Wine fermentation predictive monitoring system based on machine learning and digital twinning

Abstract

The invention relates to a grape wine fermentation predictive monitoring system based on machine learning and digital twinning, and belongs to the field of automation control. Real-time operation parameters and system configuration parameters of a fermentation process are acquired through a data acquisition module; when the power failure of the power grid is monitored, the system determines net heat flux formed by yeast metabolism heat production and environment incoming heat based on the fermentation liquid temperature, the environment temperature and the sugar degree; based on the temperature of the fermentation liquid and the temperature of the cooling medium, total energy absorbable by the wine liquid and the cooling system is determined. The system divides the total absorbable energy by the net heat flux to solve a toughness time window so as to predict the time length required for the fermentation liquid temperature to reach the safety upper limit; the time window is compared with a preset threshold to determine a risk level. Through the accurate thermodynamic coupling model and the self-optimization capability, the prospective prediction of the fermentation risk is realized, and a decision basis and an intervention opportunity are provided for operators.

Description

Technical Field

[0001] This invention relates to the field of automation control, specifically to a predictive monitoring system for wine fermentation based on machine learning and digital twins. Background Technology

[0002] The quality of wine largely depends on temperature control during fermentation. Yeasts are most active within a specific temperature range, efficiently converting sugar into alcohol and producing desirable flavor compounds. Excessive temperature can cause premature yeast death, resulting in undesirable flavors or even halting fermentation; conversely, excessively low temperatures can inhibit yeast activity, prolonging the fermentation cycle or causing fermentation to stop altogether.

[0003] Existing wine fermentation temperature monitoring systems typically employ PID control logic, using temperature sensors to monitor the wine temperature in the fermentation tank in real time and compare it with the set temperature, thereby controlling the start and stop of cooling / heating equipment. Such systems perform well under conditions of stable power supply and gentle environmental changes.

[0004] However, in certain special brewing scenarios, such as production areas with high altitudes, variable climates, and unstable power grids, existing technical solutions have revealed significant shortcomings: In regions with unstable power grids, power outages are a common risk. Traditional systems cannot assess the safe time window during a power outage when the system can still withstand the temperature rise using its own heat capacity and cooling reserves. Brewers cannot obtain quantitative risk warnings, such as how many hours the system can safely continue to operate, and therefore cannot make timely intervention decisions.

[0005] Existing systems treat the endogenous heat source and heat exchange with the external environment during fermentation as independent disturbances, failing to establish a coupling model between these two factors and the energy storage state of the cooling system. This results in the system's inability to accurately predict the temperature change trend of the fermenter under the combined effects of internal and external heat sources in extreme situations such as power outages.

[0006] The information disclosed in the background section above is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] The purpose of this invention is to provide a predictive monitoring system for wine fermentation based on machine learning and digital twins, in order to solve the problems mentioned in the background art.

[0008] The technical solution of this invention is: a predictive monitoring system for wine fermentation based on machine learning and digital twins.

[0009] In this embodiment, it includes: The data acquisition module is used to acquire real-time operating parameters of the fermentation process, as well as preset system configuration parameters; The real-time operating parameters include: fermentation broth temperature, ambient temperature, cooling medium temperature, power grid status, and sugar content. The system configuration parameters include: upper limit of safe temperature and warning time threshold; The first processing module is used to determine the net heat flux of the fermentation system based on the fermentation broth temperature, the ambient temperature, and the sugar content when the power grid status in the real-time operating parameters is a power outage. Wherein, the net heat flux is the sum of the heat generated by yeast metabolism and the heat transferred from the environment; The second processing module is used to determine the total energy that the system can absorb based on the temperature of the fermentation broth, the temperature of the cooling medium, and the upper limit of the safe temperature when the power grid is in a state of power failure. The total absorbable energy is the sum of the heat that the wine can absorb and the energy absorbed by the cooling system. The third processing module is used to divide the total absorbable energy determined by the second processing module by the net heat flux determined by the first processing module to calculate the toughness time window. The resilience time window is designed to characterize the predicted time required for the fermentation broth temperature to reach the upper limit of the safe temperature after a power outage. The risk assessment module is used to compare the resilience time window calculated by the third processing module with the preset early warning time threshold in the system configuration parameters to determine the risk level of the current fermentation process.

[0010] Preferably, the first processing module determines the net heat flux by: Calculate the metabolic heat production power of yeast; Calculate the thermal power transferred from the environment; The yeast metabolic heat production power is added to the environmental heat input power to generate the net heat flux, wherein the net heat flux is intended to quantify the total heat flow rate that drives the temperature rise of the fermentation broth under power outage conditions.

[0011] Preferably, the calculation of the yeast metabolic heat production power is based on the fermentation liquid temperature and sugar content obtained by the data acquisition module, combined with preset wine liquid density, wine liquid volume, yeast heat production rate per unit mass and yeast activity correction coefficient. The yeast metabolic heat production power is accurately calculated by multiplying the density of the wine liquid, the volume of the wine liquid, the heat production rate per unit mass of yeast, and the yeast activity correction coefficient.

[0012] Preferably, the calculation of the environmental heat transfer power is based on the temperature of the fermentation liquid, the ambient temperature, and the preset total heat transfer coefficient and outer surface area of ​​the fermentation tank. The ambient heat transfer power is calculated by calculating the difference between the ambient temperature and the fermentation broth temperature, and then multiplying the difference by the total heat transfer coefficient of the fermenter and the outer surface area of ​​the fermenter.

[0013] Preferably, the second processing module determines the total absorbable energy by: Calculate the calories that the wine can absorb; Calculate the energy absorbed by the cooling system's buffer. The heat that the liquid can absorb is added to the energy absorbed by the cooling system to generate the total absorbable energy, wherein the total absorbable energy is intended to quantify the total heat capacity of the system before it heats up to the upper limit of the safe temperature under power failure conditions.

[0014] Preferably, the calculation of the buffer absorption energy of the cooling system is based on the temperature of the fermentation broth, the temperature of the cooling medium, and the preset total mass and specific heat capacity of the cooling medium. The passive heat absorption capacity that the residual cooling medium in the cooling jacket can provide at the moment of power failure is quantified by calculating the difference between the temperature of the fermentation broth and the temperature of the cooling medium, and then multiplying the difference by the total mass of the cooling medium and the specific heat capacity of the cooling medium.

[0015] Preferably, the method for determining the yeast activity correction coefficient and the total heat transfer coefficient of the fermenter; The yeast activity correction coefficient is determined by applying a machine learning algorithm to perform regression fitting on the fermentation temperature and sugar content change curves of historical batches. The overall heat transfer coefficient of the fermenter is calibrated online during non-fermentation periods by recording the rate of temperature change of the tank under controlled heating or cooling conditions.

[0016] Preferably, it also includes a human-computer interaction and early warning module, which is used to output the predicted value of the resilience time window to the user terminal in response to the risk level determined by the risk assessment module; and when the resilience time window is lower than the early warning time threshold, trigger an audible and visual alarm or send a remote notification to prompt the operator to take intervention measures.

[0017] This invention provides an improved predictive monitoring system for wine fermentation based on machine learning and digital twins, which has the following improvements and advantages compared with the prior art: I. Achieved Proactive Prediction and Management of Fermentation Process Risks: This solution acquires fermentation broth temperature and power grid status through a data acquisition module. Upon detecting a power outage, the first and second processing modules immediately collaborate. By comprehensively analyzing thermodynamic parameters, the third processing module calculates a specific resilience time window. This prediction duration transforms the abstract risk of fermentation runaway into a concrete and perceptible time indicator, providing winemakers with an unprecedented decision-making window. This represents a shift from traditional passive response to proactive risk warning, and is particularly suitable for production areas with unstable power grids.

[0018] Second, a more accurate and comprehensive thermodynamic coupling model was established: The first processing module of this scheme couples the heat generated by yeast metabolism with the heat transferred from the environment into a net heat flux, accurately quantifying the total heat load driving the temperature rise during power outages. This overcomes the shortcomings of traditional technologies that treat internal and external heat sources as independent interference items, and avoids the problem of inaccurate predictions in extreme scenarios where internal and external heat sources act together violently. At the same time, the second processing module creatively incorporates the low-temperature medium remaining in the cooling system into the calculation of the total absorbable energy, greatly improving the accuracy of the initial buffer capacity assessment of the system.

[0019] Third, a dynamic adaptive model with self-optimization capability was constructed: To ensure the accuracy of long-term prediction, this solution can perform online calibration and self-optimization of two key model parameters: yeast activity correction coefficient and total heat transfer coefficient of fermenter. The system can continuously correct the parameters using historical batch data or through physical experiments, so that it can adapt to the differences of different grape varieties and yeast strains, as well as the changes in heat transfer performance caused by equipment aging and scaling. This intelligent and adaptive capability solves the problem of traditional model parameters being fixed and unable to adapt to changes in actual working conditions.

[0020] Fourth, it provides an intuitive and executable human-computer interaction and early warning mechanism: This solution transforms the complex model calculation results into an intuitive risk level through the risk assessment module, and presents it to the user in the form of a safety countdown by the human-computer interaction and early warning module. When the predicted resilience time window is lower than the preset threshold, the system will trigger a clear alarm, prompting the operator to take immediate intervention measures, such as starting the backup generator. This clear instruction avoids decision panic or intervention delays caused by unclear information, and constitutes a complete closed loop of prediction, assessment, decision and execution. Attached Figure Description

[0021] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the system of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0023] Example 1: Please see Figure 1 This invention provides a technical solution: a predictive monitoring system for wine fermentation based on machine learning and digital twins, comprising: The data acquisition module is used to acquire real-time operating parameters of the fermentation process, as well as preset system configuration parameters; wherein, the real-time operating parameters include: fermentation broth temperature, ambient temperature, cooling medium temperature, power grid status, and sugar content; the system configuration parameters include: upper limit of safe temperature and warning time threshold.

[0024] The first processing module is used to determine the net heat flux of the fermentation system based on the fermentation broth temperature, the ambient temperature, and the sugar content when the power grid status in the real-time operating parameters is a power outage; the net heat flux is the sum of the yeast metabolic heat production power and the ambient heat input power.

[0025] The second processing module is used to determine the total energy that the system can absorb based on the temperature of the fermentation liquid, the temperature of the cooling medium, and the upper limit of the safe temperature when the power grid is in a power outage state; the total energy that can be absorbed is the sum of the heat that the wine can absorb and the energy absorbed by the cooling system.

[0026] The third processing module is used to divide the total absorbable energy determined by the second processing module by the net heat flux determined by the first processing module to calculate the resilience time window; the resilience time window is intended to characterize the predicted time required for the fermentation broth temperature to reach the upper limit of the safe temperature after a power outage.

[0027] The risk assessment module is used to compare the resilience time window calculated by the third processing module with the preset early warning time threshold in the system configuration parameters to determine the risk level of the current fermentation process.

[0028] The present invention provides a predictive monitoring system for wine fermentation based on machine learning and digital twins, which is particularly suitable for wine-producing areas with high altitude, variable climate and unstable power grid. It can overcome the shortcomings of existing PID control systems that can only adjust after the fact and cannot predict the risk of power outages.

[0029] In this embodiment, the data acquisition module continuously acquires real-time operating parameters such as the temperature of the fermentation broth inside the fermenter, the ambient temperature outside the tank, the temperature of the cooling medium inside the cooling jacket, the key power grid status, and the sugar content measured by online sensors. When the module detects that the power grid status has switched from normal to power outage, the system immediately activates the first processing module and the second processing module. These two modules calculate in parallel the total heat load of the fermentation system during the power outage, i.e., the net heat flux, and the total heat capacity that the system can withstand to resist temperature rise, i.e., the total energy that can be absorbed.

[0030] The third processing module correlates energy and power through simple division operations to calculate a resilience time window with clear physical meaning. This window intuitively answers the question that winemakers are most concerned about: how long can the fermentation tank be safely maintained without taking any measures?

[0031] The risk assessment module compares this predicted duration with the winemaker's preset warning time threshold, such as 2 hours, and outputs a clear risk level. This transforms the abstract risk of fermentation runaway into a specific, perceptible, and highly instructive quantitative indicator, providing winemakers with an unprecedented decision-making window and realizing the transformation from passive response to proactive risk warning.

[0032] Furthermore, the first processing module determines the net heat flux, including: Calculate the metabolic heat production power of yeast; Calculate the thermal power transferred from the environment; The heat production power generated by yeast metabolism is added to the heat power introduced from the environment to generate the net heat flux.

[0033] This net heat flux is designed to quantify the total heat flow rate that drives the temperature rise of the fermentation broth under power outage conditions.

[0034] The first processing module determines the net heat flux through a comprehensive thermodynamic sub-model that precisely couples the two key physical processes of internal heat generation and external heat transfer.

[0035] The specific implementation depends on the following formula: in: Net heat flux represents the total heat flow rate that drives the temperature of the fermentation broth to rise during a power outage. It is the total heat load of the system during the power outage and is measured in watts. The metabolic heat production power of yeast, generated by the biochemical reactions of yeast, is measured in watts. : Environmental heat transfer power, that is, the heat exchange between the fermenter and the external environment, measured in watts.

[0036] The principle of this model is based on the first law of thermodynamics, namely the conservation of energy. Its technical motivation is that existing technologies usually treat yeast metabolism and environmental heat exchange as independent interference terms, which leads to serious inaccuracies in predictions under extreme scenarios where internal and external heat sources act together violently, such as strong afternoon sunlight and high yeast activity in high-altitude production areas. This invention achieves accurate quantification of the total heat load of the system by coupling the two into a net heat flux, thus solving the technical problem of lack of model coupling.

[0037] Furthermore, the formula for calculating the metabolic heat production power of yeast is as follows: in: : Wine liquid density, a known physical property parameter input by the user, in kilograms per cubic meter; : Volume of wine liquid, which is the equipment specification parameter entered by the user, in cubic meters; The heat production rate per unit mass of wine liquid, which depends on the temperature of the fermentation broth. and sugar content The internal function or lookup table value of this parameter, which serves as a comprehensive empirical coefficient, has data sourced from microbiology literature or experiments. For those skilled in the art, it can be determined in the following two typical ways: Literature review and fitting: Consult professional literature in the fields of bioengineering or food science to obtain calorific data of specific yeast strains at different temperatures and sugar concentrations. Based on this publicly available data, an empirical function suitable for this system can be fitted or a data lookup table can be generated.

[0038] Experimental calibration: Under laboratory conditions, a bioreactor equipped with a precision calorimeter was used. Calibration was performed at different set temperatures. and sugar content Small-batch fermentation experiments were conducted, and the heat production rate of the fermentation broth was measured in real time. Dividing this rate by the volume and density of the fermentation broth yielded the result. The experimental values ​​are obtained. Through multiple sets of experimental data, an accurate function model or lookup table can be established for system use.

[0039] This calibration method is a standard technique in the field of fermentation engineering, ensuring the availability and accuracy of this parameter. Yeast activity correction coefficient is a dimensionless, adjustable parameter used to correct for differences between theoretical models and actual strains, grape varieties, and batches. The time variable indicates that the parameters change dynamically over time.

[0040] Furthermore, the heat production rate per unit mass of the wine The determination of the value does not solely rely on literature references. This invention preferably employs a combination of experimental calibration and data processing, and the specific data processing procedure includes the following steps: Step 1: Experimental calibration of the basic heat production database Referring to existing technologies, a high-precision calorimetric fermentation reactor was used under laboratory conditions to calibrate the heat production performance of a specific yeast strain under different operating conditions. Specifically, a series of discrete fermentation broth temperatures were set. Control points (e.g., 18°C, 20°C, 22°C...) and sugar content Small-batch fermentation experiments were conducted at control points (e.g., 25°Bx, 20°Bx, 15°Bx...) at these crossover operating conditions. The total heat production of the fermentation broth was measured in real time using a calorimeter. Record the corresponding fermentation broth volume. and density According to the formula Calculate each The baseline heat production rate corresponding to the operating point. All calibration data are organized into a two-dimensional "temperature-saccharin content-heat production rate" baseline database.

[0041] Step 2: Real-time calculation of heat production rate and data smoothing processing During actual system operation, the data acquisition module obtains the real-time fermentation broth temperature. and sugar content The results often do not fall exactly on the discrete operating point marked in step one. To obtain the heat generation rate under any real-time operating condition, this invention uses a two-dimensional interpolation algorithm for calculation.

[0042] Preferably, a bilinear interpolation algorithm is used. Specifically, the method involves finding the value corresponding to the real-time operating point in the "temperature-saccharide content-heat production rate" basic database. Four adjacent known data points. First, perform two linear interpolations on the temperature dimension to obtain two temporary points; then, perform another linear interpolation on the sugar content dimension formed by these two temporary points, finally calculating the accurate value of the current real-time operating point. value.

[0043] Furthermore, the calculation of the ambient heat transfer power is based on the fermentation broth temperature, the ambient temperature, and the preset total heat transfer coefficient and surface area of ​​the fermenter. The difference between the ambient temperature and the fermentation broth temperature is calculated, and then this difference is multiplied by the total heat transfer coefficient and surface area of ​​the fermenter to solve for the ambient heat transfer power.

[0044] The formula is as follows: in: The total heat transfer coefficient of the fermenter is an adjustable parameter that varies depending on factors such as tank material, aging, and scaling. The unit is watts per square meter in Kelvin. : The outer surface area of ​​the fermentation tank is a user-inputted equipment specification parameter, in square meters; : Ambient temperature obtained by the data acquisition module, in Kelvin or degrees Celsius; Temperature of the fermentation broth obtained by the data acquisition module, in Kelvin or degrees Celsius.

[0045] This net heat flux formula provides a denominator term in the calculation of the resilience time window, and its technical advantage lies in enabling a dynamic and coupled assessment of the system's heat load; for example, when the ambient temperature in the afternoon... The formula accurately calculates the enormous heat load generated by the combined effects of peak activity at night and yeast activity. A sharp drop, even below At that time, it can also calculate that the environment had a slight cooling effect. The fact that the value is negative, and this dynamic adaptability to positive and negative heat flow, is the physical basis for achieving accurate all-weather prediction.

[0046] Furthermore, the second processing module determines the total absorbable energy through the following steps: Calculate the calories that the wine can absorb; Calculate the energy absorbed by the cooling system's buffer. The heat that the wine can absorb is added to the energy absorbed by the cooling system to generate the total absorbable energy.

[0047] This total absorbable energy aims to quantify the total heat capacity that the system can hold before it heats up to the upper limit of the safe temperature under power failure conditions. This calculation is achieved by quantifying the heat capacity of all components within the system that have passive heat absorption capabilities, accurately assessing the system's inherent buffering capacity against temperature rise. Specifically, it relies on the following formula: in: The total energy that the system can absorb is quantified as the total heat that the entire system can hold from the moment the power grid is cut off until the temperature of the fermentation broth rises to the preset safe limit. The unit is joules. The heat that the wine can absorb represents the heat capacity of the wine itself as the main component, and is measured in joules. Cooling system buffer absorption energy represents the passive heat absorption capacity provided by the residual cooling medium in the cooling jacket at the moment of power failure, measured in joules.

[0048] The principle of this model is based on the physical definition of heat capacity. Its technical motivation is that the traditional approach only considers the heat capacity of the liquid itself when assessing the risk of power outage, while ignoring the important and quantifiable thermal buffer of the low-temperature cooling medium remaining in the cooling system jacket or coil. This invention creatively incorporates this part of the energy into the calculation, especially in the early stage of power outage, when the heat absorption effect of this buffer is very significant. Ignoring it will seriously underestimate the initial toughness of the system and cause misjudgment.

[0049] Formula for calculating the heat that alcohol can absorb. in: Specific heat capacity of wine liquid, a known physical property parameter input by the user, in joules per kilogram Kelvin; The upper limit of safe temperature in the system configuration parameters, in Kelvin or degrees Celsius.

[0050] Furthermore, the calculation of the cooling system's buffer absorption energy is based on the fermentation broth temperature, the cooling medium temperature, and the preset total mass and specific heat capacity of the cooling medium. By calculating the difference between the fermentation broth temperature and the cooling medium temperature, and then multiplying this difference by the total mass and specific heat capacity of the cooling medium, the passive heat absorption capacity that the residual cooling medium in the cooling jacket can provide at the moment of power failure is quantified.

[0051] Formula for calculating the energy absorbed by the cooling system's buffer in: Total mass of cooling medium, which is the equipment specification parameter or estimated value entered by the user, in kilograms; Specific heat capacity of the cooling medium, a known physical property parameter input by the user, in joules per kilogram Kelvin; : The temperature of the cooling medium obtained by the data acquisition module, in Kelvin or degrees Celsius.

[0052] This formula for total absorbable energy provides the numerator in the calculation of the resilience time window. Its technical advantage lies in its ability to more accurately assess the initial buffering capacity of a system, thus reflecting the system's current energy storage state, for example, that of a system that has just completed a refrigeration cycle. Very low, therefore A larger value indicates a longer calculated resilience time window, which perfectly aligns with physical reality. Conversely, a system that has not been cooled for an extended period will have a shorter resilience time window. It is close , As the energy level approaches zero, the system's thermal buffering capacity is greatly reduced. This precise quantification of the energy storage state of the cooling system solves the problem of incomplete assessment of system resilience.

[0053] Furthermore, to ensure the long-term accuracy of model predictions, the system has the capability for online parameter calibration and self-optimization.

[0054] Yeast activity correction factor The determination method: To reflect the differences between different grape varieties, yeast batches, and specific fermentation stages, this system employs a data-driven self-optimization method. Specifically, the system uses machine learning algorithms (such as least squares regression and neural networks) to regress and fit the curves of fermentation temperature and sugar content changes over time recorded in historical batches, thereby back-calibrating the optimal [fermentation process]. Value. The initial value of this coefficient can be based on data provided by the yeast supplier or set to 1.0.

[0055] Furthermore, yeast activity correction coefficient based on machine learning algorithm The data-driven self-optimization method is implemented as follows: Step 1: Align historical batch data Select one or more completed historical fermentation batches with complete records. Extract time-series data for the entire process, including: actual measured fermentation broth temperature. Ambient temperature Cooling medium temperature and sugar content At the same time, key process information such as the grape variety and yeast strain used in this batch was recorded.

[0056] Step 2: Temperature curve prediction based on the initial model set up The initial value is usually set to 1.0. The ambient temperature of this historical batch... , sugar content Using these as inputs, the thermodynamic model established by this invention (including the calibrated heat production rate) is applied. and the overall heat transfer coefficient of the fermenter The simulation calculates the theoretical temperature rise without cooling intervention, or the complete temperature change process with cooling intervention, thereby generating a predicted temperature curve. .

[0057] Step 3: Error Calculation and Objective Function Establishment Predict the temperature curve Temperature curve compared with the actual measured temperature of this historical batch A comparison is made. The root mean square error (RMSE) is used as a quantitative indicator to evaluate prediction accuracy; its calculation formula is as follows: The optimization objective here is to find an optimal... The value that minimizes RMSE.

[0058] Step 4: Parameter optimization based on gradient descent The gradient descent optimization algorithm is used to automatically find the optimal solution. Specifically, the algorithm treats RMSE as a measure of... The objective function is calculated by evaluating the objective function. The gradient (i.e., derivative) is used to determine the adjustment direction that results in the fastest decrease in RMSE. The algorithm iteratively updates the gradient in the reverse direction with a set learning rate as the step size. The value of is determined, and steps two and three are repeated until the RMSE converges to its minimum or the preset number of iterations is reached. At this point, the obtained RMSE is... This is the optimal correction coefficient for that specific process (same grape variety, same strain).

[0059] Step 5: Parameter Storage and Application The calibrated optimal The value is associated with the corresponding process information (grape variety, yeast strain) and stored in the system configuration parameter library. When starting a new fermentation batch with the same process information in the future, the system will automatically call this optimized value. This allows the model to adapt and make high-precision predictions.

[0060] Overall heat transfer coefficient of fermenter Determination method: Considering that factors such as tank aging and internal scaling can cause heat transfer performance to change over time, this system is designed to... The system includes an online calibration function. The specific method involves injecting a fluid (usually water) of known mass and specific heat capacity into the tank during non-fermentation periods (e.g., after washing, before the next feeding). The fluid is then heated or cooled at a constant power using internal heating rods or an external jacket, while the rate of temperature change over time is recorded. This is based on the law of conservation of energy. It can accurately calculate the current Value. This method is the standard operating procedure for calibrating the thermal performance of chemical equipment.

[0061] The online parameter calibration function in this invention enables the system to transform from passive prediction based on universal model parameters to proactive self-optimization monitoring based on specific batches and equipment status. This intelligence and adaptability can adapt to changes in heat transfer performance caused by different grape varieties, yeast strains, and factors such as equipment aging and scaling, achieving true intelligent and adaptive monitoring and solving the problem of traditional model parameters being fixed and unable to adapt to changes in operating conditions.

[0062] Furthermore, it also includes a human-computer interaction and early warning module, which is used to respond to the risk level determined by the risk assessment module and output the predicted value of the resilience time window to the user terminal.

[0063] When the resilience time window is lower than the preset warning time threshold, it is determined to be high risk. The module will trigger an audible and visual alarm or send a remote notification to prompt the operator to take immediate intervention measures, such as starting a backup generator.

[0064] When the risk level is safe, meaning the resilience time window is not lower than the warning time threshold, the module will execute regular status instructions and continuously display the predicted value of the safety countdown on the user terminal.

[0065] This explicit definition and response to all logical intervals not only provides timely warnings in times of danger, but also provides continuous and quantifiable confidence in times of safety, thereby completely avoiding panic or delays in intervention caused by unclear information, and forming a complete closed loop of prediction-assessment-decision-execution.

[0066] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A machine learning and digital twin based wine fermentation predictive monitoring system, characterized in that, The method comprises the following steps: a data acquisition module is used to obtain real-time operating parameters of the fermentation process and preset system configuration parameters; wherein the real-time operating parameters include fermentation broth temperature, ambient temperature, cooling medium temperature, power grid state and brix; the system configuration parameters include upper limit of safety temperature and early warning time threshold value; a first processing module is used to determine the net heat flux of the fermentation system based on the fermentation broth temperature, the ambient temperature and the brix when the power grid state in the real-time operating parameters is power failure; wherein the net heat flux is the sum of yeast metabolic heat power and environmental incoming heat power; a second processing module is used to determine the total energy that can be absorbed by the system based on the fermentation broth temperature, the cooling medium temperature and the upper limit of safety temperature when the power grid state is power failure; wherein the total energy that can be absorbed is the sum of wine-absorbable heat and cooling system buffer absorption energy; a third processing module is used to divide the total energy that can be absorbed determined by the second processing module by the net heat flux determined by the first processing module to calculate the resilience time window; the resilience time window aims to represent the predicted length of time required for the fermentation broth temperature to reach the upper limit of safety temperature after power failure; a risk assessment module is used to compare the resilience time window calculated by the third processing module with the early warning time threshold value preset in the system configuration parameters to determine the risk level of the current fermentation process.

2. The system of claim 1, wherein, The first processing module determines the net heat flux, which comprises the following steps: calculating yeast metabolic heat power; calculating environmental incoming heat power; adding the yeast metabolic heat power and the environmental incoming heat power to generate the net heat flux, wherein the net heat flux aims to quantify the total heat flow rate driving the fermentation broth temperature to rise in the power failure state.

3. The system of claim 2, wherein, The calculation of the yeast metabolic heat power is based on the fermentation broth temperature and the brix obtained by the data acquisition module, combined with the preset grape wine density, grape wine volume, unit mass yeast heat production rate and yeast activity correction coefficient; the yeast metabolic heat power is accurately calculated by multiplying the grape wine density, the grape wine volume, the unit mass yeast heat production rate and the yeast activity correction coefficient.

4. The system of claim 2, wherein, The calculation of the environmental incoming heat power is based on the fermentation broth temperature, the ambient temperature, and the preset total heat transfer coefficient of the fermentation tank and the outer surface area of the fermentation tank; the environmental incoming heat power is calculated by calculating the difference between the ambient temperature and the fermentation broth temperature, and then multiplying the difference by the total heat transfer coefficient of the fermentation tank and the outer surface area of the fermentation tank.

5. The system of claim 1, wherein, The second processing module determines the total energy that can be absorbed, which comprises the following steps: calculating wine-absorbable heat; calculating cooling system buffer absorption energy; adding the wine-absorbable heat and the cooling system buffer absorption energy to generate the total energy that can be absorbed, wherein the total energy that can be absorbed aims to quantify the total heat capacity before the system is warmed to the upper limit of safety temperature in the power failure state.

6. The system of claim 5, wherein, The calculation of the cooling system buffer absorbing energy is based on the fermentation broth temperature, the cooling medium temperature, and the preset total mass of the cooling medium and the specific heat capacity of the cooling medium; By calculating the difference between the fermentation broth temperature and the cooling medium temperature, and then multiplying the difference by the total mass of the cooling medium and the specific heat capacity of the cooling medium, the passive heat absorption capacity of the residual cooling medium in the cooling jacket at the moment of power failure can be quantified.

7. The system of claim 4, wherein, The determination method of the yeast activity correction coefficient and the total heat transfer coefficient of the fermenter; The yeast activity correction coefficient is calibrated by applying a machine learning algorithm to regress and fit the historical batch fermentation temperature and sugar content change curve data; The total heat transfer coefficient of the fermenter is online calibrated by recording the temperature change rate of the tank body under controlled heating or cooling conditions during the non-fermentation period.

8. The system of claim 1, wherein, It also includes a human-computer interaction and early warning module, which is used to output the predicted value of the resilience time window to the user terminal in response to the risk level determined by the risk assessment module; and when the resilience time window is lower than the early warning time threshold, triggering an audible and visual alarm or sending a remote notification to prompt the operator to take intervention measures.

Images