Gas water heater self-learning and self-starting zero cold water method based on deep learning and sound detection
Through deep learning and sound detection technology, users' voice data are collected using microphone matrix, convolutional neural network model training is carried out to predict hot water demand, and the high energy consumption and resource waste caused by the zero-cold water start strategy of gas water heater is solved, and the problem of high energy consumption and resource waste is achieved. Accurate prediction of user hot water needs and resource conservation.
Patent Information
- Application Number
- CN202411054393.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2025-01-17
AI Technical Summary
The existing zero-cold water start strategy of gas water heaters has problems such as frequent start-up, resulting in high energy consumption and waste of resources.
Using self-learning and self-start methods based on deep learning and sound detection, users' voice data are collected through microphone matrix, data preprocessing and convolutional neural network model training, predict hot water demand, and control zero-cold water of gas water heater based on prediction results. Start and stop, combine time data for self-learning of the model.
Accurate prediction of user hot water needs is achieved, resources are saved, and prediction accuracy is improved.
Smart Images
Figure CN119321620A_ABST
Abstract
Description
Technical Field
[0001] The invention discloses a self-learning and self-starting zero cold water method for a gas water heater based on deep learning and sound detection, and belongs to the technical field of water heaters. Background Art
[0002] The current gas water heater zero cold water function has problems such as high energy consumption and resource waste during use. It is necessary to improve its startup strategy to save resources and user experience. Summary of the invention
[0003] The purpose of the present invention is to provide a self-learning and self-starting zero cold water method for gas water heaters based on deep learning and sound detection, so as to solve the problems in the prior art that the zero cold water starting strategy of gas water heaters has frequent starting, resulting in high energy consumption and waste of resources.
[0004] A self-learning and self-starting zero cold water method for gas water heaters based on deep learning and sound detection includes collecting user voice data, data preprocessing, convolutional neural network model training, hot water demand prediction, starting gas water heaters, water heater control, model self-learning and time data integration.
[0005] Collecting user voice data includes forming a microphone matrix through microphones at different locations in the room or microphones integrated into home appliances. User voice data includes speaking sounds, walking sounds, and door opening and closing sounds;
[0006] The user sound data includes the time data corresponding to the sound. The user sound data is recorded with two labels: used and unused. The unused label is the 15-second continuous sound other than one minute before and after the use of the gas water heater, and the used label is the 15-second continuous sound before the use of the gas water heater.
[0007] Data preprocessing includes feature extraction of user voice data, which includes Fourier transforming the voice time domain data into voice frequency domain signals:
[0008] F(ω)=∫[g(t)*e -jωt ]dt;
[0009] Where, F(ω) is the sound frequency domain signal, ∫ is the integral, t is the sound time domain data, g(t) is the transformation function, e is the natural logarithm, j is the imaginary unit, and ω is the period;
[0010] The audio frequency domain signal is converted into a Mel spectrum of Mel scale, and the logarithm of the Mel spectrum is taken to obtain a logarithmic Mel spectrum m;
[0011]
[0012] The training of the convolutional neural network model includes taking the time data and the logarithmic Mel spectrum as input and performing One-Hot encoding on the labels;
[0013] Set the convolutional neural network model parameters, number of iterations, and number of inputs per batch, match the logarithmic Mel spectrum and the label one by one, input them into the neural network model for iterative training, and obtain the trained convolutional neural network model after the iteration.
[0014] Hot water demand prediction involves inputting user voice data into a trained convolutional neural network model and outputting prediction results.
[0015] Starting the gas water heater includes: the controller of the gas water heater sends a signal of whether to start zero cold water to the gas water heater according to the prediction result, so as to instruct the gas water heater whether to start zero cold water.
[0016] The control logic of the gas water heater controller is that when the zero cold water is not started, the microphone continuously collects user voice data. After the user voice data is feature extracted, it is transmitted to the convolutional neural network model for processing. The convolutional neural network model determines whether to turn on the zero cold water according to the processing results. When the zero cold water is started, the user voice data stops feature extraction. Five minutes after the hot water is used, the user voice data is re-extracted.
[0017] Model self-learning includes continuously collecting user voice data, extracting features and then inputting them into the convolutional neural network model for iterative training. During the training process, the convolutional layer is frozen and the parameters of the fully connected layer are updated to enable the convolutional neural network model to self-learn based on user usage habits.
[0018] The integration of time data includes integrating the time data corresponding to the sound into the convolutional neural network model for training, so that the convolutional neural network model can learn the time pattern of users using hot water.
[0019] After the convolution of the user voice data channel is completed, data splicing is performed to splice the user voice data and the time data corresponding to the sound.
[0020] Compared with the prior art, the present invention has the following beneficial effects: the present invention realizes accurate prediction of users' hot water demand, saves resources, and improves the prediction accuracy for individual users. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Figure 1 is a microphone arrangement diagram of an embodiment of the present invention;
[0022] Figure 2 It is a convolutional neural network structure diagram of the present invention;
[0023] Figure 3 It is a technical flow chart of the present invention. DETAILED DESCRIPTION
[0024] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention is described clearly and completely below. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0025] A self-learning and self-starting zero cold water method for gas water heaters based on deep learning and sound detection includes collecting user voice data, data preprocessing, convolutional neural network model training, hot water demand prediction, starting gas water heaters, water heater control, model self-learning and time data integration.
[0026] Collecting user voice data includes forming a microphone matrix through microphones at different locations in the room or microphones integrated into home appliances. User voice data includes speaking sounds, walking sounds, and door opening and closing sounds;
[0027] The user sound data includes the time data corresponding to the sound. The user sound data is recorded with two labels: used and unused. The unused label is the 15-second continuous sound other than one minute before and after the use of the gas water heater, and the used label is the 15-second continuous sound before the use of the gas water heater.
[0028] Data preprocessing includes feature extraction of user voice data, which includes Fourier transforming the voice time domain data into voice frequency domain signals:
[0029] F(ω)=∫[g(t)*e -jωt ]dt;
[0030] Where, F(ω) is the sound frequency domain signal, ∫ is the integral, t is the sound time domain data, g(t) is the transformation function, e is the natural logarithm, j is the imaginary unit, and ω is the period;
[0031] The audio frequency domain signal is converted into a Mel spectrum of Mel scale, and the logarithm of the Mel spectrum is taken to obtain a logarithmic Mel spectrum m;
[0032]
[0033] The training of the convolutional neural network model includes taking the time data and the logarithmic Mel spectrum as input and performing One-Hot encoding on the labels;
[0034] Set the convolutional neural network model parameters, number of iterations, and number of inputs per batch, match the logarithmic Mel spectrum and the label one by one, input them into the neural network model for iterative training, and obtain the trained convolutional neural network model after the iteration.
[0035] Hot water demand prediction involves inputting user voice data into a trained convolutional neural network model and outputting prediction results.
[0036] Starting the gas water heater includes: the controller of the gas water heater sends a signal of whether to start zero cold water to the gas water heater according to the prediction result, so as to instruct the gas water heater whether to start zero cold water.
[0037] The control logic of the gas water heater controller is that when the zero cold water is not started, the microphone continuously collects user voice data. After the user voice data is feature extracted, it is transmitted to the convolutional neural network model for processing. The convolutional neural network model determines whether to turn on the zero cold water according to the processing results. When the zero cold water is started, the user voice data stops feature extraction. Five minutes after the hot water is used, the user voice data is re-extracted.
[0038] Model self-learning includes continuously collecting user voice data, extracting features and then inputting them into the convolutional neural network model for iterative training. During the training process, the convolutional layer is frozen and the parameters of the fully connected layer are updated to enable the convolutional neural network model to self-learn based on user usage habits.
[0039] The integration of time data includes integrating the time data corresponding to the sound into the convolutional neural network model for training, so that the convolutional neural network model can learn the time pattern of users using hot water.
[0040] After the convolution of the user voice data channel is completed, data splicing is performed to splice the user voice data and the time data corresponding to the sound.
[0041] The microphone arrangement of the embodiment of the present invention is as follows Figure 1 As shown, microphones are placed in open and unobstructed locations in the bedroom, kitchen, bathroom, balcony door, and beside the washbasin and vegetable sink. The scenario of using hot water is simulated, that is, walking from various locations to the bathroom and turning on the faucet in the kitchen, and the sound of this time period is recorded and saved using microphones in various locations to obtain sound data and time data. The time of hot water use is saved, and only the hour bit is retained.
[0042] The convolutional neural network structure diagram of the present invention is as follows Figure 2 As shown, by combining a multi-point microphone matrix and a convolutional neural network algorithm, the sound data of users using hot water is collected. After extracting features from these sound data, the convolutional neural network model is trained to predict whether hot water will be needed in the next period of time based on the sound data, and the zero cold water function of the gas water heater can be started in advance based on the prediction results. Figure 2In the figure, Input represents 7 input situations, Conv represents convolution layer, Relu represents activation function, MaxPoling represents maximum pooling, concat represents feature connection, timedata represents time data, softmax represents activation function, output represents output, and Dense represents weighted sum.
[0043] The technical process of the present invention is as follows Figure 3 As shown, the present invention guides the gas water heater startup strategy through model prediction results to save resources and user experience, and the convolutional neural network model builds a multi-channel network to adapt to multi-point sound data in the room to improve the prediction accuracy; the time data is integrated into the network for training, so that the neural network can learn the time pattern of users using hot water, thereby improving the prediction accuracy; after the model is deployed, the user's daily data is collected, and the data is input into the model after preprocessing, the convolution layer of the model is frozen for training, and the fully connected layer is parameter updated to achieve continuous learning of the model and improve the prediction accuracy for individual users.
[0044] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features may be replaced by equivalents, and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A self-learning and self-starting zero cold water method for gas water heaters based on deep learning and sound detection, characterized in that: It includes collecting user voice data, data preprocessing, convolutional neural network model training, hot water demand prediction, starting gas water heaters, water heater control, model self-learning and time data integration.
2. The method for self-learning and self-starting zero cold water for gas water heater based on deep learning and sound detection according to claim 1 is characterized in that: Collecting user voice data includes forming a microphone matrix through microphones at different locations in the room or microphones integrated into home appliances. User voice data includes speaking sounds, walking sounds, and door opening and closing sounds; The user sound data includes the time data corresponding to the sound. The user sound data is recorded with two labels: used and unused. The unused label is the 15-second continuous sound other than one minute before and after the use of the gas water heater, and the used label is the 15-second continuous sound before the use of the gas water heater.
3. The method for self-learning and self-starting zero cold water of a gas water heater based on deep learning and sound detection according to claim 2 is characterized in that: Data preprocessing includes feature extraction of user voice data, which includes Fourier transforming the voice time domain data into voice frequency domain signals: F(ω)=∫[g(t)*e -jωt ]dt; Where, F(ω) is the sound frequency domain signal, ∫ is the integral, t is the sound time domain data, g(t) is the transformation function, e is the natural logarithm, j is the imaginary unit, and ω is the period; The audio frequency domain signal is converted into a Mel spectrum of Mel scale, and the logarithm of the Mel spectrum is taken to obtain a logarithmic Mel spectrum m; 4. The method for self-learning and self-starting zero cold water for gas water heater based on deep learning and sound detection according to claim 3 is characterized in that: The training of the convolutional neural network model includes taking the time data and the logarithmic Mel spectrum as input and performing One-Hot encoding on the labels; Set the convolutional neural network model parameters, number of iterations, and number of inputs per batch, match the logarithmic Mel spectrum and the label one by one, input them into the neural network model for iterative training, and obtain the trained convolutional neural network model after the iteration.
5. The method for self-learning and self-starting zero cold water for gas water heater based on deep learning and sound detection according to claim 4 is characterized in that: Hot water demand prediction involves inputting user voice data into a trained convolutional neural network model and outputting prediction results.
6. The method for self-learning and self-starting zero cold water for gas water heater based on deep learning and sound detection according to claim 5 is characterized in that: Starting the gas water heater includes: the controller of the gas water heater sends a signal of whether to start zero cold water to the gas water heater according to the prediction result, so as to instruct the gas water heater whether to start zero cold water.
7. The method for self-learning and self-starting zero cold water for gas water heater based on deep learning and sound detection according to claim 6 is characterized in that: The control logic of the gas water heater controller is that when the zero cold water is not started, the microphone continuously collects user voice data. After the user voice data is feature extracted, it is transmitted to the convolutional neural network model for processing. The convolutional neural network model determines whether to turn on the zero cold water according to the processing results. When the zero cold water is started, the user voice data stops feature extraction. Five minutes after the hot water is used, the user voice data is re-extracted.
8. The method for self-learning and self-starting zero cold water for gas water heater based on deep learning and sound detection according to claim 7 is characterized in that: Model self-learning includes continuously collecting user voice data, extracting features and then inputting them into the convolutional neural network model for iterative training. During the training process, the convolutional layer is frozen and the parameters of the fully connected layer are updated to enable the convolutional neural network model to self-learn based on user usage habits.
9. The method for self-learning and self-starting zero cold water for gas water heater based on deep learning and sound detection according to claim 8, characterized in that: The integration of time data includes integrating the time data corresponding to the sound into the convolutional neural network model for training, so that the convolutional neural network model can learn the time pattern of users using hot water.
10. The method for self-learning and self-starting zero cold water for gas water heater based on deep learning and sound detection according to claim 9, characterized in that: After the convolution of the user voice data channel is completed, data splicing is performed to splice the user voice data and the time data corresponding to the sound.