Self-adaptive pressure regulation system for sparkling wine secondary fermentation process
By introducing an adaptive control and single-neuron PID fusion strategy, combined with sensing acquisition, feature extraction and collaborative control modules, the problems of poor dynamic adaptability of pressure regulation and collaborative control of multi-tank fermentation systems during the secondary fermentation of sparkling wine are solved. This achieves high-precision and fast-response pressure control, improving the stability and scalability of sparkling wine production.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JUELU BIOTECHNOLOGY (JIANGXI) CO LTD
- Filing Date
- 2026-03-03
- Publication Date
- 2026-05-29
AI Technical Summary
In the existing secondary fermentation process of sparkling wine, the pressure control methods have problems such as poor dynamic adaptability, low control precision, and delayed abnormal early warning. In addition, traditional multi-tank fermentation systems have problems such as sudden pressure drop in shared gas path, command execution conflict, response delay, accelerated equipment wear, and lack of quantitative evaluation standards.
A model reference adaptive control and single neuron PID fusion strategy is introduced, which combines sensing and acquisition, feature extraction, adaptive decision-making and collaborative control modules to achieve real-time self-tuning of controller parameters and intelligent collaboration among multiple tanks. The sensing and acquisition module monitors fermentation parameters in real time, the feature extraction module analyzes feature vectors, the adaptive decision-making module generates control commands, and the collaborative control module performs conflict detection and load balancing scheduling.
It significantly improves the steady-state accuracy and dynamic quality of pressure control, enhances system response speed and resource utilization efficiency, and ensures high quality and scalability of sparkling wine production.
Smart Images
Figure CN122111122A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of automatic control technology, specifically to an adaptive pressure control system for the secondary fermentation process of sparkling wine. Background Technology
[0002] The unique flavor and delicate bubbles of sparkling wine mainly originate from the secondary fermentation process, which requires extremely high precision in controlling the pressure inside the tank: too low a pressure will result in insufficient carbon dioxide solubility, affecting the smoothness of the bubbles; too high a pressure may cause yeast autolysis, produce undesirable flavor substances, and even pose a safety risk of tank explosion. However, existing methods for controlling the pressure of secondary fermentation in sparkling wine suffer from poor dynamic adaptability, low control precision, and delayed anomaly warnings. Furthermore, in traditional multi-tank fermentation systems, each fermentation tank often operates as an independent unit. When multiple tanks operate concurrently, there are problems such as sudden drops in pressure in the shared gas path, command execution conflicts, response delays, accelerated equipment wear, and a lack of quantitative evaluation standards. Summary of the Invention
[0003] To address the aforementioned issues and overcome the shortcomings of existing technologies, this invention provides an adaptive pressure control system for the secondary fermentation process of sparkling wine. Addressing the problems of poor dynamic adaptability, low control precision, and delayed anomaly warning in existing methods for secondary fermentation pressure control of sparkling wine, this solution introduces a fusion strategy of model reference adaptive control and single-neuron PID control to achieve real-time self-tuning of controller parameters. It dynamically adjusts control behavior based on uncertainties such as changes in yeast activity and temperature fluctuations during fermentation, effectively suppressing overshoot while ensuring response speed, and significantly improving the steady-state accuracy and dynamic quality of pressure control. Furthermore, in traditional multi-tank fermentation systems, each fermentation tank often operates as an independent unit. When multiple tanks operate concurrently, problems arise such as sudden pressure drops in shared gas paths, command execution conflicts, response delays, increased equipment wear, and a lack of quantitative evaluation standards. This solution organically combines conflict detection, edge computing, closed-loop feedback, and quantitative evaluation to upgrade the secondary fermentation of sparkling wine from independent single-tank control to intelligent multi-tank collaboration. While ensuring control precision, it significantly improves system response speed, resource utilization efficiency, and scalability, providing a complete technical solution for large-scale, high-quality sparkling wine production.
[0004] The adaptive pressure control system for the secondary fermentation process of sparkling wine provided by this invention includes a sensing and acquisition module, a feature extraction module, an adaptive decision-making module, and a collaborative control module.
[0005] The sensing and acquisition module collects the process parameters of each fermenter in real time during the secondary fermentation process. Through time synchronization and filtering, it generates standardized time-series data containing timestamps, tank IDs, real-time pressure values, real-time temperature values, and real-time liquid level values. The standardized time-series data is then sent to the feature extraction module.
[0006] The feature extraction module performs multi-scale feature analysis on standardized time-series data to extract feature vectors characterizing the fermentation state, including the instantaneous rate of change of pressure, the variance of pressure fluctuation, and the temperature-pressure coupling coefficient. At the same time, it uses a sliding window algorithm to establish a dynamic baseline model and generate a state feature set, which is then sent to the adaptive decision module.
[0007] The adaptive decision-making module sets a reference model of the pressure trajectory during the secondary fermentation of the sparkling wine, compares the state feature set with the expected pressure value output by the reference model, tunes the PID control parameters online through an adaptive law, generates control instructions for the fermentation tank, and determines whether to generate early warning information based on the degree of pressure deviation, and sends the control instructions to the collaborative control module.
[0008] The collaborative control module parses the control commands into drive signals for the actuators, controls the pressure inside the fermenter, performs conflict detection and load balancing scheduling when multiple tanks execute concurrently, outputs the execution results, and feeds them back to the adaptive decision module.
[0009] Furthermore, the feature extraction module includes a rate of change calculation unit, a fluctuation analysis unit, a coupled feature unit, and a feature output unit;
[0010] The rate of change calculation unit performs first-order difference calculation on the pressure values in the standardized time-series data to obtain the instantaneous rate of change of pressure. It then calculates the mean and peak values of the instantaneous rate of change of pressure within the sliding window to characterize the dynamic change of the gas production rate. The formula used is as follows: ;
[0011] In the formula, Indicates time, express The instantaneous rate of change of pressure at any given moment. express The pressure value at any moment, express The pressure value at any moment, Indicates the sampling period;
[0012] The fluctuation analysis unit uses a sliding window algorithm to calculate the variance of pressure values within a 30-minute window as a pressure fluctuation feature, and uses the LOF (Local Outlier Factor) algorithm to identify outliers in the pressure value sequence, mark abnormal fluctuation points, and generate fluctuation feature values.
[0013] The coupling feature unit calculates the cross-correlation coefficient between the time series data of pressure and temperature values, which serves as the temperature-pressure coupling coefficient. The closer the value is to 1, the more significant the influence of temperature on pressure. A baseline model is established based on historical data, and the expected pressure range at the current moment is dynamically calculated using the exponential weighted moving average method. If the real-time pressure value deviates from the baseline by more than a preset threshold, it is marked as a potential abnormal state.
[0014] The feature output unit outputs the pressure change rate, pressure fluctuation characteristics, and temperature-pressure coupling coefficient as a set of state features.
[0015] Furthermore, the adaptive decision-making module includes a reference model unit, an adaptive law unit, a hybrid controller unit, and an early warning unit;
[0016] The reference model unit is pre-set with segmented target pressure values for different stages of secondary fermentation, and sets pressure threshold intervals for each. Using a conformal piecewise cubic interpolation method, the discrete segmented target pressure values are fitted to a continuously differentiable ideal fermentation process curve. Based on the ideal fermentation process curve, a reference model describing the pressure trajectory is set using a second-order linear system. The natural frequency and damping ratio of the reference model are set according to the type of sparkling wine. The input is a state feature set, and the output is the expected pressure value under ideal conditions. The formula used is as follows: ;
[0017] In the formula, Represents the transfer function of the reference model. Represents the Laplace operator. Represents natural frequency. The damping ratio is used to determine the degree of oscillation in the response curve;
[0018] The adaptive law unit uses the pressure deviation between the real-time pressure value and the desired pressure value as the driving quantity. Through an adaptive mechanism, it adjusts the PID control parameters of the hybrid controller unit online, setting the initial gain coefficient and adaptive learning rate to gradually approximate the desired pressure value under ideal conditions. The formula used is as follows: ; ; ;
[0019] In the formula, , and They are respectively The proportional, integral, and differential gain coefficients at different times. , and The initial gain coefficient, , , For adaptive learning rate, Indicates from 0 to At any instant between moments, for Pressure deviation value at any time This is the real-time pressure value. For the control input of the fermenter, For approximate estimation by the state feature set Jacobian information at any given moment;
[0020] The hybrid controller unit calculates the control command for the fermenter based on the real-time adjusted PID control parameters and the single-neuron PID algorithm. It maps the PID control parameters to the proportional coefficients of the neurons, realizing the coordinated adaptation of the PID control parameters and the neuron weights. The input layer of the single neuron receives the pressure deviation value and its integral and derivative signals, and smoothly adjusts the output signal of the PID controller through online weight correction.
[0021] The early warning unit sets a pressure deviation threshold. When the pressure deviation value continuously exceeds the pressure deviation threshold, it generates an early warning message that includes the abnormality level and recommended measures.
[0022] Furthermore, the calculation method for the control command in the hybrid controller unit specifically includes the following steps:
[0023] Step S1: Construct the neuron input vector, using the pressure deviation value, deviation integral, and deviation derivative as the three input components of the neuron to achieve piecewise self-adjustment of PID control parameters;
[0024] Step S2: Initialize the neuron weights, corresponding to the contribution of proportional, integral, and differential terms in the output;
[0025] Step S3: Calculate the control output command of the neuron, and use supervised Hebb learning rules to correct the neuron weights online. Set the adjustment range of the neuron's scaling factor and the learning rate of the input components. The formula used is as follows: ; ; ;
[0026] In the formula, The proportional gain of the neuron is determined by the PID control parameters. These are the normalized neuron weights. Indicates the index of the input component. For the first The specific values of each input component. For the updated number The weights of each input component, For the first The weights of each input component at the current time step. For the first The learning rate of each input component. This is the pressure deviation value. This refers to the control output command at the current moment;
[0027] Step S4: Control action conversion, converting the control output command into the specific action amount of the fermenter valve opening, the specific action amount of the fermenter valve opening is proportional to the absolute value of the control output command.
[0028] Furthermore, the collaborative control module includes an instruction parsing unit, a valve group driving unit, a conflict detection unit, and a load balancing unit;
[0029] The instruction parsing unit receives the control instruction, converts the control instruction into the valve opening value and action duration, and generates a valve control signal to send to the valve group drive unit.
[0030] The valve group drive unit drives the motors of the exhaust valve and the air supply valve according to the valve control signal, and provides real-time feedback on the actual valve opening through the valve position sensor to form a closed-loop control.
[0031] The conflict detection unit monitors the command execution status of each fermenter. When it detects that multiple tanks simultaneously request large-flow exhaust, causing a sudden drop in the pressure of the shared gas path, it avoids execution conflicts through command priority sorting and time slice rotation mechanism.
[0032] The load balancing unit counts the frequency of control and valve action of each tank. When the number of fermentation tanks in the same batch exceeds the set threshold, a dynamic scheduling algorithm is used to allocate computing tasks between edge nodes and the central server, generate a multi-tank collaborative operation report, record the control compliance rate and energy consumption index of each tank, and output the execution results.
[0033] Furthermore, the conflict detection unit performs instruction priority sorting, specifically including the following steps:
[0034] Step P1: Establish a priority calculation model and set indicators to calculate the urgency score of each request;
[0035] Step P2: Sort all requests according to their urgency score from highest to lowest, and the request with the highest score gets the highest execution priority;
[0036] Step P3: Set the maximum number of concurrent requests N based on the flow capacity of the shared gas path, prioritize the execution of the first N high-priority requests, and put the remaining requests into the waiting queue;
[0037] Step P4: When the waiting time of a request in the waiting queue exceeds a set threshold, dynamically adjust its urgency score and add a time penalty factor to the original score to prevent requests from starving.
[0038] Step P5: After each instruction is completed, re-evaluate the priority of requests in the waiting queue and release resources for the next batch of execution.
[0039] The beneficial effects achieved by the present invention using the above solution are as follows:
[0040] (1) In view of the problems of poor dynamic adaptability, low control accuracy and delayed abnormal warning in the existing secondary fermentation pressure control method of sparkling wine, this scheme introduces a model reference adaptive control and single neuron PID fusion strategy to realize real-time self-tuning of controller parameters. The control behavior is dynamically adjusted according to the changes in yeast activity and temperature fluctuations during fermentation. While ensuring the response speed, overshoot is effectively suppressed, and the steady-state accuracy and dynamic quality of pressure control are significantly improved.
[0041] (2) In traditional multi-tank fermentation systems, each fermentation tank often operates as an independent unit. When multiple tanks operate concurrently, there are problems such as sudden pressure drop in the shared gas path, command execution conflict, response delay, increased equipment wear, and lack of quantitative evaluation standards. This solution upgrades the secondary fermentation of sparkling wine from single-tank independent control to multi-tank intelligent collaboration by organically combining conflict detection, edge computing, closed-loop feedback and quantitative evaluation. While ensuring control accuracy, it significantly improves system response speed, resource utilization efficiency and scalability, providing a complete technical solution for large-scale, high-quality sparkling wine production. Attached Figure Description
[0042] Figure 1 This is a schematic diagram of the adaptive pressure control system for the secondary fermentation process of sparkling wine proposed in this invention.
[0043] Figure 2 This is a schematic diagram of the feature extraction module;
[0044] Figure 3 This is a schematic diagram of the adaptive decision-making module;
[0045] Figure 4 This is a flowchart illustrating the calculation of control commands in the hybrid controller unit.
[0046] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation
[0047] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0048] Example 1, see Figure 1 The adaptive pressure control system for the secondary fermentation process of sparkling wine provided by the present invention includes a sensing and acquisition module, a feature extraction module, an adaptive decision-making module, and a collaborative control module.
[0049] The sensing and acquisition module collects the process parameters of each fermenter in real time during the secondary fermentation process. Through time synchronization and filtering, it generates standardized time-series data containing timestamps, tank IDs, real-time pressure values, real-time temperature values, and real-time liquid level values. The standardized time-series data is then sent to the feature extraction module.
[0050] The feature extraction module performs multi-scale feature analysis on standardized time-series data to extract feature vectors characterizing the fermentation state, including the instantaneous rate of change of pressure, the variance of pressure fluctuation, and the temperature-pressure coupling coefficient. At the same time, it uses a sliding window algorithm to establish a dynamic baseline model and generate a state feature set, which is then sent to the adaptive decision module.
[0051] The adaptive decision-making module sets a reference model of the pressure trajectory during the secondary fermentation of the sparkling wine, compares the state feature set with the expected pressure value output by the reference model, tunes the PID control parameters online through an adaptive law, generates control instructions for the fermentation tank, and determines whether to generate early warning information based on the degree of pressure deviation, and sends the control instructions to the collaborative control module.
[0052] The collaborative control module parses the control commands into drive signals for the actuators, controls the pressure inside the fermenter, performs conflict detection and load balancing scheduling when multiple tanks execute concurrently, outputs the execution results, and feeds them back to the adaptive decision module.
[0053] Example 2, see Figure 1 This embodiment is based on the above embodiment, and the sensing and acquisition module includes a pressure acquisition unit, a temperature acquisition unit, an environmental variable acquisition unit, and a data preprocessing unit;
[0054] The pressure acquisition unit uses a diffused silicon pressure transmitter with a range of 0-1.6MPa and an accuracy class of 0.1. The sampling frequency is set to 1Hz. It acquires the gas phase pressure at the top of the fermenter and the liquid phase pressure at the bottom of the fermenter in real time, and generates a raw pressure signal to be sent to the data preprocessing unit.
[0055] The temperature acquisition unit uses a PT100 platinum resistance temperature sensor to collect the temperature of the wine in the fermentation tank and the ambient temperature, and generates a raw temperature signal to be sent to the data preprocessing unit.
[0056] The environmental variable acquisition unit uses a radar level gauge to collect the liquid level height of the fermentation tank, and simultaneously acquires the combination of ambient humidity and vibration status as environmental variables to help determine the impact of changes in the volume of the wine on the pressure-volume relationship. The environmental variables are then sent to the data preprocessing unit.
[0057] The data preprocessing unit performs time-series alignment on the original pressure signal, original temperature signal, and environmental variables, uses a moving average filtering algorithm to remove high-frequency noise, sets the length of the sliding window to 5 sampling points, sets the sampling period to 1 second for resampling, and generates standardized time-series data containing timestamps, tank IDs, real-time pressure values, real-time temperature values, and real-time liquid level values.
[0058] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment, and the feature extraction module includes a rate of change calculation unit, a fluctuation analysis unit, a coupled feature unit, and a feature output unit;
[0059] The rate of change calculation unit performs first-order difference calculation on the pressure values in the standardized time-series data to obtain the instantaneous rate of change of pressure. It calculates the mean and peak values of the instantaneous rate of change of pressure within a 15-minute window to characterize the dynamic change of the gas production rate. The formula used is as follows: ;
[0060] In the formula, Indicates time, express The instantaneous rate of change of pressure at any given moment. express The pressure value at any moment, express The pressure value at any moment, Indicates the sampling period;
[0061] The fluctuation analysis unit uses a sliding window algorithm to calculate the variance of pressure values within a 30-minute window as a pressure fluctuation feature, and uses the LOF algorithm to identify outliers in the pressure value sequence. When the LOF score exceeds 3, it is marked as an abnormal fluctuation point and a fluctuation feature value is generated.
[0062] The coupling feature unit calculates the time-series data of pressure and temperature values in the time delay. The cross-correlation coefficient at time is used as the temperature-pressure coupling coefficient. The closer the value is to 1, the more significant the influence of temperature on pressure. A baseline model is established based on the historical data of the first three batches. The expected pressure range at the current moment is dynamically calculated using the exponential weighted moving average method. If the real-time pressure value deviates from the baseline by more than ±15 kPa, it is marked as a potential abnormal state.
[0063] The feature output unit outputs the pressure change rate, pressure fluctuation characteristics, and temperature-pressure coupling coefficient as a set of state features.
[0064] Example 4, see Figure 1 and Figure 3 This embodiment is based on the above embodiment, and the adaptive decision module includes a reference model unit, an adaptive law unit, a hybrid controller unit, and an early warning unit;
[0065] The reference model unit is pre-set with segmented target pressure values for different stages of secondary fermentation, and sets pressure threshold intervals for each. It uses a conformal piecewise cubic interpolation method to fit the discrete segmented pressure target values into a continuously differentiable ideal fermentation process curve. Based on the ideal fermentation process curve, a reference model describing the pressure trajectory is set using a second-order linear system. The natural frequency and damping ratio of the reference model are set according to the type of sparkling wine. The input is a state feature set, and the output is the desired pressure value under ideal conditions. For the classic Champagne method of secondary fermentation, the natural frequency is set to 0.02, and the damping ratio is set to 0.8. The formulas used are as follows: ;
[0066] In the formula, Represents the transfer function of the reference model. Represents the Laplace operator. Represents natural frequency. The damping ratio determines the degree of oscillation in the response curve;
[0067] The adaptive law unit uses the pressure deviation between the real-time pressure value and the desired pressure value as the driving quantity, and adjusts the PID control parameters of the hybrid controller unit online through an adaptive mechanism, including the initial gain coefficient. , and The adaptive learning rates were set to 2.5, 0.1, and 0.5 respectively. , , The values are set to 0.01, 0.001, and 0.005 respectively, so that the real-time pressure value gradually approaches the desired pressure value under ideal conditions. The formula used is as follows: ; ; ;
[0068] In the formula, , and They are respectively The proportional, integral, and differential gain coefficients at different times. , and The initial gain coefficient, , , For adaptive learning rate, Indicates from 0 to At any instant between moments, for Pressure deviation value at any time This is the real-time pressure value. To control the input, For approximate estimation by the state feature set Jacobian information at any given moment;
[0069] The hybrid controller unit calculates the control command for the fermenter based on the real-time adjusted PID control parameters and the single-neuron PID algorithm. It maps the PID control parameters to the proportional coefficients of the neurons, realizing the coordinated adaptation of the PID control parameters and the neuron weights. The input layer of the single neuron receives the pressure deviation value and its integral and derivative signals, and smoothly adjusts the output signal of the PID controller through online weight correction.
[0070] The early warning unit sets a pressure deviation threshold of ±20 kPa. When the pressure deviation value continues to exceed the pressure deviation threshold for 5 minutes, it generates an early warning message containing the anomaly level and recommended measures, specifically:
[0071] Level 1 warning: Pressure is rising too rapidly; it is recommended to check yeast activity.
[0072] Level 2 warning: The pressure deviates from the ideal trajectory. It is recommended to fine-tune the temperature setting.
[0073] Example 5, see Figure 1 , Figure 3 and Figure 4 This embodiment is based on the above embodiment. The method for calculating the control command in the hybrid controller unit specifically includes the following steps:
[0074] Step S1: Construct the neuron input vector, using the pressure deviation value, deviation integral, and deviation derivative as the three input components of the neuron to achieve piecewise self-adjustment of PID control parameters;
[0075] Step S2: Initialize the neuron weights to 0.6, 0.3, and 0.1, respectively, which correspond to the contributions of proportional, integral, and derivative terms in the output;
[0076] Step S3: Calculate control command. Calculate the control output command of the neuron and use supervised Hebb learning rules to adjust the neuron weights online. Set the proportional gain of the neuron to a range of 0.5 to 3.0, and set the learning rates of the three input components to 0.4, 0.2, and 0.1 respectively. The formula used is as follows: ; ; ;
[0077] In the formula, The proportional gain of the neuron is determined by the PID control parameters. These are the normalized neuron weights. Indicates the index of the input component. For the first The specific values of each input component. For the updated number The weights of each input component, For the first The weights of each input component at the current time step. For the first The learning rate of each input component. This is the pressure deviation value. This refers to the control output command at the current moment;
[0078] Step S4: Control action conversion. The control output command is converted into the specific action amount of the fermenter valve opening. The specific action amount of the fermenter valve opening is proportional to the absolute value of the control output command: a positive control output command indicates that the exhaust valve needs to be opened to release pressure; a negative control output command indicates that exhaust is stopped, and the gas supply valve is opened in case of emergency.
[0079] By performing the aforementioned operation, this solution addresses the problems of poor dynamic adaptability, low control precision, and delayed anomaly warning in existing methods for controlling the pressure of secondary fermentation in sparkling wine. It introduces a fusion strategy of model reference adaptive control and single-neuron PID control to achieve real-time self-tuning of controller parameters. The control behavior is dynamically adjusted according to uncertainties such as changes in yeast activity and temperature fluctuations during fermentation. This effectively suppresses overshoot while ensuring response speed, significantly improving the steady-state precision and dynamic quality of pressure control.
[0080] Example 6, see Figure 1This embodiment is based on the above embodiment, and the collaborative control module includes an instruction parsing unit, a valve group driving unit, a conflict detection unit, and a load balancing unit;
[0081] The instruction parsing unit receives the control instruction, converts the control instruction into the valve opening value and action duration, and generates a valve control signal to send to the valve group drive unit.
[0082] The valve group drive unit drives the motors of the exhaust valve and the air supply valve according to the valve control signal, and provides real-time feedback on the actual valve opening through the valve position sensor to form a closed-loop control.
[0083] The conflict detection unit monitors the command execution status of each fermenter. When it detects that multiple tanks simultaneously request large-flow exhaust, causing a sudden drop in the pressure of the shared gas path, it avoids execution conflicts through command priority sorting and time slice rotation mechanism.
[0084] The load balancing unit counts the frequency of control and valve action of each tank. When the number of fermentation tanks in the same batch exceeds 50, a dynamic scheduling algorithm is used to allocate computing tasks between edge nodes and the central server, generate a multi-tank collaborative operation report, and record the control compliance rate and energy consumption indicators of each tank.
[0085] Example 7, see Figure 1 This embodiment is based on the above embodiment. The conflict detection unit performs instruction priority sorting, specifically including the following steps:
[0086] Step P1: Establish a priority calculation model. Calculate the urgency score for each request based on three indicators: the degree to which the pressure inside the fermenter deviates from the safety threshold, the remaining fermentation time in the tank, and the instantaneous rate of pressure change. The weights for the three indicators are 0.5, 0.3, and 0.2, respectively. The formula used is as follows: ;
[0087] In the formula, The score indicates the degree of urgency. This indicates the degree to which the pressure deviates from the safe threshold. Indicates the pressure safety threshold. Indicates the total fermentation cycle. Indicates the remaining fermentation time in the tank. Indicates the instantaneous rate of change of pressure. This represents the maximum permissible instantaneous rate of change of pressure. , , These are the weighting coefficients;
[0088] Step P2: Sort all requests according to their urgency score from highest to lowest, and the request with the highest score gets the highest execution priority;
[0089] Step P3: Based on the flow capacity of the shared gas path, set the maximum number of concurrent requests N to 5, prioritize the execution of the first 5 high-priority requests, and put the remaining requests into the waiting queue;
[0090] Step P4: When the waiting time of a request in the waiting queue exceeds 5 minutes, dynamically adjust its urgency score by adding a time penalty factor to the original score;
[0091] Step P5: After each instruction is completed, re-evaluate the priority of requests in the waiting queue and release resources for the next batch of execution.
[0092] Example 8, see Figure 1 This embodiment is based on the above embodiment. The load balancing unit adopts a dynamic scheduling algorithm, specifically: constructing a node resource table containing computing power, network bandwidth, and the number of current tasks; when a new batch of fermenters is put into operation, control tasks are allocated according to the remaining resource ratio of each edge node; when an edge node detects that its load exceeds 80%, it actively uploads non-real-time data analysis tasks to the central server to ensure that the response time of real-time control tasks does not exceed 100ms; at the same time, it periodically generates multi-tank collaborative operation reports, recording the control compliance rate, valve action frequency, and energy consumption indicators of each tank, providing data support for process optimization.
[0093] By performing the aforementioned operations, this solution addresses the problems in traditional multi-tank fermentation systems, where each fermentation tank often operates as an independent unit. When multiple tanks operate concurrently, issues arise such as sudden drops in shared gas path pressure, command execution conflicts, response delays, increased equipment wear, and a lack of quantitative evaluation standards. This solution organically combines conflict detection, edge computing, closed-loop feedback, and quantitative evaluation to upgrade the secondary fermentation of sparkling wine from independent single-tank control to intelligent multi-tank collaboration. While ensuring control accuracy, it significantly improves system response speed, resource utilization efficiency, and scalability, providing a complete technical solution for large-scale, high-quality sparkling wine production.
[0094] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0095] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
[0096] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.
Claims
1. An adaptive pressure control system for the secondary fermentation process of sparkling wine, characterized in that: It includes a perception and acquisition module, a feature extraction module, an adaptive decision-making module, and a collaborative control module; The sensing and acquisition module collects the process parameters of each fermenter in real time during the secondary fermentation process. Through time synchronization and filtering, it generates standardized time-series data containing timestamps, tank IDs, real-time pressure values, real-time temperature values, and real-time liquid level values. The standardized time-series data is then sent to the feature extraction module. The feature extraction module performs multi-scale feature analysis on standardized time-series data to extract feature vectors characterizing the fermentation state, including the instantaneous rate of change of pressure, the variance of pressure fluctuation, and the temperature-pressure coupling coefficient. At the same time, it uses a sliding window algorithm to establish a dynamic baseline model and generate a state feature set, which is then sent to the adaptive decision module. The adaptive decision-making module sets a reference model of the pressure trajectory during the secondary fermentation of the sparkling wine, compares the state feature set with the expected pressure value output by the reference model, tunes the PID control parameters online through an adaptive law, generates control instructions for the fermentation tank, and determines whether to generate early warning information based on the degree of pressure deviation, and sends the control instructions to the collaborative control module. The collaborative control module parses the control commands into drive signals for the actuators, controls the pressure inside the fermenter, performs conflict detection and load balancing scheduling when multiple tanks execute concurrently, outputs the execution results, and feeds them back to the adaptive decision module.
2. The adaptive pressure control system for the secondary fermentation process of sparkling wine according to claim 1, characterized in that: The feature extraction module includes a rate of change calculation unit, a fluctuation analysis unit, a coupled feature unit, and a feature output unit; The rate of change calculation unit performs first-order difference calculation on the pressure values in the standardized time series data to obtain the instantaneous rate of change of pressure, and calculates the mean and peak values of the instantaneous rate of change of pressure within the sliding window to characterize the dynamic change of the gas production rate. The fluctuation analysis unit uses the sliding window algorithm to calculate the variance of the pressure values within the sliding window as a pressure fluctuation feature, and uses the LOF algorithm to identify outliers in the pressure value sequence, mark abnormal fluctuation points, and generate fluctuation feature values. The coupling feature unit calculates the cross-correlation coefficient between the time-series data of pressure and temperature values, which serves as the temperature-pressure coupling coefficient. A baseline model is established based on historical data, and the expected pressure range at the current moment is dynamically calculated using the exponential weighted moving average method. If the real-time deviation from the baseline exceeds a preset threshold, it is marked as a potential abnormal state. The feature output unit outputs the pressure change rate, pressure fluctuation characteristics, and temperature-pressure coupling coefficient as a set of state features.
3. The adaptive pressure control system for the secondary fermentation process of sparkling wine according to claim 1, characterized in that: The adaptive decision-making module includes a reference model unit, an adaptive law unit, a hybrid controller unit, and an early warning unit; The reference model unit is pre-set with segmented target pressure values for different stages of secondary fermentation, and sets pressure threshold ranges for each. It uses a conformal segmented cubic interpolation method to fit the discrete segmented target pressure values into a continuous and differentiable ideal fermentation process curve. Based on the ideal fermentation process curve, a reference model for setting the pressure trajectory is set using a second-order linear system description. The natural frequency and damping ratio of the reference model are set, the state feature set is input, and the expected pressure value under the ideal state is output. The adaptive law unit uses the pressure deviation between the real-time pressure value and the desired pressure value as the driving quantity, and adjusts the PID control parameters of the hybrid controller unit online through the adaptive mechanism, sets the initial gain coefficient and adaptive learning rate, so that the real-time pressure value gradually approaches the desired pressure value under the ideal state. The hybrid controller unit calculates the control command for the fermenter based on the real-time adjusted PID control parameters and the single-neuron PID algorithm. The PID control parameters are mapped to the proportional coefficients of the neurons. The input layer of the single neuron receives the pressure deviation value and its integral and derivative signals. The output signal of the PID controller is smoothly adjusted by online weight correction. The early warning unit sets a pressure deviation threshold. When the pressure deviation value continuously exceeds the pressure deviation threshold, it generates an early warning message that includes the abnormality level and recommended measures.
4. The adaptive pressure control system for the secondary fermentation process of sparkling wine according to claim 3, characterized in that: The calculation method for the control command in the hybrid controller unit specifically includes the following steps: Step S1: Construct the neuron input vector, using the pressure deviation value, deviation integral, and deviation derivative as the three input components of the neuron; Step S2: Initialize the neuron weights, corresponding to the contribution of proportional, integral, and differential terms in the output; Step S3: Calculate control instructions, calculate the control output instructions of the neuron, and use supervised Hebb learning rules to correct the neuron weights online; Step S4: Control action conversion, converting the control output command into the specific action amount of the fermenter valve opening, the specific action amount of the fermenter valve opening is proportional to the absolute value of the control output command.
5. The adaptive pressure control system for the secondary fermentation process of sparkling wine according to claim 1, characterized in that: The collaborative control module includes an instruction parsing unit, a valve group driving unit, a conflict detection unit, and a load balancing unit; The instruction parsing unit receives the control instruction, converts the control instruction into the valve opening value and action duration, and generates a valve control signal to send to the valve group drive unit. The valve group drive unit drives the motors of the exhaust valve and the replenishment valve according to the valve control signal, and feeds back the actual opening degree of the valve in real time through the valve position sensor to form a closed-loop control. The conflict detection unit monitors the command execution status of each fermenter. When it detects that multiple tanks simultaneously request large-flow exhaust, causing a sudden drop in the pressure of the shared gas path, it uses a command priority sorting and time slice rotation mechanism. The load balancing unit counts the frequency of control and valve action of each tank. When the number of fermentation tanks in the same batch exceeds the set threshold, a dynamic scheduling algorithm is used to allocate computing tasks between edge nodes and the central server, generate a multi-tank collaborative operation report, and record the control compliance rate and energy consumption indicators of each tank.