A method for controlling the temperature overshoot of an electric cooker by gradually decreasing the power of the heating element during slow start.

Abstract

This invention discloses a power-decreasing, slow-start control method for electric cookers to prevent temperature overshoot. It relates to the field of intelligent control for electric cookers and aims to solve the problems of temperature overshoot and thermal inertia hysteresis during heating, thereby improving heating accuracy and safety. The core technical solution involves applying a low-amplitude rectangular pulse power excitation during initialization or the start of a new cooking cycle. This collects multi-dimensional time-series signals of heating element temperature, pot bottom temperature, and steam pressure, generating a data fingerprint reflecting the dynamic characteristics of the thermal response. Multi-scale feature extraction, noise suppression, and baseline correction algorithms are used to establish an individual thermal behavior model for the cooker. Combined with real-time operating conditions and prediction algorithms, the power output is dynamically adjusted to achieve smooth, gradual power reduction control without overshoot risk. The system can adaptively optimize the thermal response model, improving temperature control accuracy and effectively reducing energy consumption and temperature fluctuations.

Description

Technical Field

[0001] This invention relates to the field of temperature control technology for household electric cooking appliances, and in particular to a method for preventing temperature overshoot by gradually decreasing the power of the heating element in an electric cooker. Background Technology

[0002] Currently, in the field of household electric cooking appliances, especially temperature-controlled heating devices such as electric cookers, the widely adopted power descent control technology aims to improve the accuracy and safety of temperature during the cooking process through multi-stage power strategies, preventing temperature overshoot during heating. Mainstream technical solutions often rely on fixed threshold comparisons, empirical temperature curves, single sensor feedback, or simple power limiting algorithms to achieve step power switching. For example, existing electric cooker products generally use a stepped power control method; as the water temperature in the pot approaches the target temperature, the controller gradually reduces the power supplied to the heating element. However, limited by the pot structure, the thermal response characteristics of the heating element, changes in liquid load, and environmental coupling factors, traditional descent strategies mostly use static parameter settings, lacking dynamic recognition of individual device differences and real-time thermal inertia.

[0003] On the one hand, some high-end home appliances attempt to introduce multivariate coupled modeling, using historical data fitting, external thermal field simulation, or neural network prediction to achieve complex temperature control optimization. However, such solutions typically place high demands on hardware costs, the number of sensors, and computing resources, making them difficult to popularize in cost-sensitive home appliances. On the other hand, products in the industry have not yet universally implemented individualized power reduction strategies based on the inherent thermal behavior characteristics of cookware. The vast majority of control methods still rely on manufacturer-preset curves or simple heating rate thresholds, ignoring thermal inertia issues caused by physical differences such as water volume changes, pot wall thermal resistance, and heating element heat capacity.

[0004] Current typical applications, such as household electric cookers, smart steam cookers, and mini canteen cookware, generally face the problem of temperature overshoot during power reduction transitions. This means that after the power value changes, thermal inertia causes the water temperature in the pot to exceed the target setting, leading to risks such as food boiling over, exceeding safety thresholds, or a decline in cooking quality. While some existing technologies can mitigate this problem by increasing the number of temperature sensing points, increasing the sampling frequency, or using more complex algorithms, their cost-effectiveness, real-time performance, and reliability are insufficient to meet the needs of mass-produced products in the household market.

[0005] Specifically, existing power descent strategies have the following main technical shortcomings: First, there is a lack of prediction of the dynamic change trend of thermal inertia when switching power step. Switching is often performed with a fixed time delay or empirical threshold. It is impossible to make adaptive adjustments based on the actual heat capacity of the cookware, the response hysteresis of the heating element, and the dynamic individual differences of the water level. This leads to inaccurate prediction of the time and amplitude of the temperature peak and frequent temperature overshoot.

[0006] Secondly, the control strategies mostly rely on static settings or external modeling, requiring manual calibration or batch calibration. This makes it difficult to support closed-loop optimization based on the cookware's own thermal response characteristics in actual use scenarios, reducing the long-term stability and individual adaptability of the temperature control system.

[0007] Third, some existing advanced prediction solutions require the addition of hardware sensors or access to cloud-based big data training, which increases the manufacturing, maintenance and operating costs of the equipment, hindering the widespread application and large-scale production of home appliances.

[0008] This shows that the market urgently needs a power descent and temperature overshoot prevention technology solution that is model-independent, calibration-free, dynamically adaptive, and based on the device's own thermal response fingerprint. Summary of the Invention

[0009] This application provides a method for controlling the heating element of an electric cooker with a power-decreasing slow start to prevent temperature overshoot, which aims to solve one of the problems or issues of the prior art mentioned in the background.

[0010] This application provides a method for controlling the heating element of an electric cooker with a power-decreasing, slow-start mechanism to prevent temperature overshoot, specifically including: S1: Collect the time-series signals of the surface temperature of the heating element of the electric cooker, the time-series signals of the temperature at the center of the bottom of the pot, and the time-series signals of the slight change in water vapor pressure to obtain the original multidimensional sensor dataset. S2: Perform feature extraction processing on the original multidimensional sensor dataset to parse out the time-series feature vector in order to construct a device-level thermal response fingerprint vector; S3: Encrypt and store the device-level thermal response fingerprint vector, and combine it with the initial heating rate and pressure change slope to jointly determine the current cookware load status, so as to generate a working condition matching tag; S4: Based on the operating condition matching label, call the device-level thermal response fingerprint vector, and use the embedded inference algorithm to deduce the temperature rise envelope to generate dynamic prediction results of thermal inertia; S5: Based on the thermal inertia dynamic prediction results, calculate the time permissible interval that avoids the predicted temperature acceleration rise segment and ends before the predicted peak value, so as to generate power transition safety window parameters. S6: Calculate the correction ratio based on the device-level thermal response fingerprint vector to generate the gear attenuation compensation coefficient; S7: When it is detected that the current temperature of the electric cooker heating element is less than the target temperature difference and the current heating rate of the electric cooker heating element is less than the target heating rate, the switching start time is determined by combining the power transition safety window parameter, and the target power value is adjusted by using the gear attenuation compensation coefficient to perform the power reduction stage switching action.

[0011] The electric cooker heating element power-decreasing soft-start anti-temperature overshoot control method provided in this application has the following beneficial effects: (1) By introducing an adaptive power descent mechanism based on device-level thermal response fingerprints, the accuracy and individual adaptability of the electric cooker's temperature control at the end of cooking are significantly improved, effectively overcoming the common temperature overshoot problem caused by the use of uniform thresholds or empirical curves in traditional control strategies. In the prior art, most electric cookers rely on fixed temperature rise rate judgment or preset delay logic for power switching, which is difficult to cope with individual non-consistent factors such as the aging degree of heating tubes of different bodies, differences in heat conduction paths, and sensor offsets. They are very prone to overshoot or even overflow due to thermal inertia accumulation when approaching the target temperature. This solution innovatively applies low-delay thermal excitation and extracts multi-scale dynamic features in the initialization stage to construct a thermal response fingerprint vector with device uniqueness, so that each device can "recognize its own" thermal behavior characteristics. In the actual temperature control process, the system matches the fingerprint with the current load state, predicts the temperature rise envelope in the next few seconds in real time, thereby scientifically delineating the safe time window for power transition and dynamically adjusting the output intensity of the next level. This method does not rely on external modeling, historical data fitting, or complex neural network training, avoiding high computing power consumption and cloud interaction delays. It achieves individualized closed-loop control without increasing hardware costs, significantly reducing the risk of temperature overshoot and improving the consistency of cooking results and user experience.

[0012] (2) The proposed lightweight local inference architecture and dual-parameter collaborative control mechanism fully meet the stringent requirements of home appliances for real-time performance, low power consumption, and embedded deployment while ensuring control accuracy, demonstrating excellent engineering practicality and system robustness. Unlike existing advanced control schemes that often rely on cloud model updates or multi-sensor fusion modeling to improve prediction capabilities, this invention is based entirely on thermal response fingerprints that are embedded in the local EEPROM at the factory. All feature extraction, state discrimination, and trend inference are completed in the MCU's built-in DSP coprocessor. The time for a single inference is less than 8 milliseconds, and the control strategy is refreshed every 200 milliseconds to ensure that the control action is highly synchronized with the actual thermal dynamics. It is particularly noteworthy that the joint generation mechanism of the "power transition safety window" and the "gear attenuation compensation coefficient" not only avoids erroneous switching in the temperature acceleration range from the time dimension, but also actively reserves thermal buffer margin from the energy dimension, forming a double protection in time and space. The compensation coefficient is adaptively scaled according to the hysteresis characteristics reflected in the fingerprint, so that the nominal power gear can truly match the physical output requirements, further suppressing the energy redundancy caused by device discreteness. The entire process requires no manual calibration, no online learning and iteration, and no additional communication modules, truly achieving a maintenance-free operation mode of "one-time modeling, lifetime usability, plug and play intelligence," endowing low-cost home appliances with high-end intelligent control capabilities.

[0013] The aforementioned technologies collectively construct an endogenous intelligent temperature control system tailored to individual differences. This system transforms the physical response characteristics of the device itself into reusable decision-making knowledge, overcoming the excessive reliance of traditional control logic on universal rules. Without upgrading hardware configurations, it significantly enhances the system's personalized response capabilities and environmental adaptability. This solution is not only applicable to electric cookers but can also be adapted to small household appliances with tiered power control requirements, such as rice cookers, electric kettles, and slow cookers, demonstrating excellent scalability and promising industrial application prospects. Attached Figure Description

[0014] Figure 1 This is the main flowchart of a method for controlling the slow start of an electric cooker with a power-decreasing heating element to prevent temperature overshoot.

[0015] Figure 2 This is a sub-flowchart of a method for controlling the slow start of an electric cooker's heating element with decreasing power to prevent overheating.

[0016] Figure 3 This is another sub-flowchart of a method for controlling the heating element power of an electric cooker with a gradual power reduction and slow start to prevent overheating. Detailed Implementation

[0017] 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 denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0018] The following disclosure provides many different embodiments or examples for implementing different structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0019] like Figure 1 As shown, this application provides a method for controlling the heating element of an electric cooker with a power-decreasing slow-start mechanism to prevent temperature overshoot, specifically including: S1: Collect the time-series signals of the surface temperature of the heating element of the electric cooker, the time-series signals of the temperature at the center of the bottom of the pot, and the time-series signals of the slight change in water vapor pressure to obtain the original multidimensional sensor dataset. S2: Perform feature extraction processing on the original multidimensional sensor dataset to parse out the time-series feature vector in order to construct a device-level thermal response fingerprint vector; S3: Encrypt and store the device-level thermal response fingerprint vector, and combine it with the initial heating rate and pressure change slope to jointly determine the current cookware load status, so as to generate a working condition matching tag; S4: Based on the operating condition matching label, call the device-level thermal response fingerprint vector, and use the embedded inference algorithm to deduce the temperature rise envelope to generate dynamic prediction results of thermal inertia; S5: Based on the thermal inertia dynamic prediction results, calculate the time permissible interval that avoids the predicted temperature acceleration rise segment and ends before the predicted peak value, so as to generate power transition safety window parameters. S6: Calculate the correction ratio based on the device-level thermal response fingerprint vector to generate the gear attenuation compensation coefficient; S7: When it is detected that the current temperature of the electric cooker heating element is less than the target temperature difference and the current heating rate of the electric cooker heating element is less than the target heating rate, the switching start time is determined by combining the power transition safety window parameter, and the target power value is adjusted by using the gear attenuation compensation coefficient to perform the power reduction stage switching action.

[0020] Step S1: Acquire the time-series signals of the surface temperature of the heating element of the electric cooker, the temperature at the center of the bottom of the cooker, and the slight changes in water vapor pressure to obtain the original multidimensional sensor dataset. Specifically, this includes: S1.1: Monitor the operating status of the electric cooker controller during the initialization phase or when a new cooking cycle is started, and generate a rectangular pulse power excitation command based on the set value in the range of 10% to 15% of the rated power to drive the heating tube to perform low-deflection thermal excitation action and generate an initial thermal disturbance source.

[0021] The input conditions are the operating status signal of the electric cooker during the initialization phase or the start of a new cooking cycle, and a preset value within the range of 10% to 15% of the rated power. The status register inside the electric cooker controller CPU and the input ports of external sensors are periodically polled and sampled to generate an operating status data packet containing a start flag, power setting value, and clock reference. Logical judgment is performed on the start flag in the operating status data packet. If the judgment result is a valid start, a rated power multiplier coefficient is loaded into the power setting value domain to generate a target power value for thermal excitation. The target power value is input to the rectangular pulse power synthesis module, which uses a numerically controlled waveform generator to construct a rectangular pulse power excitation command with a fixed amplitude and duration, and applies a synchronous phase-locked loop mechanism to keep the power waveform consistent with the system clock. Hardware mapping processing is performed on the rectangular pulse power excitation command, converting it into a pulse width modulation duty cycle control sequence for the heating element power drive switch, ensuring that the drive signal amplitude is stable and the noise does not exceed the preset range. Through the above processing method, the operating status monitoring results are transformed into a precise power waveform that drives the heating tube under low disturbance conditions, thereby generating a controllable and repeatable initial thermal disturbance source for the medium inside the pot.

[0022] For example, during the initialization phase of a certain model of electric cooker, the controller detects that the start flag is 1, the rated power setting is 1200W, the multiplication factor is set to 0.12, and the calculated target power value is 144W. This power value is loaded into the rectangular pulse power synthesis module, the pulse duration is set to 1.8 seconds, the waveform amplitude is kept constant at 144W, and a phase-locked loop is used to keep the waveform synchronized with the 50Hz power supply reference. The drive command is mapped by PWM to generate a control sequence with a duty cycle of 0.12, and a short-circuit protection threshold and electromagnetic compatibility filtering are applied to finally drive the heating element to work. In the test, the surface temperature change of the heating element generated by this low-disturbance thermal excitation was 2.3℃, the temperature change at the center of the pot bottom was 1.1℃, and the water vapor pressure change was 0.004MPa, successfully forming an initial thermal disturbance source, and the noise level was lower than the specified limit, meeting the accuracy requirements of subsequent high-frequency sampling and thermal response feature extraction.

[0023] S1.2: Synchronously trigger the temperature change of the heating tube surface, the temperature change of the pot bottom center, and the micro-change of water vapor pressure under the action of the initial thermal disturbance source. Use a multi-channel data acquisition module with a sampling rate of 200 Hz to acquire the heating tube surface temperature time sequence signal, the pot bottom center temperature time sequence signal, and the water vapor pressure micro-change time sequence signal, so as to form three parallel raw sensor data streams.

[0024] With the initial thermal disturbance source already generated by low-deflection rectangular pulse power excitation, this step utilizes a multi-channel acquisition link consisting of an NTC sensor on the heating element surface, a thermistor at the bottom of the pot, and a water vapor pressure micro-change sensor. A synchronous trigger pulse is issued by the signal acquisition control unit, sending the analog output signals from the three sensors in parallel to the multi-channel analog-to-digital converter module. The raw analog signals acquired from each channel are quantized at a fixed sampling rate of 200Hz, ensuring 200 time points are sampled per second, forming a digital signal sequence with consistent frequency and constant time intervals. The heating element surface temperature change signal, the pot bottom center temperature change signal, and the water vapor pressure micro-change signal are sequentially placed into independent buffers to maintain the independence between channels and the continuity of the data stream during acquisition. Through a channel index and trigger time association mechanism, the three data streams at the current acquisition moment are encapsulated into parallel data packets, establishing a one-to-one correspondence between sampling point numbers and timestamps. During acquisition, a sampling accuracy monitoring algorithm is executed, marking valid sampling points that exceed the set range or have excessively high noise peaks, reserving a basis for subsequent data cleaning processing. By using a multi-channel parallel acquisition and synchronous triggering mechanism, the temperature changes on the surface of the heating tube, the temperature changes at the center of the pot bottom, and the micro-changes in water vapor pressure are formed into three parallel raw sensor data streams under the same physical time reference, thus realizing a complete record of the multi-dimensional dynamic response under the action of thermal disturbance sources.

[0025] For example, during the factory initialization process, a certain model of electric cooker is configured with a heating element surface temperature acquisition channel sensitivity of 0.1℃ / LSB, a pot bottom center temperature acquisition channel sensitivity of 0.05℃ / LSB, and a steam pressure acquisition channel sensitivity of 0.2Pa / LSB. All three channels use a 16-bit ADC module. During the initial thermal disturbance, the acquisition control unit is triggered to perform 200Hz synchronous sampling for 2 seconds, generating 400 samples per channel. The sampled data is arranged in a buffer according to sampling point numbers 1 to 400, with a millisecond-level timestamp appended to the corresponding position. For the 145th sampling point, the timestamp for all three data streams is 725ms, and the values ​​are: heating element surface temperature signal 35.8℃, pot bottom center temperature signal 31.4℃, and steam pressure signal 102.6Pa. During data packet encapsulation, the sampling interval is calculated. The matching relationship between the sampling interval and the timestamp is stored in a cache index table, enabling accurate retrieval of corresponding samples during subsequent timestamp alignment and feature extraction. After execution, the three parallel raw sensor data streams are fully acquired, with a unified sampling frequency, a consistent number of time points, and a strictly synchronized physical time base, ensuring that the accuracy of subsequent time alignment processing and feature extraction processes is significantly improved.

[0026] S1.3: Perform timestamp alignment processing on the three parallel raw sensor data streams, eliminate the acquisition phase difference between multiple sensors based on a unified clock reference, and generate a synchronous multidimensional sensor data sequence with strict timing consistency.

[0027] S1.4: The synchronous multidimensional sensing data sequence is subjected to noise suppression and baseline drift correction processing. A sliding window filtering algorithm is used to remove high-frequency electromagnetic interference and compensate for sensor zero-point drift, so as to output a clean multidimensional sensing dataset with a signal-to-noise ratio that meets the feature extraction requirements.

[0028] For the synchronized multidimensional sensor data sequence after timestamp alignment, the input objects are set as unified time reference data for the heating element surface temperature time series signal, the pot bottom center temperature time series signal, and the water vapor pressure micro-variation time series signal. For each signal in the input sequence, a sliding window filtering module is invoked. Under the condition that the window length and sampling rate are matched, mean filtering is performed on the data within the window to eliminate high-frequency electromagnetic interference signal components caused by the power switching of the electric cooker and the heating element PWM drive pulse. Frequency domain energy detection is performed on the filtered output data to identify the residual frequency band of external interference and perform band-stop filtering to further reduce the impact of non-target frequency band energy on subsequent feature extraction. Zero-point baseline estimation is performed on each signal after interference suppression. The constant temperature or constant pressure segment collected in the initial stage is used as the reference interval to calculate the zero-point drift. A baseline drift compensation algorithm is adopted to progressively superimpose the drift amount in reverse according to the time index to the entire signal range to correct the long-term zero-point offset. The signal-to-noise ratio (SNR) is calculated on the baseline-corrected signal. A SNR threshold parameter is set, and the signal-to-noise ratio formula is used to determine whether the signal meets the requirements for subsequent feature extraction. By using sliding window filtering and baseline drift compensation, the time-series consistent data from the previous step is transformed into a clean, multi-dimensional sensing dataset with significantly reduced noise energy and stable zero points, thus achieving a high signal-to-noise ratio input condition that satisfies the extraction of 17-dimensional thermal response features.

[0029] For example, a sliding mean filter with an input length of 11 points for three synchronous signals with a sampling rate of 200Hz is used. Mean calculation and replacement operations are performed on each window to suppress interference in the frequency band above approximately 5Hz. FFT spectrum analysis is performed on the filtering results to detect the 50Hz power supply interference component, and band-stop filtering with a center frequency of 50Hz and a bandwidth of 4Hz is performed. The average temperature and pressure values ​​during the first 3 seconds of constant power are selected as the zero-point reference. The zero-point drift is calculated to be 0.12℃ for the temperature signal and 0.8Pa for the pressure signal, and linear inverse complementation is performed on the entire data according to the time index. After the above processing, the calculated signal power of the surface temperature signal is 0.85W, the noise power is 0.02W, and the signal-to-noise ratio is 42.5, achieving a significant improvement in interference suppression and zero-point correction. The output clean multidimensional sensing dataset has been verified to support stable operation of the multi-scale feature extraction process under a mean square error of less than 0.001.

[0030] S1.5: The clean multidimensional sensing dataset is structured and encapsulated to integrate the heating tube surface temperature time series signal, the pot bottom center temperature time series signal, and the water vapor pressure micro-change time series signal into an original multidimensional sensing dataset containing dynamic thermal response features, which serves as the sole input object for subsequent multi-scale feature extraction processing.

[0031] Step S2: Feature extraction processing is performed on the original multidimensional sensor dataset to parse out the temporal feature vector, thereby constructing a device-level thermal response fingerprint vector. Specifically, this includes: S2.1: Perform wavelet denoising and baseline drift correction on the heating tube surface temperature time series signal, the pot bottom center temperature time series signal, and the water vapor pressure micro-variation time series signal in the original multidimensional sensing dataset to eliminate the influence of high-frequency noise from the sensor and environmental interference, and generate a standardized thermal response time series dataset with a high signal-to-noise ratio.

[0032] Independent data preprocessing channels were established for the heating tube surface temperature time series signal, the pot bottom center temperature time series signal, and the water vapor pressure micro-variation time series signal in the original multidimensional sensor dataset, serving as input sources for wavelet denoising. A discrete wavelet decomposition algorithm was used to separate high-frequency noise components in the multi-scale frequency domain. Zeros were embedded in the corresponding coefficient domains of the noise components, and inverse wavelet transform was performed to obtain the denoised signal. For the denoised signal, a baseline drift correction module was used to calculate the long-term trend offset based on the zero-point drift monitoring curve, and a correction formula was applied. The three processed signals were resampled and mapped using a unified sampling clock to form a standardized time series dataset with high signal amplitude consistency across the entire frequency domain and significantly reduced noise residual rate. Through this wavelet denoising and baseline drift correction processing method, the original multidimensional sensor dataset from the previous step was transformed into standardized thermal response time series data with a high signal-to-noise ratio, ensuring the accuracy of subsequent second-order differential feature extraction and steady-state slope fitting.

[0033] For example, the three parallel raw sensor data streams excited at 12% of rated power are input into a three-level discrete wavelet decomposition module based on the Daubechies-6 wavelet basis to extract detail coefficients and approximation coefficients. After the detail coefficients are zeroed in layers one and two, an inverse transform is performed to obtain the denoised signal. The baseline drift offset is calculated using a moving average method with 600 sampling points. Assuming the average value of the denoised signal of the heating tube surface temperature within this window is 45.8℃ and the target baseline should be 45.2℃, the offset is 0.6℃. After correction by the formula, the temperature value of each sampling point is reduced by 0.6℃ to obtain the corrected signal. The small-variable signals of the bottom center temperature and water vapor pressure are processed in the same way, and the baselines are corrected by 0.3℃ and 2.1Pa respectively after denoising. The three signals are uniformly resampled at a sampling rate of 200Hz to form a standardized time-series dataset. The test shows that this dataset reduces the high-frequency noise power spectral density by about 47% compared to the original signal, and the zero-point drift error is reduced to within 0.05℃, meeting the high-precision requirements of subsequent feature extraction.

[0034] S2.2: Perform first-order and second-order differential operations based on the standardized thermal response time series dataset to quantify the rate of temperature change and acceleration abrupt change points, and generate a transient response characteristic parameter set including the coordinates of the rise time inflection point, the value of the first peak delay, and the offset of the second derivative zero crossing point.

[0035] Based on the small-scale change signals of heating tube surface temperature, pot bottom center temperature, and water vapor pressure in the standardized thermal response time series dataset, a first-order difference operator is used to perform discrete-time derivative operations on the data of each channel to generate a velocity sequence that characterizes the instantaneous temperature change rate, and the independent sequence of each channel is retained as the input reference for subsequent acceleration calculation.

[0036] The second-order difference operator is applied to the velocity sequence to perform discrete-time second-order derivative calculation, thereby obtaining an acceleration sequence that characterizes the temperature curvature change, and the candidate time point set of acceleration abrupt change is obtained by analyzing the curvature sign change.

[0037] Joint peak detection is performed on temperature and velocity sequences. The sample in the velocity sequence that first reaches a local maximum value and whose amplitude exceeds a preset threshold is located as the rising time inflection point, and its coordinates on the original time axis are recorded.

[0038] The first local peak point is identified in the temperature sequence, and the time difference between the peak point and the inflection point of the rise time is calculated to obtain the value of the first peak delay, which is used to describe the response hysteresis characteristics of the system from the initial heating to the formation of the first temperature peak.

[0039] The zero-crossing point where the second derivative curve changes from positive to negative is searched in the acceleration sequence, and the time difference between the zero-crossing point and the rise time inflection point is calculated to generate the zero-crossing offset of the second derivative, which is used to quantitatively characterize the response delay of thermal inertial drive.

[0040] Through the above chain processing method, the standardized thermal response time series dataset from the previous step is transformed into a transient response characteristic parameter set containing the coordinates of the rise time inflection point, the value of the first peak delay, and the offset of the second derivative zero crossing point, thereby realizing the quantitative analysis of the instantaneous dynamic characteristics in the initial heating stage and the inertial action stage.

[0041] For example, for the standardized heating tube surface temperature sequence collected under rectangular pulse excitation at 12% of rated power, with a sampling period of 5ms and a total of 400 sampling points, the velocity sequence was calculated using the first-order difference operator formula. The result showed that the velocity sequence peak appeared at the 60th sampling point, corresponding to a time of 0.3 seconds, which was recorded as the rise time inflection point coordinate of 0.3 seconds. In the temperature sequence peak detection, the first peak appeared at the 85th sampling point, corresponding to a time of 0.425 seconds, and the first peak delay was calculated as 0.125 seconds. In the acceleration sequence curvature analysis, the second derivative zero-crossing point appeared at the 78th sampling point, corresponding to a time of 0.39 seconds, and the time difference with the rise time inflection point, i.e., the second derivative zero-crossing point offset, was 0.09 seconds. The transient response characteristic parameter set output in this embodiment is as follows: rise time inflection point coordinate 0.3 seconds, first peak delay 0.125 seconds, second derivative zero-crossing offset 0.09 seconds. Its application effect significantly improves the prediction accuracy of temperature overshoot in the subsequent power descent strategy, ensuring the stability of temperature change during the heating stage switching process.

[0042] S2.3: The slope of the steady-state interval of the standardized thermal response time series dataset is fitted using the sliding window linear regression algorithm to analyze the energy decay trend when the heat conduction reaches equilibrium and generate a steady-state slope decay rate index that characterizes the heat loss characteristics of the system.

[0043] The steady-state interval of the standardized thermal response time-series dataset is divided into windows. The start and end positions of the steady-state segment are determined based on the principle of minimizing the local variance of the temperature change rate curve, and these positions are used as the input segments for the fitting operation.

[0044] Within the defined steady-state interval, the sliding window linear regression algorithm is invoked to perform least squares fitting calculations on the temperature and time data within each window, thereby obtaining the instantaneous slope value of each window.

[0045] The slope values ​​of each window are smoothed and filtered according to the time series to eliminate the influence of local perturbations on the slope estimation.

[0046] Calculate the initial slope and the final slope of the smoothed slope sequence, and calculate the steady-state slope decay rate.

[0047] By comparing the attenuation rate value with a preset threshold, the thermal loss characteristics of the system are quantified and a steady-state slope attenuation rate index is generated.

[0048] By using sliding window linear regression fitting, slope smoothing, and decay rate calculation, the standardized thermal response data from the previous step is transformed into a steady-state slope decay rate index that characterizes the energy decay trend during the thermal conduction equilibrium stage, thus supplementing the thermal response fingerprint vector with steady-state characteristics.

[0049] For example, in a heating cycle with a capacity of 2.0L of water and an initial temperature of 25℃, the steady-state region was collected over a time span of 120 seconds, with a window length of 10 seconds and a window sliding step size of 2 seconds, resulting in 50 sample windows. Within the first window, the initial slope of the steady-state region was 0.0125℃ / s, and the final slope was 0.0098℃ / s, yielding a steady-state slope decay rate of 0.216. This value is significantly higher than the typical decay rate of 0.08 under low-load conditions, indicating a faster heat loss rate under this condition. Inputting this indicator into the subsequent thermal response fingerprint vector concatenation stage can compensate for the insufficient thermal inertia of the load in advance during power descent decisions, thereby improving temperature control stability.

[0050] S2.4: Perform multi-dimensional vector splicing and normalization mapping on the transient response feature parameter set and the steady-state slope decay rate index to unify the weight distribution of features with different dimensions and generate a device-level thermal response original fingerprint vector covering seventeen time-series features.

[0051] S2.5: Based on the principal component analysis algorithm, the original device-level thermal response fingerprint vector is subjected to feature dimensionality reduction and redundancy removal to retain the key thermal behavior patterns with the largest variance contribution, and finally generate a device-level thermal response fingerprint vector that can uniquely characterize the degree of thermal inertia hysteresis of the current electric cooker.

[0052] like Figure 2 As shown, step S3 involves: encrypting and storing the device-level thermal response fingerprint vector, and combining the initial heating rate with the pressure change slope to jointly determine the current cookware load state, thereby generating a working condition matching tag. Specifically, this includes: S3.1: Read the encrypted device-level thermal response fingerprint vector from the local non-volatile memory and decrypt and verify it to restore the standard thermal behavior identification data containing seventeen-dimensional time-series features, which will serve as the benchmark reference for subsequent operating condition judgment.

[0053] When a device-level thermal response fingerprint vector read instruction is received from the local non-volatile memory, the physical address of the non-volatile memory is mapped and matched with the corresponding fingerprint data block position in the index table to locate the complete data area of ​​the encrypted fingerprint vector.

[0054] After the location is completed, the controller calls the embedded data reading driver module to perform a buffer loading operation on the encrypted fingerprint vector in byte order, and temporarily stores the loading result in the cache area to ensure the data integrity of subsequent decryption processing.

[0055] The algorithm calls a preset symmetric key decryption algorithm on the encrypted fingerprint vector in the cache area, performs block-by-block decryption operation using the unique device key stored in the security key management unit, and outputs the decrypted fingerprint raw data stream.

[0056] The decrypted fingerprint raw data stream is sent to the integrity verification module, which compares the CRC32 checksum or SHA-256 digest value with the original checksum in the memory to confirm the lossless and consistent nature of the fingerprint data during storage, reading and decryption.

[0057] The original fingerprint data stream that has passed verification is subjected to structured parsing, and the seventeen-dimensional time-series features are mapped into a standard thermal behavior identification data matrix according to the preset field order. The physical quantity description information and dimensional identifier are added to each feature.

[0058] Through the above chain processing method, the feature extraction results of the previous step are transformed into standard thermal behavior identification data that has been decrypted and verified, thereby realizing the callability and data security of static thermal fingerprints in working condition identification.

[0059] For example, in the local EEPROM of an electric cooker, the device-level thermal response fingerprint vector is stored in 256-byte encrypted form with an offset of address 0x0800. After the controller locates this offset address according to the index table, it loads the data into the buffer via the SPI bus in 64-byte blocks, with a buffer capacity of 512 bytes to ensure redundancy. During the decryption phase, the AES-128 symmetric algorithm is called, using the key K={0x2F,0xA1,...} for block decryption, generating a plaintext sequence with seventeen features. Subsequently, a CRC32 check is performed, which matches the original CRC value in memory, confirming data integrity. After parsing, a standard thermal behavior identification data matrix containing indicators such as rise time inflection point and first peak delay is obtained, which can be directly used for subsequent real-time operating condition feature mapping calculations in S3.2. During the verification process, the system found that the total time for data reading and decryption was 6.8ms, meeting the synchronization requirements of real-time performance and security of this solution.

[0060] S3.2: At the beginning of the formal cooking cycle, the water temperature rise rate time-series signal and the water vapor pressure slight change slope time-series signal are collected simultaneously during the initial heating stage. The transient features of the time-series signals are extracted using the sliding window difference algorithm to generate a set of real-time working condition feature parameters that characterize the current cookware load state.

[0061] At the initial stage of the formal cooking cycle, to address the dynamic load discrimination input required by the heating element power reduction control strategy, the controller initiates data acquisition services from the thermistor at the bottom of the pot and the steam pressure sensor based on a synchronous acquisition mechanism. This acquires water temperature and pressure time-series signals, respectively, and inputs both signals to a preset transient feature extraction module. The acquired water temperature time-series signal undergoes sliding window differential processing to calculate the temperature change within each sampling window and form a temperature change rate sequence. This rate of change is indexed by the sampling time base for subsequent load state mapping. The acquired steam pressure time-series signal undergoes first-order differential and linear fitting processing to quantify the pressure slope corresponding to the steam generation rate in the initial heating stage, thereby generating a pressure change rate sequence to dynamically characterize the evaporation intensity of the liquid surface. The temperature and pressure change rate sequences are synchronously paired one-to-one according to timestamps and synthesized into a real-time operating condition feature vector containing both temperature rise rate and steam generation rate parameters using a feature concatenation algorithm. A normalized mapping method is used to unify the dimensions and scale the parameters of the real-time operating condition feature vector to eliminate the influence of sensor range and environmental differences on the discrimination results. Through this processing method, the decryption and verification results of the previous step and the synchronous acquisition signal of the current initial heating stage are transformed into a set of real-time operating condition feature parameters characterizing the load state of the cookware, thereby realizing the construction of input conditions for dynamic load identification.

[0062] For example, during the start of a cooking cycle in a 1.5-liter electric cooker, the controller sets the sampling time window to 1 second and the sliding step size to 0.5 seconds. The water temperature sequence collected by the thermistor at the center of the pot bottom is [35.0, 35.4, 35.9, 36.3, 36.8] degrees Celsius, and the pressure sequence collected by the steam pressure sensor is [101.325, 101.340, 101.356, 101.370, 101.385] kPa. A sliding window difference calculation is performed on the water temperature sequence to generate the temperature rise rate sequence [0.4, 0.5, 0.4, 0.5]. A first-order difference is performed on the pressure sequence to generate the pressure change rate sequence [0.015, 0.016, 0.014, 0.015]. The two sequences were aligned and concatenated according to their timestamps to form four sets of real-time operating condition feature vectors [(0.4,0.015),(0.5,0.016),(0.4,0.014),(0.5,0.015)]. After normalization and mapping to a unified dimension, these vectors were used as inputs for load status determination. In the verification results, this set of feature parameters, combined with standard thermal behavior identification data, can accurately derive the load heat capacity coefficient and thermal inertia hysteresis index, ensuring high-precision dynamic prediction capabilities during power descent switching.

[0063] S3.3: Based on the steady-state slope decay rate characteristics in the standard thermal behavior identification data and the water temperature rise rate time series signal in the real-time operating condition characteristic parameter set, perform multivariate coupling mapping calculation to quantitatively derive the heat capacity coefficient and thermal inertia hysteresis index corresponding to the current water quality.

[0064] Based on the steady-state slope decay rate characteristic parameter in the decrypted and verified standard thermal behavior identification data and the water temperature rise rate time-series signal in the real-time operating condition characteristic parameter set, a multivariate input set is established for the input of the heat capacity parameter calculation model. The steady-state slope decay rate is used as a static characteristic quantity characterizing the system's heat loss rate per unit time, and the water temperature rise rate is used as a dynamic characteristic quantity representing the current actual heat input efficiency. A coupling relationship matrix is ​​constructed to simultaneously analyze heat conduction and heat storage effects. Using a multivariate coupling mapping method, the static and dynamic characteristic quantities are processed uniformly according to the physical dimension, and the heat capacity coefficient corresponding to the current water mass is inverted using the specific heat capacity formula. Based on the product relationship between specific heat capacity and steady-state slope decay rate, the value of the heat capacity coefficient is derived. Multiplying the heat capacity coefficient by the steady-state slope decay rate yields the thermal inertia hysteresis index. The thermal inertia hysteresis index is normalized to ensure it matches the range in the preset operating condition classification mapping table. By using a multivariate coupling mapping process, the feature quantities obtained from standard thermal behavior identification data and real-time operating condition feature parameter sets are transformed into heat capacity coefficients and thermal inertia hysteresis indicators that can be directly used for operating condition matching, thereby achieving quantitative unification of static heat consumption characteristics and dynamic heating characteristics.

[0065] For example, in an electric cooker with a rated power of 1500W, the steady-state slope decay rate characteristic parameter of the decrypted standard thermal behavior identification data is 0.012K / s, and the water temperature rise rate during the initial heating stage is collected in real time as 0.045K / s. Using these two as inputs to construct a coupling relationship matrix, the heat capacity coefficient is obtained by inversion using the specific heat capacity formula as 4186J / (kg·K). Combining the water temperature rise rate and the heat capacity coefficient, the water mass is calculated to be 3.0kg. Multiplying the heat capacity coefficient by the steady-state slope decay rate yields a thermal inertia hysteresis index of 50.232J / (K·s). After normalization, this index is mapped to the third level of the preset operating condition classification mapping table, indicating that the current water level of the cooker is medium-high, and the thermal inertia hysteresis is relatively strong. This output result is used in subsequent steps to accurately define the power transition safety window and the gear attenuation compensation coefficient. During the verification process, no temperature overshoot occurred during power descent switching, and the temperature control stability was significantly improved.

[0066] S3.4: Based on the heat capacity coefficient and thermal inertia hysteresis index, consult the preset operating condition classification mapping table to perform discretization interval matching processing, so as to generate a unique operating condition matching label that identifies the specific water volume level and thermal inertia hysteresis, and complete the logical association between static fingerprint and dynamic load.

[0067] Using the heat capacity coefficient and thermal inertia hysteresis index calculated by multivariate coupling mapping as input parameters, the index retrieval module of the preset operating condition classification mapping table is called. The heat capacity coefficient is quantized and normalized, and a bucketing algorithm is used to map continuous values ​​to discrete intervals of heat capacity levels, establishing a unique index identifier in each interval for subsequent label combination. The thermal inertia hysteresis index is input into the thermal inertia classification sub-table of the mapping table, and interval boundary value comparison is performed to classify the index value into the hysteresis level interval and generate a corresponding index number. The heat capacity level index and thermal inertia hysteresis level index are merged using a Cartesian product combination algorithm to form a unique combined index code, ensuring that each pair of different level combinations corresponds to a unique operating condition matching label. The combined index code is hashed and encrypted to generate an irreversible label hash value to prevent external inference of internal level classification rules. At the same time, mapping table version information and timestamps are added to the label data structure to ensure data traceability. The combined index is mapped back to the actual operating condition level description through the label encoding and decoding function, completing the logical association between static thermal response fingerprint and dynamic load conditions. By using interval discretization and tag combination processing, the heat capacity coefficient and thermal inertia hysteresis index of the previous step are transformed into unique operating condition matching tags, enabling targeted prediction and control based on specific water volume levels and inertia levels in subsequent reasoning processes.

[0068] For example, during a heating process in an electric cooker, the heat capacity coefficient calculated by S3.3 is 2.87 J / (g·℃) and the thermal inertia hysteresis index is 0.52 s. The heat capacity coefficient is input into the capacity classification module of the operating condition classification mapping table, with a preset boundary value of [2.0, 2.5, 3.0, 3.5]. After comparison, 2.87 is classified into the capacity level range [2.5, 3.0), corresponding to capacity level index C3. The thermal inertia hysteresis index is input into the delay classification module, with a preset boundary value of [0.3, 0.5, 0.7]. After comparison, 0.52 is classified into the delay level range [0.5, 0.7), corresponding to delay level index L2. The Cartesian product combination algorithm is executed to obtain the combination index (C3, L2), which is then hashed to generate a tag hash value f8a1c4e9, along with the mapping table version V1.2 and the current system clock timestamp 1632458765. In the subsequent S4 prediction process of the control system, the tag decoding function resolves f8a1c4e9 into capacity level C3 and delay level L2, accurately matching the current operating conditions of approximately 3L of water volume and moderate thermal inertia delay. This achieves precise matching between the predicted trajectory and the power reduction strategy. The verification shows that the deviation between the predicted peak time and the actual peak time is less than 0.4 seconds, and the temperature overshoot amplitude is significantly reduced.

[0069] like Figure 3As shown, step S4 involves: calling the device-level thermal response fingerprint vector based on the operating condition matching label, and using an embedded inference algorithm to deduce the temperature rise envelope to generate a dynamic prediction result for thermal inertia. Specifically, this includes: S4.1: Based on the operating condition matching label, read the device-level thermal response fingerprint vector from the local non-volatile memory, and perform multi-dimensional feature decoupling processing on the device-level thermal response fingerprint vector to separate the first feature subset characterizing the thermal capacity characteristics of the heating tube and the second feature subset characterizing the thermal resistance characteristics of the pot body, thereby generating a decoupled thermal behavior feature parameter set for subsequent inference calculation.

[0070] Based on the operating condition matching tags generated in the previous steps, the device-level thermal response fingerprint vector is retrieved from local non-volatile memory as input. Data retrieval is performed according to the water volume level and thermal inertia hysteresis indicated by the tags to ensure that the selected fingerprint vector is consistent with the current cooking load state. The read seventeen-dimensional time-series feature vector is input to the multi-dimensional feature decoupling processing module, where column-domain normalization and covariance matrix operations are performed on the eigenvalue matrix to eliminate numerical imbalances between features of different dimensions and obtain the correlation coefficient matrix of each feature. Based on the correlation coefficient matrix, a feature grouping algorithm is applied to divide significantly correlated features with similar physical properties into the same subset. Features with high heat capacity correlation are assigned to the first feature subset, and features with high thermal resistance correlation are assigned to the second feature subset. Principal component analysis is performed on the first feature subset to reduce dimensionality, retaining the heat capacity principal component with the largest variance contribution; the same principal component analysis is performed on the second feature subset to retain the thermal resistance principal component with the largest variance contribution, thus forming feature vector pairs that independently characterize the heat capacity characteristics of the heating element and the thermal resistance characteristics of the pot body. A matrix concatenation method is used to combine the principal components of heat capacity and thermal resistance into a decoupled set of thermal behavior feature parameters, which serves as the sole physical input for constructing the discrete-time heat transfer state equation. This decoupling process transforms the original fingerprint vector associated with the operating condition matching label from the previous step into a set of thermal behavior feature parameters with separated physical attributes and no redundancy, enabling independent modeling and precise quantification of heat capacity and thermal resistance characteristics during the inference and calculation process.

[0071] S4.2: Construct a discrete-time domain heat transfer state equation using the decoupled thermal behavior characteristic parameter set, and input the current real-time collected time series signal of the pot bottom center temperature into the discrete-time domain heat transfer state equation as the initial state variable to iteratively calculate the theoretical temperature rise rate at the next moment, thereby generating an instantaneous heat flow driving vector containing the thermal inertia accumulation effect.

[0072] The first feature subset of the heating tube's heat capacity characteristics and the second feature subset of the pot's thermal resistance characteristics in the decoupled thermal behavior characteristic parameter group are paired according to the mapping rules of the heat transfer physical model to establish a discrete-time domain heat transfer state equation describing the relationship between heat flow and temperature change. The real-time acquired time-series signal of the pot bottom center temperature is timestamped and synchronized within the controller, and the latest sampled temperature value is input as the initial state variable to the starting time of the state equation. An iterative recursive algorithm is executed for the temperature evolution part of the state equation, updating the heat flow transfer coefficient based on the product of heat capacity and thermal resistance within each prediction step to reflect the intensity of thermal inertia under the residual heat release environment. The instantaneous power-heat conversion formula is used to combine the transfer coefficient with the pot bottom temperature change to calculate the corresponding theoretical temperature rise rate, ensuring the consistency of unit conversion for the heat capacity and temperature difference terms in the formula. The formula is expressed as: Where v is the theoretical temperature rise rate, Q is the heat transferred, Δt is the prediction step time, m is the water mass, and C is the specific heat capacity. A series of theoretical temperature rise rate values ​​calculated with continuous step times are combined to form an instantaneous heat flow driving vector, which is used for predicting and simulating the subsequent temperature rise trajectory. Through the above processing method, the decoupled thermal behavior characteristics of the previous step are transformed into an instantaneous heat flow driving vector with time resolution, achieving accurate quantification of the cumulative effect of thermal inertia.

[0073] For example, in a 600W rated power electric cooker, the decoupled thermal behavior characteristic parameter set includes a heat capacity value of 2500 J / K and a thermal resistance value of 0.08 K / W. The real-time collected temperature at the center of the pot bottom is 85.4 degrees Celsius, and the prediction step size Δt is set to 0.5 seconds. The controller determines the transfer coefficient to be 0.12 based on the state equation, and the heat released Q within the prediction step is 150 J. The theoretical temperature rise rate is calculated to be 0.048 degrees Celsius / second. The rate values ​​of 24 consecutive steps are used to construct the driving vector. The driving vector shows a trend of accelerated temperature rise in the early stage of the thermal inertia peak at high times. The application effect is to significantly improve the stability of the temperature rise curve fitting in the prediction time domain and provide accurate input results for the subsequent power transition safety window definition.

[0074] S4.3: Based on the instantaneous heat flow driving vector, perform forward time step extrapolation calculation, continuously superimpose the temperature increment caused by the release of residual heat from the heating tube within the prediction time domain of six to twelve seconds, in order to simulate the natural temperature rise trajectory under conditions without external power input, thereby generating original temperature rise sequence data describing the future temperature change trend.

[0075] Based on the instantaneous heat flow driving vector, the prediction time domain range is set to six to twelve seconds in the embedded lightweight inference algorithm. This driving vector is used as the input reference signal for time step iteration, and an accumulation sequence of temperature increments is established for each time step. For each time step, heat transfer increment calculation is performed. Discrete integration is used to continuously accumulate the temperature increment generated by the release of residual heat from the heating tube within the prediction time window, ensuring that each increment calculation references the product of the previous accumulated result and the current instantaneous driving vector. For each increment calculation process, the water heat capacity coefficient and the boiler thermal resistance parameter are introduced for equivalent physical quantity correction to ensure that the prediction curve reflects the actual thermal inertia characteristics of the system. During the accumulation process, the accuracy of the temperature change acquisition values ​​at each step is constrained, and a heat capacity correction factor is used to adjust the calculation output to avoid numerical drift that could lead to increased prediction errors. The accumulated temperature change data at each moment within the entire time window is encapsulated by time index into an original temperature rise sequence dataset, maintaining strict temporal continuity and increment consistency between points in the sequence. Through the aforementioned continuous superposition calculation method, the instantaneous heat flow driving vector from the previous step is transformed into raw temperature rise sequence data with physical constraints, achieving accurate simulation of the natural temperature rise trajectory under conditions without external power input. For example, with a medium load on the boiler body, a heat capacity coefficient of 4180 J / (kg·℃), and a boiler body thermal resistance of 0.12℃ / W, the initial value of the instantaneous heat flow driving vector is 2.8 W / ℃. The prediction time domain is set to eight seconds, the time step to one second, and a discrete integral formula is used. Where Q(t) is the heat flow driving force at the current moment, m is the water mass, and c is the specific heat capacity. In the first second, the instantaneous heat flow driving force is 2.8W, and the water mass is 1.5kg. Substituting these values ​​into the formula, we get ΔT(1)≈0.00045℃. After eight seconds of cumulative superposition and combined with a thermal resistance correction factor of 0.95, the generated original temperature rise sequence reaches a peak of 0.0034℃ at eight seconds. In actual verification, the deviation between this predicted curve and the measured natural temperature rise trajectory is significantly reduced, which can support the accuracy requirements of subsequent extreme value search and transition window definition.

[0076] S4.4: Perform extreme point search and curvature analysis on the original temperature rise sequence data to identify the zero crossover point where the first derivative turns from positive to negative and the inflection point where the absolute value of the second derivative is the largest, so as to determine the arrival time of the highest temperature and the maximum temperature rise under the action of thermal inertia, thereby generating the coordinate information of the predicted temperature lag peak point caused by thermal inertia.

[0077] The original temperature rise sequence data describing future temperature change trends is indexed with discrete sampling points on a unified time base, and this data is passed as an input sequence to the extreme point search module.

[0078] The first derivative is calculated on the input sequence, and the central difference algorithm is used to estimate the instantaneous rate of change between adjacent sampling points, and a continuous rate of change time series curve is generated.

[0079] Zero-crossing detection is performed on the time series curve of the rate of change to identify the sampling point where the first derivative changes from a positive value to a negative value, and the time coordinate of the location is recorded as a temporary peak candidate point.

[0080] The second derivative is calculated on the original temperature rise sequence data, the curvature change is analyzed using the second-order difference operator, and the time series curve of the absolute value of curvature is established.

[0081] Search for the global maximum point in the absolute value of curvature time series curve, and match the time coordinate corresponding to the point with the temporary peak candidate point to verify the arrival time of the true highest temperature under thermal inertia.

[0082] The time coordinates after matching and verification are combined with the corresponding temperature amplitudes in the original temperature rise sequence data to generate the coordinate information of the peak point of temperature lag caused by predicted thermal inertia.

[0083] By using extreme value search and curvature analysis, the original temperature rise sequence data from the previous step is transformed into peak point time coordinates and temperature rise amplitude indicators, enabling precise positioning of the highest temperature moment under thermal inertia and providing a basis for subsequent transition window delineation.

[0084] S4.5: Based on the coordinate information of the peak temperature lag caused by the predicted thermal inertia, the starting and ending boundaries of the temperature acceleration rise interval are defined, and the latest time allowed for power switching is derived in reverse by combining the preset safety margin time threshold, so as to define the best operating time period to avoid overshoot, and finally generate the thermal inertia dynamic prediction results of the peak temperature lag caused by the predicted thermal inertia and the transition window.

[0085] Step S5: Based on the thermal inertia dynamic prediction results, calculate the time-permissible interval that avoids the predicted temperature acceleration rise segment and whose endpoint is set before the predicted peak value, to generate power transition safety window parameters. Specifically, this includes: S5.1: Perform first-order derivative calculation on the temperature rise envelope in the thermal inertia dynamic prediction results to obtain the time series curve of temperature change rate, which characterizes the instantaneous rate of change of the water temperature in the pot, as the input benchmark data for identifying the accelerated temperature rise segment.

[0086] The input for processing the first derivative of the temperature rise envelope in the thermal inertia dynamic prediction results is the original temperature rise sequence data generated in the previous step S4.3. This sequence has strict timestamp synchronization attributes and records the continuous values ​​of the boiler water temperature changing with time in the future prediction time domain. The original temperature rise sequence data is input into the numerical differential calculation module, and a sliding window difference algorithm with matched sampling rate is used to divide the difference between adjacent sampling points by the sampling interval to generate a preliminary time-series curve of the water temperature change rate at each time point. The preliminary time-series curve is then smoothed at its boundaries. Cubic spline interpolation is used to eliminate rate abrupt changes caused by the finite sampling window in the curve endpoint region to ensure the stability of subsequent critical region detection. The smoothed curve is further subjected to noise suppression. A Savitzky-Golay filter is used to fit the local waveform and replace the original value under the condition that the window length matches the polynomial order, thereby reducing the impact of high-frequency disturbances on the rate of change determination. The rate of temperature change at any given moment is calculated using the formula for the rate of temperature change. The results of this calculation are then concatenated to form a complete time-series curve of the rate of temperature change, and the timestamp information of each calculation point is associated with it. This serves as the input baseline data for identifying the accelerated temperature rise phase. Through this chained processing method, the temperature rise sequence data generated in the previous step is transformed into a time-series curve of the rate of temperature change that meets the requirements for subsequent second-derivative zero-crossing detection, thus achieving a quantitative description of the temperature change trend.

[0087] S5.2: Based on the time series curve of temperature change rate, execute the second derivative zero-crossing point detection algorithm to analyze the critical moment when the temperature change rate changes from increasing to decreasing, thereby determining the starting and ending boundary points of the predicted temperature acceleration rise segment.

[0088] Based on the input temperature change rate time series curve, the second derivative calculation module is called to perform global second derivative operation on the curve to obtain the second-order rate of change time series signal describing the acceleration of temperature change rate.

[0089] The zero-crossing detection algorithm is used on the second-order rate of change time series signal to calculate the sign change of each sampling point and identify the sign change events from positive to negative or from negative to positive as a potential critical moment candidate set.

[0090] For each zero-crossing point in the candidate set, curvature verification is performed. The absolute value of the second derivative and the magnitude of the first derivative are used to jointly determine whether the point satisfies the physical constraint that the rate of temperature change changes from increasing to decreasing.

[0091] The zero-crossing points that meet the conditions are sorted by time index, and the time coordinate of the first zero-crossing point is defined as the starting boundary point of the temperature acceleration rising segment, and the time coordinate of the last zero-crossing point that meets the conditions is defined as the ending boundary point of the temperature acceleration rising segment.

[0092] The zero-crossing determination value is calculated using a zero-crossing determination index formula based on the second derivative. The condition for zero-crossing is a sign change and ZC exceeding a preset threshold.

[0093] By using zero-cross detection and curvature confirmation processing, the dynamic change characteristics in the temperature change rate time series curve are transformed into the start and end boundary point data of the temperature acceleration rise segment, thereby realizing the time positioning of the key interval.

[0094] For example, in a certain test scenario, the real-time temperature change rate time-series curve was sampled at a frequency of 200Hz. The second-order rate of change time-series signal changed from positive to negative at 14.28 seconds, with the absolute peak reaching 2.3℃ / s², and changed from negative to positive at 18.72 seconds, with the absolute peak reaching 2.1℃ / s². After curvature confirmation with a threshold set at 2.0℃ / s², these two zero-crossing points were used as the starting and ending boundary points of the accelerated temperature rise segment, respectively. On the first derivative curve of the temperature change rate, the heating rate increased significantly before the starting boundary point and slowed down after the ending boundary point. The determination value for the starting point was calculated to be 2.3, and the determination value for the ending point was 2.1, both exceeding the set threshold, verifying that this interval can be used as the input data for the left boundary of the power transition safety window.

[0095] S5.3: Perform time backtracking calculation based on the time coordinate of the peak point of temperature lag caused by predicted thermal inertia to generate a power switching forced termination time located between 1.8 seconds and 3.2 seconds before the predicted peak, which serves as the right boundary constraint condition of the time-permitted interval.

[0096] Based on the time coordinates of the peak point of temperature lag caused by predicted thermal inertia, the peak moment data output by the embedded inference is used as the input reference for time backtracking calculation.

[0097] The peak moment data is compared with the preset safety margin time threshold by performing a difference calculation. The lower limit of the margin interval is set to 1.8 seconds and the upper limit is set to 3.2 seconds. The candidate set of power switching forced termination time is obtained by calculating the difference.

[0098] Boundary constraint checks are performed on each time point in the candidate set to ensure that it is within the prediction time domain and does not exceed the upper limit margin, and time points that do not meet the conditions are removed.

[0099] The forced termination time is derived, and the safety margin threshold is taken in the range of 1.8 seconds to 3.2 seconds.

[0100] The derived forced termination time is synchronized and mapped with the current system clock reference to generate the right boundary constraint data of the time-permitted interval.

[0101] By using the above processing method, the predicted peak point coordinate information from the previous step is transformed into a power switching forced termination time that meets the safety margin requirements, thereby achieving strict constraints on the right boundary of the time-permitted interval.

[0102] For example, in a real-world cooking scenario, the predicted peak time of temperature lag due to thermal inertia is 245.6 seconds. The safety margin threshold is set with a lower limit of 1.8 seconds and an upper limit of 3.2 seconds, and a margin value of 2.5 seconds is selected for calculation. The forced termination time of power switching is obtained as 243.1 seconds. This value is time-domain synchronized and compared with the current system clock of 245.0 seconds, resulting in a right boundary constraint of 243.1 seconds for the output time permissible interval. In this scenario, since the predicted termination boundary of the accelerated temperature rise segment is 242.8 seconds, the intersection of the intervals calculated using the right boundary constraint ensures that the power switching occurs after the accelerated temperature rise segment and within the safety margin before the peak. Experimental results show that, when power switching is executed at this forced termination time, the maximum temperature rise of the pot is significantly lower than in the unconstrained switching scenario, and both temperature control accuracy and system stability are significantly improved.

[0103] S5.4: Perform interval intersection operation by using the termination boundary point of the predicted temperature acceleration rise segment and the forced termination time of power switching to construct the original set of time-permitted intervals that avoid the temperature acceleration rise segment and meet the safety margin requirements before the peak.

[0104] Using the predicted temperature acceleration rise segment termination point and power switching forced termination time obtained from the previous steps as input parameters, time domain correlation analysis is performed on the two on a unified time axis to establish a dual-time-marked interval boundary reference system.

[0105] The termination boundary point of the predicted temperature acceleration rise segment and the forced termination time of power switching are mapped as two candidate boundaries of the time-permitted interval, respectively. A multi-boundary priority determination algorithm is used to determine their relative positional relationship on the time axis to avoid intersection or misalignment during the interval calculation process.

[0106] For the two candidate boundaries mentioned above, an interval intersection operation is performed. Based on the time axis order, the absolute time difference between the termination boundary point and the forced termination time of power switching is limited to the intersection length. Through mathematical calculation formulas, it is ensured that the intersection interval only includes the time period that satisfies the safety margin requirement and avoids the temperature acceleration rise period.

[0107] Apply a safety margin time compensation to the boundaries of the intersection interval, shift the left boundary backward by the amount of time corresponding to the safety margin coefficient, and shift the right boundary forward by the same coefficient, to ensure that the final interval has a buffer space in time to prevent temperature overshoot.

[0108] By using interval intersection and safety margin compensation processing, the termination boundary point of the temperature acceleration rise segment and the forced termination time of power switching in the previous step are transformed into the original set of time-permitted intervals that avoid risks and retain safety margins, thus realizing the calculation basis for the power transition safety window parameters.

[0109] For example, in a test of a 1500W electric cooker, the predicted termination boundary of the temperature acceleration phase was 42.6 seconds, the starting boundary of the temperature acceleration was 38.4 seconds, the forced termination time of the power switching was 46.5 seconds, the estimated end time of the temperature rise phase was 47.2 seconds, and the safety margin coefficient was set to 1.2 seconds. Substituting these parameters into the formula, the intersection interval boundary was obtained as [42.6 seconds, 46.5 seconds]. Applying a 1.2-second translation compensation to the left and right boundaries respectively, the corrected interval was obtained as [43.8 seconds, 45.3 seconds]. This interval, as the original set of time-permitted intervals, can be directly used for subsequent power transition safety window parameter encapsulation. When applied to the power descent phase switching, the test verified that after the power switching was performed in this window, the water temperature in the pot transitioned smoothly before the predicted peak, without temperature overshoot, and the temperature change trajectory after switching was highly consistent with the model prediction curve, significantly improving the temperature control accuracy and system stability.

[0110] S5.5: Perform validity verification and parameter encapsulation processing on the original set of time permission intervals to generate power transition safety window parameters containing clear start and end times, which serve as the basis for the control module to execute action instructions for switching the power descent stage without overshoot risk.

[0111] Step S6: Based on the device-level thermal response fingerprint vector, calculate the correction ratio to generate the gear attenuation compensation coefficient. Specifically, this includes: S6.1: The device-level thermal response fingerprint vector stored in the local non-volatile memory is parsed and processed to extract the steady-state slope decay rate characteristic parameter and the second derivative zero-crossing point offset characteristic parameter, which characterize the time-series response difference between the heating tube surface temperature and the pot bottom center temperature, in order to obtain the thermal response hysteresis index that quantifies the thermal transfer delay characteristics under the current pot load.

[0112] The device-level thermal response fingerprint vector stored in local non-volatile memory is decrypted and read to obtain standard thermal behavior identification data containing seventeen-dimensional temporal features, which serves as the input source for subsequent parsing.

[0113] The feature set related to steady-state heat transfer in the standard thermal behavior identification data is filtered and calculated to lock the steady-state slope decay rate parameter that characterizes the temperature change trend of the heating tube surface, so as to reflect the rate of thermal energy decay under continuous low power input conditions.

[0114] The feature set related to transient overheating in the same dataset is filtered and processed to extract the parameter that characterizes the zero-crossing shift of the second derivative of the temperature change curve at the bottom of the pot, which is used to characterize the delay at the moment of change of the sign of the temperature acceleration.

[0115] The steady-state slope decay rate parameter and the second derivative zero-crossing offset parameter are dimensionally normalized to ensure that features of different units and scales can be comprehensively analyzed within a unified computational framework.

[0116] A linear combination operator is used to construct a heat transfer delay evaluation function to calculate the thermal response hysteresis index. The multidimensional features obtained from the previous step are transformed into a thermal response hysteresis index that can be directly used for subsequent power correction calculations, thereby realizing a quantitative characterization of the heat transfer delay characteristics under the current cookware load.

[0117] For example, in an electric cooker with a rated power of 1200W, the steady-state slope decay rate parameter in the decrypted device-level thermal response fingerprint vector is 0.082, and the second derivative zero-crossing offset parameter is 0.135. After normalization, both are mapped to the [0,1] interval, resulting in a normalized steady-state slope decay rate parameter of 0.63 and a normalized second derivative zero-crossing offset parameter of 0.79. Substituting these parameters into the heat transfer delay evaluation function, the calculated output thermal response hysteresis index is 0.71. This value belongs to a relatively high level of hysteresis in the preset evaluation system. It will be mapped to the corresponding thermal inertia excess energy estimate in the subsequent power correction calculation (S6.2 and subsequent sub-steps) to guide the setting of the gear decay ratio. The verification results show that after applying this hysteresis index to the power descent strategy, the temperature overshoot amplitude is significantly reduced, and the system temperature control stability is greatly improved.

[0118] S6.2: Perform nonlinear mapping operation based on the thermal response hysteresis index, and use a preset thermal inertia-power coupling lookup table to convert the thermal response hysteresis index into a thermal inertia excess energy estimate that characterizes the trend of excess heat accumulation in the next six to twelve seconds, so as to generate a thermal inertia excess energy estimate for guiding power correction.

[0119] The extracted thermal response hysteresis index is initialized as input, and this index is set as the independent variable of the nonlinear mapping operation. The locally stored thermal inertia-power coupling lookup table is called as the core data source of the mapping function.

[0120] The thermal response hysteresis index is processed by interval positioning. The corresponding thermal inertia excess energy segment base value is determined by looking up the discrete hysteresis level index in the table, forming the initial matching node for the mapping operation.

[0121] An interpolation mechanism is introduced based on the matching nodes. For cases where the hysteresis index is between two discrete levels, bilinear or spline interpolation is performed to generate a continuous estimate of excess thermal inertia energy, ensuring the smoothness and accuracy of the mapping output.

[0122] The continuous thermal inertia excess energy estimate is used as a reference value for the cumulative heat increment in the prediction time domain from six to twelve seconds in the future, and nonlinear correction is completed by combining the power output adaptation coefficient in the lookup table.

[0123] In the nonlinear correction process, the coupling equation between thermal inertia energy estimation and power attenuation is adopted to achieve a quantitative conversion from hysteresis index to thermal inertia excess energy estimation.

[0124] Through mapping and correction processing, the thermal response hysteresis index is transformed into a thermal inertia excess energy estimate that characterizes the trend of excess heat accumulation in the future prediction time domain, thus realizing the energy benchmark input for subsequent power correction calculations.

[0125] For example, in an electric cooker with a rated power of 1500W, the thermal response hysteresis index calculated by step S6.1 is 0.42. Looking up the base value in the table, the base value for level 0.4 is 220J, and the base value for level 0.5 is 290J. Using bilinear interpolation, f(δ) = 248J is obtained. The cooker's current power setting is 900W, corresponding to a power coupling correction coefficient C(P) = 1.075, resulting in an estimated thermal inertia excess energy of 266.6J. This value will be directly used as the input for the energy equivalence conversion processing in S6.3. In actual testing, the power setting attenuation compensation calculation using this estimate makes the actual output power physically lower than the nominal power setting value by approximately 6.3%, demonstrating a significant improvement in temperature control accuracy and suppression of temperature overshoot.

[0126] S6.3: Based on the estimated excess thermal inertia energy and the target nominal level value of the current power reduction stage, perform energy equivalent conversion processing, and use the specific heat capacity integral inversion algorithm to calculate the absolute value of power reduction required to offset the estimated excess thermal inertia energy, so as to generate the absolute value of power reduction.

[0127] Using the thermal inertia excess energy estimate obtained from the previous steps and the target nominal level value of the current power reduction stage as input objects, the thermal inertia excess energy estimate is mapped from the heat dimension to the power dimension to form an equivalent energy dataset that can be directly used for power reduction calculation.

[0128] The equivalent energy dataset is solved in the time domain using the specific heat capacity integral inversion algorithm. Based on the correspondence between the water body heat capacity coefficient and the predicted temperature lag peak, the accumulated excess heat is converted into the absolute value of the power reduction required, and a physical mapping from energy integration to power reduction is established.

[0129] In the specific heat capacity integral inversion process, the excess thermal inertia energy is divided by time and combined; the relationship between heat and mass, specific heat capacity and temperature difference is established to realize the calculation of the absolute value of power reduction.

[0130] Establish a correspondence list between the absolute value of power reduction and the target nominal gear value, and output the absolute value of power reduction after energy equivalence conversion as the direct input for the next gear power correction.

[0131] Through the above chain processing method, the estimated excess thermal inertia energy in the previous step is transformed into a reduction absolute value that conforms to physical dimensions and can be directly used for power control, thereby realizing quantitative compensation for thermal inertia delay in the power descent stage.

[0132] S6.4: By dividing the absolute value of the power reduction by the target nominal gear value of the current power reduction stage, a normalization ratio is calculated to obtain a relative correction ratio that represents the actual output power needing to be lower than the nominal gear value by 3% to 9%, in order to generate a preliminary gear attenuation compensation coefficient.

[0133] S6.5: Apply safety boundary constraint verification to the preliminary gear attenuation compensation coefficient to determine whether it is within the preset effective compensation range of 3% to 9%. If it exceeds the range, cut it off to the boundary value and re-standardize it. Finally, output the determined gear attenuation compensation coefficient for subsequent power command adjustment.

[0134] Under the condition that the initial gear attenuation compensation coefficient is obtained by normalization ratio calculation, the coefficient is used as the input object for safety boundary verification processing. The preset upper and lower limit values ​​of the effective compensation range are called and loaded into the comparator to establish the benchmark for range determination.

[0135] An interval inclusion test is performed on the initial gear attenuation compensation coefficient. A logical comparison function is used to calculate the amplitude difference between the coefficient and the lower limit of 3% and the upper limit of 9%, respectively, to generate a symbolic quantity representing whether the coefficient exceeds or falls below the boundary.

[0136] For cases where the sign value is positive and exceeds the upper limit, a truncation operation is applied, directly replacing the initial compensation coefficient with the upper limit value of 9%. For cases where the sign value is negative and below the lower limit, a reverse truncation operation is applied, replacing the initial compensation coefficient with the lower limit value of 3%.

[0137] The truncated compensation coefficients are subjected to a standardized mapping operation using a normalization formula, which maps the coefficients to a standardized proportional space from zero to one.

[0138] The standardized compensation coefficients are loaded into the register of the power output correction module so that they can be directly called during subsequent power command scaling, thus achieving synchronization with the power transition safety window parameters.

[0139] By using safety boundary constraints and standardized processing methods, the initial gear attenuation compensation coefficient from the previous step is transformed into a deterministic coefficient that meets the effective range of 3% to 9% and has proportional compatibility, thereby achieving real-time and accurate correction of output power during the power descent stage.

[0140] For example, in the power reduction phase of a certain model of smart electric cooker, the target nominal setting value is 850W. The initial setting attenuation compensation coefficient calculated in the previous steps is 0.025. After interval inclusion verification, it is determined to be 3% below the lower limit, so it is truncated to 0.03. The effective interval lower limit is set to 0.03 and the effective interval upper limit is set to 0.09. Substituting into the standardization formula, the calculated standardization coefficient is 0.0, indicating that the coefficient is at the lower limit boundary and there is no proportional space offset. This coefficient is written to the power correction module register, and power output adjustment is performed. The actual output power value is 824.5W. During the verification phase, the water temperature rise rate is maintained at 0.28℃ / s and there is no temperature overshoot when approaching the target temperature, and the system operates stably.

[0141] Step S7: When the difference between the current temperature and the target temperature is detected to be less than or equal to eight degrees Celsius and the heating rate is less than 0.3 degrees Celsius per second, the switching start time is determined in conjunction with the power transition safety window parameter, and the target power value is adjusted using the gear attenuation compensation coefficient to perform a power descent phase switching action without overshoot risk. Specifically, this includes: S7.1: When it is detected that the current temperature of the electric cooker heating element is less than the target temperature difference and the current heating rate of the electric cooker heating element is less than the target heating rate, the switching start time is determined in combination with the power transition safety window parameter, and the target power value is adjusted by using the gear attenuation compensation coefficient to perform the power reduction phase switching action.

[0142] A unified clock reference acquisition buffer is established for the real-time water temperature timing signal of the electric cooker controller to ensure the time consistency of subsequent difference and differential operations.

[0143] The target temperature setpoint is loaded into the comparison register of the arithmetic unit, and point-by-point difference calculation is performed on the water temperature time sequence signal to generate a difference sequence representing the instantaneous temperature difference for critical interval determination.

[0144] A sliding window differential algorithm is applied to the difference sequence to calculate the heating rate time series curve under the conditions of a window length of 1.5 seconds and a step size of 0.5 seconds, ensuring that the rate calculation has noise resistance and transient response sensitivity.

[0145] The heating rate time series curve is logically compared with the preset rate threshold of 0.3℃ / s to form a Boolean sequence that either meets or does not meet the critical slow start rate condition.

[0146] Perform a logical AND operation on the difference sequence of instantaneous temperature difference and the Boolean sequence of rate. If the temperature difference is not greater than 8℃ and the heating rate is not greater than the threshold, the output condition discrimination flag is true; otherwise, the discrimination flag is false.

[0147] By using the above-mentioned method of combining difference and rate judgment, the real-time temperature acquisition results from the previous step are transformed into a clear critical slow-start interval condition discrimination flag, thus providing the triggering basis for the power descent phase switching action.

[0148] For example, in an electric cooker with a rated capacity of 2 liters, the MCU collects the water temperature of the pot body at 74.8℃, and the target temperature setting is 82.0℃. A 6-second acquisition buffer is established to store the water temperature data. The difference calculation yields a current temperature difference of 7.2℃. The sliding window length is set to 1.5 seconds with a step size of 0.5 seconds, and the calculated heating rate is 0.21℃ / s. The rate threshold is set to 0.3℃ / s and compared, resulting in a Boolean sequence of all true values. The difference condition (temperature difference ≤ 8℃) and the rate condition are jointly judged, and the output condition discrimination flag is set to true. In the same device, when the water temperature is 79.5℃ and the target temperature setting is 82.0℃, the temperature difference is 2.5℃. The heating rate, calculated using the same process, is 0.28℃ / s, which is still below the threshold, and the final output condition discrimination flag is set to true. This discrimination result can accurately identify the critical soft-start interval in multiple cooking processes, significantly improving the trigger accuracy of subsequent power switching window matching and power correction.

[0149] S7.2: Based on the operating condition discrimination flag, trigger the time window matching mechanism to perform interval inclusion verification between the current system clock reference and the start and end times in the power transition safety window parameters generated in the previous steps, so as to generate a valid time permission token that indicates that the power level switching action is allowed to be performed.

[0150] Based on the operating condition discrimination flag input, the discrimination result is used as the trigger condition for the time window matching mechanism to start the pairing retrieval operation of the system clock reference and power transition safety window parameters.

[0151] The start and end times in the power transition safety window parameters are converted into numerical formats with the same precision as the current system clock reference, and interval mapping is performed within a unified time scale to eliminate comparison errors caused by time base differences.

[0152] An interval inclusion check algorithm is used to determine whether the current system clock reference is within the time closed interval defined by the start and end times of the power transition safety window parameter; if the detection condition is met, the token generation logic is entered.

[0153] Using the token generation module, a valid time-limited permission token is constructed based on the inclusion verification result, which identifies the power level switching action. The token state is set to executable and maintained until the end of the safety window to ensure the continued validity of the instruction.

[0154] The generated valid time-permit tokens are subjected to integrity signing and state locking processing. Encryption is used to prevent the tokens from being mixed with those triggered by unauthorized conditions, ensuring that the power descent phase switching action is accurately implemented within the time safety window.

[0155] By using a time window matching mechanism and an interval inclusion verification method, the operating condition judgment result of the previous step is transformed into a valid time permission token that can be used to trigger the power level switching action, thereby achieving safe control of the power descent start timing.

[0156] S7.3: In response to the valid time permission token, read the target nominal gear value of the current power descent stage and the gear attenuation compensation coefficient generated in the previous step, and use the multiplier to perform numerical scaling operation to calculate the actual output power command value after thermal inertia correction, so as to generate a corrected power setpoint that eliminates the effect of thermal hysteresis.

[0157] Upon receiving a valid time permission token as the trigger signal for power switching, the target nominal gear value of the current power descent stage is read as the first input parameter and cached in the arithmetic unit. The gear attenuation compensation coefficient output from the previous step S6.5 is read as the second input parameter and cached in the same arithmetic unit. The first and second input parameters are fed into a multiplier for numerical scaling to form a candidate output power value corrected for thermal inertia hysteresis. This candidate value is then normalized to ensure that the calculation accuracy remains consistent with the subsequent control signal generation chain. The actual output power command value corrected for thermal inertia is obtained by multiplication. Upper and lower limit constraints are checked on the multiplier's calculation result to avoid abnormal power commands due to calculation overflow. The result that passes the check is output as the corrected power setpoint to the power control register, forming direct input data for subsequent pulse width modulation mapping processing. Through the above numerical scaling and parameter verification processes, the result of the previous step is transformed into accurate power setpoint data that can directly drive hardware execution, achieving synchronous fusion of thermal inertia compensation and power descent actions.

[0158] For example, during a certain cooking cycle, when the valid time permit token trigger condition is met, the controller reads the target nominal power level value of 620W for the current power reduction phase and the power attenuation compensation coefficient calculated in the previous steps as 0.94. 620W and 0.94 are fed into a multiplier to obtain a corrected power setpoint of 582.8W. Upper and lower limit constraints are checked on 582.8W; this value is within the preset power range and safety margin, so no truncation is required. 582.8W is written to the power control register and then used as input for pulse width modulation duty cycle mapping, resulting in a drive signal sequence with a duty cycle of 58.28%. After hardware execution, the actual output power of the heating element in the next power level decreases to 582.8W, significantly improving temperature control accuracy and effectively suppressing temperature overshoot caused by thermal hysteresis. During verification, the temperature peak deviation from the target temperature was reduced to within 1.2℃, and system stability was significantly improved.

[0159] S7.4: Perform pulse width modulation signal mapping processing on the corrected power setpoint and convert it into a control duty cycle sequence for driving the power switching device of the heating tube, so as to generate a low-level hardware drive pulse waveform with anti-overshoot characteristics.

[0160] S7.5: Based on the underlying hardware drive pulse waveform, execute the power reduction stage switching action to control the heating tube to smoothly transition from the current high power level to the next low power level, so as to complete the power reduction control closed-loop operation without the risk of temperature overshoot.

[0161] Furthermore, the present invention also includes step S8: after the power descent phase switching is completed, the deviation between the actual change trajectory of the pot body water temperature and the predicted peak point of temperature lag caused by thermal inertia is continuously monitored. If the deviation exceeds a preset fault tolerance threshold, the weight coefficients of the device-level thermal response fingerprint vector are updated to complete the adaptive closed-loop optimization of the thermal response model. Specifically, this includes: S8.1: Perform high-frequency time-series sampling processing on the actual temperature change trajectory of the boiler body after the power reduction phase switching is completed, and perform spatiotemporal alignment mapping operation in combination with the predicted temperature lag peak points caused by thermal inertia generated in the previous steps to obtain a real-time deviation monitoring dataset containing timestamp matching relationships.

[0162] S8.2: Calculate the root mean square error value between the actual temperature curve and the theoretical envelope based on the real-time deviation monitoring dataset, and logically compare the root mean square error value with the preset fault tolerance threshold to generate a deviation over-limit trigger signal that indicates the degree of model mismatch.

[0163] S8.3: In response to the deviation exceeding the limit trigger signal, the recursive least squares algorithm is used to perform sensitivity analysis on the seventeen-dimensional time-series feature vector in the device-level thermal response fingerprint vector to extract the key thermal response hysteresis correction factor that causes the prediction deviation.

[0164] S8.4: Construct an adaptive gain adjustment function based on the key thermal response hysteresis correction factor, and perform iterative update operations on the weight coefficients of the device-level thermal response fingerprint vector stored in the local non-volatile memory to generate an updated device-level thermal response fingerprint vector with self-learning capability.

[0165] The input conditions include a key thermal response hysteresis correction factor obtained by the recursive least squares algorithm in step S8.3, and the existing weight coefficients of the device-level thermal response fingerprint vector stored in local non-volatile memory. Based on this correction factor, the adaptive gain adjustment function construction module is invoked, employing a piecewise linear-exponential hybrid mapping strategy to determine the amplification or attenuation ratio of the gain adjustment curve within different weight ranges. The correction factor is input as an independent variable to the gain adjustment curve calculation module to obtain the corresponding gain coefficient value, and a safety boundary constraint is applied to ensure that the gain coefficient does not introduce system oscillation risk. This gain coefficient is used to iteratively update the weight coefficients of the original thermal response fingerprint vector, using an exponential smoothing weight update method. The above iterative operation is performed on the weights of each temporal feature dimension, and the norm of the weight coefficient matrix is ​​calculated in real time during the update process. If the norm exceeds a preset stability threshold, the global gain coefficient is adjusted to maintain the updated fingerprint vector within a numerically stable range. All updated weight coefficients are recombined into a new seventeen-dimensional thermal response fingerprint vector, outputting a self-learning updated version for subsequent inference. By using an iterative update process based on an adaptive gain adjustment function, the results of the previous step are transformed into optimized fingerprint data containing key hysteresis correction information, enabling the thermal response model to continuously adapt to individual differences.

[0166] For example, in a practical application, the key thermal response hysteresis correction factor was resolved to 0.0045, the value of the 5th dimension in the original weight coefficient matrix was 0.62, and the weight adjustment of the 5th dimension was mapped to 0.013. The correction factor was input into the gain adjustment function, resulting in a gain coefficient of 0.85, and the new weight coefficient of the 5th dimension was calculated to be 0.63105. The same update operation was performed on the seventeen-dimensional weight coefficients, and the 2-norm of the weight vector was monitored in each iteration during the calculation, maintaining it within 2.5. After the update, the newly generated thermal response fingerprint vector was reloaded into the inference algorithm. The verification results showed that the deviation range of the predicted peak temperature hysteresis time was significantly reduced during the subsequent power descent process, and the system temperature control stability was improved.

[0167] S8.5: Reload the updated device-level thermal response fingerprint vector into the input of the embedded lightweight inference algorithm, replace the original reference parameters to complete the adaptive closed-loop optimization of the thermal response model, and ensure that the calculation accuracy of the subsequent power transition safety window parameters and gear attenuation compensation coefficient meets the overshoot-free control requirements.

[0168] The updated device-level thermal response fingerprint vector is loaded into the cache of the embedded lightweight inference algorithm as input, and feature dimension consistency verification is performed during the loading process to ensure that the seventeen-dimensional time-series feature parameters are completely matched with the format of the algorithm's preset input template.

[0169] The updated device-level thermal response fingerprint vector is loaded and a weight coefficient replacement operation is performed. Each feature weight in the original baseline parameter set is replaced with the corresponding new weight to eliminate the impact of model bias accumulation on prediction accuracy.

[0170] After the weight coefficients are replaced, the algorithm's internal state reset instruction is triggered to clear the intermediate state variables and recursive cache remaining from the previous inference process, so as to avoid mixed calculations between old and new parameters.

[0171] The initialization interface of the thermal inertia dynamic prediction module is called, the updated fingerprint vector is used as the initial thermal behavior feature input, and an idle inference verification is performed based on the current operating condition and the matching label. The mean square error and stability index of the predicted temperature rise trajectory are calculated to confirm the effectiveness of the new parameters under the condition of no overshoot prediction.

[0172] Based on the verification results, the time step and power correction resolution of the algorithm are adjusted to optimize the prediction delay and peak advance to meet the accuracy requirements for calculating the power transition safety window and gear attenuation compensation coefficient.

[0173] By using the above processing method, the results of the previous step are transformed into thermal response feature data that can be directly used for inference of the updated model, thereby ensuring the stability of the calculation accuracy of the subsequent power transition safety window parameters and gear attenuation compensation coefficient.

[0174] For example, in a 2.5L electric cooker, the updated device-level thermal response fingerprint vector contains seventeen parameters, including a steady-state slope decay rate of 0.042, a second derivative zero-crossing offset of 1.8 seconds, and a heat capacity coefficient of 4200J / ℃. The embedded lightweight inference algorithm's input template has a dimension setting and cache capacity of 17. During the consistency check during loading, all feature values ​​match the template requirements. The weight coefficient replacement operation replaces the original decay rate weight of 0.038 with 0.042 and the second derivative offset weight of 1.6 seconds with 1.8 seconds. After executing the state reset command, the algorithm's internal prediction delay cache is cleared, ensuring the new parameters act independently. The thermal inertia dynamic prediction module initialization interface is called, and the updated fingerprint vector is input. The no-load inference verification results show the mean square error between the predicted temperature rise trajectory and the test trajectory. Based on the verification results, the algorithm's time step is optimized from 250ms to 200ms, and the power correction resolution is improved from 0.5W to 0.25W. Ultimately, the calculation accuracy of the power transition safety window and gear attenuation compensation coefficient under the new parameter conditions is significantly improved, and the subsequent descent control process remains free of overshoot risk during long-term operation, while the temperature control stability is greatly improved.

[0175] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this invention.

[0176] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains. The terms “first,” “second,” “third,” and similar terms used in this patent application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an” or “a” and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms “comprising” or “including” and similar terms mean that the element or object preceding “comprising” or “including” encompasses the element or object listed following “comprising” or “including” and its equivalents, and do not exclude other elements or objects. The “multiple” mentioned in the embodiments of this application refers to two or more. A and / or B indicate three possibilities: A; B; and A and B.

[0177] The above description is merely an exemplary embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and such modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for controlling the heating element power of an electric cooker with a gradual power reduction type to prevent temperature overshoot, characterized in that, Includes the following steps: S1: Collect the time-series signals of the surface temperature of the heating element of the electric cooker, the time-series signals of the temperature at the center of the bottom of the pot, and the time-series signals of the slight change in water vapor pressure to obtain the original multidimensional sensor dataset. S2: Perform feature extraction processing on the original multidimensional sensor dataset to parse out the time-series feature vector in order to construct a device-level thermal response fingerprint vector; S3: Encrypt and store the device-level thermal response fingerprint vector, and combine it with the initial heating rate and pressure change slope to jointly determine the current cookware load status, so as to generate a working condition matching tag; S4: Based on the operating condition matching label, call the device-level thermal response fingerprint vector, and use the embedded inference algorithm to deduce the temperature rise envelope to generate dynamic prediction results of thermal inertia; S5: Based on the thermal inertia dynamic prediction results, calculate the time permissible interval that avoids the predicted temperature acceleration rise segment and ends before the predicted peak value, so as to generate power transition safety window parameters. S6: Calculate the correction ratio based on the device-level thermal response fingerprint vector to generate the gear attenuation compensation coefficient; S7: When it is detected that the current temperature of the electric cooker heating element is less than the target temperature difference and the current heating rate of the electric cooker heating element is less than the target heating rate, the switching start time is determined by combining the power transition safety window parameter, and the target power value is adjusted by using the gear attenuation compensation coefficient to perform the power reduction stage switching action.

2. The method for controlling the heating element power of an electric cooker with a gradual power reduction and slow start to prevent temperature overshoot, as described in claim 1, is characterized in that... The following step after S7 is: S8: After the power descent phase switching is completed, continuously monitor the deviation between the actual change trajectory of the water temperature in the pot and the predicted temperature lag peak point caused by thermal inertia. If the deviation exceeds the preset fault tolerance threshold, update the weight coefficient of the device-level thermal response fingerprint vector.

3. The method for controlling the heating element power of an electric cooker with a gradual power reduction and slow start to prevent temperature overshoot, as described in claim 1, is characterized in that... The S4 specifically includes: Based on the operating condition matching tag, the device-level thermal response fingerprint vector is read from the local memory, and multi-dimensional feature decoupling processing is performed on the device-level thermal response fingerprint vector to separate the first feature subset characterizing the heat capacity characteristics of the heating tube and the second feature subset characterizing the thermal resistance characteristics of the pot body, thereby generating a decoupled thermal behavior feature parameter set. The discrete-time domain heat transfer state equation is constructed using the decoupled thermal behavior characteristic parameter set, and the current real-time collected time series signal of the bottom center temperature is input into the discrete-time domain heat transfer state equation as the initial state variable to iteratively calculate the theoretical temperature rise rate at the next moment, thereby generating an instantaneous heat flow driving vector containing the thermal inertia accumulation effect. Based on the instantaneous heat flow driving vector, a forward time step extrapolation calculation is performed, and the temperature increment caused by the release of residual heat from the heating tube is continuously superimposed in the prediction time domain of six to twelve seconds to simulate the natural temperature rise trajectory under the condition of no external power input, thereby generating the original temperature rise sequence data describing the temperature change trend. The original temperature rise sequence data is subjected to extreme point search and curvature analysis to identify the inflection point position in the sequence, so as to determine the arrival time of the highest temperature and the maximum temperature rise under the action of thermal inertia, thereby generating the coordinate information of the predicted temperature lag peak point caused by thermal inertia. Based on the coordinate information of the peak temperature lag caused by the predicted thermal inertia, the starting and ending boundaries of the temperature acceleration rise interval are defined. Combined with the preset safety margin time threshold, the time when power switching is allowed is deduced in reverse to define the operation period to avoid overshoot. The result is a dynamic prediction of the thermal inertia, including the peak temperature lag caused by the predicted thermal inertia and the transition window.

4. The method for controlling the heating element power of an electric cooker with a gradual power reduction and slow start to prevent temperature overshoot, as described in claim 1, is characterized in that... Specifically, S5 includes: performing first-order derivative calculation on the temperature rise envelope in the thermal inertia dynamic prediction results to obtain a time-series curve of temperature change rate characterizing the instantaneous rate of change of the water temperature in the pot, which serves as input reference data for identifying the accelerated temperature rise segment.

5. The method for controlling the heating element power of an electric cooker with a gradual power reduction and slow start to prevent temperature overshoot, as described in claim 4, is characterized in that... Specifically, S5 also includes: executing a second derivative zero-crossing detection algorithm based on the temperature change rate time series curve to analyze the critical moment when the temperature change rate changes from increasing to decreasing, thereby determining the starting and ending boundary points of the predicted temperature acceleration rise segment.

6. The method for controlling the heating element power of an electric cooker with a gradual power reduction and slow start to prevent temperature overshoot, as described in claim 5, is characterized in that... S5 specifically includes: performing time backtracking calculation processing based on the time coordinate of the peak point of temperature lag caused by predicted thermal inertia, so as to generate a power switching forced termination time located in the range of 1.8 seconds to 3.2 seconds before the predicted peak, as the right boundary constraint condition of the time permit interval.

7. The method for controlling the heating element power of an electric cooker with a gradual power reduction and soft start to prevent temperature overshoot, as described in claim 6, is characterized in that... S5 specifically includes: performing an interval intersection operation using the termination boundary point of the predicted temperature acceleration rise segment and the forced termination time of power switching to construct an original set of time-permitted intervals that avoid the temperature acceleration rise segment and meet the safety margin requirements before the peak.

8. The method for controlling the heating element power of an electric cooker with a gradual power reduction and slow start to prevent temperature overshoot, as described in claim 7, is characterized in that... Specifically, S5 also includes: validating and encapsulating the original set of time permission intervals to generate power transition safety window parameters containing clear start and end times, which serve as the basis for the control module to execute action instructions.

9. The method for controlling the heating element power of an electric cooker with a gradual power reduction and slow start to prevent temperature overshoot, as described in claim 1, is characterized in that... S6 specifically includes: parsing the device-level thermal response fingerprint vector stored in the local memory, extracting the steady-state slope decay rate characteristic parameter and the second derivative zero-crossing point offset characteristic parameter that characterize the time-series response difference between the heating tube surface temperature and the pot bottom center temperature, so as to obtain a thermal response hysteresis index that quantifies the thermal transfer delay characteristics under the current pot load.

10. The method for controlling the heating element power of an electric cooker with a gradual power reduction and slow start to prevent temperature overshoot, as described in claim 9, is characterized in that... S6 specifically includes: performing a nonlinear mapping operation based on the thermal response hysteresis index, and using a preset thermal inertia-power coupling lookup table to convert the thermal response hysteresis index into a thermal inertia excess energy estimate that characterizes the trend of excess heat accumulation in the next six to twelve seconds, so as to generate a thermal inertia excess energy estimate for guiding power correction.

Images