process through the mobile phone APP remote control interface includes: confirming the warning information and executing to stop the current cooking task, reduce the equipment workload or increase the ventilation volume.

Images