Temperature self-adaptive control method for douchi fermentation tank based on PID controller

Abstract

The present application relates to food fermentation temperature control technical field, specifically to a kind of temperature self-adaptive regulation and control method of bean paste fermentation tank based on PID controller, comprising: array type temperature sensor network is arranged in the multilayer space of fermentation tank, and different depth real-time temperature field data are periodically collected.Spatiotemporal fusion processing is carried out to the collected data, and temperature gradient atlas is generated, which is input into improved multi-model PID controller, and parameter self-tuning is completed in combination with fermentation process and microbial metabolic heat dynamic model.Controller dynamic output cooperates control instruction, adjusts heating power and cooling medium flow.Microbial metabolite generation rate is monitored in real time, and fermentation process stage is dynamically identified and updated, and parameter online self-adaptive adjustment is realized by feedback to controller, and closed-loop control loop is constructed.This method can adapt to the dynamic change of fermentation environment, realize the accurate regulation and control of fermentation tank global temperature field, optimize the spatial temperature distribution state, and adapt to the change of microbial metabolic rhythm.

Description

Technical Field

[0001] This invention relates to the field of food fermentation temperature control technology, and in particular to a method for adaptive temperature control of fermentation tanks for fermented soybeans based on a PID controller. Background Technology

[0002] Fermentation of fermented soybeans relies on microbial metabolism to complete the material transformation. The operating state of the temperature field directly affects the fermentation process and the stability of the finished product quality. Existing temperature control methods for fermentation tanks generally adopt a single-point temperature acquisition mode, combined with a conventional PID control structure with fixed parameters, to separately control the operating status of the heating and cooling devices, thereby achieving simple adjustment of the basic temperature inside the fermentation tank.

[0003] Single-point temperature acquisition can only obtain temperature information for a local area. Fermentation tanks have a multi-layered, three-dimensional spatial structure with significant internal temperature distribution differences. Single monitoring data cannot reflect the overall temperature change pattern, resulting in insufficient comprehensiveness of temperature monitoring. Conventional PID control parameters remain fixed over a long period, failing to adapt to the dynamic changes in the fermentation cycle. Microorganisms continuously generate heat through metabolism, and the metabolic heat production characteristics differ at different fermentation stages, making it difficult for static control logic to match the dynamically changing fermentation environment. Independent control and operation of heating and cooling equipment lacks a coordinated mechanism, leading to a lag in temperature regulation response and easily causing localized temperature imbalances in the fermentation tank, making it difficult to meet the precise temperature control requirements of fermentation production.

[0004] To address the existing problems of limited temperature monitoring dimensions, inability to dynamically update control parameters, and insufficient coordination of temperature control systems, it is necessary to optimize the temperature acquisition architecture and control operation mode, and establish a closed-loop temperature control scheme that adapts to changes in the fermentation process. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a method for adaptive temperature control of fermentation tanks for fermented soybeans based on a PID controller.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for adaptive temperature control of a fermentation tank for fermented soybeans based on a PID controller, comprising: An array of temperature sensors is deployed in the multi-layered space of the fermentation tank to periodically collect real-time temperature field data at different depths inside the fermentation tank. The collected real-time temperature field data is subjected to spatiotemporal fusion processing to generate a temperature gradient map that reflects the overall and local temperature distribution characteristics of the fermentation tank. The temperature gradient map is input into the improved multi-model PID controller, which performs parameter self-tuning based on the fermentation process stage and the microbial metabolic heat dynamics model. Using the improved multi-model PID controller, the heating and cooling systems are dynamically calculated and output according to the current stage of the fermentation process and the measured temperature field. According to the coordinated control command, the control actuator adjusts the power of the heating element and the flow rate of the cooling medium to regulate the temperature field of the fermentation tank; During the regulation process, the generation rate of key microbial metabolites is continuously monitored and used as a feedback signal to the fermentation process stage identification module. The fermentation process stage identification module is used to dynamically update the current fermentation process stage based on the changing trend of the metabolite generation rate. The updated fermentation process stage information is fed back to the improved multi-model PID controller in real time to trigger online adaptive adjustment of the controller parameters, forming a closed-loop control loop.

[0007] As a further aspect of the present invention, the step of performing spatiotemporal fusion processing on the collected real-time temperature field data to generate a temperature gradient map reflecting the overall and local temperature distribution characteristics of the fermentation tank includes: The temperature readings of all sensors in the current sampling period are obtained from the array-type temperature sensor network to form the original temperature data matrix; The original temperature data matrix is ​​subjected to outlier detection and correction, and abnormal temperature readings that exceed the reasonable physical range are replaced by interpolation using data from spatially adjacent sensors. The corrected temperature data matrix is ​​smoothed and filtered in the time dimension, and the moving average method is used to suppress random fluctuation noise in the sensor readings. Based on the three-dimensional geometric model of the fermentation tank and the spatial coordinates of the sensors, spatial interpolation is performed on the smoothed and filtered temperature data to reconstruct a continuous three-dimensional temperature distribution field covering the entire internal space of the fermentation tank. Calculate the rate of temperature change of the continuous three-dimensional temperature distribution field in the vertical depth direction, and identify the hot and cold spots in the horizontal direction where the temperature is higher or lower than the surrounding area. By combining the continuous three-dimensional temperature distribution field, the temperature change rate in the vertical depth direction, and the location information of hot and cold spots, a temperature gradient map containing isothermal surfaces, temperature gradient vectors, and anomaly region annotations is generated.

[0008] As a further aspect of the present invention, the improved multi-model PID controller performs parameter self-tuning based on the fermentation process stage and the microbial metabolic heat dynamic model, and its working principle includes: Multiple baseline PID parameter sets are preset to correspond to different fermentation process stages. Each baseline PID parameter set includes the nominal values ​​of proportional coefficient, integral coefficient, and derivative coefficient. A dynamic correlation model was established between the rate of microbial metabolic heat generation and the average temperature of the fermentation tank and the concentration of fermentation materials. The dynamic correlation model was used to predict the natural temperature change trend of the fermentation tank in the next control cycle. At the beginning of each control cycle, the current fermentation process stage identifier is received from the fermentation process stage identification module. Based on the current fermentation process stage identifier, the corresponding baseline PID parameter group is called as the starting point for parameter adjustment; Based on the deviation between the current temperature field and the set temperature field provided by the temperature gradient map, and the natural temperature change trend predicted by the dynamic correlation model, the mismatch between the current control requirements and the reference control parameters is calculated. Based on the magnitude and direction of the mismatch, the values ​​of the proportional coefficient, integral coefficient, and derivative coefficient are dynamically adjusted through an online optimization algorithm to form the optimal PID control parameters suitable for the current transient process. The optimal PID control parameters are applied to the deviation calculation of the current control cycle to generate a preliminary control output.

[0009] As a further aspect of the present invention, the establishment of a dynamic correlation model between the microbial metabolic heat generation rate and the average temperature of the fermentation tank and the concentration of fermentation materials includes: During the fermentation experiment, the cumulative curves of key microbial metabolic products were recorded simultaneously under different temperatures and material concentrations. The cumulative curves of the key metabolic products of the microorganisms were differentiated to obtain the instantaneous generation rate of the metabolites at different times, and this rate was used as a proxy indicator of the metabolic activity of the microorganisms. The correlation data between the instantaneous generation rate of the metabolites and the average temperature of the fermentation tank and the concentration of fermentation materials at the same time were analyzed, and an empirical model of the metabolic heat generation rate was obtained by fitting the data using a multivariate nonlinear regression method. The empirical model expresses the functional relationship between the rate of metabolic heat generation and the average temperature of the fermentation tank at a given concentration of fermentation material, and the functional relationship between the rate of metabolic heat generation and the concentration of fermentation material at a given temperature. The empirical model is coupled with the heat transfer equation of the fermentation tank to construct the dynamic correlation model that predicts the change in internal temperature of the fermentation tank due to the heat generated by the microorganisms themselves.

[0010] As a further aspect of the present invention, the improved multi-model PID controller dynamically calculates and outputs coordinated control commands for the heating and cooling systems based on the current stage of the fermentation process and the measured temperature field, including: The deviation between the overall average temperature of the current fermentation tank and the set temperature, as well as the temperature distribution uniformity index, are extracted from the temperature gradient map. The deviation between the overall average temperature and the set temperature, the temperature distribution uniformity index, and the current fermentation process stage identifier are all input into the control logic of the improved multi-model PID controller. Within the improved multi-model PID controller, independent proportional, integral, and derivative control components are calculated for both heating and cooling regulation methods. The heating control component and the cooling control component are compared in direction and magnitude. When the heating control component and the cooling control component act in opposite directions, they are canceled out to generate a net control demand. When they act in the same direction, they are superimposed and enhanced. The processed net control requirements or the superimposed and enhanced control requirements are converted into specific actuator instructions, including control instructions for the duty cycle of the heating element and control instructions for the opening of the cooling medium regulating valve. The control commands for the heating element and the cooling medium are integrated to form a unified coordinated control command, which includes both heating power adjustment and cooling flow rate adjustment.

[0011] As a further aspect of the present invention, the comparison of the direction and magnitude of the heating control component and the cooling control component, and the cancellation process performed when the heating control component and the cooling control component act in opposite directions to generate a net control requirement, includes: Setting the heating control component to the positive direction indicates that more heat energy input is needed; setting the cooling control component to the negative direction indicates that more cooling heat dissipation is needed. Calculate the algebraic sum of the heating control component and the cooling control component to obtain the preliminary net control quantity; When the initial net control quantity is positive, the current net demand is determined to be heating demand, and its value is equal to the initial net control quantity. At the same time, the cooling control component is set to zero. When the initial net control quantity is negative, the current net demand is determined to be cooling demand, the value of which is equal to the absolute value of the initial net control quantity, and the heating control component is set to zero. When the initial net control quantity is close to zero and within the preset dead zone range, it is determined that no heating or cooling adjustment is needed at present, and both the heating control component and the cooling control component are set to zero. The determined heating or cooling demand value is used as the final net control demand output.

[0012] As a further aspect of the present invention, the continuous monitoring of the generation rate of key microbial metabolites and inputting it as a feedback signal to the fermentation process stage identification module includes: The concentration of specific gaseous components directly related to the metabolic activities of key microorganisms is monitored online at the exhaust pipe of the fermentation tank or at a pre-set sampling point. These specific gaseous components include carbon dioxide and specific flavor ester gases. Alternatively, micro-fermentation samples can be collected automatically at regular intervals, and the concentrations of key metabolites, including specific organic acids or amino acids, in the samples can be determined using an online near-infrared spectrometer or electrochemical sensor. The generation rate of the key microbial metabolites is calculated based on the rate of change of the concentration of the specific gas component or the rate of change of the concentration of the key metabolites. The calculated data stream of the generation rate changing over time is input into the fermentation process stage identification module.

[0013] As a further aspect of the present invention, the fermentation process stage identification module is used to dynamically update the current fermentation process stage based on the changing trend of the metabolite generation rate, including: The fermentation process stage identification module pre-stores typical change patterns of the generation rate of key microbial metabolites in different fermentation process stages. The typical change patterns include rate characteristics of the start-up acceleration phase, logarithmic growth phase, stationary phase, and decline phase. The generation rate data of the key microbial metabolites input in real time are dynamically matched and the degree of consistency is calculated with the typical change patterns of each stage that are stored in advance. Identify the typical change pattern that best matches the current real-time data, and select the fermentation process stage corresponding to the typical change pattern as a candidate stage. Based on the fermentation process stages identified in the previous moment, the candidate stages are smoothly corrected using a state transition probability model to avoid frequent jumps between stages. Output the fermentation process stage identifier after smoothing correction, as the result of the current dynamically updated fermentation process stage judgment.

[0014] As a further aspect of the present invention, the step of feeding back the updated fermentation process stage information to the improved multi-model PID controller in real time to trigger online adaptive adjustment of the controller parameters includes: The improved multi-model PID controller continuously monitors the fermentation process stage identifier output by the fermentation process stage identification module. When a change in the fermentation process stage identifier is detected, the controller parameter self-tuning process is triggered; The controller parameter self-tuning process uses a baseline PID parameter set corresponding to the new fermentation process stage to replace the current PID control parameters. After parameter replacement, based on the temperature field deviation reflected by the temperature gradient map at the replacement time and the temperature trend predicted by the dynamic correlation model, a fast local parameter optimization is performed to fine-tune the benchmark PID parameter set so that it better fits the process characteristics of the instantaneous stage switching. The PID control parameters, after self-tuning and fine-tuning, are loaded into the controller and applied to subsequent temperature control calculations.

[0015] As a further aspect of the present invention, the method of spatially interpolating and reconstructing the smoothed and filtered temperature data based on the three-dimensional geometric model of the fermentation tank and the spatial coordinates of the sensor to generate a continuous three-dimensional temperature distribution field covering the entire internal space of the fermentation tank includes: The three-dimensional geometric model characterizing the internal structural dimensions of the fermentation tank is obtained, as well as the spatial coordinates of each sensor in the array-type temperature sensor network, wherein the spatial coordinates define the three-dimensional position of each sensor within the three-dimensional geometric model. The smoothed and filtered temperature data is associated with the spatial coordinates of each sensor to form a set of temperature data points discretely distributed in three-dimensional space. A spatial interpolation algorithm based on radial basis functions is adopted, using the spatial coordinates of all sensors as interpolation nodes and the smoothed and filtered temperature data as node values ​​to construct a spatially continuous scalar field interpolation function. Using the spatially continuous scalar field interpolation function, the temperature value of each virtual grid point is calculated within the entire fermentation tank space volume defined by the three-dimensional geometric model, according to a preset spatial resolution. By integrating the temperature values ​​of all virtual grid points, a continuous three-dimensional temperature distribution field is formed that is spatially continuous and numerically smooth. This continuous three-dimensional temperature distribution field completely describes the temperature state at any location inside the fermentation tank.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: Utilizing a multi-layered spatial array temperature sensor network, comprehensive data acquisition is achieved. Real-time temperature data from multiple regions and depths undergoes spatiotemporal fusion processing to generate standardized temperature gradient maps. This large-scale distributed data acquisition method covers all three-dimensional spatial levels of the fermentation tank, integrating temperature change information across time and space to fully present the overall and local temperature distribution characteristics within the space. This broadens the coverage of temperature monitoring, overcoming the limitations of single-point acquisition, accurately reflecting the differentiated characteristics of temperature distribution within the three-dimensional space, improving the data dimensions of temperature monitoring, and fully restoring the true temperature field distribution within the fermentation tank, making the temperature-related monitoring data more closely reflect actual operating conditions.

[0017] Continuous collection of data related to the generation rate of microbial metabolites allows for dynamic updates of the fermentation process based on data trends. Combined with a dynamic model of microbial metabolic heat, the parameters of a multi-model PID controller are autonomously tuned and adjusted online. Real-time fermentation status outputs coordinated control commands to synchronously regulate the operating power of heating elements and the flow rate of cooling media, constructing a complete closed-loop control structure. This overcomes the limitations of constant control parameters, adapts to the dynamic changes in metabolic heat production throughout fermentation, achieves synchronized and coordinated operation of heating and cooling structures, unifies the operating rhythm of temperature regulation, reduces operational deviations caused by independent control, improves the overall adaptability of temperature field regulation, and maintains a balanced temperature distribution within the fermentation tank. Attached Figure Description

[0018] Figure 1 The flowchart shows the adaptive temperature control method for fermentation tank of broad beans based on PID controller according to the present invention. Figure 2 A flowchart for generating temperature gradient maps through spatiotemporal fusion processing of temperature field data; Figure 3 A flowchart for establishing a dynamic correlation model of microbial metabolic heat. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] See Figure 1 This invention provides a method for adaptive temperature control of a fermentation tank for fermented soybeans based on a PID controller, the overall implementation of which is as follows: An array of temperature sensors is deployed within the multi-layered space of the fermentation tank, periodically collecting real-time temperature field data at different depths within the tank. The collected real-time temperature field data undergoes spatiotemporal fusion processing to generate a temperature gradient map reflecting the overall and local temperature distribution characteristics of the fermentation tank. This temperature gradient map is input into an improved multi-model PID controller, which self-tunes its parameters based on the fermentation process stage and a microbial metabolic thermal dynamic model. Using the improved multi-model PID controller, based on the current stage of the fermentation process and the measured temperature field, it dynamically calculates and outputs coordinated control commands for the heating and cooling systems. According to these coordinated control commands, the actuators adjust the power of the heating elements and the flow rate of the cooling medium, thereby regulating the temperature field of the fermentation tank. During the regulation process, the generation rate of key microbial metabolites is continuously monitored, and this generation rate is input as a feedback signal to the fermentation process stage identification module. Using this module, the current fermentation process stage is dynamically updated based on the changing trend of the metabolite generation rate. The updated fermentation process stage information is fed back to the improved multi-model PID controller in real time. This information is used to trigger online adaptive adjustment of the controller parameters, thereby forming a closed-loop control loop.

[0022] In one embodiment of the present invention, after the array-type temperature sensor network completes data acquisition, refer to... Figure 2The system acquires temperature readings from all sensors within the current sampling period from an array-type temperature sensor network, forming an original temperature data matrix. Outlier detection and correction are performed on this original temperature data matrix, replacing abnormal temperature readings outside the reasonable physical range using interpolation methods based on data from spatially adjacent sensors. The corrected temperature data matrix undergoes time-dimensional smoothing filtering, employing a moving average method to suppress random fluctuation noise in the sensor readings. A three-dimensional geometric model characterizing the internal structural dimensions of the fermentation tank, along with the spatial coordinates of each sensor in the array-type temperature sensor network, are obtained. These spatial coordinates define the three-dimensional position of each sensor within the three-dimensional geometric model. Based on the three-dimensional geometric model of the fermentation tank and the spatial coordinates of the sensors, spatial interpolation reconstruction is performed on the smoothed temperature data. This process associates the smoothed temperature data with the spatial coordinates of each sensor, forming a set of discretely distributed temperature data points in three-dimensional space. A spatial interpolation algorithm based on radial basis functions is used, with the spatial coordinates of all sensors as interpolation nodes and the smoothed temperature data as node values, to construct a spatially continuous scalar field interpolation function. Using the spatially continuous scalar field interpolation function, the temperature value of each virtual grid point is calculated within the entire fermentation tank volume defined by the three-dimensional geometric model, according to a preset spatial resolution. The temperature values ​​of all virtual grid points are integrated to form a spatially continuous and numerically smooth continuous three-dimensional temperature distribution field, which completely describes the temperature state at any location within the fermentation tank. The rate of temperature change in the vertical depth direction of the continuous three-dimensional temperature distribution field is calculated, and hot and cold spots with temperatures higher or lower than the surrounding area in the horizontal direction are identified. Combining the continuous three-dimensional temperature distribution field, the rate of temperature change in the vertical depth direction, and the information on hot and cold spot locations, a temperature gradient map containing isothermal surfaces, temperature gradient vectors, and anomaly region annotations is generated.

[0023] In the specific implementation, after periodic data acquisition from the array-type temperature sensor network, temperature readings from all sensors within the current sampling period are obtained. These temperature readings are constructed into an original temperature data matrix according to the sensor's placement within the fermentation tank space. In the specific implementation, outlier detection and correction are performed on the original temperature data matrix. The detection criteria are the physical reasonable range and the spatial continuity of adjacent sensor data. For abnormal temperature readings exceeding the reasonable physical range, data from spatially adjacent sensors are used to replace them using the Kriging interpolation method, generating a corrected temperature data matrix. In the specific implementation, the corrected temperature data matrix undergoes time-dimensional smoothing filtering using a moving average method with a window length of five sampling periods to suppress random fluctuation noise in the sensor readings, resulting in smoothed temperature data.

[0024] In some embodiments, spatial interpolation reconstruction is performed on the smoothed and filtered temperature data based on the three-dimensional geometric model of the fermentation tank and the spatial coordinates of each sensor in the array-type temperature sensor network. In a specific implementation, the three-dimensional geometric model of the fermentation tank is a cuboid model. The spatial coordinates of each sensor in the array-type temperature sensor network define the three-dimensional position of each sensor within the three-dimensional geometric model. The smoothed and filtered temperature data is associated with the spatial coordinates of each sensor to form a set of temperature data points discretely distributed in three-dimensional space. A spatial interpolation algorithm based on radial basis functions is used for reconstruction, constructing a spatially continuous scalar field interpolation function with the spatial coordinates of all sensors as interpolation nodes and the smoothed and filtered temperature data as node values. This interpolation function can be understood to have the following general form:

[0025] Where: symbol Represents spatial coordinates The temperature value obtained by interpolation, sign Represents the total number of sensors in an array-type temperature sensor network, symbol [symbol missing]. Indicates the first Spatial coordinates of each sensor, symbol It is the first The weighting coefficients corresponding to each sensor are to be determined, with the symbol... These are the selected radial basis functions; here, the Gaussian function is used, with the symbol... Indicates the calculation of Euclidean distance, symbol This is a low-order polynomial term used to guarantee the uniqueness of the solution. Using a spatially continuous scalar field interpolation function, the temperature value of each virtual grid point is calculated within the entire fermentation tank volume defined by the three-dimensional geometric model, at a preset spatial resolution of 10 cm x 10 cm x 10 cm. The temperature values ​​of all virtual grid points are integrated to form a spatially continuous and numerically smooth continuous three-dimensional temperature distribution field, which completely describes the temperature state at any location within the fermentation tank.

[0026] In practical implementation, the rate of temperature change in the vertical depth direction of the continuous three-dimensional temperature distribution field is calculated. The temperature difference between adjacent virtual grid layers is calculated using the central difference method, and then divided by the vertical distance between layers to obtain the vertical temperature gradient field. Hot spots with temperatures higher than the surrounding area and cold spots with temperatures lower than the surrounding area are identified in the horizontal direction. The identification method compares the temperature of each virtual grid point with the average temperature of its eight adjacent grid points at the same depth. If the difference exceeds a positive threshold, it is marked as a hot spot; if the difference is lower than a negative threshold, it is marked as a cold spot. Integrating the continuous three-dimensional temperature distribution field, the rate of temperature change in the vertical depth direction, and the location information of hot and cold spots, a temperature gradient map is generated. The temperature gradient map, in the form of a three-dimensional data field, includes isothermal surfaces, temperature gradient vectors, and anomaly region labels. It can be understood that the temperature gradient map is stored as a composite data structure containing spatial coordinates, temperature values, gradient vectors, and attribute labels, for the improved multi-model PID controller to read.

[0027] In one embodiment of the present invention, the improved multi-model PID controller performs parameter self-tuning based on the fermentation process stage and the microbial metabolic heat dynamic model. Its working principle includes the following: Multiple baseline PID parameter sets corresponding to different fermentation process stages are preset, each baseline PID parameter set containing nominal values ​​for proportional coefficient, integral coefficient, and derivative coefficient. A dynamic correlation model is established between the microbial metabolic heat generation rate and the average temperature of the fermentation tank and the concentration of fermentation materials. This dynamic correlation model is used to predict the natural temperature change trend of the fermentation tank in the next control cycle. At the beginning of each control cycle, the current fermentation process stage identifier provided by the fermentation process stage identification module is received. Based on the current fermentation process stage identifier, the corresponding baseline PID parameter set is called as the starting point for parameter adjustment. Based on the deviation between the current temperature field and the set temperature field provided by the temperature gradient map, and the natural temperature change trend predicted by the dynamic correlation model, the mismatch between the current control requirement and the baseline control parameters is calculated. Based on the magnitude and direction of the mismatch, the values ​​of the proportional coefficient, integral coefficient, and derivative coefficient are dynamically adjusted through an online optimization algorithm to form the optimal PID control parameters suitable for the current transient process. The optimal PID control parameters are applied to the deviation calculation of the current control cycle to generate a preliminary control output. The process of establishing a dynamic correlation model between the microbial metabolic heat generation rate and the average temperature of the fermentation tank and the concentration of fermentation materials is as follows. (See reference...) Figure 3During the fermentation experiment, the cumulative curves of key microbial metabolic products were recorded simultaneously under different temperatures and material concentrations. The cumulative curves of these key metabolic products were differentiated to obtain the instantaneous generation rate of metabolites at different times, which was then used as a proxy indicator of microbial metabolic activity. The correlation data between the instantaneous generation rate of metabolites and the average temperature of the fermentation tank and the concentration of fermentation materials at the same time were analyzed. An empirical model of metabolic heat generation rate was fitted using a multivariate nonlinear regression method. This empirical model expresses the functional relationship between the metabolic heat generation rate and the average temperature of the fermentation tank at a given fermentation material concentration, and the functional relationship between the metabolic heat generation rate and the concentration of fermentation materials at a given temperature. The empirical model was coupled with the heat transfer equation of the fermentation tank to construct a dynamic correlation model predicting the change in internal temperature of the fermentation tank due to microbial heat production.

[0028] In practical implementation, the improved multi-model PID controller performs parameter self-tuning based on the fermentation process stages and the microbial metabolic heat dynamic model. Its working principle includes pre-setting multiple baseline PID parameter sets corresponding to different fermentation process stages. Each baseline PID parameter set contains nominal values ​​for proportional coefficient, integral coefficient, and derivative coefficient. In some embodiments, the fermentation process stages are divided into the start-up phase, the main fermentation phase, and the late maturation phase. The baseline PID parameter set preset for the start-up phase has a higher nominal value for the integral coefficient, the baseline PID parameter set preset for the main fermentation phase has a moderate nominal value for the proportional coefficient, and the baseline PID parameter set preset for the late maturation phase has lower nominal values ​​for both the proportional coefficient and the derivative coefficient. In practical implementation, a dynamic correlation model is established between the microbial metabolic heat generation rate and the average temperature of the fermentation tank and the concentration of fermentation materials. The correlation model is used to predict the natural temperature change trend of the fermentation tank in the next control cycle. At the beginning of each control cycle, the improved multi-model PID controller receives the current fermentation process stage identifier provided by the fermentation process stage identification module. Based on the current fermentation process stage identifier, it calls the corresponding benchmark PID parameter set as the starting point for parameter adjustment. Based on the deviation between the current temperature field and the set temperature field provided by the temperature gradient map, and the natural temperature change trend predicted by the dynamic correlation model, it calculates the mismatch between the current control requirements and the benchmark control parameters. According to the magnitude and direction of the mismatch, the values ​​of the proportional coefficient, integral coefficient and derivative coefficient are dynamically adjusted through an online optimization algorithm to form the optimal PID control parameters suitable for the current transient process. The optimal PID control parameters are applied to the deviation calculation of the current control cycle to generate the initial control output.

[0029] In practical implementation, establishing a dynamic correlation model between the microbial metabolic heat generation rate and the average temperature of the fermentation tank and the concentration of fermentation materials includes simultaneously recording the cumulative curves of key microbial metabolic products under different temperature and material concentration conditions during the fermentation experiment. The cumulative curves of key microbial metabolic products are differentiated to obtain the instantaneous generation rate of metabolites at different times. This instantaneous generation rate of metabolites is used as a proxy indicator of microbial metabolic activity. The correlation data between the instantaneous generation rate of metabolites and the average temperature of the fermentation tank and the concentration of fermentation materials at the same time are analyzed. An empirical model of the metabolic heat generation rate is obtained by fitting using a multivariate nonlinear regression method. This empirical model expresses the functional relationship between the metabolic heat generation rate and the average temperature of the fermentation tank at a given fermentation material concentration, and the functional relationship between the metabolic heat generation rate and the concentration of fermentation materials at a given temperature. In some embodiments, the empirical model obtained by fitting using multivariate nonlinear regression has the following form:

[0030] Where: symbol This represents the rate of heat generation from microbial metabolism, with dimensions of power, and the symbol... This refers to the pre-factor, whose dimensions are related to the rate of metabolic heat generation. Same, symbol Indicates the concentration of fermentation materials, symbol It is a dimensionless exponent characterizing the dependence of the rate of metabolic heat generation on the concentration of the material, with the symbol... It is a natural constant, symbol It is the activation energy parameter that characterizes the reaction process, with the symbol... It is the universal gas constant, symbol The exponent represents the average thermodynamic temperature of the fermentation tank. The entire equation is dimensionless. An empirical model is coupled with the heat transfer equation of the fermentation tank, which describes the relationship between heat generation, conduction, and dissipation within the tank. This couples the model with the heat transfer equation of the fermentation tank to construct a dynamic correlation model that predicts the temperature changes within the fermentation tank due to microbial heat production. Optionally, the output of the dynamic correlation model is the predicted temperature of the fermentation tank in the next time period without external heating or cooling intervention.

[0031] In one embodiment of the present invention, the process of dynamically calculating and outputting coordinated control commands for the heating and cooling systems within the improved multi-model PID controller, based on the current stage of the fermentation process and the measured temperature field, is as follows: The deviation between the overall average temperature of the current fermentation tank and the set temperature, as well as the temperature distribution uniformity index, are extracted from the temperature gradient map. The deviation between the overall average temperature and the set temperature, the temperature distribution uniformity index, and the current fermentation stage identifier are input together into the control logic of the improved multi-model PID controller. Within the improved multi-model PID controller, independent proportional, integral, and derivative control components are calculated for both heating and cooling regulation methods. The directions and magnitudes of the heating control component and the cooling control component are compared. When the heating control component and the cooling control component act in opposite directions, they are canceled out to generate a net control demand; when they act in the same direction, they are superimposed and enhanced. The processed net control demand or the superimposed and enhanced control demand is converted into specific actuator commands, including control commands for the duty cycle of the heating element and control commands for the opening of the cooling medium regulating valve. The integrated control commands for heating elements and cooling media form a unified coordinated control command, which includes both heating power adjustment and cooling flow rate adjustment. The process of comparing the directions and magnitudes of the heating and cooling control components, and canceling them out when their directions are opposite, to generate a net control demand includes the following steps: The heating control component is set to a positive direction, indicating a need for increased heat input; the cooling control component is set to a negative direction, indicating a need for increased cooling dissipation. The algebraic sum of the heating and cooling control components is calculated to obtain a preliminary net control quantity. When the preliminary net control quantity is positive, the current net demand is determined to be a heating demand, and its value equals the preliminary net control quantity; simultaneously, the cooling control component is set to zero. When the preliminary net control quantity is negative, the current net demand is determined to be a cooling demand, and its value equals the absolute value of the preliminary net control quantity; simultaneously, the heating control component is set to zero. When the preliminary net control quantity is close to zero and within a preset dead zone, it is determined that no heating or cooling adjustment is needed, and both the heating and cooling control components are set to zero. The determined heating or cooling demand value is used as the final net control demand output.

[0032] In practical implementation, an improved multi-model PID controller dynamically calculates and outputs coordinated control commands for the heating and cooling systems based on the current stage of the fermentation process and the measured temperature field. It extracts the deviation between the overall average temperature of the fermentation tank and the set temperature, as well as the temperature distribution uniformity index, from the temperature gradient map. The temperature distribution uniformity index is obtained by calculating the standard deviation of the temperatures at all virtual grid points within the fermentation tank. In practical implementation, the deviation between the overall average temperature and the set temperature, the temperature distribution uniformity index, and the current fermentation stage identifier are input into the control logic of the improved multi-model PID controller. Internally, the improved multi-model PID controller calculates independent proportional control components, integral control components, and derivative control components for both heating and cooling regulation methods. In some embodiments, within a specific control cycle, the fermentation process is in the main fermentation phase, the temperature setpoint is 30 degrees Celsius, the extracted overall average temperature is 28.5 degrees Celsius, and the temperature distribution uniformity index is 2.1 degrees Celsius. The improved multi-model PID controller calculates the independent control components for heating and cooling based on these inputs.

[0033] The heating control component and the cooling control component are compared in direction and magnitude. When their directions are opposite, they are canceled out to generate a net control demand. When their directions are the same, they are superimposed to enhance the control. In specific implementations, the heating control component is set as positive, indicating a need for increased heat input, and the cooling control component is set as negative, indicating a need for increased cooling dissipation. The algebraic sum of the heating and cooling control components is calculated to obtain the initial net control quantity. In some embodiments, the heating and cooling control components calculated within one control cycle are shown in Table 1. Table 1: Calculation Table for Heating and Cooling Control Components

[0034] It is understandable that the algebraic sum of the heating control component and the cooling control component... Calculated using the following formula:

[0035] Where: symbol Indicates the initial net control amount, symbol The proportional component representing the heating control component, symbol The integral component representing the heating control component, symbol The derivative component representing the heating control component, symbol [symbol missing]. The proportional component representing the cooling control component, symbol The integral component representing the cooling control component, symbol This represents the differential component of the cooling control component. Substituting the values ​​into the table, the preliminary net control quantity is calculated. When the initial net control value is positive, the current net demand is determined to be heating demand, and its value equals the initial net control value. Simultaneously, the cooling control component is set to zero. When the initial net control value is negative, the current net demand is determined to be cooling demand, and its value equals the absolute value of the initial net control value. Simultaneously, the heating control component is set to zero. When the initial net control value is close to zero and within the preset dead zone, it is determined that no heating or cooling adjustment is needed, and both the heating and cooling control components are set to zero. The determined heating or cooling demand value is used as the final net control demand output.

[0036] In practical implementation, the processed net control requirements or the superimposed and enhanced control requirements are converted into specific actuator instructions. The control instruction for the heating element's duty cycle is calculated using a linear mapping relationship based on the final heating demand value. Similarly, the control instruction for the cooling medium regulating valve opening is calculated using a linear mapping relationship based on the final cooling demand value. These control instructions for the heating element and cooling medium are integrated to form a unified coordinated control instruction, which includes both heating power adjustment and cooling flow rate adjustment. Optionally, the coordinated control instruction is sent as a digital signal via a fieldbus to the power controller of the heating element and the actuator of the cooling medium regulating valve.

[0037] In one embodiment of the present invention, during the control process, the generation rate of key microbial metabolites is continuously monitored and input as a feedback signal to the fermentation process stage identification module. This monitoring process includes the following methods: Online monitoring of the concentration of specific gas components directly related to the metabolic activities of key microorganisms at the exhaust pipe of the fermentation tank or at a preset sampling point; these specific gas components include carbon dioxide and specific flavor ester gases. Alternatively, periodically and automatically collecting trace fermentation samples, and measuring the concentration of key metabolites, including specific organic acids or amino acids, in the samples using an online near-infrared spectrometer or electrochemical sensor. The generation rate of the key microbial metabolites is calculated based on the rate of change of the concentration of the specific gas components or the rate of change of the concentration of the key metabolites. The data stream of the calculated generation rate over time is input to the fermentation process stage identification module. The fermentation process stage identification module dynamically updates the current fermentation process stage based on the changing trend of the metabolite generation rate. The fermentation process stage identification module pre-stores typical change patterns of the generation rate of key microbial metabolites in different fermentation process stages, including rate characteristics of the acceleration phase, logarithmic growth phase, stationary phase, and decline phase. The real-time input data on the generation rate of the key microbial metabolites is dynamically matched and the degree of agreement is calculated with pre-stored typical change patterns for each stage. The typical change pattern with the highest degree of agreement with the current real-time data is identified, and the fermentation process stage corresponding to the typical change pattern is selected as a candidate stage. Combining the fermentation process stages identified at the previous moment, a state transition probability model is applied to smooth the candidate stages, avoiding frequent jumps between stages. The smoothed fermentation process stage identifier is output as the current dynamically updated fermentation process stage judgment result.

[0038] In practice, the generation rate of key microbial metabolites is continuously monitored and used as a feedback signal to the fermentation process stage identification module. An online gas analyzer is installed in the exhaust pipe of the fermentation tank to monitor the concentration of specific gas components directly related to the metabolic activities of key microorganisms. These specific gas components include carbon dioxide and specific flavor esters. Alternatively, by periodically and automatically collecting trace fermentation samples, the concentration of key metabolites, including specific organic acids, in the samples is determined using an online near-infrared spectrometer. The generation rate of key microbial metabolites is calculated based on the rate of change of the concentration of specific gas components or the rate of change of the concentration of key metabolites. The generation rate is calculated by dividing the concentration difference between two adjacent sampling times by the sampling time interval. The data stream of the calculated generation rate of key microbial metabolites over time is then input into the fermentation process stage identification module.

[0039] In some embodiments, the fermentation process stage identification module pre-stores typical change patterns of the generation rate of key microbial metabolites at different fermentation process stages. These typical change patterns include rate characteristics of the initiation acceleration phase, logarithmic growth phase, stationary phase, and decline phase. These characteristic patterns are stored in the module's database in the form of time series functions or discrete data points. The fermentation process stage identification module dynamically matches and calculates the degree of agreement between the real-time input generation rate data of key microbial metabolites and the pre-stored typical change patterns for each stage. Dynamic matching is achieved by calculating the correlation coefficient between the real-time data sliding window and each pre-stored pattern sequence; the result of the degree of agreement calculation is a quantified similarity score. The typical change pattern with the highest degree of agreement with the current real-time data is identified, and the fermentation process stage corresponding to this typical change pattern is selected as a candidate stage. In a specific implementation, a matching calculation process is shown in Table 2, which illustrates the calculated correlation coefficient values ​​between the real-time data window and each pre-stored pattern at a certain time point. Table 2: Calculation of Correlation Coefficients Between Real-Time Generation Rate Data and Typical Patterns at Each Stage

[0040] In the table above, the correlation coefficient The calculation formula is:

[0041] Where: symbol The Pearson correlation coefficient represents the relationship between a real-time data sequence and a pre-stored pattern sequence. (Symbol: Pearson) Indicates the number of data points within the sliding window, symbol Indicating the first data in the real-time generation rate data sequence The value of each data point, sign This represents the average value of all data points in the real-time generation rate data sequence, with the symbol […]. This represents the first element in a pre-stored sequence of typical change patterns. The value of each data point, sign This represents the average value of all data points in the typical change pattern sequence. Based on the calculation results, the typical change pattern with the highest consistency with real-time data is the logarithmic growth phase pattern, which has the highest correlation coefficient. Therefore, the logarithmic growth phase is selected as a candidate phase.

[0042] In some embodiments, a state transition probability model is applied to smooth the candidate stages, combining the previously identified fermentation process stages. This model defines the probability of transitioning from one stage to another. If a candidate stage differs from the previous stage but its transition probability is below a threshold, the previous stage's judgment is maintained. Optionally, the state transition probability model can be constructed based on historical fermentation process stage transition statistics. The smoothed fermentation process stage identifier is output as the currently dynamically updated fermentation process stage judgment result. It can be understood that the stage identifier output by the fermentation process stage identification module is a discrete state variable, for example, using numerical codes 1, 2, 3, and 4 to represent the start-up acceleration phase, logarithmic growth phase, stable phase, and decline phase, respectively.

[0043] In one embodiment of the present invention, updated fermentation process stage information is fed back to the improved multi-model PID controller in real time to trigger online adaptive adjustment of controller parameters. The improved multi-model PID controller continuously monitors the fermentation process stage identifier output by the fermentation process stage identification module. When a change in the fermentation process stage identifier is detected, a controller parameter self-tuning process is triggered. The controller parameter self-tuning process uses a baseline PID parameter set corresponding to the new fermentation process stage to replace the current PID control parameters. After parameter replacement, based on the temperature field deviation reflected by the temperature gradient spectrum at the replacement time and the temperature trend predicted by the dynamic correlation model, a rapid local parameter optimization is performed to fine-tune the baseline PID parameter set to better match the process characteristics at the moment of stage switching. The PID control parameters after self-tuning and fine-tuning are loaded into the controller and applied to subsequent temperature control calculations.

[0044] In practical implementation, the updated fermentation process stage information is fed back to the improved multi-model PID controller in real time to trigger online adaptive adjustment of controller parameters. The improved multi-model PID controller continuously monitors the fermentation process stage identifier output by the fermentation process stage identification module. In some embodiments, this monitoring is achieved through software interrupts or polling data-sharing variables. When a change in the fermentation process stage identifier is detected, the controller parameter self-tuning process is triggered. For example, a change in the stage identifier from "start-up" code "1" to "main fermentation" code "2" is considered a valid change event. The controller parameter self-tuning process uses the baseline PID parameter set corresponding to the new fermentation process stage to replace the current PID control parameters. The replacement operation directly updates the values ​​of the current proportional coefficient, integral coefficient, and derivative coefficient stored in memory to the nominal values ​​defined in the new baseline PID parameter set.

[0045] After parameter replacement, based on the temperature field deviation reflected in the temperature gradient map at the replacement time and the temperature trend predicted by the dynamic correlation model, a rapid local parameter optimization is performed to fine-tune the baseline PID parameter set to better match the process characteristics at the moment of stage switching. In specific implementation, the rapid local parameter optimization is based on a simplified cost function. It can be understood that the cost function J combines the relative magnitudes of the changes in the current control deviation and the predicted deviation, and its form is as follows:

[0046] Where: symbol The cost function value represents the cost of local optimization; it is a dimensionless scalar, symbolized by... , , These represent the proportional, integral, and derivative coefficients to be fine-tuned, respectively. The initial values ​​of these three parameters are the nominal values ​​of the new baseline PID parameter set, with the following symbols: Indicates the moment of phase transition. The deviation between the overall average temperature calculated from the temperature gradient map and the set temperature, sign... It is a predefined reference temperature deviation constant used to normalize deviations, with the sign... This represents the predicted natural temperature change trend due to microbial metabolic heat during the next control period, as determined by the dynamic correlation model. (Symbol: ) It is a dimensionless adjustment factor used to balance the weights of instantaneous bias and predictive trend. The local optimization algorithm iterates several times using gradient descent within a small neighborhood of the nominal value of the baseline parameter to find the factor that makes the cost function... Decrease the parameter to adjust the direction, and fine-tune the parameter.

[0047] In some embodiments, the fine-tuning process can be described as an incremental update of the baseline parameter, such as updating the scaling factor to... ,in This is a small incremental value obtained through local optimization calculation. The PID control parameters, after self-tuning and fine-tuning, are loaded into the controller and applied to subsequent temperature control calculations, marking the completion of online adaptive adjustment. Optionally, the loading operation means that the improved multi-model PID controller will use new, fine-tuned proportional, integral, and derivative coefficients in the next control cycle calculation.

[0048] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for adaptive temperature control of a fermentation tank for fermented soybeans based on a PID controller, characterized in that, The method includes: An array of temperature sensors is deployed in the multi-layered space of the fermentation tank to periodically collect real-time temperature field data at different depths inside the fermentation tank. The collected real-time temperature field data is subjected to spatiotemporal fusion processing to generate a temperature gradient map that reflects the overall and local temperature distribution characteristics of the fermentation tank. The temperature gradient map is input into the improved multi-model PID controller, which performs parameter self-tuning based on the fermentation process stage and the microbial metabolic heat dynamics model. Using the improved multi-model PID controller, the heating and cooling systems are dynamically calculated and output according to the current stage of the fermentation process and the measured temperature field. According to the coordinated control command, the control actuator adjusts the power of the heating element and the flow rate of the cooling medium to regulate the temperature field of the fermentation tank; During the regulation process, the generation rate of key microbial metabolites is continuously monitored and used as a feedback signal to the fermentation process stage identification module. The fermentation process stage identification module is used to dynamically update the current fermentation process stage based on the changing trend of the metabolite generation rate. The updated fermentation process stage information is fed back to the improved multi-model PID controller in real time to trigger online adaptive adjustment of the controller parameters, forming a closed-loop control loop.

2. The method for adaptive temperature control of a fermentation tank for fermented soybeans based on a PID controller according to claim 1, characterized in that, The process of performing spatiotemporal fusion processing on the collected real-time temperature field data to generate a temperature gradient map reflecting the overall and local temperature distribution characteristics of the fermentation tank includes: The temperature readings of all sensors in the current sampling period are obtained from the array-type temperature sensor network to form the original temperature data matrix; The original temperature data matrix is ​​subjected to outlier detection and correction, and abnormal temperature readings that exceed the reasonable physical range are replaced by interpolation using data from spatially adjacent sensors. The corrected temperature data matrix is ​​smoothed and filtered in the time dimension, and the moving average method is used to suppress random fluctuation noise in the sensor readings. Based on the three-dimensional geometric model of the fermentation tank and the spatial coordinates of the sensors, spatial interpolation is performed on the smoothed and filtered temperature data to reconstruct a continuous three-dimensional temperature distribution field covering the entire internal space of the fermentation tank. Calculate the rate of temperature change of the continuous three-dimensional temperature distribution field in the vertical depth direction, and identify the hot and cold spots in the horizontal direction where the temperature is higher or lower than the surrounding area. By combining the continuous three-dimensional temperature distribution field, the temperature change rate in the vertical depth direction, and the location information of hot and cold spots, a temperature gradient map containing isothermal surfaces, temperature gradient vectors, and anomaly region annotations is generated.

3. The method for adaptive temperature control of a fermentation tank for fermented soybeans based on a PID controller according to claim 1, characterized in that, The improved multi-model PID controller performs parameter self-tuning based on the fermentation process stages and the microbial metabolic heat dynamic model. Its working principle includes: Multiple baseline PID parameter sets are preset to correspond to different fermentation process stages. Each baseline PID parameter set includes the nominal values ​​of proportional coefficient, integral coefficient, and derivative coefficient. A dynamic correlation model was established between the rate of microbial metabolic heat generation and the average temperature of the fermentation tank and the concentration of fermentation materials. The dynamic correlation model was used to predict the natural temperature change trend of the fermentation tank in the next control cycle. At the beginning of each control cycle, the current fermentation process stage identifier is received from the fermentation process stage identification module. Based on the current fermentation process stage identifier, the corresponding baseline PID parameter group is called as the starting point for parameter adjustment; Based on the deviation between the current temperature field and the set temperature field provided by the temperature gradient map, and the natural temperature change trend predicted by the dynamic correlation model, the mismatch between the current control requirements and the reference control parameters is calculated. Based on the magnitude and direction of the mismatch, the values ​​of the proportional coefficient, integral coefficient, and derivative coefficient are dynamically adjusted through an online optimization algorithm to form the optimal PID control parameters suitable for the current transient process. The optimal PID control parameters are applied to the deviation calculation of the current control cycle to generate a preliminary control output.

4. The method for adaptive temperature control of a fermentation tank for fermented soybeans based on a PID controller according to claim 3, characterized in that, The establishment of a dynamic correlation model between the rate of microbial metabolic heat generation and the average temperature of the fermentation tank and the concentration of fermentation materials includes: During the fermentation experiment, the cumulative curves of key microbial metabolic products were recorded simultaneously under different temperatures and material concentrations. The cumulative curves of the key metabolic products of the microorganisms were differentiated to obtain the instantaneous generation rate of the metabolites at different times, and this rate was used as a proxy indicator of the metabolic activity of the microorganisms. The correlation data between the instantaneous generation rate of the metabolites and the average temperature of the fermentation tank and the concentration of fermentation materials at the same time were analyzed, and an empirical model of the metabolic heat generation rate was obtained by fitting the data using a multivariate nonlinear regression method. The empirical model expresses the functional relationship between the rate of metabolic heat generation and the average temperature of the fermentation tank at a given concentration of fermentation material, and the functional relationship between the rate of metabolic heat generation and the concentration of fermentation material at a given temperature. The empirical model is coupled with the heat transfer equation of the fermentation tank to construct the dynamic correlation model that predicts the change in internal temperature of the fermentation tank due to the heat generated by the microorganisms themselves.

5. The method for adaptive temperature control of a fermentation tank for fermented soybeans based on a PID controller according to claim 1, characterized in that, Using the improved multi-model PID controller, based on the current stage of the fermentation process and the measured temperature field, it dynamically calculates and outputs coordinated control commands for the heating and cooling systems, including: The deviation between the overall average temperature of the current fermentation tank and the set temperature, as well as the temperature distribution uniformity index, are extracted from the temperature gradient map. The deviation between the overall average temperature and the set temperature, the temperature distribution uniformity index, and the current fermentation process stage identifier are all input into the control logic of the improved multi-model PID controller. Within the improved multi-model PID controller, independent proportional, integral, and derivative control components are calculated for both heating and cooling regulation methods. The heating control component and the cooling control component are compared in direction and magnitude. When the heating control component and the cooling control component act in opposite directions, they are canceled out to generate a net control demand. When they act in the same direction, they are superimposed and enhanced. The processed net control requirements or the superimposed and enhanced control requirements are converted into specific actuator instructions, including control instructions for the duty cycle of the heating element and control instructions for the opening of the cooling medium regulating valve. The control commands for the heating element and the cooling medium are integrated to form a unified coordinated control command, which includes both heating power adjustment and cooling flow rate adjustment.

6. The method for adaptive temperature control of a fermentation tank for fermented soybeans based on a PID controller according to claim 5, characterized in that, The comparison of the direction and magnitude of the heating control component and the cooling control component, and the cancellation process when the heating control component and the cooling control component act in opposite directions to generate a net control requirement, includes: Setting the heating control component to the positive direction indicates that more heat energy input is needed; setting the cooling control component to the negative direction indicates that more cooling heat dissipation is needed. Calculate the algebraic sum of the heating control component and the cooling control component to obtain the preliminary net control quantity; When the initial net control quantity is positive, the current net demand is determined to be heating demand, and its value is equal to the initial net control quantity. At the same time, the cooling control component is set to zero. When the initial net control quantity is negative, the current net demand is determined to be cooling demand, the value of which is equal to the absolute value of the initial net control quantity, and the heating control component is set to zero. When the initial net control quantity is close to zero and within the preset dead zone range, it is determined that no heating or cooling adjustment is needed at present, and both the heating control component and the cooling control component are set to zero. The determined heating or cooling demand value is used as the final net control demand output.

7. The method for adaptive temperature control of a fermentation tank for fermented soybeans based on a PID controller according to claim 1, characterized in that, The continuous monitoring of the generation rate of key microbial metabolites, and the input of this rate as a feedback signal to the fermentation process stage identification module, includes: The concentration of specific gaseous components directly related to the metabolic activities of key microorganisms is monitored online at the exhaust pipe of the fermentation tank or at a pre-set sampling point. These specific gaseous components include carbon dioxide and specific flavor ester gases. Alternatively, micro-fermentation samples can be collected automatically at regular intervals, and the concentrations of key metabolites, including specific organic acids or amino acids, in the samples can be determined using an online near-infrared spectrometer or electrochemical sensor. The generation rate of the key microbial metabolites is calculated based on the rate of change of the concentration of the specific gas component or the rate of change of the concentration of the key metabolites. The calculated data stream of the generation rate changing over time is input into the fermentation process stage identification module.

8. The method for adaptive temperature control of a fermentation tank for fermented soybeans based on a PID controller according to claim 7, characterized in that, Using the fermentation process stage identification module, the current fermentation process stage is dynamically updated based on the changing trend of metabolite generation rate, including: The fermentation process stage identification module pre-stores typical change patterns of the generation rate of key microbial metabolites in different fermentation process stages. The typical change patterns include rate characteristics of the start-up acceleration phase, logarithmic growth phase, stationary phase, and decline phase. The generation rate data of the key microbial metabolites input in real time are dynamically matched and the degree of consistency is calculated with the typical change patterns of each stage that are stored in advance. Identify the typical change pattern that best matches the current real-time data, and select the fermentation process stage corresponding to the typical change pattern as a candidate stage. Based on the fermentation process stages identified in the previous moment, the candidate stages are smoothly corrected using a state transition probability model to avoid frequent jumps between stages. Output the fermentation process stage identifier after smoothing correction, as the result of the current dynamically updated fermentation process stage judgment.

9. The method for adaptive temperature control of a fermentation tank for fermented soybeans based on a PID controller according to claim 8, characterized in that, The step of feeding back the updated fermentation process stage information to the improved multi-model PID controller in real time to trigger online adaptive adjustment of controller parameters includes: The improved multi-model PID controller continuously monitors the fermentation process stage identifier output by the fermentation process stage identification module. When a change in the fermentation process stage identifier is detected, the controller parameter self-tuning process is triggered; The controller parameter self-tuning process uses a baseline PID parameter set corresponding to the new fermentation process stage to replace the current PID control parameters. After parameter replacement, based on the temperature field deviation reflected by the temperature gradient map at the replacement time and the temperature trend predicted by the dynamic correlation model, a fast local parameter optimization is performed to fine-tune the benchmark PID parameter set so that it better fits the process characteristics of the instantaneous stage switching. The PID control parameters, after self-tuning and fine-tuning, are loaded into the controller and applied to subsequent temperature control calculations.

10. The method for adaptive temperature control of a fermentation tank for fermented soybeans based on a PID controller according to claim 2, characterized in that, The three-dimensional geometric model of the fermentation tank and the spatial coordinates of the sensors are used to perform spatial interpolation reconstruction on the smoothed and filtered temperature data to generate a continuous three-dimensional temperature distribution field covering the entire internal space of the fermentation tank, including: The three-dimensional geometric model characterizing the internal structural dimensions of the fermentation tank is obtained, as well as the spatial coordinates of each sensor in the array-type temperature sensor network, wherein the spatial coordinates define the three-dimensional position of each sensor within the three-dimensional geometric model. The smoothed and filtered temperature data is associated with the spatial coordinates of each sensor to form a set of temperature data points discretely distributed in three-dimensional space. A spatial interpolation algorithm based on radial basis functions is adopted, using the spatial coordinates of all sensors as interpolation nodes and the smoothed and filtered temperature data as node values ​​to construct a spatially continuous scalar field interpolation function. Using the spatially continuous scalar field interpolation function, the temperature value of each virtual grid point is calculated within the entire fermentation tank space volume defined by the three-dimensional geometric model, according to a preset spatial resolution. By integrating the temperature values ​​of all virtual grid points, a continuous three-dimensional temperature distribution field is formed that is spatially continuous and numerically smooth. This continuous three-dimensional temperature distribution field completely describes the temperature state at any location inside the fermentation tank.

Images