Intelligent temperature control method for heating kettle

Abstract

The invention provides an intelligent temperature control method for a heating kettle. The method comprises the steps that various sensors are integrated inside and outside the kettle to collect various data of the inside and the environment of the kettle in real time; constructing and training a hybrid deep learning model based on the combination of a convolutional neural network and a long-short-term memory network, and deploying the trained hybrid deep learning model into a control system of the kettle; carrying out data fusion and synchronization and real-time denoising, and carrying out real-time feature extraction on the denoised sensor data; dynamically adjusting heating power, and detecting water temperature abnormity; and a user-friendly interface design and a remote control function are designed. According to the intelligent temperature control kettle system combining deep learning, fuzzy logic control and sound wave sensing technologies, high-precision temperature control, intelligent analysis and personalized service are achieved through comprehensive application of multiple advanced technologies, and therefore the user experience is remarkably improved.

Description

Technical Field

[0001] The invention belongs to the technical field of intelligent temperature control, and in particular relates to an intelligent temperature control method for a heating kettle. Background Art

[0002] With the rapid development of smart home products, smart kettles have gradually entered people's daily lives. Traditional kettles usually only have a simple heating function. After the user manually sets the temperature, the kettle will heat according to the preset time and power. However, this method has many shortcomings and is difficult to meet the personalized needs of users. First, traditional kettles cannot accurately control the temperature according to different types of drinks and user preferences, resulting in the heated water temperature often not being able to achieve the best drinking effect. Secondly, traditional kettles cannot monitor and feedback water temperature changes in real time, resulting in frequent over-temperature or under-temperature phenomena, affecting user experience. Thirdly, traditional kettles lack intelligent analysis and learning functions and cannot optimize the heating process based on historical usage data. In addition, the temperature control system of traditional kettles is usually relatively simple, using a single temperature sensor and a fixed control algorithm, which is difficult to adapt to complex environments and changing usage scenarios. Even some more advanced smart kettles are often limited to remote control through mobile phone applications, lacking real intelligent and personalized services. Based on the above deficiencies in the existing technologies, the market urgently needs an intelligent kettle system that can provide higher-precision temperature control, more intelligent analysis, and more personalized services to meet the growing needs of users.

[0003] At present, some smart kettles have begun to try to combine different technologies to improve the user experience. For example, some smart kettles have introduced Internet of Things technology and are remotely monitored and controlled through mobile phone applications. However, these smart kettles still have great limitations, which are mainly manifested in the following aspects: First, although remote control improves convenience, it does not fundamentally solve the problem of temperature control accuracy. Users still need to manually adjust the temperature and heating time. Secondly, most of the existing smart kettles use simple temperature sensors and lack multi-sensor fusion technology, and cannot monitor and dynamically adjust the heating process in real time. In addition, the control algorithms of these smart kettles are relatively simple and cannot be optimized and adjusted according to the user's personalized needs and different usage scenarios, resulting in poor user experience. Summary of the invention

[0004] The purpose of the present invention is to design an intelligent temperature control method for heating a kettle, an intelligent temperature control kettle system combining deep learning, fuzzy logic control and acoustic wave sensing technology, and through the comprehensive application of multiple advanced technologies, high-precision temperature control, intelligent analysis and personalized services are achieved, thereby significantly improving the user experience.

[0005] In order to achieve the above object, the present invention provides an intelligent temperature control method for a heating kettle, the method comprising the following steps:

[0006] S1. Integrate multiple sensors inside and outside the kettle to collect various data inside and outside the kettle in real time, and transmit the data to the cloud through a wireless communication module for storage and processing; wherein the multiple sensors include temperature sensors, humidity sensors, pressure sensors and acoustic wave sensors;

[0007] S2. According to the data processed in step S1, a hybrid deep learning model based on a combination of a convolutional neural network and a long short-term memory network is constructed and trained, and the trained hybrid deep learning model is deployed to the control system of the kettle to receive sensor data in real time;

[0008] S3, using an improved dynamic time warping algorithm to fuse and synchronize the data processed in step S1 and the sensor data collected in real time, and using a real-time adaptive wavelet denoising algorithm to denoise the sensor data in real time, performing real-time feature extraction on the denoised sensor data, using a real-time adaptive anomaly detection algorithm to detect and process outliers, and then inputting the real-time feature vector into a trained deep learning model to predict and control the temperature of the heating kettle; wherein the improved dynamic time warping algorithm IDTW is expressed as follows:

[0009]

[0010] Among them, D IDTW (t) represents the data after time t alignment, D i (t) represents the data of the i-th sensor, γ represents the weight coefficient, t represents time t, j represents the j-th sensor, and N represents the total number of sensors;

[0011] The real-time adaptive wavelet denoising algorithm is expressed as follows:

[0012]

[0013] Among them, F denoise (t) represents the denoised data, c k represents the wavelet decomposition coefficient, ψ k (t) represents the wavelet basis function, λ k represents the adaptive threshold, 1 represents the indicator function, K represents the number of decomposition levels, and k represents the spatial index;

[0014] The real-time adaptive anomaly detection algorithm includes:

[0015] The mean μ and standard deviation σ of the eigenvector are calculated based on historical data to calculate the difference between the current eigenvalue and the historical mean. If the difference exceeds the set threshold, it is determined to be an outlier, as shown below:

[0016] |X real (t)-μ|>k·σ

[0017] Among them, k represents the set threshold coefficient, X real (t) represents the real-time feature vector; the detected outliers are processed and the linear interpolation method is used to fill the outlier data, which is expressed as follows:

[0018]

[0019] in, Indicates the abnormal data after processing;

[0020] S4, construct a fuzzy logic control algorithm, dynamically adjust the heating power of the kettle through fuzzy logic rules and real-time sensor data to ensure that the temperature is stable within the set range;

[0021] S5. Use an acoustic wave sensor to monitor the sound wave propagation speed and frequency changes in the water in real time, then use a physical model to establish the relationship between the sound wave propagation speed and the water temperature, calculate the real-time water temperature based on the formula for the relationship between the sound wave propagation speed and the water temperature, and then use the extended Kalman filter algorithm to fuse the real-time data of the acoustic wave sensor and the sensor to perform multi-sensor data fusion to predict the state of the water temperature, and then dynamically adjust the heating power according to the real-time water temperature and the water temperature change rate, and perform water temperature anomaly detection at the same time;

[0022] S6. Design user-friendly interface design and remote control function, allowing users to monitor and adjust the working status of the kettle.

[0023] Furthermore, the temperature sensor calibration adopts a standard temperature source for comparative calibration; the humidity sensor calibration adopts a standard humidity source for comparative calibration; the pressure sensor calibration adopts a standard pressure source for comparative calibration; the acoustic wave sensor calibration adopts a standard acoustic wave source for comparative calibration;

[0024] The calibrated data is time synchronized and aligned using linear interpolation, and then the wavelet transform denoising method is used to extract the key features of temperature, humidity, pressure and acoustic characteristics of the denoised sensor data. The weighted fusion algorithm is used to fuse the multi-sensor data to obtain the fused feature data F. fused (t).

[0025] Furthermore, the S2 specifically includes:

[0026] S201, the fused feature data Ffused (t) Divide into training set, validation set and test set, and add random noise and time window smoothing to the training set, validation set and test set for data enhancement;

[0027] S202, from the fused feature data F fused (t), which is expressed as follows:

[0028]

[0029] in, Respectively represent the average values ​​of temperature, humidity, pressure and acoustic wave characteristics, ΔT(t), ΔH(t), ΔP(t), and ΔW(t) represent the corresponding rates of change, which are defined as follows:

[0030]

[0031] ΔT(t)=T cal (t)-T cal (t-1)

[0032] ΔH(t)=H cal (t)-H cal (t-1)

[0033] Δp(t)=P cal (t)-P cal (t-1)

[0034] ΔW(t)=W cal (t)-W cal (t-1)

[0035] Where N is the number of sampling points, t is the time t, i is the current sampling point, T cal (i) represents the temperature of the current sampling point i, H cal (i) Humidity at current sampling point i, P cal (i) The pressure at the current sampling point i, W cal (i) The acoustic wave characteristics of the current sampling point i, T cal (t) represents the temperature at time t, T cal (t-1) represents the temperature at time t-1, H cal (t) represents the humidity at time t, H cal (t-1) represents the humidity at time t-1, P cal (t) represents the pressure at time t, P cal (t-1) represents the pressure at time t-1, W cal (t) represents the acoustic wave characteristics at time t, W cal (t-1) represents the characteristics of the sound wave at time t-1;

[0036] S203. Construct a deep learning model based on the combination of convolutional neural network and long short-term memory network;

[0037] S204. According to the different effects of different features on temperature prediction, a weighted mean square error loss function L with a context bias term E(t) is designed, which is expressed as follows:

[0038]

[0039] Among them, w i represents the weight coefficient, indicating the importance of feature i, y i Indicates the actual temperature, Represents the predicted temperature, N represents the number of samples; the Adam optimizer is used to update the deep learning model parameters, which is expressed as follows:

[0040]

[0041] Among them, θ is the model parameter and α is the learning rate;

[0042] S205. Evaluate the model performance on the validation set, adjust the hyperparameters to prevent overfitting, perform a final evaluation on the test set, calculate the mean square error and mean absolute error, and ensure the generalization ability of the model;

[0043] S206. Deploy the trained model to the control system of the kettle, receive sensor data in real time for temperature prediction and control, and after the model is deployed, update the model regularly according to actual usage to ensure that it adapts to changes in user needs.

[0044] Furthermore, the structure of the deep learning model is expressed as follows:

[0045] Input layer: input feature vector X(t);

[0046] Convolution layer: Perform a one-dimensional convolution operation on the input data to extract local features, which are expressed as follows:

[0047]

[0048] Among them, C i (t) represents the output feature of the convolutional layer, s k represents the convolution kernel weight, b represents the bias, and σ represents the activation function;

[0049] Pooling layer: Perform the maximum pooling operation to reduce the feature dimension, as shown below:

[0050] Pi(t)=max(C i (t))

[0051] Among them, Pi (t) represents the output feature of the pooling layer;

[0052] LSTM layer: processes the pooled time series data and captures long-term dependencies, as shown below:

[0053] h t =LSTM(P(t),h t-1 )

[0054] Among them, h t represents the hidden state of the LSTM layer, P(t) represents the output of the pooling layer, and h t-1 Represents the hidden state of the previous moment;

[0055] Fully connected layer: The output of the LSTM layer is connected to the fully connected layer for the final temperature prediction, which is expressed as follows:

[0056]

[0057] in, represents the predicted temperature, W represents the weight matrix, and b represents the bias.

[0058] Furthermore, in S3, real-time feature extraction is performed on the denoised sensor data, as shown below:

[0059]

[0060] Among them, X real (t) represents the real-time feature vector, Respectively represent the average values ​​of temperature, humidity, pressure and acoustic wave characteristics after denoising, ΔT denoise (t), ΔH denoise (t), ΔP denoise (t), ΔW denoise (t) respectively represent their rates of change;

[0061] The real-time feature vector is input into the trained deep learning model for temperature prediction and control. The prediction formula is as follows:

[0062]

[0063] in, represents the real-time predicted temperature, f(·) represents the deep learning model, and n represents the time window length.

[0064] Furthermore, the S4 specifically includes:

[0065] S401, defining input fuzzy variables and output fuzzy variables, and then defining fuzzy sets and membership functions for each fuzzy variable; the fuzzy variables include temperature error e(t) and temperature error change rate Δe(t), and the output fuzzy variable is heating power adjustment ΔP(t);

[0066] S402, establishing a fuzzy rule base based on experience and control objectives, wherein each rule of the fuzzy rule base is a fuzzy value for adjusting the heating power under specific temperature error and temperature error change rate;

[0067] S403, using Mamdani fuzzy reasoning method to perform rule matching and reasoning, according to the current temperature error e(t) and temperature error change rate Δe(t), activate the relevant rules in the fuzzy rule base, and calculate the activation degree α of each rule i , the activation formula is expressed as follows:

[0068]

[0069] in, and They represent the membership functions of the input fuzzy variables in rule i respectively;

[0070] S404, synthesize the output fuzzy sets of all activated rules to obtain a comprehensive fuzzy set, use the centroid method to defuzzify, and convert the comprehensive fuzzy set into a specific heating power adjustment value ΔP(t), where the membership function μ of the comprehensive fuzzy set is output (x), which is represented as follows:

[0071]

[0072] in, represents the membership function of the output fuzzy set of rule i;

[0073] The defuzzification formula is expressed as follows:

[0074]

[0075] Wherein, the ΔP(t) represents the heating power adjustment value after defuzzification;

[0076] S405, superimposing the defuzzified heating power adjustment value ΔP(t) onto the current heating power P(t-1) to obtain a new heating power P(t), and adjusting the heating power in real time to ensure that the water temperature is stable within the set range;

[0077] S406. In combination with the deep learning model trained in S2, dynamically adjust the fuzzy rules and membership functions, and dynamically adjust the fuzzy rule base and membership function parameters according to the actual usage data and the predicted temperature results; wherein the dynamic adjustment of the fuzzy rule base and membership function parameters specifically includes:

[0078] The fuzzy rule weight adjustment is expressed as follows:

[0079]

[0080] Among them, w i (t) represents the weight of rule i, η represents the learning rate, represents the predicted temperature, and y(t) represents the actual temperature;

[0081] The membership function parameters are adjusted as follows:

[0082]

[0083] Among them, a i (t) represents the membership function parameter, β represents the adjustment coefficient;

[0084] S407. Adding the regularization term R of fuzzy rule weights to the deep learning model w and the regularization term R of the fuzzy rule weight a , adjust the objective function L of the deep learning model last , avoiding overfitting and overadjustment; wherein the objective function L last It is expressed as follows:

[0085]

[0086] Where N represents the number of samples.

[0087] Furthermore, the fuzzy rule base is expressed as follows:

[0088]

[0089] Among them, NL is negative and large, NM is negative and medium, NS is negative and small, ZO is zero, PS is positive and small, PM is positive and medium, and PL is positive and large.

[0090] Furthermore, in S5, the relationship between the sound wave propagation speed and the water temperature is expressed as follows:

[0091] v(t)=v0+α·T(t)

[0092] Among them, v0 represents the reference propagation speed of sound waves in water, and α represents the influence coefficient of temperature on the propagation speed.

[0093] Furthermore, in S5, multi-sensor data fusion based on the extended Kalman filter algorithm is adopted, and the update rule is as follows:

[0094] State prediction, expressed as follows:

[0095]

[0096] The forecast error covariance update is expressed as follows:

[0097] P k|k-1 =F·P k-1|k-1 ·F T +Q

[0098] Kalman gain, expressed as follows:

[0099] K k =P k|k-1 ·H T ·(H·P k|k-1 ·H T +R) -1

[0100] The status is updated as follows:

[0101]

[0102] The error covariance update is expressed as follows:

[0103] P k|k =(IK k ·H)·P k|k-1

[0104] in, Represents the predicted state, Indicates the update status, P k|k-1 and P k|k represents the error covariance, F represents the state transfer matrix, B represents the control input matrix, u k represents the control input, Q represents the process noise covariance, H represents the observation matrix, R represents the observation noise covariance, z k Represents the observed value.

[0105] Furthermore, the user-friendly interface supports IOS and Android platforms, and the remote control function integrates Wi-Fi and Bluetooth modules.

[0106] The beneficial technical effects of the present invention are at least as follows:

[0107] (1) The present invention introduces a deep learning algorithm, which analyzes the user's historical usage data and preferences, intelligently adjusts the kettle's heating curve, and provides the temperature setting that best meets the user's needs. The deep learning module can continuously learn and optimize to adapt to the personalized needs of different users, thus solving the problem that traditional kettles cannot accurately control temperature.

[0108] (2) The present invention adopts a fuzzy logic control algorithm to accurately adjust the heating power according to the target temperature range set by the user, achieve a smooth transition from heating to maintaining temperature, and avoid temperature overshoot and excessive fluctuations. The fuzzy logic control module can monitor water temperature changes in real time and dynamically adjust the heating power to ensure that the temperature is stable within the user-set range, thereby overcoming the problem of unstable temperature control in existing smart kettles.

[0109] (3) The present invention combines acoustic wave sensing technology to monitor the propagation speed and frequency changes of acoustic waves in water, detect water temperature in real time, and intelligently adjust heating power. The acoustic wave sensing module can provide higher-precision temperature measurement results, and through data fusion technology, combined with other sensor data, further improve temperature control accuracy. This innovation effectively solves the problem of inaccurate measurement caused by a single sensor in traditional kettles.

[0110] (4) The intelligent temperature control kettle system of the present invention also integrates a wireless communication module and a user interface module. Users can remotely monitor and control the kettle through a mobile phone application, view the real-time temperature, set the target temperature, and obtain personalized temperature control suggestions. The system also supports multi-sensor fusion technology, comprehensively analyzes temperature, humidity, pressure and other data, and provides more accurate and stable temperature control.

[0111] (5) The present invention solves the problems of inaccurate temperature control, lack of personalized services and low intelligence in the prior art by comprehensively utilizing deep learning, fuzzy logic control and acoustic wave sensing technology, and provides a high-precision, intelligent and personalized smart temperature control kettle system, which significantly improves the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0112] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0113] Figure 1 The present invention is a flow chart of an intelligent temperature control method for heating a kettle. DETAILED DESCRIPTION

[0114] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0115] In one or more embodiments, Figure 1 As shown, an intelligent temperature control method for heating a kettle is disclosed, and the method comprises the following steps S1-S6:

[0116] S1. Integrate multiple sensors inside and outside the kettle to collect various data inside and outside the kettle in real time, and transmit the data to the cloud for storage and processing through a wireless communication module; wherein the multiple sensors include temperature sensors, humidity sensors, pressure sensors and acoustic wave sensors.

[0117] Specifically, multiple sensors are integrated inside and outside the kettle, including a temperature sensor (T s )、Humidity sensor (H s )、Pressure sensor (P s ) and acoustic wave sensor (W s ).

[0118] These sensors collect real-time data on the water temperature inside the kettle, ambient temperature, humidity, pressure, and sound wave propagation. The sensor data is recorded in the form of a time series, and the data stream of each sensor contains a timestamp (t), a measurement value (T(t), H(t), P(t), W(t), etc.).

[0119] The data is transmitted to the cloud server through the wireless communication module (Wi-Fi or Bluetooth) for storage and processing. The data of all sensors are recorded synchronously according to the unified timestamp to form a multi-dimensional time series data set.

[0120] Furthermore, the temperature sensor is calibrated by comparing with a standard temperature source, and the correction formula is:

[0121] T cal (t) = T(t) - 6 T (1)

[0122] Among them, T cal (t) is the calibrated temperature data, T(t) is the original temperature data, δ T is the system deviation of the temperature sensor. The humidity sensor is calibrated by comparing with the standard humidity source, and the correction formula is:

[0123] H cal (t) = H(t) - 6 H (2)

[0124] Among them, H cal (t) is the calibrated humidity data, H(t) is the original humidity data, δ H is the system deviation of the humidity sensor. The pressure sensor is calibrated by comparing with the standard pressure source, and the correction formula is:

[0125] P cal (t) = P(t) - 6 P (3)

[0126] Among them, P cal (t) is the calibrated pressure data, P(t) is the original pressure data, δ P is the system deviation of the pressure sensor. The acoustic wave sensor is calibrated by comparing with a standard acoustic wave source, and the correction formula is:

[0127] W cal (t)=W(t)-δ W (4)

[0128] Among them, W cal (t) is the calibrated sound wave data, W(t) is the original sound wave data, δ W is the system deviation of the acoustic wave sensor.

[0129] It is understandable that the data collection frequency and timestamp of different sensors may be different, and the data needs to be synchronized and aligned. The linear interpolation method is used to align the data of different sensors. The formula is as follows:

[0130]

[0131] Among them, D aligned (t) is the data after time t alignment, D i (t1) and D i (t2) is the data value of two adjacent time points, and t1 and t2 are the timestamps of two adjacent time points.

[0132] Furthermore, in order to eliminate high-frequency noise in the data, the wavelet transform denoising method is used. The specific steps are as follows:

[0133] Perform wavelet decomposition on sensor data to obtain multi-scale details and approximation coefficients;

[0134] Threshold processing is performed on detail coefficients to remove high-frequency noise;

[0135] The coefficients after threshold processing are reconstructed by wavelet to obtain the denoised data.

[0136] Furthermore, feature extraction is performed on the denoised sensor data to extract key features of temperature, humidity, pressure and acoustic wave characteristics. The feature extraction formula is as follows:

[0137]

[0138] Among them, T mean , H mean , P mean and W mean are the average values ​​of temperature, humidity, pressure and acoustic wave characteristics respectively, and N is the number of sampling points.

[0139] Furthermore, a weighted fusion algorithm is used to fuse multi-sensor data, and the fused data is used for subsequent deep learning model training and fuzzy logic control. The fusion formula is as follows:

[0140] F fused (t) = w T ·T cal (t)+w H ·H cal (t)+w P ·P cal (t)+w W ·W cal (t) (7)

[0141] Among them, F fused (t) is the fused feature data, w T 、w H 、w P and w W They are the weight coefficients of temperature, humidity, pressure and sound wave data respectively. The weight coefficients are determined through training data according to the importance and reliability of the sensor.

[0142] S2. According to the data processed in step S1, a hybrid deep learning model based on a combination of a convolutional neural network and a long short-term memory network is constructed and trained, and the trained hybrid deep learning model is deployed to the control system of the kettle to receive sensor data in real time.

[0143] Among them, the preprocessed multi-sensor fusion data F obtained in step 1 fused (t), including temperature, humidity, pressure and acoustic properties.

[0144] The data are divided into training set, validation set and test set, with the proportions of 70%, 15% and 15% respectively.

[0145] In order to enhance the generalization ability of the model, data enhancement is performed, including random noise addition and time window smoothing.

[0146] Random noise addition formula:

[0147]

[0148] Among them, ∈ is a matrix with mean 0 and variance σ 2 Random noise sampled from a Gaussian distribution.

[0149] Time window smoothing formula:

[0150]

[0151] Where w is the width of the time window.

[0152] Furthermore, the feature vector X(t) is extracted from the fused data. The features include the current temperature, humidity, pressure, acoustic wave characteristics and their historical values, average values, and change rates. The feature vector formula is as follows:

[0153]

[0154] in, Respectively represent the average values ​​of temperature, humidity, pressure and acoustic wave characteristics, ΔT(t), ΔH(t), ΔP(t), ΔW(t) represent their rates of change, and are defined as follows:

[0155]

[0156] Furthermore, a hybrid model combining convolutional neural network (CNN) and long short-term memory network (LSTM) is selected to process time series data and extract complex features. The model structure is designed as follows:

[0157] Input layer: Input feature vector X(t).

[0158] Convolution layer: Perform one-dimensional convolution operation on the input data to extract local features.

[0159]

[0160] Among them, C i (t) is the output feature of the convolutional layer, w k is the convolution kernel weight, b is the bias, and σ is the activation function.

[0161] Pooling layer: Perform maximum pooling operation to reduce feature dimensions.

[0162] P i (t) = max(C i (t)) (13)

[0163] Among them, P i (t) is the output feature of the pooling layer.

[0164] LSTM layer: processes the pooled time series data and captures long-term dependencies.

[0165] h t =LSTM(P(t),h t-1 ) (14)

[0166] Among them, h t is the hidden state of the LSTM layer, P(t) is the output of the pooling layer, and h t-1 is the hidden state at the previous moment.

[0167] Fully connected layer: Connect the output of the LSTM layer to the fully connected layer for final temperature prediction.

[0168]

[0169] in, To predict temperature, W is the weight matrix and b is the bias.

[0170] Furthermore, loss function: In order to better serve the patent scenario, we designed a weighted mean square error (WMSE) loss function, which gives different weights to different features considering their different impacts on temperature prediction.

[0171]

[0172] Among them, w i is the weight coefficient, indicating the importance of feature i, y i is the actual temperature, To predict temperature, N is the number of samples.

[0173] In order to adapt to the use of the kettle in different environments, the present invention also adds an environmental bias term E(t) to take into account the influence of external factors such as ambient temperature and humidity:

[0174]

[0175] Among them, E(t) is calculated from the data of the environmental sensor, for example:

[0176] E(t)=αT env (t)+βH env (t) (18)

[0177] Among them, T env (t) and H env (t) are the ambient temperature and humidity, respectively, and α and β are adjustment coefficients.

[0178] Furthermore, the Adam optimizer is used to update the model parameters.

[0179]

[0180] Among them, θ is the model parameter and α is the learning rate.

[0181] Furthermore, the model performance is evaluated on the validation set, and the hyperparameters are adjusted to prevent overfitting. The final evaluation is performed on the test set, and the mean square error (MSE) and mean absolute error (MAE) are calculated to ensure the generalization ability of the model.

[0182]

[0183] Furthermore, the trained model is deployed to the control system of the kettle, which receives sensor data in real time for temperature prediction and control. After the model is deployed, it is regularly updated according to actual usage to ensure that it adapts to changes in user needs.

[0184] S3. The data processed in step S1 and the sensor data collected in real time are fused and synchronized using an improved dynamic time warping algorithm, and the sensor data is denoised in real time using a real-time adaptive wavelet denoising algorithm. Real-time feature extraction is performed on the denoised sensor data, and anomalies are detected and processed using a real-time adaptive anomaly detection algorithm. The real-time feature vector is then input into a trained deep learning model to predict and control the temperature of the heating kettle.

[0185] Specifically, the preprocessed multi-sensor fusion data F obtained in step 1 fused (t), including temperature, humidity, pressure and sound wave characteristics. A variety of sensors are integrated inside the kettle to collect real-time data on the water temperature, ambient temperature, humidity, pressure and sound wave propagation inside the kettle to ensure the comprehensiveness and real-time nature of the data.

[0186] Furthermore, the data collection frequency and timestamp of different sensors may be different, so the data needs to be synchronized and fused. The improved dynamic time warping (IDTW) algorithm is used to synchronize and align the data of different sensors to ensure data consistency. The IDTW algorithm formula is as follows:

[0187]

[0188] Among them, D IDTW (t) is the data after time t alignment, D i (t) is the data of the i-th sensor, and γ is the weight coefficient.

[0189] Since sensor data may contain high-frequency noise and outliers, in order to ensure the accuracy of the data, the real-time adaptive wavelet denoising (RAWD) algorithm is used to denoise the sensor data in real time. The steps of the RAWD algorithm are as follows:

[0190] Perform wavelet decomposition on sensor data to obtain multi-scale detail coefficients and approximation coefficients;

[0191] Adaptive threshold processing is performed on detail coefficients to remove high-frequency noise;

[0192] Perform wavelet reconstruction on the coefficients after threshold processing to obtain denoised data. The RAWD denoising formula is:

[0193]

[0194] Among them, F denoise (t) is the denoised data, c k is the wavelet decomposition coefficient, ψ k (t) is the wavelet basis function, λ k is the adaptive threshold and 1 is the indicator function.

[0195] Furthermore, the denoised sensor data is subjected to real-time feature extraction to extract key features of temperature, humidity, pressure and acoustic characteristics, as well as historical values, average values ​​and rates of change. The feature vector formula is as follows:

[0196]

[0197] in, Respectively represent the average values ​​of temperature, humidity, pressure and acoustic wave characteristics after denoising, ΔT denoise (t), ΔH denoise (t), ΔP denoise (t), ΔW denoise (t) represent their changing rates respectively.

[0198] Furthermore, outliers in sensor data are detected in real time to ensure data reliability. Adaptive Anomaly Detection (AAD) algorithm is used to detect and process outliers in real time. The steps of AAD algorithm are as follows:

[0199] Calculate the mean μ and standard deviation σ of the feature vector based on historical data;

[0200] Calculate the difference between the current feature value and the historical mean. If the difference exceeds the set threshold, it is determined to be an outlier. Anomaly detection formula:

[0201] |X real(t)-μ|>k·σ (25)

[0202] Among them, k is the set threshold coefficient;

[0203] The detected outliers are processed and the abnormal data are filled using linear interpolation:

[0204]

[0205] Furthermore, the real-time feature vector is input into the trained deep learning model for temperature prediction and control. The model predicts the real-time temperature based on the current feature value, historical feature value and environmental factors. Prediction formula:

[0206]

[0207] in, To predict temperature in real time, f(·) is the deep learning model and n is the time window length.

[0208] According to the prediction results, the heating power and time are adjusted in real time to ensure that the kettle temperature is stable within the set range.

[0209] S4. Construct a fuzzy logic control algorithm to dynamically adjust the heating power of the kettle through fuzzy logic rules and real-time sensor data to ensure that the temperature is stable within the set range, including:

[0210] S401, define input fuzzy variables and output fuzzy variables, and then define fuzzy sets and membership functions for each fuzzy variable; the fuzzy variables include temperature error e(t) and temperature error change rate Δe(t), and the output fuzzy variable is heating power adjustment ΔP(t); wherein, temperature error e(t):

[0211] e(t)=T set -T current (t) (28)

[0212] Among them, T set is the set temperature, T current (t) is the current water temperature.

[0213] Temperature error change rate Δe(t):

[0214] Δe(t)=e(t)-e(t-1) (29)

[0215] Furthermore, fuzzy sets and membership functions are defined for each fuzzy variable. Assume that the fuzzy sets of temperature error e(t) and temperature error change rate Δe(t) are {NL (negative large), NM (negative medium), NS (negative small), ZO (zero), PS (positive small), PM (positive medium), PL (positive large)}, and the membership function is represented by a triangular or trapezoidal function. The mathematical expression of the membership function is as follows:

[0216] Triangular membership function:

[0217]

[0218] Trapezoidal membership function:

[0219]

[0220] S402. A fuzzy rule base is established based on experience and control objectives, wherein each rule of the fuzzy rule base is a fuzzy value for adjusting the heating power under specific temperature errors and temperature error change rates; example rules:

[0221] If e(t) is PL and Δe(t) is PL, then ΔP(t) is NL.

[0222] If e(t) is NS and Δe(t) is ZO, then ΔP(t) is PS.

[0223] The specific fuzzy rule table is as follows:

[0224]

[0225] S403, using Mamdani fuzzy reasoning method to perform rule matching and reasoning, according to the current temperature error e(t) and temperature error change rate Δe(t), activate the relevant rules in the fuzzy rule base, and calculate the activation degree α of each rule i , the activation formula is expressed as follows:

[0226]

[0227] in, and They represent the membership functions of the input fuzzy variables in rule i respectively;

[0228] S404, synthesize the output fuzzy sets of all activated rules to obtain a comprehensive fuzzy set, use the centroid method to defuzzify, and convert the comprehensive fuzzy set into a specific heating power adjustment value ΔP(t), where the membership function μ of the comprehensive fuzzy set is output (x), which is represented as follows:

[0229]

[0230] in, represents the membership function of the output fuzzy set of rule i;

[0231] The defuzzification formula is expressed as follows:

[0232]

[0233] Wherein, the ΔP(t) represents the heating power adjustment value after defuzzification;

[0234] S405. Superimpose the defuzzified heating power adjustment value ΔP(t) on the current heating power P(t-1) to obtain a new heating power P(t), and adjust the heating power in real time to ensure that the water temperature is stable within the set range; superimpose the defuzzified heating power adjustment value ΔP(t) on the current heating power P(t-1) to obtain a new heating power P(t).

[0235] P(t)=P(t-1)+ΔP(t) (35)

[0236] Adjust the heating power in real time to ensure that the water temperature is stable within the set range;

[0237] S406. In combination with the deep learning model trained in S2, dynamically adjust the fuzzy rules and membership functions, and dynamically adjust the fuzzy rule base and membership function parameters according to the actual usage data and the predicted temperature results; wherein the dynamic adjustment of the fuzzy rule base and membership function parameters specifically includes:

[0238] The fuzzy rule weight adjustment is expressed as follows:

[0239]

[0240] Among them, w i (t) represents the weight of rule i, η represents the learning rate, represents the predicted temperature, and y(t) represents the actual temperature;

[0241] The membership function parameters are adjusted as follows:

[0242]

[0243] Among them, a i (t) represents the membership function parameter, β represents the adjustment coefficient;

[0244] S407. Adding the regularization term R of fuzzy rule weights to the deep learning model w and the regularization term R of the fuzzy rule weight a , adjust the objective function L of the deep learning model last, to avoid overfitting and overadjustment; among them, the regularization term of the fuzzy rule weight is:

[0245]

[0246] Among them, λ w is the regularization parameter.

[0247] Regularization term for membership function parameters:

[0248]

[0249] Among them, λ a is the regularization parameter;

[0250] Among them, the objective function L last It is expressed as follows:

[0251]

[0252] Where N represents the number of samples.

[0253] S5. Use an acoustic wave sensor to monitor the sound wave propagation speed and frequency changes in the water in real time, then use a physical model to establish the relationship between the sound wave propagation speed and the water temperature. Calculate the real-time water temperature based on the formula for the relationship between the sound wave propagation speed and the water temperature. Then use the extended Kalman filter algorithm to fuse the real-time data of the acoustic wave sensor with other sensors to predict the water temperature state. Then, according to the real-time water temperature and the water temperature change rate, dynamically adjust the heating power, and perform water temperature anomaly detection at the same time.

[0254] Specifically, a high-precision acoustic sensor is integrated inside the kettle to monitor the propagation speed and frequency changes of acoustic waves in the water in real time. The acoustic sensor sends out acoustic signals and receives reflected signals.

[0255] Data collection includes timestamp t, sound wave propagation time τ(t) and sound wave frequency f(t), forming a multidimensional time series data set.

[0256] Furthermore, the physical model is used to establish the relationship between the propagation speed of sound waves and the water temperature. The relationship between the propagation speed of sound waves in water v(t) and the water temperature T(t) is as follows:

[0257] v(t)=v0+α·T(t) (41)

[0258] Among them, v0 is the reference propagation speed of sound waves in water, and α is the influence coefficient of temperature on the propagation speed.

[0259] By measuring the sound wave propagation time τ(t) and the distance d between the sound wave sensors in the kettle, the real-time sound wave propagation speed v(t) is calculated:

[0260]

[0261] Furthermore, the real-time water temperature T(t) is calculated according to the relationship between the sound wave propagation speed and the water temperature:

[0262]

[0263] Where v(t) is calculated using the formula in step 2.

[0264] It is understandable that in order to improve the accuracy of water temperature measurement, the acoustic sensor data is combined with other sensors (such as temperature sensor T s 、Humidity sensor H s 、Pressure sensor P s ) data. A multi-sensor data fusion method based on the extended Kalman filter (EKF) algorithm is used. The state update formula of the extended Kalman filter is as follows:

[0265] Status prediction:

[0266]

[0267] Forecast error covariance update:

[0268] P k|k-1 =F·P k-1|k-1 ·F T +Q (45)

[0269] Kalman gain:

[0270] K k =P k|k-1 ·H T ·(H·P k|k-1 ·H T +R) -1 (46)

[0271] Status Update:

[0272]

[0273] Error covariance update:

[0274] P k|k =(IK k ·H)·P k|k-1 (48)

[0275] in, For the predicted state, To update the status, P k|k-1 and P k|k is the error covariance, F is the state transfer matrix, B is the control input matrix, u kis the control input, Q is the process noise covariance, H is the observation matrix, R is the observation noise covariance, z k is the observed value.

[0276] Furthermore, the heating power P(t) is dynamically adjusted according to the real-time water temperature T(t) and the water temperature change rate ΔT(t). The calculation formula of the water temperature change rate ΔT(t) is as follows:

[0277] ΔT(t)=T(t)-T(t-1) (49)

[0278] Dynamic heating power adjustment formula:

[0279] P(t)=P(t-1)+K p ·e(t)+K d ΔT(t) (50)

[0280] Among them, K p is the proportional gain, K d is the differential gain, e(t) = T set -T(t) is the temperature error.

[0281] Furthermore, the Adaptive Anomaly Detection (AAD) algorithm is used to detect and process outliers in real time to ensure data reliability. The algorithm includes:

[0282] Calculate the mean μ and standard deviation σ of the feature vector based on historical data;

[0283] Calculate the difference between the current feature value and the historical mean. If the difference exceeds the set threshold, it is determined to be an outlier. Anomaly detection formula:

[0284] |T(t)-μ|>k·σ (51)

[0285] Among them, k is the set threshold coefficient;

[0286] The detected outliers are processed and the abnormal data are filled using linear interpolation:

[0287]

[0288] S6. Design user-friendly interface design and remote control function, allowing users to monitor and adjust the working status of the kettle.

[0289] Specifically, we developed a mobile application (App) that supports IOS and Android platforms. The interface design follows the principles of simplicity, intuitiveness and ease of use. The main functions include real-time temperature display, target temperature setting, heating mode selection and historical data viewing.

[0290] Furthermore, the main interface elements are designed as follows:

[0291] (1) Real-time temperature display: The current water temperature and set temperature are displayed on the main interface using a dynamic thermometer or digital display.

[0292] (2) Target temperature setting: Users can set the target temperature using the slider or numeric input box. The setting range is 40°C to 100°C with an accuracy of 0.1°C.

[0293] (3) Heating mode selection: A variety of heating modes are provided, such as rapid heating, heat preservation mode, timed heating, etc. Users can choose different heating modes according to their needs.

[0294] (4) Historical data viewing: Displays the temperature change curve for the past 24 hours, 7 days, or 30 days to help users understand usage and optimize settings.

[0295] Furthermore, the remote control function integrates Wi-Fi and Bluetooth modules to achieve wireless connection between the kettle and the mobile phone app.

[0296] Users can remotely control the kettle through the App, including starting / stopping heating, adjusting the target temperature, selecting the heating mode, etc.

[0297] Among them, the specific steps of remote control are:

[0298] Users log in to their personal accounts through the App to ensure the security of devices and user data;

[0299] The app communicates with the kettle through the cloud server, sending control commands and receiving status feedback;

[0300] The kettle responds to the user's control commands in real time and feeds back the latest status (such as current water temperature, heating power, working mode, etc.) to the App.

[0301] Understandably, to ensure safe use, real-time notification and alarm functions are designed. The main functions include:

[0302] (1) Temperature arrival notification: When the water temperature reaches the set temperature, the App pushes a notification to remind the user.

[0303] (2) Low water level alarm: When the water level in the kettle is lower than the safety line, the App will issue an alarm to prevent dry boiling.

[0304] (3) Equipment fault alarm: When an equipment fault is detected (such as sensor abnormality, heating element failure, etc.), the App pushes a fault alarm and provides fault diagnosis and handling suggestions.

[0305] Furthermore, it provides personalized settings and optimization suggestions based on user usage habits and historical data. The main functions include:

[0306] (1) Intelligent recommendation: Based on the user's historical usage data, intelligently recommend the best heating mode and set temperature.

[0307] (2) Personalized settings: Users can save commonly used temperature settings and heating modes for quick access.

[0308] (3) Usage Report: Generate usage reports regularly to analyze user usage habits and provide energy-saving and safety recommendations.

[0309] Understandably, in order to ensure data security and continuity, data synchronization and backup functions are designed. The main functions include:

[0310] (1) Cloud synchronization: User settings, historical data, and usage records are automatically synchronized to the cloud server to ensure data security and continuity.

[0311] (2) Local backup: Regularly back up local data in the App, and users can manually restore data.

[0312] (3) Data export: Users can export usage data to CSV or Excel format for further analysis and management.

[0313] Furthermore, we collect user feedback through the App to continuously improve system performance and user experience. The main functions include:

[0314] (1) Feedback collection: Users can submit usage feedback and suggestions through the App, and the development team will regularly organize and analyze the feedback information.

[0315] (2) System Update: Regularly release App and kettle firmware updates to fix known issues, optimize system performance, and add new features.

[0316] (3) User Community: Establish a user community where users can exchange usage experiences, share insights, and interact with the development team.

[0317] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. An intelligent temperature control method for heating a kettle, characterized in that: The method comprises the following steps: S1. Integrate multiple sensors inside and outside the kettle to collect various data inside and outside the kettle in real time, and transmit the data to the cloud through a wireless communication module for storage and processing; wherein the multiple sensors include temperature sensors, humidity sensors, pressure sensors and acoustic wave sensors; S2. According to the data processed in step S1, a hybrid deep learning model based on a combination of a convolutional neural network and a long short-term memory network is constructed and trained, and the trained hybrid deep learning model is deployed to the control system of the kettle to receive sensor data in real time; S3, the data processed in step S1 and the sensor data collected in real time are fused and synchronized using an improved dynamic time warping algorithm, and the sensor data is denoised in real time using a real-time adaptive wavelet denoising algorithm, and real-time feature extraction is performed on the denoised sensor data, and an abnormal value is detected and processed using a real-time adaptive anomaly detection algorithm, and then the real-time feature vector is input into the trained deep learning model to predict and control the temperature of the heating kettle; wherein the improved dynamic time warping algorithm IDTW is expressed as follows: Among them, D IDTW (t) represents the data after time t alignment, D i (t) represents the data of the i-th sensor, D j (t) represents the data of the jth sensor, γ represents the weight coefficient, t represents time t, j represents the jth sensor, and N represents the total number of sensors; The real-time adaptive wavelet denoising algorithm is expressed as follows: Among them, F denoise (t) represents the denoised data, c k represents the wavelet decomposition coefficient, ψ k (t) represents the wavelet basis function, λ k represents the adaptive threshold, 1 represents the indicator function, K represents the number of decomposition levels, and k represents the spatial index; The real-time adaptive anomaly detection algorithm includes: The mean μ and standard deviation σ of the eigenvector are calculated based on historical data to calculate the difference between the current eigenvalue and the historical mean. If the difference exceeds the set threshold, it is determined to be an outlier, as shown below: |X real (t)-μ|>k·σ Among them, k represents the set threshold coefficient, X real (t) represents the real-time feature vector; the detected outliers are processed and the linear interpolation method is used to fill the outlier data, which is expressed as follows: in, Indicates the abnormal data after processing; S4, construct a fuzzy logic control algorithm, dynamically adjust the heating power of the kettle through fuzzy logic rules and real-time sensor data to ensure that the temperature is stable within the set range; S5. Use an acoustic wave sensor to monitor the sound wave propagation speed and frequency changes in the water in real time, then use a physical model to establish the relationship between the sound wave propagation speed and the water temperature, calculate the real-time water temperature based on the formula for the relationship between the sound wave propagation speed and the water temperature, and then use the extended Kalman filter algorithm to fuse the real-time data of the acoustic wave sensor and the sensor to perform multi-sensor data fusion to predict the state of the water temperature, and then dynamically adjust the heating power according to the real-time water temperature and the water temperature change rate, and perform water temperature anomaly detection at the same time; S6. Design user-friendly interface design and remote control function, allowing users to monitor and adjust the working status of the kettle.

2. The intelligent temperature control method for heating a kettle according to claim 1, characterized in that: The temperature sensor calibration uses a standard temperature source for comparison calibration; the humidity sensor calibration uses a standard humidity source for comparison calibration; the pressure sensor calibration uses a standard pressure source for comparison calibration; the acoustic wave sensor calibration uses a standard acoustic wave source for comparison calibration; The calibrated data is time synchronized and aligned using linear interpolation, and then the wavelet transform denoising method is used to extract the key features of temperature, humidity, pressure and acoustic characteristics of the denoised sensor data. The weighted fusion algorithm is used to fuse the multi-sensor data to obtain the fused feature data F. fused (t).

3. The intelligent temperature control method for heating a kettle according to claim 2, characterized in that: The S2 specifically includes: S201, the fused feature data F fused (t) Divide into training set, validation set and test set, and add random noise and time window smoothing to the training set, validation set and test set for data enhancement; S202, from the fused feature data F fused (t), which is expressed as follows: in, Respectively represent the average values ​​of temperature, humidity, pressure and acoustic wave characteristics, ΔT(t), ΔH(t), AP(t), and ΔW(t) represent the corresponding rates of change, which are defined as follows: ΔT(t)=T cal (t)-T cal (t-1) ΔH(t)=H cal (t)-H cal (t-1) Δp(t)=P cal (t)-P cal (t-1) ΔW(t)=W cal (t)-W cal (t-1) Where N is the number of sampling points, t is the time t, i is the current sampling point, T cal (i) represents the temperature of the current sampling point i, H cal (i) Humidity at current sampling point i, P cal (i) The pressure at the current sampling point i, W cal (i) The acoustic wave characteristics of the current sampling point i, T cal (t) represents the temperature at time t, T cal (t-1) represents the temperature at time t-1, H cal (t) represents the humidity at time t, H cal (t-1) represents the humidity at time t-1, P cal (t) represents the pressure at time t, P cal (t-1) represents the pressure at time t-1, W cal (t) represents the acoustic wave characteristics at time t, W cal (t-1) represents the characteristics of the sound wave at time t-1; S203. Construct a deep learning model based on the combination of convolutional neural network and long short-term memory network; S204. According to the different effects of different features on temperature prediction, a weighted mean square error loss function L with a context bias term E(t) is designed, which is expressed as follows: Among them, w i represents the weight coefficient, indicating the importance of feature i, y i Indicates the actual temperature, Represents the predicted temperature, N represents the number of samples; the Adam optimizer is used to update the deep learning model parameters, which is expressed as follows: Among them, θ is the model parameter and α is the learning rate; S205. Evaluate the model performance on the validation set, adjust the hyperparameters to prevent overfitting, perform a final evaluation on the test set, calculate the mean square error and mean absolute error, and ensure the generalization ability of the model; S206. Deploy the trained model to the control system of the kettle, receive sensor data in real time for temperature prediction and control, and after the model is deployed, update the model regularly according to actual usage to ensure that it adapts to changes in user needs.

4. The intelligent temperature control method for heating a kettle according to claim 3, characterized in that: The structure of the deep learning model is represented as follows: Input layer: input feature vector X(t); Convolution layer: Perform a one-dimensional convolution operation on the input data to extract local features, which are expressed as follows: Among them, C i (t) represents the output feature of the convolutional layer, w k represents the convolution kernel weight, b represents the bias, and σ represents the activation function; Pooling layer: Perform the maximum pooling operation to reduce the feature dimension, as shown below: Pi(t)=max(C i (t)) Among them, P i (t) represents the output feature of the pooling layer; LSTM layer: processes the pooled time series data and captures long-term dependencies, as shown below: h t =LSTM(P(t),h t-1 ) Among them, h t represents the hidden state of the LSTM layer, P(t) represents the output of the pooling layer, and h t-1 Represents the hidden state of the previous moment; Fully connected layer: The output of the LSTM layer is connected to the fully connected layer for the final temperature prediction, which is expressed as follows: in, represents the predicted temperature, W represents the weight matrix, and b represents the bias.

5. The intelligent temperature control method for heating a kettle according to claim 1, characterized in that: In S3, real-time feature extraction is performed on the denoised sensor data, as shown below: Among them, X real (t) represents the real-time feature vector, Respectively represent the average values ​​of temperature, humidity, pressure and acoustic wave characteristics after denoising, ΔT denoise (t), AH denoise (t), ΔP denoise (t), ΔW denoise (t) respectively represent their rates of change; The real-time feature vector is input into the trained deep learning model for temperature prediction and control. The prediction formula is as follows: in, represents the real-time predicted temperature, f(·) represents the deep learning model, and n represents the time window length.

6. The intelligent temperature control method for heating a kettle according to claim 1, characterized in that: The S4 specifically includes: S401, defining input fuzzy variables and output fuzzy variables, and then defining fuzzy sets and membership functions for each fuzzy variable; the fuzzy variables include temperature error e(t) and temperature error change rate Δe(t), and the output fuzzy variable is heating power adjustment ΔP(t); S402, establishing a fuzzy rule base based on experience and control objectives, wherein each rule of the fuzzy rule base is a fuzzy value for adjusting the heating power under specific temperature error and temperature error change rate; S403, using Mamdani fuzzy reasoning method to perform rule matching and reasoning, according to the current temperature error e(t) and temperature error change rate Δe(t), activate the relevant rules in the fuzzy rule base, and calculate the activation degree α of each rule i , the activation formula is expressed as follows: in, and They represent the membership functions of the input fuzzy variables in rule i respectively; S404, synthesize the output fuzzy sets of all activated rules to obtain a comprehensive fuzzy set, use the centroid method to defuzzify, and convert the comprehensive fuzzy set into a specific heating power adjustment value ΔP(t), where the membership function μoutput(x) of the comprehensive fuzzy set is expressed as follows: in, represents the membership function of the output fuzzy set of rule i; The defuzzification formula is expressed as follows: Wherein, ΔP(t) represents the heating power adjustment value after defuzzification, and dx represents the differentiation of x; S405, superimposing the defuzzified heating power adjustment value ΔP(t) onto the current heating power P(t-1) to obtain a new heating power P(t), and adjusting the heating power in real time to ensure that the water temperature is stable within the set range; S406. In combination with the deep learning model trained in S2, dynamically adjust the fuzzy rules and membership functions, and dynamically adjust the fuzzy rule base and membership function parameters according to the actual usage data and the predicted temperature results; wherein the dynamic adjustment of the fuzzy rule base and membership function parameters specifically includes: The fuzzy rule weight adjustment is expressed as follows: Among them, w i (t) represents the weight of rule i, η represents the learning rate, represents the predicted temperature, and y(t) represents the actual temperature; The membership function parameters are adjusted as follows: Among them, a i (t) represents the membership function parameter, β represents the adjustment coefficient; S407. Adding the regularization term R of fuzzy rule weights to the deep learning model W and the regularization term R of the fuzzy rule weight a , adjust the objective function L of the deep learning model last , avoiding overfitting and overadjustment; wherein the objective function L last It is expressed as follows: Where N represents the number of samples.

7. The intelligent temperature control method for heating a kettle according to claim 6, characterized in that: The fuzzy rule base is expressed as follows: Among them, NL is negative and large, NM is negative and medium, NS is negative and small, ZO is zero, PS is positive and small, PM is positive and medium, and PL is positive and large.

8. The intelligent temperature control method for a heating kettle according to claim 1, characterized in that: In S5, the relationship between the sound wave propagation speed and the water temperature is expressed as follows: v(t)=V0+α·T(t) Among them, v0 represents the reference propagation speed of sound waves in water, and α represents the influence coefficient of temperature on the propagation speed.

9. The intelligent temperature control method for heating a kettle according to claim 8, characterized in that: In S5, multi-sensor data fusion based on the extended Kalman filter algorithm is adopted, and the update rules are as follows: State prediction, expressed as follows: The forecast error covariance update is expressed as follows: P k|k-1 =F·P k-1|k-1 ·F T +Q Kalman gain, expressed as follows: K k =P k|k-1 ·H T ·(H·P k|k-1 ·H T +R) -1 The status is updated as follows: The error covariance update is expressed as follows: P k|k =(I-K k ·H)·P k|k-1 in, Represents the predicted state, Indicates the update status, P k|k-1 and P k|k represents the error covariance, F represents the state transfer matrix, B represents the control input matrix, u k represents the control input, Q represents the process noise covariance, H represents the observation matrix, R represents the observation noise covariance, z k Represents the observed value.

10. The intelligent temperature control method for a heating kettle according to claim 1, characterized in that: The user-friendly interface supports IOS and Android platforms, and the remote control function integrates Wi-Fi and Bluetooth modules.

Images