Steviol glycoside production system based on dynamic weight adaptive mpc algorithm

Abstract

The present application relates to the technical field of steviol production, and discloses a steviol production system based on a dynamic weight self-adaptive MPC algorithm, which comprises a raw material pretreatment module, a solution extraction module, a purification module, a decolorization and post-treatment module, a crystallization and drying module, a central control module and an acquisition unit. The raw material pretreatment module crushes and enzymatically hydrolyzes sweet wormwood leaves, removes impurities and obtains an enzymatic hydrolysis solution. The solution extraction module extracts glycosides in the enzymatic hydrolysis solution. The purification module separates and purifies the glycosides. The decolorization and post-treatment module decolorizes the purified glycosides. The crystallization and drying module crystallizes and dries the decolorized glycosides. The acquisition unit acquires production data of all modules. The central control module uses an STM32 to process data and control production of all modules. The present application realizes real-time data acquisition, accurate algorithm control and improvement of the automation level of steviol production. The STM32 is used for control, thereby realizing full-process automation.

Description

Technical Field

[0001] This invention relates to the field of stevia production technology, and more specifically to a stevia production system. Background Technology

[0002] Stevia is a glycoside extracted from the leaves of the stevia plant, which is native to Paraguay and Brazil. It is characterized by its high sweetness and low calorie content. Its sweetness is 200-300 times that of sucrose, and its calorie value is only 1 / 300 of that of sucrose. Stevia is a new type of natural sweetener refined from the herbaceous plant stevia. Stevia has been used as a herb and sugar substitute for hundreds of years. Stevia has been widely used in the production of food, beverages and seasonings.

[0003] Traditional stevia extraction technology suffers from low raw material processing efficiency and loss of active ingredients. For example, it often results in high impurity content, which is difficult to remove efficiently using pulverization and acid / alkali methods. Furthermore, because steviol glycosides are highly heat-sensitive, high-temperature pretreatment can lead to decomposition, making temperature control a significant challenge. In addition, traditional processes present five major problems: First, uneven physical pulverization can affect penetration efficiency; second, the extraction time and temperature with water are difficult to control, resulting in low yield; third, the purification process is complex, relying on precipitation, centrifugation, and resin adsorption, making time control difficult; fourth, metal salt precipitation methods may lead to heavy metal contamination; fifth, the decolorizing agent has low glycoside adsorption efficiency during decolorization; and sixth, traditional crystallization methods suffer from uneven crystal particle size and hygroscopic product absorption. Therefore, a new automated stevia production system is urgently needed. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a stevia production system based on a dynamic weight adaptive MPC algorithm to solve the technical problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a stevia production system based on a dynamic weight adaptive MPC algorithm, comprising a raw material pretreatment module, a solution extraction module, a purification module, a decolorization and post-treatment module, a crystallization and drying module, a central control module, and a data acquisition unit. The raw material pretreatment module crushes and enzymatically hydrolyzes stevia leaves to remove impurities and obtain an enzymatic hydrolysate. The solution extraction module extracts glycosides from the enzymatic hydrolysate. The purification module separates and purifies the glycosides. The decolorization and post-treatment module decolorizes the purified glycosides and regenerates the adsorbent during decolorization. The crystallization and drying module crystallizes and dries the decolorized glycosides to complete the production of stevia. The data acquisition unit collects production data from all modules. The central control module uses an STM32 microcontroller for data processing and production control of all modules.

[0006] The data acquisition unit collects pressure data (YL) during stevia leaf pulverization and temperature data (WD) and pH data during enzymatic hydrolysis within the raw material pretreatment module. The acquisition unit then sends the collected data to the central control module. The central control module receives the data and calculates the predicted temperature (WY) and predicted pH (SY). The formula for calculating the predicted temperature (WY) is as follows: In the formula, n represents the number of seconds in which the data acquisition unit acquires data, WD N This is the latest temperature data collected in the nth second, WD N+1 WD represents the latest temperature data collected at the (n+1)th second, WD represents the latest temperature data inside the enzymatic hydrolysis vessel, and the calculated temperature prediction value WY represents the temperature data inside the enzymatic hydrolysis vessel five minutes later.

[0007] Furthermore, the formula for calculating the predicted pH value SY is as follows: PH N The pH value is the latest pH value collected at the nth second. N+1 The pH value is the latest pH data collected at the (n+1)th second. The calculated pH prediction value SY is the pH data in the enzymatic hydrolysis tank five minutes later. The central control module receives the temperature prediction value WY and the pH prediction value SY and controls the flow rate of the enzyme solution.

[0008] Furthermore, the raw material pretreatment module includes two steps in crushing stevia leaves: coarse crushing and fine crushing. During coarse crushing, an objective function J1 is used to control the crushing chamber. The formula for calculating the objective function J1 is as follows: In the formula W grain As the weight for particle size uniformity, D is set to 0.8. 50 Let be the median particle size of the powder's average particle size, and T be the temperature inside the grinding chamber. During fine grinding, the grinding chamber is controlled by an objective function J2. The formula for calculating the objective function J2 is as follows: In the formula W temp The temperature control weight is set to 0.6.

[0009] Furthermore, the acquisition unit uses a conductivity sensor and a near-infrared spectrometer to monitor the glycoside concentration in the extract within the solution extraction module in real time. The solution extraction module performs a heating phase and a steady-state phase. The heating phase lasts for ten minutes, and the objective function J3 is used to control the solution temperature. In the formula, T represents the temperature of the solution, and pH represents the acidity or alkalinity of the solution during extraction. During the steady-state phase, the objective function J4 is used to control the temperature and pH of the solution for extraction. The formula for calculating the objective function J4 is as follows: .

[0010] Furthermore, the purification module is controlled by a central control module during the purification process. This central control module uses simulated moving bed chromatography to control the purification module, and the simulated moving bed chromatography employs a dynamic weighted adaptive model predictive control algorithm. The dynamic weighted adaptive model predictive control algorithm first uses the system state equation C... k+1 Define a dynamic model of component concentration within the chromatographic column, with the system state equation C. k+1 The calculation formula is: In the formula, Ck is the glycoside concentration vector at the current time, uk is the control input value, wk is the external disturbance value, and A, B, and D are all system matrices identified from offline experimental data.

[0011] Furthermore, the dynamic weighted adaptive model predictive control algorithm within the central control module is processed through the system state equation C. k+1 After definition, time-varying weights are introduced through a dynamic weight objective function J. The formula for calculating the objective function J is as follows: In the formula To dynamically adjust the prediction step size weights based on the concentration gradient, The value is an adaptive adjustment based on the smoothness of historical control inputs, where i represents the current i-th dynamic adjustment of the concentration gradient, and N... P This represents the total number of times the concentration gradient was dynamically adjusted. This represents the dynamic changes in component concentration. Here, represents the target value for the component concentration; j represents the j-th historical control input smoothness adjustment; and Ne represents the total number of historical control input smoothness adjustments. The target value for adaptive adjustment. Adjustment values ​​for adaptive adjustment.

[0012] Furthermore, the steps for dynamic weight adjustment are as follows: First, real-time feedback optimization: dynamically increase the separation accuracy weight based on the glycoside concentration deviation detected by the UV-Vis spectrometer. Second, historical data learning: use historical process data stored in STM32 to generate a weight adjustment rule library through offline training, and match the optimal weight combination online. Third, disturbance compensation: automatically reduce the control quantity change weight when the pressure or pH sensor detects a sudden disturbance.

[0013] Furthermore, the central control module introduces time-varying weights through a dynamic weight objective function J, and finally applies constraints. These constraints include mobile phase ratio constraints, solenoid valve switching frequency constraints, and concentration stability constraints. The mobile phase ratio constraint requires that the water / ethanol ratio be adjusted between 0.3% and 0.7% each time. The solenoid valve switching frequency constraint requires that the solenoid valve switching frequency be greater than or equal to ten seconds each time. The concentration stability constraint requires that the glycoside concentration change be within 0.5% every five minutes.

[0014] Furthermore, the dynamic weights within the central control module are used for reinforcement learning with a reward function r, the calculation formula for which is: In the formula, NH represents the current energy consumption, 500 represents the standard energy consumption, and the unit of energy consumption is kilowatt-hour.

[0015] Furthermore, the crystallization and drying module includes a constant-rate drying stage and a falling-rate drying stage. The constant-rate drying stage is controlled by a dynamic function J5, and the calculation formula for the dynamic function J5 is as follows: In the formula W rate k is the drying rate weight. target To achieve the target drying efficiency, M represents the moisture content, and t represents time. The falling-rate drying stage is controlled using a dynamic function J6, calculated using the following formula: In the formula W uniform The temperature uniformity weight is Tmax, the maximum temperature is Tmin, the minimum temperature is M, and the water content is M.

[0016] The technical effects and advantages of this invention are as follows:

[0017] 1. This invention performs real-time data acquisition and precise algorithm control to improve the automation level of stevia production. It uses STM32 for control, thereby achieving full-process automation, reducing manual intervention, significantly shortening the production cycle, and allowing real-time adjustment of process parameters to maximize production efficiency. The real-time control capability of STM32 ensures the stability of key parameters and higher product purity.

[0018] 2. By calculating the predicted temperature value WY and the predicted pH value SY, this invention helps to understand the temperature and pH in the enzymatic hydrolysis tank five minutes later. Therefore, the STM32 in the central control module of this application will predict the change in enzyme activity based on the trend of temperature and pH value changes, and adjust the enzyme liquid flow valve 5 minutes in advance to keep the enzyme activity at its best.

[0019] 3. This invention uses a dynamic weighting mechanism to adaptively adjust the weights of the objective function based on the concentration gradient. It prioritizes separation accuracy during rapid component changes and reduces control energy consumption during steady-state conditions. Combined with modeling of temperature and flow velocity disturbances, it enhances the algorithm's anti-interference capability and optimizes matrix operations and iterative algorithms to ensure real-time operation on the central control module. This solves the problems of lag response in traditional PID control and insufficient adaptability of fixed-weight MPC. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall system composition of the present invention. Detailed Implementation

[0021] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The stevia production system based on the dynamic weight adaptive MPC algorithm involved in the present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Reference Figure 1 This invention provides a stevia production system based on a dynamic weighted adaptive MPC algorithm, including a raw material pretreatment module, a solution extraction module, a purification module, a decolorization and post-treatment module, a crystallization and drying module, a central control module, and a data acquisition unit. The raw material pretreatment module crushes and enzymatically hydrolyzes stevia leaves to remove impurities and obtain an enzymatic hydrolysate. The solution extraction module extracts glycosides from the enzymatic hydrolysate. The purification module separates and purifies the glycosides. The decolorization and post-treatment module decolorizes the purified glycosides and regenerates the adsorbent during decolorization. The crystallization and drying module crystallizes and dries the decolorized glycosides to complete the production of stevia. The data acquisition unit collects production data from all modules. The central control module uses an STM32 microcontroller for data processing and production control of all modules.

[0023] In this embodiment, STM32 is used for control, thereby achieving full-process automation, reducing manual intervention, significantly shortening the production cycle, and allowing for real-time adjustment of process parameters to maximize production efficiency, increase capacity flexibility, and enable 24-hour uninterrupted operation, greatly increasing output. The real-time control capability of STM32 ensures the stability of key parameters and higher product purity. Through sensor feedback and algorithm control, glycoside retention rate is improved, glycoside loss is reduced, the crystallization process is precisely controlled, and uniform crystals are generated to improve product quality. By optimizing the crushing, enzymatic hydrolysis, and extraction processes, raw material waste is reduced. Real-time monitoring of waste liquid ion concentration and automatic initiation of neutralization reactions reduce the risk of environmental penalties. Solvent recovery and recycling are controlled by STM32 to reduce solvent emissions and lower environmental pollution.

[0024] Reference Figure 1 The data acquisition unit collects pressure data (YL) during stevia leaf pulverization and temperature data (WD) and pH data during enzymatic hydrolysis within the raw material pretreatment module. The acquisition unit then sends the collected data to the central control module. The central control module receives the data and calculates the predicted temperature (WY) and predicted pH (SY). The formula for calculating the predicted temperature (WY) is as follows: In the formula, n represents the number of seconds in which the data acquisition unit acquires data, WD N This is the latest temperature data collected in the nth second, WD N+1 The temperature data is the latest collected at the (n+1)th second. WD is the latest temperature data inside the enzymatic hydrolysis vessel. The calculated temperature prediction value WY is the temperature data inside the enzymatic hydrolysis vessel five minutes later. The pH prediction value SY is calculated using the following formula: PH N The pH value is the latest pH value collected at the nth second. N+1 The pH value is the latest pH data collected at the (n+1)th second. The calculated pH prediction value SY is the pH data in the enzymatic hydrolysis tank five minutes later. The central control module receives the temperature prediction value WY and the pH prediction value SY and controls the flow rate of the enzyme solution.

[0025] In this embodiment, during the enzymatic hydrolysis of shredded stevia leaves, temperature data (WD) and pH data (PH) are collected, and a predicted temperature value (WY) and a predicted pH value (SY) are calculated. These predicted values ​​represent the temperature and pH within the hydrolysis tank five minutes later. Therefore, the STM32 microcontroller in the central control module predicts changes in enzyme activity based on the trends in temperature and pH, adjusting the enzyme flow valve five minutes in advance to maintain optimal enzyme activity. Furthermore, this application calculates the predicted temperature value (WY). During the calculation, the difference in temperature change within each second is first calculated, and the total data of changes within sixty seconds, or one minute, is accumulated. The cumulative calculation method is used instead of directly collecting data from the first second and the sixtieth second to avoid the problem of inaccurate temperature prediction caused by sudden changes within the sixtieth second during enzymatic hydrolysis. In addition, the temperature change within a total of five minutes is used to predict the temperature five minutes later to ensure the accuracy of the temperature prediction. That is, the temperature prediction value WY calculated by this application is accurate enough, and the pH prediction value SY is calculated in the same way as the temperature prediction value WY, and the calculation results are relatively accurate.

[0026] Reference Figure 1The central control module receives pressure data information YL, and monitors the particle distribution in the stevia leaf crushing chamber through the pressure data information YL. The central control module adjusts the ultrasonic power and frequency, and the crushing particle size of the stevia leaf is 50μm-100μm.

[0027] The raw material pretreatment module includes two steps in pulverizing stevia leaves: coarse crushing and fine crushing. During coarse crushing, the crushing chamber is controlled by an objective function J1. The formula for calculating the objective function J1 is as follows: In the formula W grain As the weight for particle size uniformity, D is set to 0.8. 50 Let be the median particle size of the powder's average particle size, and T be the temperature inside the grinding chamber. During fine grinding, the grinding chamber is controlled by an objective function J2. The formula for calculating the objective function J2 is as follows: In the formula W temp Temperature control weight is set to 0.6. For both coarse and fine crushing, an ultrasonic pulverization dynamic model is used to predict the degree of powder pulverization. The ultrasonic pulverization dynamic model is... In the formula, k is the raw material characteristic coefficient, t is the crushing time, and P is the pressure inside the crushing chamber during crushing.

[0028] In this embodiment, a pressure sensor is used to monitor the particle distribution in the stevia leaf grinding chamber in real time, and the ultrasonic power and frequency are dynamically adjusted to ensure uniform particle size. The target particle size is 50μm-100μm, which reduces the oxidation loss of glycosides due to excessive grinding. The grinding of stevia leaves in this application includes two steps: coarse grinding and fine grinding. The objective function J1 is used for coarse grinding to quickly reduce the particle diameter, which is convenient for subsequent fine grinding. During fine grinding, the diameter of the particles after grinding is precisely controlled. In both coarse and fine grinding, an ultrasonic grinding dynamic model is used to predict the degree of powder grinding to ensure the accuracy of the prediction.

[0029] Reference Figure 1 The acquisition unit uses a conductivity sensor and a near-infrared spectrometer to monitor the glycoside concentration in the extract in the solution extraction module in real time. The acquisition unit transmits the acquired glycoside concentration data to the central control module. The central control module receives the acquired concentration data, calculates the concentration change rate K per unit minute, and compares the calculated concentration change rate K with a threshold of 1% / min. When the concentration change rate K is less than the threshold, the central control module sends a command to terminate heating and start the cooling process. When the concentration change rate K is greater than or equal to the threshold, the operation continues.

[0030] The acquisition unit uses a conductivity sensor and a near-infrared spectrometer to monitor the glycoside concentration in the extract solution in the solution extraction module in real time. The solution extraction module performs two phases during extraction: a heating phase and a steady-state phase. The heating phase lasts for ten minutes, and the objective function J3 controls the temperature rise. In the formula, T represents the temperature of the solution, and pH represents the acidity or alkalinity of the solution during extraction. During the steady-state phase, the objective function J4 is used to control the temperature and pH of the solution for extraction. The formula for calculating the objective function J4 is as follows: Furthermore, the thermodynamic equilibrium equation is used for prediction during solution extraction. The thermodynamic equilibrium equation is: In the formula, C glycoside The current concentration of the solution is given, Ea is the activation energy, k is the dissolution rate constant, and C is the concentration of the solution. solid This represents the initial concentration of steviol glycosides.

[0031] In this embodiment, an integrated conductivity sensor, near-infrared spectroscopy module, temperature sensor, and liquid level sensor are used to collect extraction liquid data in real time and transmit it to the central control module. The central control module detects the rate of change in glycoside concentration and controls the heating and cooling processes. If the temperature exceeds the limit or the liquid level is abnormal, heating is immediately cut off and an alarm is triggered to avoid over-extraction and safety hazards, and to improve batch consistency. In this application, different methods are used to control the temperature and pH during the heating and steady-state stages when extracting glycoside concentration from the extraction liquid, thereby ensuring that the final extracted glycoside concentration is higher.

[0032] Reference Figure 1 The purification module is controlled by a central control module during purification processing. This central control module uses simulated moving bed chromatography to control the purification module, and the simulated moving bed chromatography employs a dynamic weighted adaptive model predictive control algorithm. The dynamic weighted adaptive model predictive control algorithm first uses the system state equation C... k+1 Define a dynamic model of component concentration within the chromatographic column, with the system state equation C. k+1 The calculation formula is: In the formula, Ck is the glycoside concentration vector at the current time, uk is the control input value, wk is the external disturbance value, and A, B, and D are the system matrices identified from offline experimental data. Then, time-varying weights are introduced through a dynamic weight objective function J. The formula for calculating the objective function J is as follows: In the formula To dynamically adjust the prediction step size weights based on the concentration gradient, The value is an adaptive adjustment based on the smoothness of historical control inputs, where i represents the current i-th dynamic adjustment of the concentration gradient, and N... P This represents the total number of times the concentration gradient was dynamically adjusted. This represents the dynamic changes in component concentration. Here, represents the target value for the component concentration; j represents the j-th historical control input smoothness adjustment; and Ne represents the total number of historical control input smoothness adjustments. The target value for adaptive adjustment. To ensure adaptive adjustment, the final adjustment values ​​are constrained by several conditions, including mobile phase ratio constraints, solenoid valve switching frequency constraints, and concentration stability constraints. The mobile phase ratio constraint requires the water / ethanol ratio to be adjusted between 0.3% and 0.7% each time. The solenoid valve switching frequency constraint requires the solenoid valve to switch at least once every ten seconds. The concentration stability constraint requires the glycoside concentration to change by less than 0.5% every five minutes. The dynamic weights within the central control module are reinforced using a reward function r. The formula for calculating the reward function r is... In the formula, NH represents the current energy consumption, 500 represents the standard energy consumption, and the unit of energy consumption is kilowatt-hour.

[0033] In this embodiment, the dynamic weighting mechanism can adaptively adjust the weights of the objective function through the concentration gradient, prioritizing separation accuracy during rapid component changes and reducing control energy consumption during steady-state conditions. Combined with modeling disturbances such as temperature and flow rate, it enhances the algorithm's anti-interference capability and optimizes matrix operations and iterative algorithms to ensure real-time operation on the central control module. This formula and algorithm solve the problems of lag response in traditional PID control and insufficient adaptability of fixed-weight MPC by integrating model predictive control, dynamic weight optimization, and machine learning compensation mechanisms.

[0034] Reference Figure 1 The decolorization and post-treatment module uses magnetic nano-adsorbent for dosage control. The decolorization and post-treatment module uses an online spectrometer to detect the pigment content and sends it to the central control module. The central control module uses a neural network algorithm to calculate the optimal adsorbent dosage. The neural network algorithm first uses historical pigment content, adsorbent dosage and glycoside loss rate data as training set to train the neural network. The value corresponding to the lowest glycoside loss rate is the optimal adsorbent dosage.

[0035] In this embodiment, after the online spectrometer detects the pigment content, the STM32 in the central control module uses a neural network algorithm to calculate the optimal adsorbent dosage. Historical pigment content, adsorbent dosage, and glycoside loss rate data are used as a training set to train the neural network. During actual operation, the current pigment content data is input, and the neural network outputs the optimal dosage. After each adsorbent addition, a 5-minute wait is taken, and the pigment content is detected again. If the target decolorization effect is not achieved, adsorbent is added proportionally (20% of the deviation) according to the deviation value. Simultaneously, the glycoside concentration is monitored. If the glycoside loss rate approaches 3%, the addition is immediately stopped. The online spectrometer detects the pigment content and transmits it to the STM32. The STM32 uses the least squares method to calculate the optimal adsorbent dosage. After addition, the pigment content is detected again, and adsorbent is added proportionally according to the deviation. The glycoside concentration is monitored simultaneously. If the loss rate approaches the threshold, the addition is stopped. After each batch of production, the dosage calculation model is optimized based on the pigment removal rate and glycoside loss rate.

[0036] Furthermore, during post-treatment, namely magnetic field separation and adsorbent regeneration, the central control module controls the on / off cycle of the electromagnet. For the first separation, a frequency of 1 second on and 1 second off is used, for a total of 3 times. If the separation effect is not good, the frequency is adjusted to 2 seconds on and 1 second off until the adsorbent is separated quickly. During the adsorbent regeneration process, the flow rate of the regeneration solution is dynamically adjusted according to the number of times the adsorbent is used and the degree of contamination. When the new adsorbent is regenerated for the first time, the flow rate of dilute hydrochloric acid is 0.5 L / min. For every 5 additional uses, the flow rate is increased by 0.1 L / min to ensure the regeneration effect of the adsorbent and extend its reuse count.

[0037] Reference Figure 1 The crystallization and drying module includes a crystallization unit and a drying unit. The crystallization unit adjusts the temperature gradient and stirring rate inside the crystallizer through a central control module. The central control module uses a PID algorithm for adjustment. When adjusting the temperature gradient and stirring rate inside the crystallizer using the PID algorithm, a temperature integral separation coefficient is set. When the temperature deviation is less than ±0.05℃, the integral action is canceled. When the deviation is greater than ±0.05℃, the integral action is restored. The stirring rate is adjusted using variable speed control. The first 30 minutes are the initial crystallization period, and the stirring rate is increased from 100 rpm to 150 rpm. From 30 to 90 minutes is the middle crystallization period, and the stirring rate is maintained at 200 rpm. After 90 minutes is the later crystallization period, and the stirring rate is reduced to 150 rpm.

[0038] In this embodiment, when the central control module uses the PID algorithm to adjust the temperature gradient and stirring rate within the crystallizer, it sets a temperature integral separation coefficient. When the temperature deviation is less than ±0.05℃, the integral action is canceled to avoid overshoot caused by integral saturation. When the deviation exceeds this range, the integral action is restored. The stirring rate adjustment adopts variable speed control. In the first 30 minutes, which is the initial stage of crystallization, the stirring rate is increased from 100rpm to 150rpm to promote crystal nucleus formation. From 30 to 90 minutes, which is the middle stage of crystallization, the stirring rate is maintained at 200rpm to facilitate crystal growth. After 90 minutes, which is the later stage of crystallization, the stirring rate is reduced to 150rpm to prevent crystal breakage. The crystal particle size distribution is detected every 15 minutes, and the temperature gradient and stirring rate are fine-tuned based on the D90 value. The central control module collects data through temperature and stirring rate sensors and uses an adaptive PID algorithm to adjust the crystallizer parameters. The PID parameters are automatically adjusted at different stages of crystallization. At the same time, the temperature gradient and stirring rate are fine-tuned based on the crystal particle size distribution feedback. The crystal growth data is analyzed every 30 minutes to optimize the PID parameters and control strategy.

[0039] Reference Figure 1 The crystallization and drying module includes a constant-rate drying stage and a falling-rate drying stage. The constant-rate drying stage is controlled by a dynamic function J5, and the calculation formula for the dynamic function J5 is as follows: In the formula W rate k is the drying rate weight. target To achieve the target drying efficiency, M represents the moisture content, and t represents time. The falling-rate drying stage is controlled using a dynamic function J6, calculated using the following formula: In the formula W uniform Assuming a weighted average for temperature uniformity, Tmax represents the highest temperature, Tmin the lowest temperature, and M the moisture content. The crystallization and drying module uses Fick's diffusion law and heat transfer equation for prediction. Fick's diffusion law and heat transfer equation are as follows: In the formula, D is the moisture diffusion coefficient. Let h be the spatial second derivative of the moisture content, h be the convective heat transfer coefficient, ρ be the material density, and C be the density of the material. P T represents the specific heat capacity of the material. air For air temperature, T material The temperature is the material temperature.

[0040] In this embodiment, fluidized bed drying is optimized by integrating a humidity sensor and a hot air flow meter. The central control module adjusts the drying temperature and air speed according to the humidity change rate. When the humidity change rate is greater than 0.1% / min, the drying temperature is increased by 2°C, and the air speed is increased by 0.1m / s. If the humidity change rate is less than 0.1% / min, the temperature is decreased by 1°C, and the air speed is decreased by 0.05m / s. When the finished product moisture content is 0.3%, it enters the fine drying stage. The moisture content is detected every 10 minutes, and the temperature is adjusted by 0.5°C and the air speed by 0.02m / s each time until the moisture content reaches the standard, at which point nitrogen filling and sealing are automatically triggered. Therefore, the crystallization and drying module of this application includes a constant-rate drying stage and a falling-rate drying stage. The constant-rate drying stage can rapidly reduce the moisture content, and the falling-rate drying stage allows for precise control of the moisture content, ultimately making the final product moisture content more in line with the requirements.

[0041] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A stevia production system based on a dynamic weight adaptive MPC algorithm, characterized in that: The system includes a raw material pretreatment module, a solution extraction module, a purification module, a decolorization and post-treatment module, a crystallization and drying module, a central control module, and a data acquisition unit. The raw material pretreatment module crushes and enzymatically hydrolyzes stevia leaves to remove impurities and obtain an enzymatic hydrolysate. The solution extraction module extracts glycosides from the enzymatic hydrolysate. The purification module separates and purifies the glycosides. The decolorization and post-treatment module decolorizes the purified glycosides and regenerates the adsorbent during decolorization. The crystallization and drying module crystallizes and dries the decolorized glycosides to complete the production of stevia. The data acquisition unit collects production data from all modules. The central control module uses an STM32 microcontroller for data processing and production control of all modules. The central control module receives the data sent by the acquisition unit and calculates the predicted temperature value WY and the predicted pH value. The formula for calculating the predicted temperature value WY is as follows: In the formula, n represents the number of seconds in which the data acquisition unit acquires data, WD N This is the latest temperature data collected in the nth second, WD N+1 WD represents the latest temperature data collected at the (n+1)th second, WD represents the latest temperature data inside the enzymatic hydrolysis tank, and the calculated temperature prediction value WY represents the temperature data inside the enzymatic hydrolysis tank five minutes later. The dynamic weighted adaptive model predictive control algorithm within the central control module, after being defined by the system state equations, introduces time-varying weights through the dynamic weighted objective function J. The formula for calculating the objective function J is as follows: In the formula To dynamically adjust the prediction step size weights based on the concentration gradient, The value is an adaptive adjustment based on the smoothness of historical control inputs, where i represents the current i-th dynamic adjustment of the concentration gradient, and N... P This represents the total number of times the concentration gradient was dynamically adjusted. This represents the dynamic changes in component concentration. Here, represents the target value for the component concentration; j represents the j-th historical control input smoothness adjustment; and Ne represents the total number of historical control input smoothness adjustments. The target value for adaptive adjustment. Adjustment values ​​for adaptive adjustment; The steps for dynamic weight adjustment are as follows: First, real-time feedback optimization: dynamically increase the separation accuracy weight based on the glycoside concentration deviation detected by the UV-Vis spectrometer. Second, historical data learning: use historical process data stored in STM32 to generate a weight adjustment rule library through offline training and match the optimal weight combination online. Third, disturbance compensation: automatically reduce the control quantity change weight when the pressure or pH sensor detects a sudden disturbance. After introducing time-varying weights through the dynamic weight objective function J, the central control module finally applies constraints, including mobile phase ratio constraints, solenoid valve switching frequency constraints, and concentration stability constraints. The mobile phase ratio constraint requires that the water / ethanol ratio be adjusted between 0.3% and 0.7% each time. The solenoid valve switching frequency constraint requires that the solenoid valve switching frequency be greater than or equal to ten seconds each time. The concentration stability constraint requires that the glycoside concentration change be within 0.5% every five minutes.

2. The stevia production system based on dynamic weight adaptive MPC algorithm according to claim 1, characterized in that: The formula for calculating the predicted pH value SY is as follows: PH N The pH value is the latest pH value collected at the nth second. N+1 The pH value is the latest pH data collected at the (n+1)th second. The calculated pH prediction value SY is the pH data in the enzymatic hydrolysis tank five minutes later. The central control module receives the temperature prediction value WY and the pH prediction value SY and controls the flow rate of the enzyme solution.

3. The stevia production system based on dynamic weight adaptive MPC algorithm according to claim 1, characterized in that: The raw material pretreatment module includes two steps in pulverizing stevia leaves: coarse crushing and fine crushing. During coarse crushing, the crushing chamber is controlled by an objective function J1. The formula for calculating the objective function J1 is as follows: In the formula W grain As the weight for particle size uniformity, D is set to 0.

8. 50 Let be the median particle size of the powder's average particle size, and T be the temperature inside the grinding chamber. During fine grinding, the grinding chamber is controlled by an objective function J2. The formula for calculating the objective function J2 is as follows: In the formula W temp The temperature control weight is set to 0.

6.

4. The stevia production system based on dynamic weight adaptive MPC algorithm according to claim 1, characterized in that: The acquisition unit uses a conductivity sensor and a near-infrared spectrometer to monitor the glycoside concentration in the extract solution in the solution extraction module in real time. The solution extraction module performs two phases during extraction: a heating phase and a steady-state phase. The heating phase lasts for ten minutes, and the objective function J3 controls the temperature rise. In the formula, T represents the temperature of the solution, and pH represents the acidity or alkalinity of the solution during extraction. During the steady-state phase, the objective function J4 is used to control the temperature and pH of the solution for extraction. The formula for calculating the objective function J4 is as follows: .

5. The stevia production system based on dynamic weight adaptive MPC algorithm according to claim 1, characterized in that: The purification module is controlled by a central control module during purification processing. This central control module uses simulated moving bed chromatography to control the purification module, and the simulated moving bed chromatography employs a dynamic weighted adaptive model predictive control algorithm. The dynamic weighted adaptive model predictive control algorithm first uses the system state equation C... k+1 Define a dynamic model of component concentration within the chromatographic column, with the system state equation C. k+1 The calculation formula is: In the formula, Ck is the glycoside concentration vector at the current time, uk is the control input value, wk is the external disturbance value, and A, B, and D are all system matrices for offline experimental data identification.

6. The stevia production system based on dynamic weight adaptive MPC algorithm according to claim 1, characterized in that: The dynamic weights within the central control module are used for reinforcement learning with a reward function r. The formula for calculating the reward function r is as follows: In the formula, NH represents the current energy consumption, 500 represents the standard energy consumption, and the unit of energy consumption is kilowatt-hour.

7. The stevia production system based on dynamic weight adaptive MPC algorithm according to claim 1, characterized in that: The crystallization and drying module includes a constant-rate drying stage and a falling-rate drying stage. The constant-rate drying stage is controlled by a dynamic function J5, and the calculation formula for the dynamic function J5 is as follows: In the formula W rate k is the drying rate weight. target To achieve the target drying efficiency, M represents the moisture content, and t represents time. The falling-rate drying stage is controlled using a dynamic function J6, calculated using the following formula: In the formula W uniform The temperature uniformity weight is Tmax, the maximum temperature is Tmin, the minimum temperature is M, and the water content is M.

Images