Control method of integrated stove fume extraction fan based on Bayesian network

Through Bayesian network control method, the intelligent control of integrated stove fan solves the problem of mismatch between the speed of the fan and the actual needs in the existing technology, achieving more accurate smoke removal effect and noise reduction, while adapting to user habits and public smoke. The wind pressure of the duct has changed, achieving the effect of energy conservation and emission reduction.

CN115163532BActive Publication Date: 2024-11-26INNOVATION RES INST OF ZHEJIANG UNIV OF TECH SHENGZHOU
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210271767.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-03-18
Publication Date
2024-11-26
Estimated Expiration
2042-03-18

AI Technical Summary

Technical Problem

The existing integrated stove lacks intelligence in the control of the range-sucking fume fan, which leads to mismatch between the fan speed and actual demand, resulting in poor smoke removal and noise problems, especially when the wind pressure of the public flue is too high.

Method used

Using a Bayesian network-based control method, the initial Bayesian network model is established and trained to build a fan control module by measuring related variables such as public flue air pressure, smoke concentration, pot bottom temperature and fan speed at the previous moment. Adjust the fan motor speed in real time for more precise control.

Benefits of technology

It realizes intelligent control based on user habits and on-site conditions, improves smoke removal effect, reduces noise interference, and continuously optimizes the model to adapt to individual habits during long-term use, solving the fan speed caused by public flue air pressure. Match the problem and achieve the purpose of energy conservation and emission reduction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115163532B_ABST
    Figure CN115163532B_ABST
Patent Text Reader

Abstract

The present invention provides an integrated stove range hood fan control method based on a Bayesian network, which realizes intelligent control of the motor speed of the integrated stove fan according to the on-site conditions during the user's use, and can continuously collect data in real time and correct the model during the user's use, and continuously improve the accuracy of the system judgment result to achieve more in line with the user's personal habits and more accurate control, solve the problem that the fan speed does not match the actual demand due to excessive wind pressure in the public flue, achieve the purpose of effective smoke removal, reduce the noise trouble when the integrated stove is removing smoke, and achieve the purpose of energy saving and emission reduction. In the control method: during the user's use, the integrated stove control system controls the operation of the range hood fan according to the following steps: S1, the measured relevant variable data is processed and input into the Bayesian network model for reasoning and judgment; S2, the judgment result is transmitted to the fan control module, and the motor speed of the integrated stove fan is automatically adjusted according to the result.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of kitchen appliances and artificial intelligence, and in particular to a method for controlling an integrated stove fume exhaust fan based on a Bayesian network. Background Art

[0002] In recent years, with the proposal of "dual carbon economy", integrated stoves have gradually occupied a dominant position in kitchen appliances. Excellent fume extraction effect is the main reason why integrated stoves stand out, but achieving more intelligent control while having excellent smoke removal effect is still the primary problem to be solved. At present, the integrated stoves on the market basically control the range hood fan with simple manual high and low gears, which cannot reach the level of intelligence. Most users do not frequently adjust the gear after determining the gear during use. However, during use, users often use different cooking methods, such as (frying, frying, deep-frying, boiling), etc. Different methods require different gears. If the user forgets to adjust when changing the cooking method, the low gear will always be used, which will cause the range hood to have poor removal effect, and frequent gear changes are relatively troublesome. Similarly, if the user always uses a high gear, although the best smoke removal effect is achieved, the high decibel noise generated by the fan when working will also cause great trouble to the user.

[0003] Although there is currently a control system that uses a single smoke sensor as a detection element to detect smoke concentration and automatically adjust the air volume of the range hood, the smoke exhaust ducts of most users belong to public flues. When nearby users cook at the same time, the wind pressure in the public flue will increase, causing the fan gear to not match actual needs, and poor smoke removal will still occur, and sometimes even the problem of oil smoke backflow. Summary of the invention

[0004] In order to overcome the above-mentioned deficiencies in the prior art, the present invention provides an integrated stove range hood fan control method based on a Bayesian network, which realizes the intelligent control of the integrated stove fan motor speed according to the on-site conditions during user use, and can continuously collect data in real time and correct the model during normal user use, and continuously improve the accuracy of the system judgment results to achieve more in line with the user's personal habits and more precise control, solve the problem that the fan speed does not match the actual demand due to excessive wind pressure in the public flue, achieve the purpose of effective smoke removal, and minimize the noise trouble during integrated stove smoke removal, while achieving the purpose of energy saving and emission reduction.

[0005] The technical solution of the present invention to solve the technical problem is as follows:

[0006] Integrated stove oil fume extraction fan control method based on Bayesian network. In this control method: measurable relevant variables affecting the user's adjustment of the integrated stove fan motor speed are defined; an initial Bayesian network model between the relevant variables and the integrated stove fan motor speed is established, and the Bayesian network model is trained and learned according to the initial general average data; a fan control module composed of the Bayesian network model is constructed and integrated into the integrated stove control system;

[0007] During the user's use, the integrated stove control system controls the operation of the oil fume extraction fan according to the following steps:

[0008] S1, Process the measured relevant variable data and input it into the Bayesian network model for inference and judgment;

[0009] S2, Transmit the judgment result to the fan control module, and the integrated stove control system automatically adjusts the speed of the integrated stove fan motor according to the result.

[0010] Furthermore, in the aforementioned integrated stove oil fume extraction fan control method based on Bayesian network, the relevant variables may include the wind pressure of the common flue, the smoke concentration, the bottom temperature of the pot, and the speed of the integrated stove fan motor at the previous moment (where the speed of the integrated stove fan motor at the previous moment includes the adjusted speed of the integrated stove fan motor and the speed of the integrated stove fan motor adjusted by the user himself / herself when not satisfied with the adjustment result).

[0011] Furthermore, in the aforementioned integrated stove oil fume extraction fan control method based on Bayesian network, the Bayesian network is a directed acyclic graph, where nodes represent random variables, and the edges between nodes represent the direct dependence relationship between variables; each node Xi is attached with a probability distribution table p(X i 丨parents(Xi)), where parents(X i ) refers to all the parent nodes of X i , and the parent node of X i refers to the node with an edge shooting at X i . The Bayesian network expresses two meanings: ① Express the dependence and independence relationship between variables through directed edges; ② Express the degree of dependence of the node on its parent nodes through the conditional probability of each node. It is necessary to determine not only the influence degree of the wind pressure of the common flue, the smoke concentration, the bottom temperature of the pot and the gas stove usage time on the integrated stove fan according to the relevant variables, but also consider the air volume of the integrated stove fan at the previous moment. This model belongs to a Bayesian network model with feedback.

[0012] Furthermore, in the aforementioned integrated stove oil fume extraction fan control method based on Bayesian network, the training and learning steps of the Bayesian network model specifically include:

[0013] Perform mean filtering and median filtering on the original data set D to obtain the training data set D t and validation dataset D v ;

[0014] For the training data set D t and validation dataset D v Perform straight line fitting, and then discretize through binning to obtain the discrete value of the change trend of each data;

[0015] Use the FastICA algorithm on the discrete training set to obtain the separation matrix W;

[0016] Use the separation matrix W to transform the discrete data set into another feature space to obtain the training set D ti and validation set D vi ;

[0017] In the training set D ti Learning Bayesian network model parameters.

[0018] Furthermore, in the aforementioned integrated stove range hood fan control method based on Bayesian network, due to different personal habits of users, the integrated stove control system has achieved more accurate judgment and control through long-term relevant variable data collection and continuous correction of the Bayesian network model; the specific steps are as follows:

[0019] S1.1: Real-time and continuous collection of sample data of each variable node;

[0020] S1.2: Process the variable data and transfer the sample data to the data storage device for database update;

[0021] S1.3: Use the real-time Bayesian network parameter learning algorithm to continuously train the Bayesian network model to achieve more precise control of the wind turbine that is more in line with user habits.

[0022] Furthermore, in the aforementioned integrated stove range hood fan control method based on Bayesian network, the Bayesian network model is optimized and adjusted using a real-time Bayesian network parameter learning algorithm. The specific implementation steps are:

[0023] (1) Discretize the relevant variables as follows: Discretize the wind pressure P of the public flue into (p1, p2, p3···p A ), discretize the speed N of the integrated stove fan motor at the previous moment into (n1, n2, n3···n B ), the smoke density L is discretized into (l1,l2,l3···l C ), discretize the pot bottom temperature R into (p1, p2, p3···p D ), the motor speed adjustment value W is discretized into (w1,w2,w3···wE );

[0024] (2) Use the array Count[A][B][C][D][E] to count the frequency of occurrence of each variable combination, where A, B, C, D, and E are the number of discrete values ​​of P, N, L, R, and W, respectively, and are used to record the frequency of occurrence of each discrete value; the initial value of the array element is 0, and the subscript starts at 1;

[0025] (3) Read the combined value of variables P, R, L, N, and W, and obtain the N saved at the last moment. Compare the values ​​of each variable and their discreteness. Set P to p f , N is n g , R is r h , L takes l i , W takes w j , then the sampled combined value (p f , n g , r h , l i , w j ), f∈(1···A), g∈(1···B), i∈(1···C), h∈(1···D), j∈(1···E);

[0026] (4) Check whether the speed of the integrated stove fan motor at the previous moment meets the speed of the integrated stove fan motor based on the user average data. If not, return to step (3). If satisfied, save the current N as N' at the next moment, N'=N, and execute step (5);

[0027] (5) The number of sampling times K is equal to K+1; the array elements Count[f][g][h][i][j] with subscripts f, g, h, i, j = Count[f][g][h][i][j]+1. At this time, the conditional probability can be calculated:

[0028]

[0029] When the learning time is long enough and the learning samples are sufficient:

[0030]

[0031] The meaning is that only the first four subscripts of the Count array elements can be f, g, h, i, and the value of the last subscript x can be the sum of all array elements from 1 to E, that is, to satisfy (P = p f ,R=r h ,L=l i ,N=n g ) The total number of samples; Divide the right side of the above two equations to get the conditional probability:

[0032]

[0033] According to the above formula, the current real-time conditional probability of each node variable can be calculated;

[0034] (6) Output the Bayesian network learned by current real-time sampling;

[0035] (7) Return to step (3) and repeat (3) to (7);

[0036] The real-time Bayesian network parameter learning algorithm starts running when the integrated stove fan is started, updates the conditional variables of each node in real time, and outputs the updated Bayesian network in real time.

[0037] Furthermore, in the aforementioned integrated stove range fume exhaust fan control method based on Bayesian network, the integrated stove control system adopts a programmable controller to collect data and control the integrated stove control system.

[0038] Compared with the prior art, the technical solution of the present invention has made the following significant progress: the device initially makes judgments based on the average data of the general public; due to the different personal habits of users, the control system has achieved more accurate judgments and controls after a long period of relevant variable data collection and continuous correction of the Bayesian network model, that is, it has achieved intelligent control that conforms to the user's usage habits, helping users solve the problem of untimely fume cleaning caused by not timely adjusting the fan gear position during the use of the integrated stove and the trouble of continuous noise caused by the high gear position, and also solved the problem of "backflow" caused by the change of public flue wind pressure caused by multiple users using the range hood at the same time. At the same time, while realizing intelligent control, it also achieves the purpose of energy saving and emission reduction. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 : is a Bayesian network structure diagram of the integrated stove fan control in an embodiment of the present invention;

[0040] Figure 2 This is a flow chart of integrated stove fan control in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The present invention is further described below in conjunction with the accompanying drawings and specific implementation methods (including examples), but is not intended to be limiting of the present invention. The contents not described in detail in the following examples are all common technical knowledge in the art.

[0042] In the control method of the integrated stove's oil extraction fan based on the Bayesian network of the present invention: measurable relevant variables that affect the user's adjustment of the integrated stove fan motor speed are identified; an initial Bayesian network model between the relevant variables and the integrated stove fan's motor speed is established, and the Bayesian network model is trained and learned based on the initial general average data; a fan control module composed of the Bayesian network model is constructed and integrated into the integrated stove control system;

[0043] During the user's use, the integrated stove control system controls the operation of the oil extraction fan according to the following steps:

[0044] S1, process the measured relevant variable data and input it into the Bayesian network model for inference and judgment;

[0045] S2, send the judgment result to the fan control module, and the integrated stove control system automatically adjusts the speed of the integrated stove fan motor according to the result.

[0046] As a specific implementation manner, the relevant variables include the wind pressure of the common flue, the smoke concentration, the bottom temperature of the pot, and the speed of the integrated stove fan motor at the previous moment.

[0047] As a specific implementation manner, the Bayesian network is a directed acyclic graph, where the nodes represent random variables and the edges between the nodes represent the direct dependence relationship between the variables; each node Xi is attached with a probability distribution table p(X i 丨parents(Xi)), where parents(X i ) refers to all the parent nodes of X i , and the parent node of X i refers to the node with an edge shooting at X i . The Bayesian network expresses two meanings: ① expressing the dependence and independence relationship between variables through directed edges; ② expressing the degree of dependence of the node on its parent nodes through the conditional probability of each node. According to the relevant variables, determine the influence degree of the common flue wind pressure, smoke concentration, bottom temperature of the pot and the gas stove usage time on the integrated stove fan, and also consider the air volume of the integrated stove fan at the previous moment. This model belongs to a Bayesian network model with feedback.

[0048] As a specific implementation manner, the training and learning steps of the Bayesian network model specifically include:

[0049] (1) Perform mean filtering and median filtering on the original data set D to obtain the training data set D t and the validation data set D v ;

[0050] (2) For the training data set D t and the validation data set D vPerform straight line fitting, and then discretize through binning to obtain the discrete value of the change trend of each data;

[0051] (3) Use the FastICA algorithm on the discrete training set to obtain the separation matrix W;

[0052] (4) Use the separation matrix W to transform the discrete data set into another feature space to obtain the training set D ti and validation set D vi ;

[0053] (5) In the training set D ti Learning Bayesian network model parameters.

[0054] As a specific implementation method, due to the different personal habits of users, the integrated stove control system achieves more accurate judgment and control through long-term collection of relevant variable data and continuous correction of the Bayesian network model; the specific steps are as follows:

[0055] S1.1: Real-time and continuous collection of sample data of each variable node;

[0056] S1.2: Process the variable data and transfer the sample data to the data storage device for database update;

[0057] S1.3: Use the real-time Bayesian network parameter learning algorithm to continuously train the Bayesian network model to achieve more precise control of the wind turbine that is more in line with user habits.

[0058] As a specific implementation method, a real-time Bayesian network parameter learning algorithm is used to optimize and adjust the parameters of the Bayesian network model. The specific implementation steps are:

[0059] (1) Discretize the relevant variables as follows: Discretize the wind pressure P of the public flue into (p1, p2, p3···p A ), discretize the speed N of the integrated stove fan motor at the previous moment into (n1, n2, n3···n B ), the smoke density L is discretized into (l1,l2,l3···l C ), discretize the pot bottom temperature R into (p1, p2, p3···p D ), the motor speed adjustment value W is discretized into (w1,w2,w3···w E );

[0060] (2) Use the array Count[A][B][C][D][E] to count the frequency of occurrence of each variable combination, where A, B, C, D, and E are the number of discrete values ​​of P, N, L, R, and W, respectively, and are used to record the frequency of occurrence of each discrete value; the initial value of the array element is 0, and the subscript starts at 1;

[0061] (3) Read the combined value of variables P, R, L, N, and W, and obtain the N saved at the last moment. Compare the values ​​of each variable and their discreteness. Set P to p f , N is n g , R is r h , L takes l i , W takes w j , then the sampled combined value (p f , n g , r h , l i , w j ), f∈(1···A), g∈(1···B), i∈(1···C), h∈(1···D), j∈(1···E);

[0062] (4) Check whether the fan motor speed at the previous moment meets the fan motor speed of the integrated stove based on the user average data. If not, return to step (3). If satisfied, save the current N as N' at the next moment, N'=N, and execute step (5);

[0063] (5) The number of sampling times K is equal to K+1; the array elements Count[f][g][h][i][j] with subscripts f, g, h, i, j = Count[f][g][h][i][j]+1. At this time, the conditional probability can be calculated:

[0064]

[0065] When the learning time is long enough and the learning samples are sufficient:

[0066]

[0067] The meaning is that only the first four subscripts of the Count array elements can be f, g, h, i, and the value of the last subscript x can be the sum of all array elements from 1 to E, that is, to satisfy (P = p f ,R=r h ,L=l i ,N=n g ) The total number of samples; Divide the right side of the above two equations to get the conditional probability:

[0068]

[0069] According to the above formula, the current real-time conditional probability of each node variable can be calculated;

[0070] (6) Output the Bayesian network learned by current real-time sampling;

[0071] (7) Go back to step (3) and repeat steps (3) - (7);

[0072] The real - time Bayesian network parameter learning algorithm starts running when the integrated stove fan starts, updates the conditional variables of each node in real - time, and outputs the updated Bayesian network in real - time.

[0073] As a specific implementation, the integrated stove control system uses a programmable logic controller to be responsible for collecting data and controlling the integrated stove control system. Using a single - chip microcomputer to build the control system has good reliability and strong anti - interference ability.

[0074] The following further illustrates the present invention with reference to embodiments:

[0075] In this embodiment, specifically, as shown in the appendix Figure 1 and Figure 2 , the present invention provides a Bayesian network structure for real - time update, which is realized by a real - time Bayesian network parameter learning algorithm. The specific implementation method: According to the measured P, R, L and the values of N' saved at the previous moment (p f , r h , l i , n g ), querying the array Count will obtain a set of values Count[f][g][h][i][x], X ∈ (1···E). For this set of values, find the maximum Max(Count[f][g][h][i][x], X ∈ (1···E)), then obtain x = x' such that Count[f][g][h][i][x'] is the largest, that is, P(W = w x' , 丨P = p f , R = r h , L = l i , N' = n g ) is the largest. Then the adjustment amount of the fan motor speed W = W x' . This algorithm starts running when the integrated stove fan starts, obtains the value combination of each variable from devices such as sensors in real - time and statistically calculates the probability of the value combination, updates the variables of each node in real - time, and updates the Bayesian network model. In this way, by constructing a real - time update Bayesian network model through the real - time parameter learning algorithm, the work of measuring uncertain variables and complex feedback control by increasing the number of sensors can be transformed into a simple probability calculation process. As the user's usage time increases continuously, the learning samples based on user habits increase continuously, and the Bayesian network model will gradually stabilize and the control will be more accurate.

[0076] In this embodiment, specifically, the present invention provides an intelligent control method for an integrated stove fan based on a Bayesian network, and the specific control process is as follows: when the integrated stove is turned on, sensors and other equipment begin to collect data of various variables in real time, and after discretization, the data is input into a pre-trained Bayesian network model for judgment, and the judgment result is transmitted to the fan control submodule in the integrated stove control system. The control module adjusts the speed of the integrated stove fan, and the adjusted motor speed is fed back to the Bayesian network model in the form of related variables as the motor speed N at the previous moment. If N is basically consistent with the actual requirement judged by the model, the control system will maintain the current motor speed. If not, the above process will continue until it is satisfied.

[0077] The above general description of the invention involved in this application and the description of its specific implementation methods should not be understood as limiting the technical solutions of the invention. Based on the disclosure of this application, those skilled in the art can add, reduce or combine the disclosed technical features in the above general description and / or specific implementation methods (including examples) without violating the constituent elements of the invention involved, to form other technical solutions within the scope of protection of this application.

Claims

1. A Bayesian network-based integrated stove range hood fan control method, characterized in that: In this control method: the measurable related variables that affect the user's adjustment of the integrated stove fan motor speed are identified; an initial Bayesian network model between the related variables and the integrated stove fan motor speed is established, and the Bayesian network model is trained and learned based on the initial mass average data; a fan control module composed of the Bayesian network model is constructed and integrated into the integrated stove control system; During user use, the integrated stove control system controls the operation of the range hood fan according to the following steps: S1, the measured relevant variable data is processed and input into the Bayesian network model for reasoning and judgment; S2, transmitting the judgment result to the fan control module, and the integrated stove control system automatically adjusts the speed of the integrated stove fan motor according to the result; The relevant variables include the wind pressure of the public flue, smoke concentration, pot bottom temperature, and the speed of the integrated stove fan motor at the last moment; The Bayesian network is a directed acyclic graph, in which nodes represent random variables and edges between nodes represent direct dependencies between variables; each node Xi is attached with a probability distribution table p(X i |parents(Xi)), where parents(X i ) means X i All parent nodes of X i The parent node is the one with an edge to X i Nodes; The training and learning steps of the Bayesian network model specifically include: Perform mean filtering and median filtering on the original data set D to obtain the training data set D t and validation dataset D v ; For the training data set D t and validation dataset D v Perform straight line fitting, and then discretize through binning to obtain the discrete value of the change trend of each data; Use the FastICA algorithm on the discrete training set to obtain the separation matrix W; Use the separation matrix W to transform the discrete data set into another feature space to obtain the training set D ti and validation set D vi ; In the training set D ti Learning Bayesian network model parameters.

2. The integrated stove range fume exhaust fan control method based on Bayesian network according to claim 1 is characterized in that: In this method, due to the different personal habits of users, the integrated stove control system achieves more accurate judgment and control through long-term collection of relevant variable data and continuous correction of the Bayesian network model; the specific steps are as follows: S1.1: Real-time and continuous collection of sample data of each variable node; S1.2: Process the variable data and transfer the sample data to the data storage device for database update; S1.3: Use the real-time Bayesian network parameter learning algorithm to continuously train the Bayesian network model to achieve more precise control of the wind turbine that is more in line with user habits.

3. The integrated stove range fume exhaust fan control method based on Bayesian network according to claim 2 is characterized in that: The real-time Bayesian network parameter learning algorithm is used to optimize the parameters of the Bayesian network model. The specific implementation steps are as follows: (1) Discretize the relevant variables as follows: Discretize the wind pressure P of the public flue into (p1, p2, p3···p A ), discretize the speed N of the integrated stove fan motor at the previous moment into (n1, n2, n3···n B ), the smoke density L is discretized into (l1, l2, l3···l C ), discretize the pot bottom temperature R into (p1, p2, p3···p D ), the motor speed adjustment value W is discretized into (w1, w2, w3···w E ); (2) Use the array Count[A][B][C][D][E] to count the frequency of occurrence of each variable combination, where A, B, C, D, and E are the number of discrete values ​​of P, N, L, R, and W, respectively, and are used to record the frequency of occurrence of each discrete value; the initial value of the array element is 0, and the subscript starts at 1; (3) Read the combined value of variables P, R, L, N, and W, and obtain the N saved at the last moment. Compare the values ​​of each variable and their discreteness. Set P to p f , N is n g , R is r h , L takes l i , W takes w j , then the sampled combined value (p f , n g , r h , l i , w j ), f∈(1···A), g∈(1···B), i∈(1···C), h∈(1···D), j∈(1···E); (4) Check whether the speed of the integrated stove fan motor at the previous moment meets the speed of the integrated stove fan motor based on the user average data. If not, return to step (3). If satisfied, save the current N as N' at the next moment, N'=N, and execute step (5); (5) The number of sampling times K is equal to K+1; the array elements Count[f][g][h][i][j] with subscripts f, g, h, i, j = Count[f][g][h][i][j]+1. At this time, the conditional probability can be calculated: When the learning time is long enough and the learning samples are sufficient: The meaning is that only the first four subscripts of the Count array elements have values ​​of f, g, h, i, and the value of the last subscript x is the sum of all array elements from 1 to E, that is, to find the value that satisfies (P = p f , R = r h , L = l i , N = n g ) The total number of samples; Divide the right side of the above two equations to get the conditional probability: According to the above formula, the current real-time conditional probability of each node variable can be calculated; (6) Output the Bayesian network learned by current real-time sampling; (7) Return to step (3) and repeat (3) to (7); The real-time Bayesian network parameter learning algorithm starts running when the integrated stove fan is started, updates the conditional variables of each node in real time, and outputs the updated Bayesian network in real time.

4. The integrated stove range fume exhaust fan control method based on Bayesian network according to any one of claims 1 to 3, characterized in that: The integrated stove control system adopts a programmable controller to collect data and control the integrated stove control system.