Intelligent temperature control method and system for baking equipment

Abstract

The present invention relates to the technical field of baking equipment and discloses a method and system for intelligent temperature control of baking equipment. The method comprises: real-time acquisition of dough parameters, internal state parameters, and environmental parameters in the baking equipment with timestamps, and synchronously calling regional characteristic parameters and historical data of the baking equipment to form a multi-dimensional original data set including time-modal tags. The present invention accurately captures the actual state changes of dough in different regional environments, providing a detailed basis for the subsequent formulation of temperature control strategies. The generated dynamic control parameter set can accurately adjust the baking temperature according to the real-time regional environment, so that the baking equipment can achieve precise temperature control of the baking process in different regions, ensuring that the dough is heated evenly and consistently, and ultimately significantly improving the quality stability of the baked products, ensuring that baked products can achieve ideal baking effects under different regional climatic conditions, and improving the baking yield rate. The system is suitable for chain bakery workshops.

Description

Technical Field

[0001] The present invention relates to the technical field of baking equipment, and in particular to an intelligent temperature control method and system for baking equipment. Background Art

[0002] In the field of baking equipment, current temperature control methods are mainly based on preset temperature control curves or the use of single sensors for feedback. This static temperature control method is widely used in the industry, and a fixed temperature-time curve is preset based on historical experience. This type of static temperature control solidifies historical experience data and sets the corresponding relationship between temperature and time as a fixed program to achieve the goal of intelligent temperature control.

[0003] Baking equipment is suitable for different baking needs and is often widely used in various scenarios, such as chain bakery workshops. Chain bakery workshops are often located in different cities. When traditional static temperature control methods are used for temperature control, baking is often affected by the climate in different regions, making it difficult to achieve adaptive adjustment of intelligent temperature control during the baking process, ultimately resulting in poor baking results. Summary of the Invention

[0004] In view of the deficiencies in the prior art, the present invention provides an intelligent temperature control method and system for baking equipment to solve the above problems.

[0005] The above technical objectives of the present invention are achieved through the following technical solutions:

[0006] An intelligent temperature control method for baking equipment, comprising:

[0007] S1: Real-time collection of dough parameters, internal state parameters, and environmental parameters within the baking equipment with timestamps, and simultaneous call of the regional characteristic parameters and historical data of the baking equipment to form a multi-dimensional raw data set;

[0008] S2: Analyze the preprocessed multi-dimensional raw data set to generate a dough state evolution feature matrix, and construct time-synchronized real-time multimodal time series data based on the dough state evolution feature matrix;

[0009] S3: Analyze real-time multimodal time series data to generate temperature control parameters, mine the correlation between temperature control parameters, and generate a temperature control strategy matrix;

[0010] S4: Based on the real-time environmental parameters and regional characteristic parameters, the temperature control strategy matrix is ​​modified to generate a modified strategy that adapts to the regional climate characteristics;

[0011] S5: Solve the correction strategy and generate a dynamic control parameter set;

[0012] S6: Analyze the dynamic control parameter set and build an adaptive intelligent temperature control mechanism based on deviation accumulation.

[0013] Furthermore, the pre-processed multi-dimensional raw data set is analyzed to generate a dough state evolution feature matrix, including:

[0014] Extract the preprocessed multi-dimensional original data set to obtain absorption peak features, and form the absorption peak features into an initial feature matrix;

[0015] Calculate the initial feature matrix to generate regression parameters;

[0016] The regression parameters were verified to obtain the surface water activity index;

[0017] The surface water activity index is combined with environmental parameters and dough parameters to generate the dough state evolution characteristic matrix.

[0018] Furthermore, based on the dough state evolution feature matrix, time-synchronized real-time multimodal time series data is constructed, including:

[0019] The temporal correlation analysis of the dough state evolution feature matrix is ​​performed to generate a three-dimensional spatiotemporal data cube containing temporal correlation. Specifically, the time stamps of the dough state feature matrix at each moment are aligned, the dynamic time warping algorithm is used to calculate the temporal similarity matrix of the feature dimension between adjacent matrices, the autocorrelation coefficient matrix is ​​used to represent the cross-time feature dependency, and then the feature matrices with temporal correlation weights are stacked in sequence along the time axis to construct a feature matrix. Number of features A three-dimensional space-time data cube of the number of time steps;

[0020] Dynamically weight the three-dimensional spatiotemporal data cube to generate a temporal correlation structure;

[0021] According to the time series correlation structure, the dough parameters, internal state parameters and environmental parameters are synchronized and processed in a time-aligned manner to construct time-synchronized real-time multimodal time series data.

[0022] Furthermore, the real-time multimodal time series data is analyzed to generate temperature control parameters, including:

[0023] Performing standardization processing on the real-time multimodal time series data to obtain standardized real-time multimodal time series data;

[0024] Perform importance evaluation on the standardized real-time multimodal time series data to obtain the importance score of each parameter;

[0025] The importance threshold is calculated based on the internal state parameters of the baking equipment. Specifically, the kernel density estimation is used to construct the probability of the sliding window data of the internal state parameters of the baking equipment, and the 95% quantile of the distribution function is directly extracted as the importance threshold.

[0026] Compare and filter the importance scores of the parameters with the importance threshold to obtain a set of key parameters;

[0027] The key parameter set is associated with the temperature control parameters of the internal state parameters to generate temperature control parameters.

[0028] Furthermore, the correlation between temperature control parameters is mined to generate a temperature control strategy matrix, including:

[0029] The correlation degree of the temperature control parameters is calculated to obtain the correlation degree distribution characteristic matrix. Specifically, for each type of temperature control parameter, the linear correlation between each two parameters is calculated to obtain the correlation coefficient. By analyzing the correlation coefficient, a parameter correlation degree heat map is drawn to observe the correlation strength and positive and negative correlation distribution between the parameters, and finally a complete correlation degree distribution characteristic matrix is ​​formed.

[0030] According to the correlation distribution feature matrix, the strong correlation threshold is calculated;

[0031] Compare and screen the correlation value in the correlation distribution feature matrix with the strong correlation threshold to obtain a set of strong correlation parameter pairs;

[0032] Construct the initial temperature control strategy matrix based on the set of strongly correlated parameter pairs;

[0033] The initial temperature control strategy matrix is ​​optimized to generate the final temperature control strategy matrix.

[0034] Furthermore, the temperature control strategy matrix is ​​modified according to the real-time environmental parameters and regional characteristic parameters to generate a modified strategy adapted to the regional climate characteristics, including:

[0035] Mining real-time environmental parameters and regional characteristic parameters to construct an initial correlation matrix;

[0036] The initial correlation matrix is ​​combined with the dough water activity data in the real-time dough parameters to obtain the humidity deviation compensation coefficient and the temperature fluctuation compensation gradient;

[0037] The humidity deviation compensation coefficient is integrated with the temperature fluctuation compensation gradient to generate a water activity change curve;

[0038] The temperature control strategy matrix is ​​modified using the water activity change curve to generate a modified strategy that adapts to regional climate characteristics.

[0039] Furthermore, the correction strategy is solved to generate a dynamic control parameter set, including:

[0040] Implement multi-parameter decoupling analysis based on the correction strategy to generate a set of climate adaptability rules;

[0041] By constraining the climate adaptability rule set through geographic feature constraint vectors, a regional adaptive optimization framework is formed;

[0042] Perform global optimization on the regional adaptive optimization framework to generate a candidate set of dynamic compensation parameters;

[0043] The thermodynamic characteristics of the dynamic compensation parameter candidate set are verified to generate a dynamic control parameter set.

[0044] Furthermore, the dynamic control parameter set is analyzed to construct an adaptive intelligent temperature control mechanism based on deviation accumulation, including:

[0045] Extract the dynamic control parameter set to generate a real-time fluctuation feature sequence;

[0046] Determine the quantile baseline based on kernel density estimation in each window of the real-time fluctuation feature sequence and generate a dynamic fluctuation threshold;

[0047] Establish a comparison mechanism between real-time fluctuation values ​​and dynamic fluctuation thresholds, and record abnormalities exceeding the threshold and corresponding deviations;

[0048] The statistical distribution characteristics of the cumulative deviation are calculated based on the statistical distribution algorithm to obtain the warning threshold.

[0049] Furthermore, the dynamic control parameter set is analyzed to build an adaptive intelligent temperature control mechanism based on deviation accumulation, which also includes:

[0050] When the accumulated deviation exceeds the warning threshold, an adaptive intelligent temperature control mechanism based on the accumulated deviation is constructed.

[0051] An intelligent temperature control system for baking equipment, comprising:

[0052] Collection unit: This collects dough parameters, internal state parameters, and environmental parameters of baking equipment in real time with timestamps, and simultaneously calls the regional characteristic parameters and historical data of the baking equipment to form a multi-dimensional raw data set;

[0053] Analysis unit: Analyzes the pre-processed multi-dimensional raw data set to generate a dough state evolution feature matrix, and constructs time-synchronized real-time multi-modal time series data based on the dough state evolution feature matrix;

[0054] Mining unit: Analyzes real-time multimodal time series data to generate temperature control parameters, mines the correlation between temperature control parameters, and generates a temperature control strategy matrix;

[0055] Correction unit: Corrects the temperature control strategy matrix based on real-time environmental parameters and regional characteristic parameters, and generates a correction strategy that adapts to regional climate characteristics;

[0056] Solving unit: solves the correction strategy and generates a dynamic control parameter set;

[0057] Construction unit: Analyze the dynamic control parameter set and build an adaptive intelligent temperature control mechanism based on deviation accumulation.

[0058] In summary, the present invention mainly has the following beneficial effects:

[0059] By real-time collection of multi-dimensional original data sets such as dough parameters, internal state parameters, and environmental parameters in baking equipment, and integrating regional characteristic parameters with historical data, we can comprehensively and dynamically perceive the changes in various factors in the baking process. The dough state evolution characteristic matrix and time-synchronized real-time multimodal time series data constructed based on this will lay a solid foundation for the subsequent generation of accurate temperature control parameters. Further correlation analysis and strategy matrix construction will enable the temperature control strategy to closely fit the actual baking situation. The strategy matrix will then be corrected in combination with real-time environmental parameters and regional characteristic parameters to generate a correction strategy adapted to regional climate characteristics. The final dynamic control parameter set obtained can flexibly adapt to baking needs under different regional climate conditions, realize adaptive adjustment of intelligent temperature control, effectively solve the problem of poor temperature control effect of traditional methods affected by regional climate, and improve the stability and consistency of baking quality.

[0060] By collecting multi-dimensional parameters such as dough, internal equipment state, and external environment in real time with timestamps, the comprehensiveness and accuracy of the data are ensured. Simultaneously accessing regional characteristic parameters and historical data enriches the data context. In subsequent processing, an initial feature matrix is ​​constructed by extracting absorption peak features. The surface water activity index is then calculated and verified, and a dough state evolution feature matrix is ​​generated. This process fully utilizes intelligent data processing techniques to deeply mine dough state information. When constructing time-synchronized, real-time multimodal time series data, methods such as dynamic time warping algorithms are used to accurately capture the temporal correlations between parameters, making data processing more closely aligned with the dynamic characteristics of the actual baking process. For the analysis and correlation mining of temperature control parameters, techniques such as importance assessment and kernel density estimation are used to screen key parameters and generate temperature control parameters. A temperature control strategy matrix is ​​then constructed and optimized. The entire process is highly intelligent, reducing reliance on manual experience and making the generation of temperature control strategies more rational. Furthermore, the construction of an adaptive intelligent temperature control mechanism based on deviation accumulation further enhances the intelligent control capabilities of baking equipment.

[0061] The importance threshold is determined by the probability construction method of kernel density estimation, and the key parameters are screened out and associated with the temperature control parameters to generate temperature control parameters. This series of processes ensures the accuracy of the temperature control parameters. In the process of constructing the temperature control strategy matrix, not only the correlation distribution feature matrix is ​​calculated, but also a heat map is drawn to intuitively present the parameter correlation strength and the positive and negative correlation distribution. Based on this, a strong correlation threshold is set to screen the set of strongly correlated parameter pairs, and the temperature control strategy matrix is ​​constructed and optimized so that the temperature control strategy can accurately reflect the actual correlation between the parameters. At the same time, the temperature control strategy matrix is ​​modified based on real-time environmental parameters and regional characteristic parameters, fully considering the impact of regional climate factors on baking. By constructing the initial correlation matrix and combining it with the dough water activity data, the compensation coefficient and gradient are obtained, and then the water activity change curve is generated for strategy correction. The revised strategy is made closer to the actual situation and the temperature control accuracy is improved. Steps such as multi-parameter decoupling analysis, geographical feature constraint vector constraint and global optimization have generated a strictly verified dynamic control parameter set to ensure the accuracy of temperature control parameters in actual applications. The adaptive intelligent temperature control mechanism based on deviation accumulation can monitor the fluctuation of dynamic control parameters in real time, determine the dynamic fluctuation threshold using kernel density estimation, accurately record over-threshold anomalies and deviations, calculate the cumulative deviation characteristics based on the statistical distribution algorithm to obtain the early warning threshold, and make timely adaptive adjustments when the cumulative deviation exceeds the early warning threshold, further enhancing the accuracy and reliability of the temperature control process, ensuring that baked products can achieve ideal baking effects under different regional climatic conditions, improving the baking yield and quality stability, and is suitable for chain baking workshops. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a flow chart of the intelligent temperature control method for baking equipment of the present invention;

[0063] Figure 2 This is a block diagram of the intelligent temperature control system for baking equipment of the present invention. DETAILED DESCRIPTION

[0064] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0065] refer to Figure 1 , an intelligent temperature control method for baking equipment, comprising:

[0066] S1: Real-time collection of dough parameters, internal state parameters, and environmental parameters within the baking equipment with timestamps, and simultaneous call of the regional characteristic parameters and historical data of the baking equipment to form a multi-dimensional raw data set;

[0067] S2: Analyze the preprocessed multi-dimensional raw data set to generate a dough state evolution feature matrix, and construct time-synchronized real-time multimodal time series data based on the dough state evolution feature matrix;

[0068] S3: Analyze real-time multimodal time series data to generate temperature control parameters, mine the correlation between temperature control parameters, and generate a temperature control strategy matrix;

[0069] S4: Based on the real-time environmental parameters and regional characteristic parameters, the temperature control strategy matrix is ​​modified to generate a modified strategy that adapts to the regional climate characteristics;

[0070] S5: Solve the correction strategy and generate a dynamic control parameter set;

[0071] S6: Analyze the dynamic control parameter set and build an adaptive intelligent temperature control mechanism based on deviation accumulation.

[0072] By collecting dough parameters, internal state parameters, and environmental parameters in baking equipment in real time with timestamps, and synchronously calling regional characteristic parameters and historical data, a multi-dimensional original data set is formed, laying a solid data foundation for precise temperature control. It can fully grasp all kinds of key information in the baking process, analyze the preprocessed data set, generate a dough state evolution feature matrix, and construct real-time multimodal time series data. It can accurately capture the evolution law of dough state, so that the temperature control strategy is more in line with the actual change needs of the dough. Temperature control parameters are generated and their correlations are mined to form a temperature control strategy matrix. Combining the real-time environment and regional characteristic parameter correction strategy, it can adapt to the climate characteristics of different regions, solving the problem that traditional temperature control is greatly affected by regional climate, ensuring that intelligent temperature control can be achieved in different environments, solving the correction strategy to generate a dynamic control parameter set, and constructing an adaptive intelligent temperature control mechanism based on deviation accumulation. It can dynamically adjust the control parameters according to real-time data to achieve precise and adaptive control of the temperature of baking equipment, effectively improving baking quality and efficiency.

[0073] In one case of this embodiment, the pre-processed multi-dimensional original data set is analyzed to generate a dough state evolution feature matrix, including:

[0074] Extracting the preprocessed multidimensional raw data set to obtain absorption peak features, and forming the absorption peak features into an initial feature matrix, specifically including: analyzing the preprocessed multidimensional raw data set using a first-order derivative method to locate the positions of multiple absorption peaks, and recording key characteristic values ​​such as the wavelength position and absorbance of each absorption peak, and then arranging and combining these extracted absorption peak characteristic values ​​in order to form an initial feature matrix, where the rows of the matrix represent different samples and the columns of the matrix correspond to each absorption peak;

[0075] The initial feature matrix is ​​calculated to generate regression parameters. Specifically, the initial feature matrix is ​​used as the independent variable and the measured surface water activity index is used as the dependent variable. The latent variables are extracted using the PLSR method. The covariance between the independent and dependent variables is maximized. The weight vector of the independent variables is calculated through iterative optimization. The leave-one-out cross-validation method is used. The optimal number of latent variables is determined by evaluating the prediction error. From the optimal number of latent variables after dimensionality reduction, the weight coefficient of each original feature (corresponding to the absorption peak) is extracted to generate regression parameters reflecting the strength of its association with the dependent variable.

[0076] The regression parameters were verified to obtain the surface water activity index. The specific calculation formula is as follows:

[0077] ;

[0078] Where, represents the surface water activity index, The value range is , The smaller the value, the lower the surface water activity of the dough. The larger the value is, the higher the surface water activity of the dough is. n represents the number of absorption peak characteristics. Represents the first The absorption peak characteristics, Indicates the first step in calculating regression parameters The regression coefficient corresponding to the absorption peak characteristics is is a comprehensive indicator of environmental parameters. It is a comprehensive indicator of dough parameters;

[0079] The surface water activity index is spliced ​​with the environmental parameters and dough parameters to generate a dough state evolution feature matrix, specifically including: based on the row alignment principle, according to the sample correspondence, the surface water activity index, environmental parameters, and dough parameters of each sample are sequentially spliced ​​using the column vector horizontal splicing method to form a new matrix column to generate a dough state evolution feature matrix, wherein the matrix rows still represent the samples, and the columns correspond to the dimensional features of the surface water activity index, environmental parameters, and dough parameters respectively;

[0080] Absorption peak features are extracted through the first-order derivative method to accurately analyze dough spectral data. The first-order derivative method can accurately locate the position of spectral absorption peaks and obtain key characteristic values ​​such as wavelength and absorbance. This overcomes the subjective errors of traditional manual identification and ensures the objectivity and consistency of feature extraction. The constructed initial feature matrix uses samples as rows and absorption peak features as columns to fully retain spectral information. When calculating regression parameters, the PLSR method effectively solves the problem of multivariate collinearity. By maximizing the covariance of latent variables and combining it with leave-one-out cross-validation to determine the optimal number of latent variables, overfitting is avoided, so that the calculated surface water activity index can accurately reflect changes in dough state.

[0081] By aligning and splicing the surface water activity index with the environment and dough parameters, a three-dimensional state assessment system was constructed. By integrating feature weights, environment and dough parameters, the dynamics of water migration on the dough surface can be keenly captured. Matrix column-wise splicing breaks down data barriers, integrating spectral features, environmental conditions and basic dough parameters to provide a full-dimensional monitoring perspective, and further explores the potential correlations between parameters, such as the impact of environmental factors on water activity and the relationship between basic dough parameters and spectral features, to ensure real-time monitoring of the dough state during baking.

[0082] In one case of this embodiment, time-synchronized real-time multimodal time series data is constructed based on the dough state evolution feature matrix, including:

[0083] The temporal correlation analysis of the dough state evolution feature matrix is ​​performed to generate a three-dimensional spatiotemporal data cube containing temporal correlation. Specifically, the time stamps of the dough state feature matrix at each moment are aligned, the dynamic time warping algorithm is used to calculate the temporal similarity matrix of the feature dimension between adjacent matrices, the autocorrelation coefficient matrix is ​​used to represent the cross-time feature dependency, and then the feature matrices with temporal correlation weights are stacked in sequence along the time axis to construct a dimension (feature number Number of features A three-dimensional spatiotemporal data cube (time steps), where each spatiotemporal unit stores the feature correlation strength at the corresponding moment, achieving time synchronization and spatiotemporal coupling representation of multimodal time series data;

[0084] Dynamically weight the three-dimensional spatiotemporal data cube to generate a temporal association structure, specifically including: constructing a query-key-value matrix in the time dimension, calculating the weight coefficient of the feature association of each time step based on scaled dot product attention, generating a temporal attention matrix after Softmax normalization, multiplying the temporal attention matrix by the feature association matrix of the corresponding time step in the cube element by element, realizing dynamic adjustment of the cross-time association strength, arranging the dynamically adjusted feature association matrices of each time step in chronological order, and generating a dynamic graph structure containing temporal dependency weights, namely the temporal association structure, in which the nodes of the temporal association structure represent the feature dimensions, and the edge weights represent the spatiotemporal coupling strength after dynamic weighting;

[0085] According to the temporal association structure, the dough parameters, internal state parameters and environmental parameters are synchronously processed in time alignment to construct time-synchronized real-time multimodal time series data. Specifically, based on the dynamic edge weight of the temporal association structure, a weighted dynamic time warping algorithm is used to align the multi-source parameter sequences, and the original time series of the dough parameters, internal state parameters and environmental parameters are matched with the corresponding feature nodes in the association structure respectively. The edge weight is used as the path constraint weight to calculate the optimal warping path of each modal sequence to the reference time axis, and the high-correlation feature information of adjacent time steps is aggregated through the graph attention network. The missing values ​​after alignment are filled by cubic spline interpolation, and the multimodal data are then integrated into a unified time axis (time axis) using tensor splicing. Parameter Type parameter values) three-dimensional matrix, where the time dimension is synchronized at the millisecond level through Newton interpolation, forming a real-time, updateable and time-synchronized multimodal time series dataset;

[0086] By constructing time-synchronized, real-time, multimodal time series data of the dough state evolution feature matrix, we can accurately monitor and characterize the dough state during the baking process. The generated three-dimensional spatiotemporal data cube accurately depicts the dynamic changes in the dough state along the time axis and the correlations between features, providing strong data support for a deeper understanding of the physical and chemical changes in the dough. Secondly, when constructing the time series correlation structure, the dynamic weighting and attention mechanism are used to effectively capture the changes in the correlation strength of each dough feature dimension at different time steps, more accurately reflecting the actual evolution of the dough. This is of great significance for optimizing temperature control in baking and improves the stability of intelligent temperature control.

[0087] By integrating multiple modal data such as dough parameters, internal state parameters and environmental parameters, a time-synchronized real-time multimodal time series dataset is constructed. On the one hand, the weighted dynamic time warping algorithm and graph attention network technologies are used to solve the problem of time alignment of multi-source parameters, ensure the accurate matching of different modal data on the time axis, and completely preserve the spatiotemporal coupling relationship between data. On the other hand, the use of cubic spline interpolation and tensor splicing not only fills the missing values ​​after alignment, but also realizes efficient integration of data and millisecond-level synchronous updates, providing a high-quality data foundation for real-time monitoring and analysis of dough status. This provides comprehensive and real-time dough status information for intelligent temperature control of baking equipment, assisting it in making quick decisions, thereby improving baking efficiency.

[0088] In one case of this embodiment, analyzing real-time multimodal time series data to generate temperature control parameters includes:

[0089] Performing standardization processing on the real-time multimodal time series data to obtain standardized real-time multimodal time series data;

[0090] The importance of the standardized real-time multimodal time series data is evaluated to obtain the importance score of each parameter. Specifically, the importance of the standardized real-time multimodal time series data is evaluated using a local weighted linear regression algorithm. The weight value is first determined based on the sample time series distance of the real-time multimodal time series data. The closer the distance to the target sample, the greater the weight. Then, a Gaussian kernel function is introduced, and the parameters are estimated by the least squares method. The objective function of minimizing the weighted mean square error is solved to obtain the regression coefficient of each parameter. The recursive feature elimination algorithm is used to select features according to the size of the regression coefficient. The feature with a larger absolute value of the regression coefficient has a higher importance score, thereby obtaining the importance score of each parameter.

[0091] The importance threshold is calculated based on the internal state parameters of the baking equipment. Specifically, kernel density estimation is used to construct a probability structure for the sliding window data of the baking equipment's internal state parameters (temperature, pressure), and the 95% quantile of the distribution function is directly extracted as the importance threshold. This threshold dynamically captures the boundaries of the high-density interval of the state parameters, ensuring that only significant features are retained at a 95% confidence level. This achieves a strong correlation and self-adaptation between the threshold and the real-time operating conditions of the baking equipment.

[0092] The importance scores of the parameters are compared with the importance threshold and screened to obtain a key parameter set. Specifically, the importance score of each parameter is compared with the importance threshold. If the importance score of the parameter is greater than the importance threshold, it indicates that the parameter is a significant feature and is included in the key parameter set. If the importance score of the parameter is less than or equal to the importance threshold, the parameter is excluded. After this screening, a key parameter set that is adapted to the real-time working conditions of the baking equipment is finally obtained.

[0093] The key parameter set is associated with the temperature control parameters of the internal state parameters to generate the temperature control parameters. The specific calculation formula is as follows:

[0094] ;

[0095] Where, Indicates temperature control parameters, Indicates the The importance score of each parameter, K represents the key parameter set, Temperature control parameters representing internal state parameters, Indicates the internal state parameters of the baking equipment;

[0096] By standardizing and dynamically filtering real-time multimodal time series data, data quality is improved and key parameters are adaptively extracted. Standardization eliminates dimensional differences and distribution offsets in multimodal parameters, making heterogeneous time series data such as temperature and pressure comparable. This establishes a unified benchmark for subsequent analysis and avoids evaluation bias caused by differences in parameter units or magnitudes. On this basis, a combined method of local weighted linear regression and recursive feature elimination is adopted. Sample weights are dynamically assigned using a Gaussian kernel function. While retaining local time series features, the regression coefficients are iteratively optimized using the least squares method to accurately capture the dynamic influence of parameters at different time points on temperature control. Compared with traditional fixed-window feature selection, the weighted mechanism strengthens the temporal correlation of neighboring samples, reducing the error in feature importance assessment. Combined with the recursive feature elimination algorithm, the absolute value of the regression coefficient is used as the judgment basis to effectively eliminate redundant parameter interference, ensuring a strong correlation between the dynamic update of the key parameter set and the operating status of the baking equipment, laying a high-precision data foundation for subsequent temperature control parameter generation.

[0097] Through dynamic threshold modeling and multi-dimensional parameter fusion calculation of kernel density estimation, adaptive optimization of temperature control parameters and precise matching of equipment working conditions are achieved. Based on the kernel density estimation method of sliding window, dynamic probability modeling of state parameters such as temperature and pressure is performed. By extracting the 95% quantile threshold, the limitations of traditional fixed thresholds are broken through, so that the feature screening boundary can be automatically adjusted with the changes in the working conditions of the baking equipment. This threshold mechanism significantly improves the robustness of feature screening under abnormal working conditions such as high temperature and high pressure by capturing the high-density distribution interval of state parameters in real time. In the temperature control parameter generation stage, the importance score of key parameters is coupled with the internal state parameters of the baking equipment for calculation, which not only strengthens the decision weight of high-importance parameters, but also balances the influence of the real-time state of the baking equipment through the normalization of the denominator, achieving the coordinated improvement of energy efficiency optimization of the baking process and product quality.

[0098] In one case of this embodiment, the correlation between temperature control parameters is mined to generate a temperature control strategy matrix, including:

[0099] Calculate the correlation of temperature control parameters to obtain a correlation distribution feature matrix. Specifically, the linear correlation between each parameter in the temperature control parameters, such as temperature, humidity, and equipment power, is calculated to obtain the correlation coefficient. By analyzing the correlation coefficient, a parameter correlation heat map is drawn to observe the correlation strength and positive and negative correlation distribution between the parameters, ultimately forming a complete correlation distribution feature matrix.

[0100] For any two parameters XA and YA, calculate the correlation coefficient between the two parameters , the calculation formula is as follows:

[0101] ;

[0102] in, represents the number of data samples, and Represents the first Sample values, express The index variable, and Represent the sample means of any two parameters XA and YA, The value range is between -1 and 1;

[0103] Setting the Color Mapping Rules: Establishing the Correlation Coefficient The corresponding relationship with the color specifies the positive correlation coefficient (0 1) Use red, and the closer the correlation coefficient is to 1, the darker the red; negative correlation coefficient (-1 0) uses blue. The closer the correlation coefficient is to -1, the darker the blue. The closer the correlation coefficient is to 0 (such as 0.2), use light color or white;

[0104] Create a two-dimensional table and arrange the temperature control parameters (such as temperature, humidity, and equipment power) in the horizontal rows and vertical columns of the table respectively. For each cell in the table, the two parameters corresponding to the intersection of the rows and columns are filled with the corresponding colors according to the calculated correlation coefficient and the color mapping rules, thus forming a parameter correlation heat map. By observing the color distribution of the heat map, the correlation strength and positive and negative correlation between the parameters can be intuitively judged. Darker cells indicate a high correlation strength between the corresponding parameters, red cells indicate a positive correlation between the two parameters, blue cells indicate a negative correlation between the two parameters, and lighter cells indicate a weak correlation between the corresponding parameters. Recreate a two-dimensional matrix, and use the temperature control parameters as the identifier for the rows and columns. Fill the previously calculated correlation coefficient between each two parameters into the corresponding cells of the matrix to form a complete correlation distribution feature matrix;

[0105] Calculating a strong correlation threshold according to the correlation distribution feature matrix, specifically including: obtaining correlation data of all parameter pairs from the correlation distribution feature matrix, calculating the mean and standard deviation of the correlation of all parameter pairs, and setting the strong correlation threshold as the mean plus two times the standard deviation according to the normal distribution characteristics to obtain the strong correlation threshold;

[0106] Comparing and screening the correlation values ​​in the correlation distribution feature matrix with the strong correlation threshold to obtain a set of strongly correlated parameter pairs, specifically comprising: traversing the correlation distribution feature matrix, extracting the correlation values ​​of each parameter pair one by one, comparing them with the strong correlation threshold one by one, screening out parameter pairs with correlation values ​​greater than or equal to the strong correlation threshold, integrating these parameter pairs that meet the conditions into a set, and constructing a set of strongly correlated parameter pairs;

[0107] Based on the set of strongly correlated parameter pairs, an initial temperature control strategy matrix is ​​constructed. Specifically, the matrix dimensions are determined, with rows representing temperatures and columns corresponding to parameter pairs. The correlation characteristics of the parameter pairs in the set of strongly correlated parameter pairs are extracted, with positive values ​​assigned to positive correlations, negative values ​​assigned to negative correlations, and zero assigned to no correlations. The correlation strength is graded, with strong correlations assigned high weights, medium correlations assigned medium weights, and weak correlations discarded. The matrix elements are then filled in to construct the initial temperature control strategy matrix.

[0108] The initial temperature control strategy matrix is ​​optimized to generate a final temperature control strategy matrix, specifically including: when optimizing the initial temperature control strategy matrix, based on the actual operation data of the baking equipment, continuously collecting temperature-related data, including the set temperature, actual temperature, temperature fluctuation amplitude, and equipment operating parameters, comparing these data with the target temperature of the initial temperature control strategy matrix to determine the temperature deviation, using the mean square error to calculate the error between the actual temperature and the target temperature, and using this as a basis to determine the adjustment direction and step size of the matrix weights, using the gradient descent algorithm to calculate the gradient of the error with respect to each weight, thereby determining the direction and step size of the weight adjustment, and then updating each weight in the initial temperature control strategy matrix, applying the updated matrix to the actual operation of the baking equipment, observing whether the temperature of the baking equipment can approach the target temperature faster and whether the temperature fluctuation is reduced, collecting operation data again, and repeating the above process continuously, adjusting the matrix weights until the temperature control achieves the expected stability and response speed, and then using the initial temperature control strategy matrix that shows good temperature control effect as the temperature control strategy matrix to achieve long-term stable temperature control;

[0109] By deeply exploring the complex relationships between temperature control parameters, comprehensive and accurate correlation coefficients are obtained from the pairwise linear correlation calculations of various parameters such as temperature, humidity, and equipment power, forming a correlation distribution feature matrix. Strong correlation thresholds are set according to the normal distribution characteristics, and a set of strongly correlated parameter pairs is screened out, making the construction of the temperature control strategy matrix targeted. Furthermore, an initial temperature control strategy matrix is ​​constructed based on the set of strongly correlated parameter pairs, and corresponding weights are assigned according to different correlation characteristics, preliminarily achieving effective integration of the mutual influence between temperature control parameters. In the optimization stage, combined with actual operation data, advanced optimization methods such as gradient descent algorithm are used to continuously adjust the matrix weights so that it can accurately adapt to actual temperature control requirements. The final generated temperature control strategy matrix can significantly improve the stability and response speed of the temperature control system, effectively reduce temperature fluctuations, make the temperature control of baking equipment more accurate and efficient, provide a stable and reliable temperature environment for the baking process, and improve the quality of baked products and production efficiency.

[0110] By setting color mapping rules, the abstract correlation coefficient is converted into intuitive color information, allowing staff to quickly and clearly observe the correlation strength and positive and negative correlation between each temperature control parameter. The changes in color depth intuitively reflect the closeness of the correlation between the parameters, and the sharp contrast between red and blue clearly indicates the positive and negative correlation. This greatly reduces the difficulty and complexity of data analysis, facilitates technicians to quickly identify key parameter pairs, and provides a strong visual basis for formulating effective temperature control strategies. At the same time, from the correlation distribution feature matrix to the screening process of the set of strongly correlated parameter pairs, and then to the construction and optimization of the temperature control strategy matrix, the entire process builds a systematic and intelligent temperature control decision support system, which can dynamically adjust the strategy according to actual data and realize adaptive adjustment of temperature control during the operation of baking equipment.

[0111] In one case of this embodiment, the temperature control strategy matrix is ​​modified according to the real-time environmental parameters and the regional characteristic parameters to generate a modified strategy adapted to the regional climate characteristics, including:

[0112] Mining real-time environmental parameters and regional characteristic parameters to construct an initial correlation matrix, specifically including: pre-processing real-time environmental parameters (such as temperature, humidity, air pressure, etc.) and regional characteristic parameters (such as annual average temperature, monsoon intensity, precipitation distribution, etc.), removing noise and outliers, using principal component analysis algorithm to mine key features, and standardizing the data to Interval, to eliminate dimensional differences, the initial correlation matrix is ​​constructed with regions as rows and parameters as columns, the correlation value of each parameter and the target climate characteristics is calculated by Pearson correlation coefficient, and the matrix is ​​filled, thereby constructing an initial correlation matrix that can reflect the correlation between regional climate characteristics and environmental parameters;

[0113] Combining the initial correlation matrix with dough water activity data in the real-time dough parameters to obtain a humidity deviation compensation coefficient and a temperature fluctuation compensation gradient, specifically comprising: normalizing the real-time dough water activity data to obtain a normalized dough water activity matrix, performing matrix multiplication operations on humidity-related rows in the initial correlation matrix and the normalized dough water activity matrix to obtain a humidity deviation compensation coefficient, and performing gradient descent calculations on temperature-related rows in the initial correlation matrix and the normalized dough water activity matrix to obtain a temperature fluctuation compensation gradient;

[0114] The humidity deviation compensation coefficient and the temperature fluctuation compensation gradient are integrated to generate a water activity change curve. Specifically, the method includes: using a weighted average method, multiplying the humidity deviation compensation coefficient by its corresponding weight coefficient (0.6), and multiplying the temperature fluctuation compensation gradient by its corresponding weight coefficient (0.4), adding the two results to obtain a fusion value, with time as the horizontal axis and the fusion value as the vertical axis, using a cubic spline interpolation algorithm to interpolate the discrete fusion values ​​on the time axis to generate a continuous water activity change curve;

[0115] The temperature control strategy matrix was modified using the water activity change curve to generate a correction strategy adapted to the regional climate characteristics. Specifically, the strategy involved extracting the fused value at each time point in the water activity change curve and using a linear mapping algorithm to decompose the fused value into temperature correction and humidity correction values ​​according to preset proportional coefficients (temperature correction coefficient 0.3, humidity correction coefficient 0.7). Through matrix dot multiplication, the temperature correction and humidity correction values ​​at the corresponding time point were applied to the temperature control column and humidity control column of the temperature control strategy matrix, respectively. This completed the row-by-row correction of the matrix elements and generated a correction strategy adapted to the regional climate characteristics.

[0116] By deeply integrating real-time environmental parameters with regional characteristic parameters, an initial correlation matrix that accurately reflects the relationship between regional climate and environmental parameters is constructed, and a water activity change curve is further generated, providing a strong basis for revising the temperature control strategy matrix. During the data mining and analysis stage, algorithms such as principal component analysis and Pearson correlation coefficient are used to effectively remove noise and outliers, accurately extract key features, and ensure the accuracy and reliability of the data. Based on this, the generated water activity change curve can accurately capture the dynamic trend of dough water activity changing with environmental changes, making the revised temperature control strategy matrix more suitable for dough fermentation requirements under actual regional climatic conditions, significantly improving the pertinence and effectiveness of the temperature control strategy, thereby providing more accurate temperature and humidity control for the dough fermentation process, and helping to improve the stability and consistency of dough fermentation quality;

[0117] By realizing intelligent dynamic adjustment of temperature control strategies, the adaptability of dough fermentation process to regional climate is greatly enhanced. The humidity deviation compensation coefficient and temperature fluctuation compensation gradient are integrated to generate a water activity change curve, which is used to correct the temperature control strategy matrix so that the temperature control strategy can respond to changes in environmental parameters and regional climate in real time. This dynamic correction mechanism avoids the possible inadaptability of traditional fixed temperature control strategies in different regional environments, reduces the risk of dough baking failure due to climate differences, and improves the success rate and efficiency of fermentation. At the same time, the use of data processing technology ensures the scientificity and rationality of the correction process, provides a stable and precise temperature and humidity environment for dough baking, helps to improve the quality and taste of dough products, and meets the requirements of dough baking technology in different regions.

[0118] In one case of this embodiment, the correction strategy is solved to generate a dynamic control parameter set, including:

[0119] Based on the correction strategy, a multi-parameter decoupling analysis is conducted to generate a climate adaptability rule set. Specifically, this involves: introducing a feature extraction algorithm to accurately extract the characteristics of each parameter in the correction strategy to identify key influencing factors; using a principal component analysis algorithm to decouple key parameters and separate independent multiple parameters; using a clustering algorithm to analyze the performance of the decoupled parameters under different climate conditions; and finally, using a decision tree algorithm to construct a climate adaptability rule set based on the parameter analysis results.

[0120] The climate adaptability rule set is constrained by a geographic feature constraint vector to form a regional adaptive optimization framework. Specifically, the framework includes: quantifying and normalizing geographic feature parameters (such as latitude, altitude, and average annual temperature), constructing a geographic feature constraint vector, and embedding the geographic feature constraint vector as a priori condition into the decision space of the climate adaptability rule set. A constrained nonlinear programming algorithm is used, with the compatibility of the control parameters output by the rule set and the geographic features as the objective function. The optimal solution is obtained through the Lagrange multiplier method, and the parameter thresholds and weight coefficients in the rule set are iteratively adjusted to ultimately form a regional adaptive optimization framework that incorporates geographic feature constraints.

[0121] A global optimization search is performed on the regional adaptive optimization framework to generate a candidate set of dynamic compensation parameters. Specifically, the optimization is performed using a genetic algorithm, with the degree of fit between the control parameters output by the regional adaptive optimization framework and actual needs as the objective function. In the genetic algorithm, the control parameters are converted into chromosome encoding. Through continuous iterations of selection, crossover, and mutation operations, the fitness of individuals in each generation is calculated, and individuals with high fitness are screened. At the same time, a simulated annealing algorithm is introduced to escape the local optimum. After multiple iterations, parameters are finally extracted from high-quality individuals to form a candidate set of dynamic compensation parameters.

[0122] The dynamic compensation parameter candidate set is thermodynamically verified to generate a dynamic control parameter set. This includes: using the candidate parameters in the dynamic compensation parameter candidate set as control variables, evaluating the thermodynamic response under different compensation parameters through a parameter response simulation algorithm, collecting actual temperature distribution data with a thermal imager, and calculating the root mean square error (RMSE < 0.5°C) between the two to verify thermal balance accuracy. Monte Carlo sampling technology (sampling times 1000) is used to perform sampling analysis on the dynamic compensation parameter candidate set to test its robustness. For parameters that do not meet the standards, a sequential quadratic programming algorithm is used to optimize and adjust them under the condition that Lagrange multiplier constraints are satisfied. Finally, a time series analysis algorithm is used to verify the thermodynamic stability under the optimized parameters, and parameters that meet the accuracy and stability requirements are screened out to generate a dynamic control parameter set.

[0123] Through the comprehensive application of sophisticated multi-parameter decoupling analysis and multiple advanced algorithms, in-depth optimization of the correction strategy is achieved. First, a feature extraction algorithm is introduced to accurately locate key influencing factors, and then the principal component analysis algorithm is used to decouple key parameters and separate independent multi-parameters. This lays the foundation for the subsequent precise construction of rule sets according to different climatic conditions. Then, a clustering algorithm is used to clearly present the performance of each parameter under different climatic conditions, and a decision tree algorithm is used to construct a climate-adaptive rule set based on the analysis results. In addition, the geographic feature constraint vector is embedded in the rule set decision space, and a constrained nonlinear programming algorithm is used to solve the optimal solution. This enables the optimization framework to fully adapt to the geographical characteristics of a specific region, greatly improving the regional adaptability and accuracy of the control parameters, and solving the problem that previous control strategies were difficult to apply evenly under different climatic and geographical environments, providing solid support for the realization of refined dynamic control.

[0124] By using genetic algorithms for global optimization, the control parameters are converted into chromosome codes, and continuous iterations such as selection, crossover, and mutation are performed, which effectively improves the fit between the control parameters and actual needs and avoids the limitations of local optimal solutions. At the same time, the simulated annealing algorithm is introduced to further break out of the local optimum, making the optimization process more comprehensive and effective. In the subsequent thermodynamic characteristic verification stage, the dynamic compensation parameter candidate set is strictly screened and robustness tested using parameter response simulation algorithms, measured data comparison, and Monte Carlo sampling technology, ensuring that the final generated dynamic control parameter set can meet high standards in terms of thermal balance accuracy and stability. For parameters that do not meet the standards, the sequential quadratic programming algorithm is used for optimization and adjustment, and finally the thermodynamic stability is verified by the time series analysis algorithm. This series of rigorous verification and optimization processes greatly enhances the reliability and practicality of the control parameters.

[0125] In one case of this embodiment, the dynamic control parameter set is analyzed to construct an adaptive intelligent temperature control mechanism based on deviation accumulation, including:

[0126] The dynamic control parameter set is extracted to generate a real-time fluctuation feature sequence, specifically including: extracting key parameters from the dynamic control parameter set, including but not limited to temperature, pressure, flow, etc., using a sliding window algorithm to calculate the mean of the key parameters, taking a window of 5 sample points, using a difference algorithm to calculate the difference between adjacent sample points, generating a difference sequence, and then calculating its standard deviation based on the difference sequence, taking the square root of the sum of the squares of each element in the difference sequence after subtracting its mean, and dividing the square root by the number of sample points minus one, so as to quantify the degree of fluctuation, and finally forming a real-time fluctuation feature sequence.

[0127] Determining a quantile baseline based on kernel density estimation within each window of the real-time fluctuation feature sequence and generating a dynamic fluctuation threshold, specifically comprising: fitting the data distribution with a Gaussian kernel function using a kernel density estimation algorithm within each window of the real-time fluctuation feature sequence to obtain a probability density function, calculating the cumulative distribution function of the probability density function using a numerical integration method, and using a binary search algorithm to determine a value corresponding to a cumulative probability of 0.95 in the cumulative distribution function as the 95% quantile baseline. Since the data in each window is different, the fitted distribution and the calculated baseline change accordingly. In combination with the standard deviation of the data within the window, upper and lower fluctuation thresholds are generated around the dynamic baseline based on the outlier judgment criterion of the normal distribution, thereby generating a dynamic fluctuation threshold;

[0128] A comparison mechanism between real-time fluctuation values ​​and dynamic fluctuation thresholds is established to record over-threshold anomalies and corresponding deviations. Specifically, the mechanism includes: using a numerical comparison algorithm to compare each fluctuation value in the real-time fluctuation feature sequence with the corresponding dynamic fluctuation threshold one by one. If the fluctuation value exceeds the upper threshold or is lower than the lower threshold, it is determined to be an over-threshold anomaly. The abnormal moment and the deviation between the fluctuation value and the threshold are recorded. At the same time, the number of over-threshold anomalies is counted. The absolute difference between the fluctuation value and the corresponding threshold (the upper threshold if it exceeds the upper limit, and the lower threshold if it is lower than the lower limit) is calculated as the deviation.

[0129] The statistical distribution characteristics of the cumulative deviation are calculated based on the statistical distribution algorithm to obtain the warning threshold. The specific calculation formula is as follows:

[0130] ;

[0131] Where, Indicates the warning threshold, Indicates the length of the real-time fluctuation characteristic sequence, Indicates the first Features, represents the cumulative deviation, and Corresponding to and The weight coefficient of Indicates the number of times the threshold value is exceeded;

[0132] Through refined real-time fluctuation feature analysis, the dynamic changes of temperature control parameters are accurately captured and quantitatively evaluated. Using a sliding window algorithm and differential calculations, the real-time fluctuation characteristics of key parameters such as temperature and pressure can be effectively extracted. The degree of fluctuation is quantified through standard deviation, forming a feature sequence that contains details of the dynamic changes in parameters. The dynamic fluctuation threshold constructed in combination with kernel density estimation can adaptively adjust the baseline based on the actual distribution of data in each window, avoiding the problem of the traditional fixed threshold's lack of adaptability to complex fluctuation scenarios. This dynamic threshold generation method based on data distribution characteristics can not only sensitively identify abnormal fluctuations, but also filter out small disturbances within the normal range, providing a scientific and reliable benchmark for subsequent anomaly judgment and significantly improving the dynamic adaptability of the temperature control system to nonlinear and time-varying baking processes.

[0133] By combining the comparison mechanism with the statistical distribution algorithm, a deviation accumulation warning system with self-learning ability was constructed. By comparing the fluctuation value with the dynamic threshold in real time, the over-threshold anomaly is captured immediately and the deviation data is recorded, providing the system with accurate anomaly positioning and quantitative evaluation. The warning threshold is calculated based on the statistical distribution characteristics, and multi-dimensional information such as real-time fluctuation characteristics, cumulative deviation and number of anomalies is weighted and calculated, so that the warning strategy can be dynamically adjusted with the actual deviation status of the baking process. This design that integrates data statistical laws and real-time deviation feedback not only realizes the cumulative correction of temperature control deviations, but also gives the system the ability of "experience learning", so that it can continuously optimize the warning strategy in long-term operation, reduce the frequency of manual intervention, and improve the equipment's autonomous control accuracy and stability in complex baking environments, providing intelligent deviation control guarantees for high-quality baking.

[0134] In one case of this embodiment, analyzing the dynamic control parameter set to construct an adaptive intelligent temperature control mechanism based on deviation accumulation further includes:

[0135] When the accumulated deviation exceeds the warning threshold, an adaptive intelligent temperature control mechanism based on the accumulated deviation is constructed, specifically including: The upper limit of the dynamic fluctuation threshold is recorded as a positive deviation (need to cool down); if The lower limit of the dynamic fluctuation threshold is recorded as a negative deviation (need to increase the temperature). Dynamic fluctuation threshold, no adjustment;

[0136] The adaptive intelligent temperature control mechanism is divided into three levels, as follows:

[0137] Mild: When the absolute value of the ratio of the cumulative deviation to the warning threshold Within 20%, fine-tune the temperature setting value by 1°C each time. When the deviation is positive, lower the set temperature to avoid over-baking. When the deviation is negative, increase the set temperature to ensure the baking progress.

[0138] Moderate: When the absolute value of the ratio of the cumulative deviation to the warning threshold is between 21% and 50%, the heating element power is adjusted according to the deviation ratio. When the deviation is positive, the power is reduced by 15% to 30% to prevent the temperature from being too high. When the deviation is negative, the power is increased by 15% to 30% to ensure that the temperature reaches the standard as soon as possible.

[0139] Severe: When the absolute value of the ratio of the cumulative deviation to the warning threshold 50%, enter emergency mode, turn off some heating elements when the deviation is positive, quickly cool down, and all heating elements run at full power when the deviation is negative, quickly heat up;

[0140] In order to solve the problem of temperature control being difficult to adapt and baking effects being poor due to climate differences when chain bakery workshops are used across regions, a graded deviation response strategy is adopted to divide the adjustment mode into light, medium and heavy levels according to the ratio of cumulative deviation to warning threshold, so as to accurately adapt to the temperature control needs in different regional environments. In the humid southern regions, if the temperature fluctuates due to high ambient humidity, a slight deviation of 1°C is used to fine-tune the dough to avoid excessive fermentation. In the dry northern environment, the heating power is adjusted proportionally when there is a moderate deviation to prevent temperature overshoot or energy waste. When extreme climate causes severe deviation, the emergency mode quickly switches the heating element on and off and quickly corrects the temperature, so that the equipment can always maintain stable temperature control accuracy under complex and changeable regional climatic conditions, effectively solving the problem of insufficient adaptability of traditional technology and ensuring that baking quality transcends regional restrictions.

[0141] refer to Figure 2 , an intelligent temperature control system for baking equipment, comprising:

[0142] Collection unit: This collects dough parameters, internal state parameters, and environmental parameters of baking equipment in real time with timestamps, and simultaneously calls the regional characteristic parameters and historical data of the baking equipment to form a multi-dimensional raw data set;

[0143] Analysis unit: Analyzes the pre-processed multi-dimensional raw data set to generate a dough state evolution feature matrix, and constructs time-synchronized real-time multi-modal time series data based on the dough state evolution feature matrix;

[0144] Mining unit: Analyzes real-time multimodal time series data to generate temperature control parameters, mines the correlation between temperature control parameters, and generates a temperature control strategy matrix;

[0145] Correction unit: Corrects the temperature control strategy matrix based on real-time environmental parameters and regional characteristic parameters, and generates a correction strategy that adapts to regional climate characteristics;

[0146] Solving unit: solves the correction strategy and generates a dynamic control parameter set;

[0147] Construction unit: Analyze the dynamic control parameter set and build an adaptive intelligent temperature control mechanism based on deviation accumulation.

[0148] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent temperature control method for baking equipment, characterized in that: include: S1: Real-time collection of dough parameters, internal state parameters, and environmental parameters within the baking equipment with timestamps. It also synchronously calls the regional characteristic parameters and historical data of the baking equipment to form a multi-dimensional raw data set. S2: Analyze the preprocessed multi-dimensional raw data set to generate a dough state evolution feature matrix, including: Extract the preprocessed multi-dimensional original data set to obtain absorption peak features, and form the absorption peak features into an initial feature matrix; Calculate the initial feature matrix to generate regression parameters; The regression parameters were verified to obtain the surface water activity index; The surface water activity index is combined with environmental parameters and dough parameters to generate a dough state evolution characteristic matrix; The real-time multimodal time series data with time synchronization is constructed based on the dough state evolution feature matrix, including: performing time series correlation analysis on the dough state evolution feature matrix, generating a three-dimensional spatiotemporal data cube containing time series correlation, specifically including: aligning the timestamps of the dough state feature matrix at each moment, using the dynamic time warping algorithm to calculate the time series similarity matrix of the feature dimension between adjacent matrices, representing the cross-time feature dependency through the autocorrelation coefficient matrix, and then stacking the feature matrices with time series correlation weights along the time axis to construct a feature matrix with the dimension of feature number. Number of features A three-dimensional space-time data cube of the number of time steps, where The product symbol representing the dimension; Dynamically weight the three-dimensional spatiotemporal data cube to generate a temporal correlation structure; Based on the time series correlation structure, the dough parameters, internal state parameters and environmental parameters are synchronized and processed in a time-aligned manner to construct time-synchronized real-time multimodal time series data. S3: Analyze real-time multimodal time series data to generate temperature control parameters, mine the correlation between temperature control parameters, and generate a temperature control strategy matrix; S4: Based on the real-time environmental parameters and regional characteristic parameters, the temperature control strategy matrix is ​​modified to generate a modified strategy that adapts to the regional climate characteristics; S5: Solve the correction strategy and generate a dynamic control parameter set; S6: Analyze the dynamic control parameter set and build an adaptive intelligent temperature control mechanism based on deviation accumulation.

2. The intelligent temperature control method for baking equipment according to claim 1, characterized in that: Analyze real-time multimodal time series data to generate temperature control parameters, including: Performing standardization processing on the real-time multimodal time series data to obtain standardized real-time multimodal time series data; Perform importance evaluation on the standardized real-time multimodal time series data to obtain the importance score of each parameter; The importance threshold is calculated based on the internal state parameters of the baking equipment. Specifically, the kernel density estimation is used to construct the probability of the sliding window data of the internal state parameters of the baking equipment, and the 95% quantile of the distribution function is directly extracted as the importance threshold. Compare and filter the importance scores of the parameters with the importance threshold to obtain a set of key parameters; The key parameter set is associated with the temperature control parameters of the internal state parameters to generate temperature control parameters.

3. The intelligent temperature control method for baking equipment according to claim 2, characterized in that: Mining the correlation of temperature control parameters to generate a temperature control strategy matrix, including: The correlation degree of the temperature control parameters is calculated to obtain the correlation degree distribution characteristic matrix. Specifically, for each type of temperature control parameter, the linear correlation between each two parameters is calculated to obtain the correlation coefficient. By analyzing the correlation coefficient, a parameter correlation degree heat map is drawn to observe the correlation strength and positive and negative correlation distribution between the parameters, and finally a complete correlation degree distribution characteristic matrix is ​​formed. According to the correlation distribution feature matrix, the strong correlation threshold is calculated; Compare and screen the correlation value in the correlation distribution feature matrix with the strong correlation threshold to obtain a set of strong correlation parameter pairs; Construct the initial temperature control strategy matrix based on the set of strongly correlated parameter pairs; The initial temperature control strategy matrix is ​​optimized to generate the final temperature control strategy matrix.

4. The intelligent temperature control method for baking equipment according to claim 3, characterized in that: Based on real-time environmental parameters and regional characteristic parameters, the temperature control strategy matrix is ​​modified to generate a correction strategy that adapts to regional climate characteristics, including: Mining real-time environmental parameters and regional characteristic parameters to construct an initial correlation matrix; The initial correlation matrix is ​​combined with the dough water activity data in the real-time dough parameters to obtain the humidity deviation compensation coefficient and the temperature fluctuation compensation gradient; The humidity deviation compensation coefficient is integrated with the temperature fluctuation compensation gradient to generate a water activity change curve; The temperature control strategy matrix is ​​modified using the water activity change curve to generate a modified strategy that adapts to regional climate characteristics.

5. The intelligent temperature control method for baking equipment according to claim 4, characterized in that: Solve the correction strategy and generate a dynamic control parameter set, including: Implement multi-parameter decoupling analysis based on the correction strategy to generate a set of climate adaptability rules; By constraining the climate adaptability rule set through geographic feature constraint vectors, a regional adaptive optimization framework is formed; Perform global optimization on the regional adaptive optimization framework to generate a candidate set of dynamic compensation parameters; The thermodynamic characteristics of the dynamic compensation parameter candidate set are verified to generate a dynamic control parameter set.

6. The intelligent temperature control method for baking equipment according to claim 5, characterized in that: Analyze the dynamic control parameter set and build an adaptive intelligent temperature control mechanism based on deviation accumulation, including: Extract the dynamic control parameter set to generate a real-time fluctuation feature sequence; Determine the quantile baseline based on kernel density estimation in each window of the real-time fluctuation feature sequence and generate a dynamic fluctuation threshold; Establish a comparison mechanism between real-time fluctuation values ​​and dynamic fluctuation thresholds, and record abnormalities exceeding the threshold and corresponding deviations; The statistical distribution characteristics of the cumulative deviation are calculated based on the statistical distribution algorithm to obtain the warning threshold.

7. The intelligent temperature control method for baking equipment according to claim 6, characterized in that: Analyze the dynamic control parameter set and build an adaptive intelligent temperature control mechanism based on deviation accumulation, which also includes: When the accumulated deviation exceeds the warning threshold, an adaptive intelligent temperature control mechanism based on the accumulated deviation is constructed.

8. An intelligent temperature control system for baking equipment, applied to an intelligent temperature control method for baking equipment according to any one of claims 1 to 7, characterized in that: include: Collection unit: This collects dough parameters, internal state parameters, and environmental parameters of baking equipment in real time with timestamps. It also synchronously calls the regional characteristic parameters and historical data of the baking equipment to form a multi-dimensional raw data set. Analysis unit: Analyzes the pre-processed multi-dimensional raw data set to generate a dough state evolution feature matrix, and constructs time-synchronized real-time multi-modal time series data based on the dough state evolution feature matrix; Mining unit: Analyzes real-time multimodal time series data to generate temperature control parameters, mines the correlation between temperature control parameters, and generates a temperature control strategy matrix; Correction unit: Corrects the temperature control strategy matrix based on real-time environmental parameters and regional characteristic parameters, and generates a correction strategy that adapts to regional climate characteristics; Solving unit: solves the correction strategy and generates a dynamic control parameter set; Construction unit: Analyze the dynamic control parameter set and build an adaptive intelligent temperature control mechanism based on deviation accumulation.

Images