A method for optimizing thermal data of a food storage cabinet

Abstract

This invention belongs to the field of data optimization technology. It discloses a method for optimizing the insulation data of a food storage cabinet, comprising: acquiring the food insulation data of the target food storage cabinet; performing data compensation on the acquired food insulation data to obtain corresponding initial insulation data, and decomposing and optimizing the corresponding initial insulation data to obtain corresponding optimized insulation data; performing short-time Fourier transform spectrum analysis on the obtained optimized insulation data, and adaptively adjusting the food storage cabinet based on the spectrum analysis results and in conjunction with the centroid error and error duration of the food insulation data. This invention achieves optimized processing and adaptive adjustment of the food storage cabinet insulation data, improving the accuracy and efficiency of the insulation process.

Description

Technical Field

[0001] This invention relates to the field of data optimization technology, and more specifically, to a method for optimizing the insulation data of a food storage cabinet. Background Technology

[0002] With the rapid development of modern society and the continuous improvement of people's living standards, the demand for food storage equipment is growing. Especially in the catering industry and home kitchens, there is a pressing need for storage cabinets that can efficiently and conveniently keep food fresh. As an important type of food storage equipment, the heat preservation performance of food storage cabinets is directly related to the preservation quality and safety of food.

[0003] However, existing food storage cabinets have some shortcomings in terms of heat preservation performance. Firstly, existing food storage cabinets have certain limitations in heat preservation effect and cannot be intelligently adjusted according to the characteristics of different food items, resulting in problems such as high energy consumption and unsatisfactory heat preservation effect. In view of this, the present invention proposes a method for optimizing the insulation data of food storage cabinets to solve the above problems. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: A method for optimizing the insulation data of a food storage cabinet includes: Step 1: Based on the pre-deployed data acquisition unit, data is collected from each storage sub-cabinet in the food storage cabinet to obtain food insulation data, which includes temperature data and humidity data; Step 2: Perform hybrid neural network compensation processing on the acquired food insulation data, and generate compensated initial insulation data by combining the inherent error characteristics and dynamic error characteristics of the data acquisition unit; Step 3: Perform wavelet packet decomposition optimization on the initial insulation data, apply adaptive thresholding to the wavelet packet coefficients during the optimization process, and then perform wavelet reconstruction to obtain the optimized insulation data; Step 4: Perform short-time Fourier transform spectrum analysis on the optimized insulation data; based on the spectrum analysis results, and combined with the centroid error and error duration of the food insulation data, make adaptive adjustments to the food storage cabinet.

[0005] Furthermore, the process of obtaining food temperature control data includes: A data acquisition unit is set up, which consists of several sensor terminal groups, including temperature sensor terminals and humidity sensor terminals. The data acquisition unit is deployed in each storage sub-cabinet of the target food storage cabinet, and data is collected from the corresponding storage sub-cabinets based on the deployed data acquisition unit to obtain the corresponding food insulation data. The food insulation data includes temperature data and humidity data.

[0006] Furthermore, the process of obtaining initial insulation data includes: Acquire the collected food insulation data and input it into the pre-built data compensation model to obtain the data compensation food insulation data and record it as the initial insulation data; The construction process of the data compensation model includes: The error behavior of each sensor terminal deployed in the storage sub-cabinet is characterized to obtain the error feature vector corresponding to each sensor terminal; the error feature vector includes inherent error features extracted based on the physical characteristics of the device and dynamic error features extracted based on the operating environment status; Acquire the sensing information corresponding to the temperature sensing terminal and humidity sensing terminal in the deployed data acquisition unit. The sensing information includes sensing type, sensing accuracy, etc. Based on the sensor information, select several identical temperature and humidity sensors; and deploy them, along with a high-precision thermometer and a high-precision hygrometer, in a pre-constructed test space. Design a test sequence based on the actual workflow of the storage cabinet; conduct sensor terminal tests based on the test sequence, and collect sensor terminal readings and corresponding device readings from high-precision thermometers and high-precision hygrometers; form a multi-condition error sample set; Repeated tests were conducted based on the test space, and several batches of sensing terminals were selected to finally form an error dataset containing samples at several time points. Based on the obtained error dataset, corresponding training data is constructed, which consists of several training samples; The backbone network of the corresponding mathematical compensation model is defined as a hybrid neural network architecture, which includes three core components: a temporal feature encoding module, an error feature fusion module, and a compensation amount prediction module. Define a loss function for the data compensation model. The loss function consists of three parts: the mean absolute error of temperature compensation, the mean absolute error of humidity compensation, and the deviation loss of temperature and humidity correlation constraints. The three parts are weighted and summed to obtain the loss function. Then, the corresponding mathematical compensation model is iteratively trained based on the constructed training data, and the degree of data loss during the construction of the corresponding data compensation model is obtained based on the loss function and recorded as the loss error. It is compared with the expected error. If the corresponding loss error is not higher than the expected error, the corresponding training process ends and the model parameters are saved. If the loss error is higher than the expected error, the weights of each output weight in the corresponding MLP neural network are updated by back-derivative based on the derta learning rule, and the model training continues to obtain a new loss error. This process is repeated until the corresponding data compensation model meets the requirements.

[0007] Furthermore, the time-series feature module is used to process the historical reading sequence of the sensing terminal within a continuous time window, and to encode the input sequence by using a bidirectional long short-term memory network to obtain a hidden feature vector containing time-series evolution information. The error feature fusion layer is used to deeply fuse the implicit feature vector containing time-series evolution information with the obtained error feature vector; The compensation prediction module is used to output the final temperature compensation value and humidity compensation value through a multi-layer fully connected network based on the fused feature vector.

[0008] Furthermore, optimizing the process of acquiring thermal insulation data includes: The initial insulation data is converted into corresponding insulation data signals using an analog-to-digital converter, and the insulation data signals include temperature data signals and humidity data signals. Construct a set of noise sequences, which consist of several different noise signals that satisfy a standard normal distribution; The noise signals in the corresponding noise sequences are added to the obtained temperature data signals to obtain the corresponding temperature noise sequences, which are composed of several different temperature noise signals. The temperature noise signal is decomposed to obtain the corresponding signal components and residual components; the signal components corresponding to different temperature noise signals at the j-th decomposition are obtained respectively, and the average signal components are calculated to obtain the corresponding temperature signal components; j is an integer. Based on the process of acquiring the temperature signal components, the temperature signal components corresponding to all temperature noise signals are acquired. The corresponding temperature signal component is decomposed using the wavelet packet decomposition algorithm to obtain several wavelet components at different scales, and the wavelet packet coefficients corresponding to the wavelet components are obtained. The wavelet packet coefficients are then combined to perform adaptive thresholding on the corresponding wavelet components, and the wavelet components after adaptive thresholding are reconstructed by wavelet to obtain the corresponding temperature optimization components. Based on the acquisition process of the temperature optimization component, the temperature optimization component corresponding to other temperature signal components in the corresponding temperature data signal is acquired; and based on the reverse process of the signal decomposition, the corresponding temperature optimization component is recombined to obtain the first temperature signal. The system acquires the first temperature signals corresponding to different temperature sensing terminals in the corresponding storage sub-cabinet, and performs weighted summation to obtain the corresponding optimized temperature signal; and performs inverse conversion on the corresponding optimized temperature signal based on the analog-to-digital converter to obtain the corresponding optimized temperature data. Based on the process of acquiring the corresponding optimized temperature data, the humidity data in the corresponding initial insulation data is optimized to obtain the corresponding optimized humidity data. The obtained optimized humidity and temperature data are statistically analyzed to obtain corresponding optimized insulation data.

[0009] Furthermore, the process of performing wavelet decomposition on the corresponding temperature signal components based on the wavelet packet decomposition algorithm to obtain several wavelet components at different scales includes: Define the Haar wavelet as the wavelet basis function corresponding to the wavelet packet decomposition algorithm; perform wavelet decomposition on the corresponding signal component based on the wavelet basis function to obtain several detail components; perform detail optimization processing on the obtained detail components to obtain corresponding detail optimized components; take the maximum value of each element of the obtained detail optimized components and detail components to obtain the corresponding transform components; and upsample the obtained transform components to make them consistent with the length of the corresponding signal component, thus obtaining several wavelet components.

[0010] Furthermore, the process of performing wavelet decomposition on the corresponding signal components based on the wavelet basis functions to obtain several detail components includes: The corresponding temperature signal component is decomposed based on the wavelet basis function to obtain the corresponding detail component; the process of obtaining the corresponding detail component is repeated until the preset number of decomposition layers or the length of the approximation coefficient is less than the set threshold; the corresponding process is stopped, and the obtained detail components are counted; at the same time, the wavelet packet coefficients corresponding to the corresponding detail component are obtained.

[0011] Furthermore, the process of adaptive thresholding of the corresponding wavelet components using wavelet packet coefficients includes: Obtain the wavelet packet coefficients corresponding to the corresponding wavelet component and compare them with a preset adaptive threshold. If the wavelet packet coefficients are less than the adaptive threshold, obtain the wavelet features corresponding to the corresponding wavelet component and reduce the wavelet packet coefficients of the corresponding wavelet component to zero by combining the wavelet features. If the wavelet packet coefficients are greater than the adaptive threshold, reduce the wavelet packet coefficients of the corresponding wavelet component to below the adaptive threshold by combining the corresponding wavelet features. Then, perform wavelet reconstruction on the wavelet component processed by the adaptive threshold to obtain the corresponding temperature optimization component.

[0012] Furthermore, the process of obtaining the adaptive threshold includes: Read the obtained wavelet packet coefficients and construct a threshold range based on them; arbitrarily select a threshold from the threshold range, obtain the threshold error corresponding to the wavelet packet coefficient based on the threshold, and determine whether it is the minimum value. If it is not the minimum value, reselect the threshold and obtain the new threshold error, and so on; if it is the minimum value, use the corresponding threshold as the adaptive threshold corresponding to the wavelet packet coefficient.

[0013] Furthermore, the process of adaptively adjusting the food storage cabinets includes: A segmentation function is constructed to divide the obtained thermal insulation data signal into several overlapping signal segments. The corresponding signal segments are then convolved with the corresponding segmentation function. A Fourier transform is performed on the convolved signal segments to obtain the corresponding local thermal insulation spectrum. The local thermal insulation spectrum includes the local temperature spectrum and the local humidity spectrum. Obtain the spectral characteristics of the corresponding local temperature spectrum, wherein the spectral characteristics include the spectral centroid and the spectral standard deviation; Obtain the desired centroid range, which includes the temperature centroid range and the humidity centroid range; The obtained temperature centroid range is compared with the corresponding spectral centroid. If the spectral centroid is within the temperature centroid range, no other operation is performed; if the spectral centroid is outside the expected centroid range, the corresponding temperature centroid error is obtained. Set a standard deviation threshold, compare the obtained standard deviation of the spectrum, if the standard deviation of the spectrum is less than or equal to the standard deviation threshold, no other operation is performed; if the standard deviation of the spectrum is greater than the standard deviation threshold, mark the time node corresponding to the corresponding standard deviation of the spectrum as the error time node; then, count the error time nodes corresponding to all local temperature spectra to obtain the corresponding temperature error duration. Using the above process for obtaining temperature centroid error and temperature error duration, the humidity centroid error and humidity error duration corresponding to the local humidity spectrum are obtained. Furthermore, based on the obtained humidity centroid error, humidity error duration, temperature centroid error, and temperature error duration, and combined with the PID control algorithm, the required equipment adjustment power is obtained; Then, the obtained equipment adjustment power is fed back to the management center of the corresponding food storage cabinet, and the management center adaptively adjusts the food insulation data in the corresponding storage cabinet based on the received equipment adjustment power.

[0014] The technical effects and advantages of the method for optimizing the insulation data of a food storage cabinet according to the present invention are as follows: 1. This invention constructs a data compensation model to compensate for the collected food insulation data. At the same time, it optimizes the compensated data, effectively reducing the data error rate caused by sensor errors and improving the accuracy of the collected food insulation data.

[0015] 2. By combining advanced technologies such as digital twin technology, wavelet packet decomposition algorithm, and PID control algorithm, intelligent management and optimization of food storage cabinet insulation data are achieved. Through intelligent management, manual intervention can be reduced, work efficiency can be improved, and energy consumption and operating costs can be reduced. Moreover, it can perform precise insulation data control according to the characteristics of food, and can more accurately control the temperature and humidity inside the storage cabinet, thereby better maintaining the freshness and quality of food. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of a method for optimizing the insulation data of a food storage cabinet according to the present invention; Figure 2 This is a schematic diagram of a food storage cabinet insulation data optimization system according to the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1 This embodiment provides a method for optimizing the insulation data of a food storage cabinet, including: Step 1: Based on the pre-deployed data acquisition unit, data is collected from each storage sub-cabinet in the food storage cabinet to obtain food insulation data, which includes temperature data and humidity data; Step 2: Perform hybrid neural network compensation processing on the acquired food insulation data, and generate compensated initial insulation data by combining the inherent error characteristics and dynamic error characteristics of the data acquisition unit; Step 3: Perform wavelet packet decomposition optimization on the initial insulation data, apply adaptive thresholding to the wavelet packet coefficients during the optimization process, and then perform wavelet reconstruction to obtain the optimized insulation data; Step 4: Perform short-time Fourier transform spectrum analysis on the optimized insulation data; based on the spectrum analysis results, and combined with the centroid error and error duration of the food insulation data, make adaptive adjustments to the food storage cabinet.

[0019] It should be further explained that, in the specific implementation process, the process of obtaining food temperature control data includes: A data acquisition unit is set up, which consists of several sensor terminal groups, including temperature sensor terminals and humidity sensor terminals. The data acquisition unit is deployed to each storage sub-cabinet in the target food storage cabinet, and data is collected from the corresponding storage sub-cabinets based on the deployed data acquisition unit to obtain the corresponding food insulation data. The food insulation data includes temperature data and humidity data.

[0020] It should be further explained that, in the specific implementation process, the initial insulation data acquisition process includes: Acquire the collected food insulation data and input it into the pre-built data compensation model to obtain the data compensation food insulation data and record it as the initial insulation data; The process of constructing the data compensation model includes: The error behavior of each sensor terminal deployed in the storage sub-cabinet is characterized. A hierarchical feature extraction method is used to capture features from the corresponding temperature and humidity sensors, resulting in an error feature vector for each sensor terminal. The error feature vector includes inherent error features extracted based on the physical characteristics of the equipment and dynamic error features extracted based on the operating environment. Inherent error features reflect the stability deviations of the sensor due to factors such as manufacturing process, material properties, and installation location. These errors exhibit repeatability and predictability under the same environmental conditions, including key dimensions such as zero-point offset, range linearity, response sensitivity, and hysteresis. These features are obtained through multi-condition comparative testing with high-precision reference equipment in a standard testing environment. Dynamic error features capture the time-varying errors caused by external factors such as accumulated operating time, changes in ambient temperature and humidity, and electromagnetic interference. These errors fluctuate with time and environment, including time-series indicators such as aging drift rate, temperature compensation coefficient, humidity cross-interference degree, and periodic fluctuation amplitude. These features are extracted through statistical analysis of long-term monitoring data of the sensors in actual working environments.

[0021] Acquire the sensing information corresponding to the temperature sensing terminal and humidity sensing terminal in the deployed data acquisition unit. The sensing information includes sensing type, sensing accuracy, etc. Based on the sensor information, select several identical temperature and humidity sensors; and deploy them, along with high-precision thermometers and high-precision hygrometers, in a pre-constructed test space. The test space refers to a twin storage cabinet built by staff based on digital twin technology and combined with storage sub-cabinets, which can be used to simulate the operation process of the corresponding storage sub-cabinets. Furthermore, a test sequence was designed based on the actual workflow of the storage cabinet. The test sequence included the complete working cycle from cold start to steady-state insulation. Sensor terminal tests were conducted based on the test sequence. During the tests, the test space moved according to a preset time curve, simulating various typical working conditions such as heating start-up, temperature rise, steady-state insulation, and cooling. Random door opening events were introduced to simulate user food retrieval operations. Simultaneously, sensor terminal readings and corresponding device readings from high-precision thermometers and high-precision hygrometers were collected. This formed a multi-condition error sample set. Each sample in the error sample set included environmental parameters, raw sensor terminal readings, high-precision reference readings, and error feature labels. The environmental parameters recorded control variables such as the target temperature, target humidity, heating power, and ventilation status of the current test space. The error feature labels were associated with the sensor's working state at that moment, such as whether it was in a rapidly changing phase or affected by external interference. Next, repeated tests were conducted based on the test space, and several batches of sensor terminals were selected to finally form an error dataset containing samples at several time points. Based on the obtained error dataset, corresponding training data is constructed, which consists of several training samples; The backbone network of the corresponding mathematical compensation model is defined as a hybrid neural network architecture, which includes three core components: a temporal feature encoding module, an error feature fusion module, and a compensation quantity prediction module. The temporal feature module processes the historical reading sequence of the sensing terminal within a continuous time window. It encodes the input sequence using a bidirectional long short-term memory network, and the LSTM network is a variant of the recurrent neural network specifically for processing sequence data, capturing long-term dependencies through a gating mechanism. The bidirectional LSTM traverses the sequence in both forward and backward directions. Forward encoding captures the causal information of how the past affects the present, and backward encoding captures the contextual information of future trends. The output is a latent feature vector containing temporal evolution information. The error feature fusion layer is used to deeply fuse the implicit feature vector containing time-series evolution information with the obtained error feature vector. It uses an attention mechanism to dynamically allocate the weight contribution of different features. For example, in the steady-state heat preservation stage, the error feature vector offset is dominant, while in the stage of rapid temperature change, the dynamic response lag is more significant, and the implicit feature vector is dominant. The compensation prediction module outputs the final temperature and humidity compensation values ​​through a multi-layer fully connected network based on the fused feature vectors. The module employs a dual-channel parallel structure, with separate temperature and humidity prediction channels. Each channel contains an independent fully connected layer; the parameters of the temperature channel's fully connected layer are specifically designed to learn the temperature compensation rules, while the parameters of the humidity channel's fully connected layer are specifically designed to learn the humidity compensation rules. This dual-channel design decouples the processing of temperature and humidity, allowing each to focus on the characteristics of its own physical quantity, thus avoiding mutual interference between the two physical quantities in a single output layer of the basic architecture. Furthermore, temperature compensation and humidity compensation are output based on the temperature prediction channel and humidity prediction channel, respectively. After the compensation is output in each of the two channels, a physical coupling constraint check is added to the compensation prediction module. The constraint check calculates the correlation between the calibrated temperature and humidity and compares it with the correlation between the true temperature and humidity in the training data. According to thermodynamic laws, the relative humidity should decrease when the temperature rises in a closed space, and the two are negatively correlated. If the predicted temperature and humidity violate this law, for example, the temperature rises but the humidity also rises, the constraint check will trigger a correction mechanism to adjust the compensation to make the output conform to physical consistency. The loss function of the data compensation model is defined. The loss function consists of three parts: the mean absolute error of temperature compensation, the mean absolute error of humidity compensation, and the deviation loss of temperature and humidity correlation constraints. The three parts are weighted and summed based on preset weight parameters to obtain the loss function. The weight parameters are preset by those skilled in the art based on expert experience. Then, based on the constructed training data, the corresponding mathematical compensation model is iteratively trained, and the degree of data loss during the construction of the corresponding data compensation model is obtained based on the loss function, which is recorded as the loss error; this is compared with the expected error. If the corresponding loss error is not higher than the expected error, the corresponding training process ends and the model parameters are saved; if the loss error is higher than the expected error, then based on... The learning rule is updated by taking the inverse derivative to update the output weights within the corresponding MLP neural network. The corresponding weight update formula is as follows: In the formula, Indicates the first term in the corresponding MLP neural network The weight adjustment parameters for each output weight; This represents a learning efficiency adjustment weight factor, used to speed up training and avoid getting stuck in local minima; Indicates learning efficiency; This indicates partial derivative operations; Indicates learning efficiency; Indicates the first term in the corresponding MLP neural network The weight coefficients corresponding to each output weight; Once the weights are updated, continue training the model to obtain new loss errors, and so on, until the corresponding data compensation model meets the requirements.

[0022] It should be further explained that, in the specific implementation process, the process of optimizing the acquisition of insulation data includes: The initial insulation data is converted into corresponding insulation data signals based on the analog-to-digital converter. The insulation data signals include temperature data signals and humidity data signals. Taking temperature data signals as an example, a set of noise sequences is constructed, which consists of several different noise signals that satisfy a standard normal distribution; The noise signals in the corresponding noise sequences are added to the obtained temperature data signals to obtain the corresponding temperature noise sequences. The temperature noise sequences are composed of several different temperature noise signals; that is, different temperature noise signals are obtained by adding different noise signals to the corresponding temperature data signals. The obtained temperature noise signal is numbered and denoted as . and It is an integer. The total number of temperature noise signals is determined by the total number of noise signals constructed. The temperature noise signal is decomposed to obtain the corresponding signal components and residual components; the corresponding signal decomposition formula is: In the formula, Indicates the first One temperature noise signal; Indicates the first The first temperature noise signal corresponding to the The signal components corresponding to the next decomposition. This represents the i-th temperature noise signal corresponding to the th... The residual components corresponding to the next decomposition; and It is an integer. This represents the total number of signal components obtained (it can also be used to represent the total number of decompositions). The specific size is set in advance by the staff based on actual needs; The signal components corresponding to different temperature noise signals at the j-th decomposition are obtained respectively, and the average signal components are calculated to obtain the corresponding temperature signal components. In the formula, This indicates the first corresponding temperature data signal. One temperature signal component; Furthermore, based on the process of acquiring temperature signal components, the temperature signal components corresponding to all temperature noise signals are obtained; and a corresponding signal feature set is constructed based on them. The obtained temperature signal components are the corresponding signal components after performing signal decomposition operations on the corresponding temperature data signals. Taking any temperature signal component in the signal feature set as an example, the corresponding temperature signal component is decomposed into wavelets based on the wavelet packet decomposition algorithm to obtain several wavelet components at different scales, and the wavelet packet coefficients corresponding to the wavelet components are obtained; the wavelet packet coefficients are combined to perform adaptive thresholding on the corresponding wavelet components, and the wavelet components after adaptive thresholding are reconstructed by wavelets to obtain the corresponding temperature optimized components. Based on the process of acquiring temperature optimization components, the temperature optimization components corresponding to other temperature signal components in the corresponding temperature data signal are acquired; and based on the reverse process of signal decomposition, the corresponding temperature optimization components are recombined to obtain the first temperature signal. The system acquires the first temperature signals corresponding to different temperature sensing terminals in the corresponding storage sub-cabinet, and performs weighted summation to obtain the corresponding optimized temperature signal; and performs inverse conversion on the corresponding optimized temperature signal based on the analog-to-digital converter to obtain the corresponding optimized temperature data; wherein, in the corresponding weighted summation process, the weight factor corresponding to the first temperature signal is determined by the number of failures and the continuous working time of the corresponding sensing terminal. Based on the process of acquiring the corresponding optimized temperature data, the humidity data in the corresponding initial insulation data is optimized to obtain the corresponding optimized humidity data. The obtained optimized humidity and temperature data are statistically analyzed to obtain corresponding optimized insulation data; It should be further explained that, in the specific implementation process, the process of performing wavelet decomposition on the temperature signal components based on the wavelet packet decomposition algorithm to obtain several wavelet components at different scales includes: definition The wavelet is the wavelet basis function corresponding to the wavelet packet decomposition algorithm, based on the preset decomposition scale. Based on wavelet basis functions, wavelet decomposition is performed on the corresponding signal components to obtain several detail components. These detail components are then optimized to obtain corresponding optimized detail components. The optimized detail components and the detail components are then element-wise maximized to obtain the corresponding transform components. Finally, the transformed components are upsampled to match the length of the corresponding signal components, thus obtaining several wavelet components. F is a constant. Detail optimization refers to extracting different phase features of the signal through cyclic translation operations. Maximizing the value preserves significant features in each phase, enhancing the signal's feature representation ability while suppressing random noise. The process of wavelet decomposition of the corresponding signal components based on wavelet basis functions includes: The number of decomposition levels of wavelet basis functions is defined as follows: Then the scaling function and wavelet function of the corresponding wavelet basis functions are: In the formula, and These represent the scaling function and the wavelet function, respectively. For time indexing, This is the index of the filter coefficients; and These represent the detail component coefficients and the approximate component coefficients, respectively. Furthermore, the corresponding temperature signal components are decomposed based on wavelet basis functions to obtain the corresponding detail components. The mathematical formula for the decomposition process is as follows: In the formula, and They represent the obtained first and second digits respectively. Detailed components and approximate components; and These represent the scale parameter and displacement parameter, respectively; where the scale function... Used to determine the width or scaling ratio of wavelet basis functions; displacement parameters Used to determine the position of the wavelet basis function along the time axis; This refers to the expansion of wavelet basis functions at different scales and locations; ; Represents the temperature signal component; , For natural numbers, This represents the total number of wavelet components; The process of acquiring the corresponding detail components is repeated until the preset number of decomposition layers L or the length of the approximation coefficients is less than a set threshold is reached; the corresponding process is stopped, and the acquired detail components are counted; the set threshold is set in advance by those skilled in the art; at the same time, the wavelet packet coefficients corresponding to the corresponding detail components are acquired, wherein the process of acquiring the wavelet packet coefficients is prior art and will not be described in detail in this invention. It should be further explained that, in the specific implementation process, the adaptive thresholding of the corresponding wavelet components based on the wavelet packet coefficients includes: The wavelet packet coefficients corresponding to the corresponding wavelet component are obtained and compared with a preset adaptive threshold. If the wavelet packet coefficients are less than the adaptive threshold, the wavelet features corresponding to the corresponding wavelet component are obtained, including the amplitude and sign of the wavelet packet coefficients. The wavelet packet coefficients of the corresponding wavelet component are then reduced to zero by combining the wavelet features. If the wavelet packet coefficients are greater than the adaptive threshold, the wavelet packet coefficients of the corresponding wavelet component are reduced to below the adaptive threshold (i.e., the wavelet packet coefficients are less than the adaptive threshold) by combining the corresponding wavelet features. The wavelet components processed by the adaptive threshold are then reconstructed by wavelet to obtain the corresponding temperature optimization components. The process of obtaining the adaptive threshold includes: Read the obtained wavelet packet coefficients and construct a threshold range based on them. ); and These are the maximum and minimum values ​​among all wavelet packet coefficients, respectively; Choose any threshold ;in, The function to obtain the target threshold retrieves the corresponding threshold. The corresponding threshold error In the formula, Indicates the first Passing the threshold Processed wavelet packet coefficients; Indicates the first The original wavelet packet coefficients; Indicates threshold The corresponding influence matrix is ​​a diagonal matrix based on the energy of wavelet packet coefficients. It is used to weight the errors of different wavelet packet coefficients during the threshold error calculation process, so that wavelet packet coefficients with larger energy have higher weights in the threshold optimization process. Indicates the first The threshold error corresponding to each original wavelet packet coefficient; and It is an integer. Used to represent the total number of original wavelet packet coefficients; Obtain the threshold error corresponding to the wavelet packet coefficients and determine if it is the minimum value. If it is not the minimum value, reselect the threshold. Calculate a new threshold, and so on; if it is the minimum value, then set the corresponding threshold. As the adaptive threshold corresponding to the wavelet packet coefficients; The process of performing detail optimization on the obtained detail components includes: The obtained detail components are cyclically shifted forward along the time axis to obtain several shifted detail components. The maximum value of each element is taken from the original detail components and the shifted detail components to obtain the corresponding optimized detail components. The number of forward cyclic shifts depends on the actual needs.

[0023] It should be further explained that, in a specific implementation process, one embodiment of the present invention also includes: constructing a two-dimensional rectangular coordinate system with time as the horizontal axis and optimized insulation data as the vertical axis, mapping the corresponding optimized insulation data to the corresponding two-dimensional rectangular coordinate lines, obtaining the corresponding insulation data curve, and storing it in the management platform of the corresponding food storage cabinet. Then, after identity verification, the user can view the insulation data curve corresponding to their own storage sub-cabinet; wherein, identity verification includes at least two of the following: facial recognition, QR code verification, and verification code verification; the device used by the user to view the corresponding insulation data curve can be a personal mobile terminal or a display equipped in the food storage cabinet.

[0024] It should be further explained that the process of adaptively adjusting the food storage cabinets includes: Taking the optimized insulation data corresponding to any storage cabinet as an example, the obtained optimized insulation data is converted to obtain the corresponding insulation data signal; A segmentation function is constructed to divide the obtained thermal insulation data signal into several overlapping signal segments. The corresponding signal segments are then convolved with the corresponding segmentation function. Finally, a Fourier transform is performed on the convolved signal segments to obtain the corresponding local thermal insulation spectrum. The local insulation spectrum is a time-frequency representation, that is, the spectrum diagram corresponding to each time point, used to analyze the frequency component changes of the insulation data signal in different time periods; among them, the local insulation spectrum includes the local temperature spectrum and the local humidity spectrum; Obtain the spectral characteristics of the corresponding local temperature spectrum, including the spectral centroid and spectral standard deviation; The system reads pre-collected images of items, performs image recognition on these images to obtain the types of meals stored by the user, and obtains corresponding expected insulation data based on these meal types. Simultaneously, it can also retrieve the user's storage insulation preferences from the locker's backend. Based on these preferences and the expected insulation data, corresponding desired insulation data is set. (If there is a significant deviation between the expected insulation data and the storage insulation preferences, i.e., a large difference between the retrieved insulation data and the user's desired temperature, the corresponding locker sends insulation setting information to the corresponding user via wired or wireless network. Upon receiving this information, the user can return their desired insulation data via their personal mobile terminal.) Based on the obtained desired insulation data, a corresponding desired centroid range is constructed, including both temperature and humidity centroid ranges. The obtained temperature centroid range is compared with the corresponding spectral centroid. If the spectral centroid is within the temperature centroid range, no further operation is performed; if the spectral centroid is outside the desired centroid range, the corresponding temperature centroid error is obtained. ; In the formula, and These represent the maximum and minimum values ​​of the expected centroid range, respectively; Indicates the centroid of the spectrum; Set a standard deviation threshold, compare the obtained standard deviation of the spectrum, if the standard deviation of the spectrum is less than or equal to the standard deviation threshold, no other operation is performed; if the standard deviation of the spectrum is greater than the standard deviation threshold, mark the time node corresponding to the corresponding standard deviation of the spectrum as the error time node; then, count the error time nodes corresponding to all local temperature spectra to obtain the corresponding temperature error duration. Using the above process for obtaining temperature centroid error and temperature error duration, the humidity centroid error and humidity error duration corresponding to the local humidity spectrum are obtained. Furthermore, based on the obtained humidity centroid error, humidity error duration, temperature centroid error, and temperature error duration, and combined with a PID control algorithm, the required equipment adjustment power is obtained. In the formula, , , These represent the proportional gain matrix, integral gain matrix, and differential gain matrix, respectively. It represents the error matrix composed of humidity centroid error, humidity error duration, temperature centroid error, and temperature error duration; Then, the obtained equipment adjustment power is fed back to the management center of the corresponding food storage cabinet. The management center adaptively adjusts the food insulation data in the corresponding storage cabinet based on the received equipment adjustment power. Through the above process, the power of the equipment can be dynamically adjusted according to real-time temperature and humidity data to achieve more precise environmental control. The mathematical formula for calculating the segmentation function is as follows: In the formula, This indicates the first element within the corresponding insulation data signal. The function value of the data point; This indicates the signal length of the insulation data signal, which is the total number of data points within the insulation data signal. ; and All are integers; The formula for obtaining the spectral centroid is: In the formula, Represents the first value within the corresponding local temperature spectrum. One frequency component; Indicates the first Fourier transform coefficients of the frequency components.

[0025] This invention, by setting up a data acquisition unit, can comprehensively acquire food insulation data within the food storage cabinet; and by using a data compensation model to preprocess the acquired insulation data, it can compensate for errors in the sensing terminal and improve the accuracy of the initial insulation data; further data optimization processing is performed on the corresponding food insulation data, which can effectively reduce data noise and improve the accuracy of the optimized insulation data; based on the optimized insulation data, data analysis is performed, and according to the analysis results, the insulation process of the food storage cabinet can be adaptively adjusted, such as adjusting the equipment power, to achieve more precise environmental control and meet the insulation requirements of different foods.

[0026] Example 2 Please see Figure 2 As shown, for parts not described in detail in this embodiment, please refer to the description in Embodiment 1. A system for optimizing the insulation data of a food storage cabinet is provided, including: Data acquisition module: Based on the pre-deployed data acquisition unit, data is collected from each storage sub-cabinet in the food storage cabinet to obtain food insulation data, which includes temperature data and humidity data; The data processing module is used to perform hybrid neural network compensation processing on the acquired food insulation data, and generate compensated initial insulation data by combining the inherent error characteristics and dynamic error characteristics of the data acquisition unit; the initial insulation data is subjected to wavelet packet decomposition optimization processing, the wavelet packet coefficients in the optimization process are subjected to adaptive threshold processing, and wavelet reconstruction is performed to obtain optimized insulation data. An adaptive adjustment module is used to perform short-time Fourier transform spectrum analysis on the optimized insulation data; based on the spectrum analysis results, and combined with the centroid error and error duration of the food insulation data, the food storage cabinet is adaptively adjusted. The modules are connected via wired and / or wireless means to enable data transmission between them.

[0027] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0028] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0029] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for optimizing the insulation data of a food storage cabinet, characterized in that, include: Step 1: Obtain the food insulation data of the target food storage cabinet; Step 2: Perform data preprocessing on the acquired food insulation data to obtain the corresponding initial insulation data, and then perform data optimization on the initial insulation data to obtain the corresponding optimized insulation data; The process of obtaining the corresponding optimized insulation data includes: The initial insulation data is converted into corresponding insulation data signals using an analog-to-digital converter, and the insulation data signals include temperature data signals and humidity data signals. Construct a set of noise sequences; add the noise signals from the corresponding noise sequences to the obtained temperature data signals to obtain the corresponding temperature noise sequences; The temperature noise signal in the temperature noise sequence is decomposed to obtain the corresponding signal components and residual components; the signal components corresponding to different temperature noise signals at the j-th decomposition are obtained respectively, and the average signal components are calculated to obtain the corresponding temperature signal components; j is an integer. Based on the process of acquiring the temperature signal components, the temperature signal components corresponding to all temperature noise signals are acquired. Define the Haar wavelet as the wavelet basis function corresponding to the wavelet packet decomposition algorithm; decompose the corresponding temperature signal component based on the wavelet basis function to obtain the corresponding detail component; repeat the process of obtaining the corresponding detail component until the preset number of decomposition layers or the length of the approximation coefficient is less than the set threshold; stop the corresponding process and count the obtained detail components; at the same time, obtain the wavelet packet coefficients corresponding to the corresponding detail component. The obtained detail components are subjected to detail optimization processing to obtain corresponding detail optimized components. The maximum value of each element of the obtained detail optimized components and detail components is taken to obtain the corresponding transform components. The obtained transform components are upsampled to make their length consistent with the corresponding signal components, thus obtaining several wavelet components. The wavelet packet coefficients corresponding to the corresponding wavelet components are obtained and compared with a preset adaptive threshold. If the wavelet packet coefficients are less than the adaptive threshold, the wavelet features corresponding to the corresponding wavelet components are obtained, and the wavelet packet coefficients of the corresponding wavelet components are reduced to zero in combination with the wavelet features. If the wavelet packet coefficients are greater than the adaptive threshold, the wavelet packet coefficients corresponding to the wavelet components are reduced to below the adaptive threshold in combination with the corresponding wavelet features. The wavelet components processed by the adaptive threshold are then reconstructed by wavelets to obtain the corresponding temperature optimized components. Based on the acquisition process of the temperature optimization component, the temperature optimization component corresponding to other temperature signal components in the corresponding temperature data signal is acquired; and based on the reverse process of the signal decomposition, the corresponding temperature optimization component is recombined to obtain the first temperature signal. The system acquires the first temperature signals corresponding to different temperature sensing terminals in the corresponding storage sub-cabinet, and performs weighted summation to obtain the corresponding optimized temperature signal; and performs inverse conversion on the corresponding optimized temperature signal based on the analog-to-digital converter to obtain the corresponding optimized temperature data. Based on the process of acquiring the corresponding optimized temperature data, the humidity data in the corresponding initial insulation data is optimized to obtain the corresponding optimized humidity data. The obtained optimized humidity and temperature data are statistically analyzed to obtain corresponding optimized insulation data; The process of obtaining the adaptive threshold includes: Read the obtained wavelet packet coefficients and construct a threshold range based on them; arbitrarily select a threshold from the threshold range, obtain the threshold error corresponding to the wavelet packet coefficient based on the threshold, and determine whether it is the minimum value. If it is not the minimum value, reselect the threshold and obtain the new threshold error, and so on; if it is the minimum value, use the corresponding threshold as the adaptive threshold corresponding to the wavelet packet coefficient. Step 3: Analyze the obtained optimized insulation data to obtain corresponding analysis results, and adaptively adjust the insulation process of the corresponding food storage cabinets based on the analysis results; The process of adaptive adjustment includes: A segmentation function is constructed to divide the obtained thermal insulation data signal into several overlapping signal segments. The corresponding signal segments are then convolved with the corresponding segmentation function. A Fourier transform is performed on the convolved signal segments to obtain the corresponding local thermal insulation spectrum. The local thermal insulation spectrum includes the local temperature spectrum and the local humidity spectrum. Obtain the spectral characteristics of the corresponding local temperature spectrum, wherein the spectral characteristics include the spectral centroid and the spectral standard deviation; Obtain the desired centroid range, which includes the temperature centroid range and the humidity centroid range; The obtained temperature centroid range is compared with the corresponding spectral centroid; if the spectral centroid is outside the expected centroid range, the corresponding temperature centroid error is obtained. Set a standard deviation threshold and compare the obtained standard deviations of the spectrum; if the standard deviation of the spectrum is greater than the standard deviation threshold, mark the time node corresponding to the corresponding standard deviation of the spectrum as the error time node; count the error time nodes corresponding to all local temperature spectra to obtain the corresponding temperature error duration. Using the above process for obtaining temperature centroid error and temperature error duration, the humidity centroid error and humidity error duration corresponding to the local humidity spectrum are obtained. Based on the obtained humidity centroid error, humidity error duration, temperature centroid error and temperature error duration, and combined with the PID control algorithm, the required equipment adjustment power is obtained; Then, the obtained equipment adjustment power is fed back to the management center of the corresponding food storage cabinet, and the management center adaptively adjusts the food insulation data in the corresponding storage cabinet based on the received equipment adjustment power.

2. The method for optimizing the insulation data of the food storage cabinet according to claim 1, characterized in that, The process of obtaining the food temperature control data of the target food storage cabinet includes: A data acquisition unit is set up, which consists of several sensor terminal groups, including temperature sensor terminals and humidity sensor terminals. The data acquisition unit is deployed in each storage sub-cabinet of the target food storage cabinet, and data is collected from the corresponding storage sub-cabinets based on the deployed data acquisition unit to obtain the corresponding food insulation data. The food insulation data includes temperature data and humidity data.

3. The method for optimizing the insulation data of the food storage cabinet according to claim 2, characterized in that, The process of preprocessing the obtained food temperature control data includes: Acquire the collected food insulation data and input it into the pre-built data compensation model to obtain the data compensation food insulation data and record it as the initial insulation data; The construction process of the data compensation model includes: The error behavior of each sensor terminal deployed in the storage sub-cabinet is characterized to obtain the error feature vector corresponding to each sensor terminal; the error feature vector includes inherent error features extracted based on the physical characteristics of the device and dynamic error features extracted based on the operating environment status; Acquire the sensing information corresponding to the temperature sensing terminal and humidity sensing terminal in the deployed data acquisition unit. The sensing information includes sensing type, sensing accuracy, etc. Based on the sensor information, select several identical temperature and humidity sensors; and deploy them, along with a high-precision thermometer and a high-precision hygrometer, in a pre-constructed test space. Design a test sequence based on the actual workflow of the storage cabinet; conduct sensor terminal tests based on the test sequence, and collect sensor terminal readings and corresponding device readings from high-precision thermometers and high-precision hygrometers; form a multi-condition error sample set; Repeated tests were conducted based on the test space, and several batches of sensing terminals were selected to finally form an error dataset containing samples at several time points. Based on the obtained error dataset, corresponding training data is constructed, which consists of several training samples; The backbone network of the corresponding mathematical compensation model is defined as a hybrid neural network architecture, which includes three core components: a temporal feature encoding module, an error feature fusion module, and a compensation amount prediction module. Define a loss function for the data compensation model. The loss function consists of three parts: the mean absolute error of temperature compensation, the mean absolute error of humidity compensation, and the deviation loss of temperature and humidity correlation constraints. The three parts are weighted and summed to obtain the loss function. Then, the corresponding mathematical compensation model is iteratively trained based on the constructed training data, and the degree of data loss during the construction of the corresponding data compensation model is obtained based on the loss function and recorded as the loss error. It is compared with the expected error. If the corresponding loss error is not higher than the expected error, the corresponding training process ends and the model parameters are saved. If the loss error is higher than the expected error, the weights of each output weight in the corresponding MLP neural network are updated by back-derivative based on the derta learning rule, and the model training continues to obtain a new loss error. This process is repeated until the corresponding data compensation model meets the requirements.

4. The method for optimizing the insulation data of the food storage cabinet according to claim 3, characterized in that, The time-series feature encoding module is used to process the historical reading sequence of the sensing terminal within a continuous time window, and to obtain a hidden feature vector containing time-series evolution information by encoding the input sequence using a bidirectional long short-term memory network. The error feature fusion module is used to deeply fuse the implicit feature vector containing time-series evolution information with the obtained error feature vector; The compensation prediction module is used to output the final temperature compensation value and humidity compensation value through a multi-layer fully connected network based on the fused feature vector.

Images