Craft IPA beer microbial population optimization method based on improved bat algorithm

Abstract

The present invention discloses a method for optimizing the microbial population of craft IPA beer based on an improved bat algorithm. The method comprises the following steps: S1. constructing an initial state space for dynamic optimization of yeast microbial population ratio; S2. generating standardized real-time process data; S3. fusing the standardized real-time process data with the initial state space to obtain an updated comprehensive state vector, and dividing the entire fermentation process into a sugar collection stage, a main fermentation stage, and a post-ripening stage in chronological order; S4. obtaining a staged entropy weight vector; S5. formulating a multi-objective optimization problem; S6. obtaining an optimal microbial population ratio parameter set and an optimal fermentation control parameter set; and S7. transmitting the optimal microbial population ratio parameter set and the optimal fermentation control parameter set to a microfluidic dosing device, a feeding control unit, and a variable frequency temperature control unit to achieve microbial population optimization for craft IPA beer. The method can effectively guide the optimization path to approach the flavor synergistic expression zone, significantly improving the spatial consistency between yeast metabolism and aroma generation.

Description

Technical Field

[0001] The present invention relates to the technical field of beer, and in particular to a method for optimizing the microbial population of craft IPA beer based on an improved bat algorithm. Background Art

[0002] With the development of digital manufacturing and intelligent optimization algorithms, the modern craft beer industry is gradually exploring the introduction of data-driven online optimization methods into the fermentation process to improve flavor consistency and production efficiency. Among them, IPA-style beer faces higher process challenges in flavor regulation due to its rich ester and terpene aromas and the use of multiple yeast species in brewing. Currently, the yeast community of IPA beer is mainly composed of cerevisiae and Brettanomyces. There are significant differences between these two types of bacteria in metabolic behavior, nutritional requirements, fermentation cycle and aroma expression. Coordinated control of their ratio plays a decisive role in the flavor of the end product.

[0003] However, the existing IPA beer yeast population ratio control method mostly adopts a static feeding strategy, that is, a fixed population ratio is set based on experience at the beginning of fermentation and remains unchanged throughout the fermentation process. The static method cannot dynamically respond to changes in the fermentation environment. Under external disturbances such as raw material fluctuations, temperature drift, and changes in dissolved oxygen levels, it is easy to cause yeast population imbalance and metabolic disorder, resulting in flavor deviation and decreased consistency between batches.

[0004] In actual processes, the yeast's absorption rate of sugar and nitrogen substrates has obvious stage characteristics, and there is also a dynamic conflict between the aroma generation and by-product inhibition goals in different stages. Traditional algorithms often lack a precise stage-by-stage weight adjustment mechanism when dealing with these timing goals, making it difficult to achieve flavor synergy throughout the entire process. In addition, existing swarm intelligence algorithms, such as the standard bat algorithm, are prone to search hysteresis or oscillation problems in high-dimensional nonlinear process parameter space, and are unable to stably output optimization decision-making plans suitable for actual brewing sites.

[0005] Therefore, it is urgent to propose a new optimization method that integrates the phased perception mechanism, dynamic weight adjustment strategy and real-time convergence control capability to effectively cope with the complexity of regulating the proportion of bacterial populations during the fermentation process and the need to maintain flavor consistency. Summary of the Invention

[0006] One purpose of the present invention is to propose a method for optimizing the craft IPA beer flora based on an improved bat algorithm. The present invention can effectively guide the optimization path close to the flavor synergistic expression area, greatly improving the spatial consistency of yeast metabolism and aroma generation.

[0007] According to an embodiment of the present invention, a method for optimizing the microbial population of craft IPA beer based on an improved bat algorithm includes the following steps:

[0008] S1. Collect initial data on IPA beer fermentation and construct an initial state space for dynamic optimization of yeast population proportions;

[0009] S2. Standardize the initial data of IPA beer fermentation to form standardized real-time process data;

[0010] S3. Fusing the standardized real-time process data with the initial state space to obtain an updated comprehensive state vector, and dividing the entire fermentation process into the sugar collection stage, the main fermentation stage, and the post-ripening stage in chronological order;

[0011] S4. Based on the comprehensive state vector, yeast growth rate, ethyl acetate concentration, isovaleric acid concentration, and fermentation time are selected as multi-objective optimization indicators. The information entropy of the indicators is used to calculate the stage-by-stage entropy weight for each fermentation stage to obtain the stage-by-stage entropy weight vector.

[0012] S5. Combine the staged entropy weight vector and the comprehensive state vector to construct a staged entropy weight embedded multi-objective fitness function. The multi-objective optimization problem is formulated using the Saccharomyces cerevisiae and Brettanomyces yeast population ratio, the sugar and nitrogen supplementation window, and the fermentation temperature as optimization variables.

[0013] S6. Apply the improved bat algorithm to the multi-objective optimization problem, search within the integrated state vector, and obtain the optimal bacterial population ratio parameter set and the optimal fermentation control parameter set;

[0014] S7. Transmit the optimal bacterial flora ratio parameter set and the optimal fermentation control parameter set to the microfluidic dosing device, feeding control unit, and variable frequency temperature control unit to adjust the ratio of brewer's yeast and wine yeast, and simultaneously adjust the sugar and nitrogen supplementation amount and fermentation temperature to achieve the optimization of the craft IPA beer flora.

[0015] Optionally, the S1 includes the following steps:

[0016] S11. Set IPA beer fermentation batch number , collect initial data of IPA beer fermentation for each fermentation batch , including wort fermentable sugar concentration, pH value, initial bacterial population ratio of Saccharomyces cerevisiae and Brettanomyces, and target aroma spectrum parameters;

[0017] S12. Constructing the initial state space for dynamic optimization of IPA beer yeast population proportions , the initial state vector in the initial state space records the five-dimensional features of the initial data of IPA beer fermentation.

[0018] Optionally, S2 includes the following steps:

[0019] S21. Number each IPA beer fermentation batch Corresponding IPA beer fermentation initial data Each feature data in is normalized to obtain the normalized feature value;

[0020] S22. Recombining all normalized eigenvalues ​​to form an IPA beer fermentation batch number The normalized state vector of ;

[0021] S23. Perform wavelet reconstruction and noise reduction on each eigenvalue in the standardized state vector to obtain the denoised eigenvalue;

[0022] S24. Recombine all the denoised eigenvalues ​​to form the IPA beer fermentation batch number The denoised normalized state vector , and the noise-reduced normalized state vector is used as the normalized real-time process data set.

[0023] Optionally, S3 includes the following steps:

[0024] S31. Denoise the normalized state vector With the initial state space Perform fusion and construct IPA beer fermentation batch number The comprehensive state vector ;

[0025] S32 sets the total fermentation time interval, the total fermentation time interval is composed of the fermentation start time and fermentation end time of the batch of IPA beer;

[0026] S33. A threshold set for stage division is set based on the changing trend of the fermentable sugar concentration rate, bacterial concentration curve, and dissolved oxygen curve in the real-time process data. The threshold set includes the end time of the sugar harvesting stage and the end time of the main fermentation stage. The end time of the sugar harvesting stage is defined as the time when the rate of decline of the fermentable sugar concentration first increases significantly. The end time of the main fermentation stage is defined as the time when the dissolved oxygen concentration stabilizes and the ethyl acetate production rate reaches a plateau.

[0027] S34. The total fermentation time interval is divided into three consecutive stages based on the stage division threshold set, namely, the sugar collection stage, the main fermentation stage, and the ripening stage;

[0028] S35. Record the start and end time of each fermentation stage, and combine the start and end times of three consecutive stages in sequence to form a stage time label vector The stage time labeling vector includes the start time and end time of the sugar collection stage, the start time and end time of the main fermentation stage, and the start time and end time of the ripening stage.

[0029] Optionally, the S4 includes the following steps:

[0030] S41. Based on the integrated state vector and stage time label vector , extracting key indicators in the sugar collection stage, main fermentation stage and after-ripening stage, wherein the key indicators include yeast growth rate, ethyl acetate concentration, isovaleric acid concentration and fermentation time, respectively constituting a stage optimization indicator set for each fermentation stage;

[0031] S42. For each stage of fermentation, the stage optimization index is concentrated on each index, and the real-time fluctuation intensity factor of the index in that stage is calculated. , the real-time volatility intensity factor is the ratio of the standard deviation of the real-time measurement value of the indicator in the current stage to the stage average value;

[0032] S43. Calculate the characteristic dynamic weight factor for the optimization index of each fermentation stage The characteristic dynamic weight factor is the real-time fluctuation intensity factor and the average absolute gradient change rate of the indicator in the current stage. The product of , where the mean absolute gradient rate of change is:

[0033] ;

[0034] in, For the first stage The real-time data of an indicator measured at the tth time, For the first stage The real-time data of the indicator measured at the t+1th time, For the The total number of real-time measurement data in the stage;

[0035] S44. Calculate the improved information entropy value for each optimization index in each fermentation stage :

[0036] ;

[0037] in, To prevent the occurrence of extremely small positive constants with zero logarithmic values;

[0038] S45. Based on the improved information entropy value, calculate the stage entropy weight of the optimization index in each fermentation stage , reflecting the dynamic optimization priority of optimization indicators in each fermentation stage;

[0039] S46. Combine the staged entropy weights of the optimization indicators of each stage into a staged entropy weight vector ,in, .

[0040] Optionally, the S5 includes the following steps:

[0041] S51. Set IPA beer fermentation batch number The optimized decision variable vector of :

[0042] ;

[0043] in, Indicates the optimized bacterial population ratio of this batch of brewer's yeast, Indicates the optimized bacterial population ratio of Brettanomyces, and Respectively represent the sugar supplementation time point and nitrogen supplementation time point of this batch, Indicates the optimized fermentation temperature;

[0044] S52. Combining the comprehensive state vector and the staged entropy weight vector , define the stage optimization objective function:

[0045] ;

[0046] in, For fermentation batch In the The stage-by-stage multi-objective loss function, Indicates the The optimization index is in the optimization variable vector The predicted value under is the target value of the indicator at this stage;

[0047] S53. Combine the optimization objective functions of the sugar harvesting stage, main fermentation stage and post-ripening stage to construct a multi-objective comprehensive optimization function for the entire IPA beer fermentation process ;

[0048] S54. Constructing a multi-objective optimization constraint set , including bacterial population ratio constraints, fermentation temperature adjustable range constraints and feeding window range constraints;

[0049] S55. In the multi-objective comprehensive optimization function and constraints Based on this, a mathematical model for dynamic optimization of the ratio of brewer's yeast and Brettanomyces yeast populations in IPA beer was constructed. .

[0050] Optionally, the S6 includes the following steps:

[0051] S61. Initialize IPA beer fermentation batch number The corresponding bat population, the first The position vector of each bat is represented as the optimization decision variable vector ;

[0052] S62. Based on the physiological characteristics of the decay of the activities of brewer's yeast and Brettanomyces yeast during IPA beer fermentation, a dynamic loudness attenuation strategy update formula is proposed:

[0053] ;

[0054] in, For the The bat in the The updated dynamic loudness value in the iteration, For the The bat in the The dynamic loudness value in the iteration, is the basic loudness attenuation coefficient, is the yeast activity sensitivity coefficient in the IPA brewing process, The real-time fluctuation intensity factor of yeast growth rate reflects the dynamic constraint effect of yeast activity fluctuation on search step length during the fermentation stage;

[0055] S63. The staged entropy weight vector Introducing bat local search position update, a phased entropy weight-guided local search strategy is constructed for optimizing the proportion of IPA beer yeast population:

[0056] ;

[0057] in, is the local search step factor, Indicates the The importance coefficient of the stage, is the predicted value of the jth optimization indicator in the nth stage corresponding to the position of the i-th bat in the t-1th iteration, is the predicted value of the jth optimization indicator in the nth stage corresponding to the global optimal bat position in the t-1th iteration;

[0058] S65. Set the convergence judgment condition of the IPA beer yeast population ratio optimization problem as follows: the change of the multi-objective comprehensive optimization function corresponding to the optimal position of the population for Q consecutive iterations is lower than the threshold. If this condition is met or the preset maximum number of iterations is reached, stop the iteration and obtain the IPA beer fermentation batch number. The optimal bacterial population ratio parameter set and the optimal fermentation control parameter set .

[0059] Optionally, during the search process of the improved bat algorithm, the flavor fluctuation trend during the IPA beer brewing stage is used to dynamically adjust the adaptive pulse frequency of the bat population. The adaptive pulse frequency parameter is composed of the minimum pulse frequency parameter and the maximum pulse frequency parameter, and changes in an exponential function manner. The exponential function attenuation of the pulse frequency is based on the function value change amplitude of the multi-objective comprehensive optimization function between two adjacent iterations. control.

[0060] Optionally, the S7 includes the following steps:

[0061] S71, the optimal bacterial population ratio parameter set and the optimal fermentation control parameter set Transfer to the microfluidic dosing device, which optimizes the bacterial population ratio according to Saccharomyces cerevisiae Optimize the bacterial flora ratio with Brettanomyces The target injection rate and duration of the two independent injection channels are controlled separately, so that the actual injection ratio reaches the optimization target;

[0062] S72, the time point of sugar supplementation and nitrogen supplementation time point They are respectively sent as instructions to the feeding control unit, and the feeding control unit starts the precision peristaltic pump at the corresponding time point to perform the nutrient material injection operation;

[0063] S73, the optimal fermentation temperature The input is sent to the variable frequency temperature control unit, which dynamically adjusts the water temperature of the fermentation tank cooling jacket and the stirring frequency at a response rate of seconds to keep the temperature inside the tank stable. ±0.3°C error margin.

[0064] Optionally, the nutrient material injection operation specifically includes:

[0065] Sugar supplementation operation: When the fermentable sugar liquid with preset mass concentration is injected into the fermentation tank, the sugar replenishment volume is determined by the current wort residual sugar concentration and fermentation trend;

[0066] Nitrogen supplementation operation: When fermenting, inject amino acid nitrogen source slow-release liquid into the fermentation tank to activate the metabolic potential of Brettanomyces yeast.

[0067] The beneficial effects of the present invention are:

[0068] (1) The present invention introduces a multi-objective weight adaptive mechanism driven by staged information entropy to solve the problems of stage objective conflict and dynamic weight imbalance. A three-stage (sugar collection, main fermentation, and post-ripening) dynamic segmentation mechanism based on the actual IPA beer fermentation process is constructed. In each stage, the real-time fluctuation intensity factor and the average gradient change rate are integrated to calculate the improved information entropy value, and then a staged entropy weight vector is generated based on this. It can accurately capture the optimization sensitivity of key indicators in different fermentation stages in different time periods, realize the temporal reconstruction of the objective function structure in the optimization process, show higher multi-objective coordination ability in complex dynamic systems, and significantly improve the stability of microbial metabolic interaction regulation and the accuracy of flavor control.

[0069] (2) The present invention proposes a pulse frequency and loudness control strategy for brewing disturbance adaptation, which enhances the robustness and directional guidance capability of optimization search. Aiming at the common problems of raw material fluctuation, temperature offset and uneven yeast metabolic activity in the IPA beer fermentation process, the flavor change trend is introduced into the pulse frequency function, and a frequency control formula based on the amplitude of the difference change of the optimization function is designed to give the bat algorithm the ability to respond dynamically to local disturbances. At the same time, a loudness attenuation mechanism is designed in combination with the real-time fluctuation factor of the yeast growth rate to suppress excessive search behavior when the yeast is unbalanced. The dual control structure for perturbation scenarios avoids the problems of frequent oscillation or premature convergence in the standard bat algorithm, and significantly enhances the convergence stability and global optimal search capability of the algorithm in nonlinear complex process space.

[0070] (3) The present invention integrates a local search update strategy guided by multi-stage indicator deviations to achieve distributed convergence optimization of flavor targets. The embedded structure of staged entropy weight-target deviation is introduced into the local search update formula to dynamically adjust the direction and amplitude of each local search. The update strategy no longer relies on traditional position difference or historical optimal path, but instead evaluates the normalized deviation between the key indicators of each stage and their target values ​​under the current decision variable, and then multiplies it by the entropy weight coefficient of each stage to form a comprehensive directional factor, which directly drives the variable adjustment. It can effectively guide the optimization path to approach the flavor synergistic expression area, greatly improving the spatial consistency of yeast metabolism and aroma generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0072] Figure 1 This is a flow chart of a method for optimizing the microbial flora of craft IPA beer based on an improved bat algorithm proposed in the present invention;

[0073] Figure 2This is a flowchart of the staged entropy weight calculation based on real-time fluctuation intensity and gradient changes in the craft IPA beer flora optimization method based on the improved bat algorithm proposed by the present invention;

[0074] Figure 3 This is a schematic diagram of the search structure of the improved bat algorithm in the craft IPA beer flora optimization method based on the improved bat algorithm proposed in the present invention. DETAILED DESCRIPTION

[0075] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0076] refer to Figure 1-Figure 3 A method for optimizing the microbial flora of craft IPA beer based on an improved bat algorithm comprises the following steps:

[0077] S1. Collect initial data for IPA beer fermentation, including wort fermentable sugar concentration, pH, initial Saccharomyces cerevisiae and Brettanomyces yeast population ratios, and target aroma spectrum parameters, to construct an initial state space for dynamic optimization of yeast population ratios.

[0078] S2. Standardize the initial data of IPA beer fermentation and eliminate measurement noise to form standardized real-time process data;

[0079] S3. The standardized real-time process data is integrated with the initial state space to obtain an updated integrated state vector. The fermentation process is then divided into the sugar harvesting stage, the main fermentation stage, and the post-ripening stage in chronological order, and the start and end times of each fermentation stage are recorded.

[0080] S4. Based on the comprehensive state vector, yeast growth rate, ethyl acetate concentration, isovaleric acid concentration, and fermentation time are selected as multi-objective optimization indicators. The information entropy of the indicators is used to calculate the stage-by-stage entropy weight for each fermentation stage to obtain the stage-by-stage entropy weight vector.

[0081] S5. Combine the staged entropy weight vector and the comprehensive state vector to construct a staged entropy weight embedded multi-objective fitness function. The multi-objective optimization problem is formulated using the Saccharomyces cerevisiae and Brettanomyces yeast population ratio, the sugar and nitrogen supplementation window, and the fermentation temperature as optimization variables.

[0082] S6. Apply the improved bat algorithm to the multi-objective optimization problem. The improved bat algorithm uses adaptive pulse frequency control, dynamic loudness attenuation strategy, and staged entropy weight synergy mechanism to search within the integrated state vector to obtain the optimal bacterial population ratio parameter set and the optimal fermentation control parameter set;

[0083] S7. Transmit the optimal bacterial flora ratio parameter set and the optimal fermentation control parameter set to the microfluidic dosing device, feeding control unit, and variable frequency temperature control unit to adjust the ratio of brewer's yeast and wine yeast, and simultaneously adjust the sugar and nitrogen supplementation amount and fermentation temperature to achieve the optimization of the craft IPA beer flora.

[0084] In this embodiment, S1 includes the following steps:

[0085] S11. Set IPA beer fermentation batch number , collect initial data of IPA beer fermentation for each fermentation batch , including wort fermentable sugar concentration, pH value, initial bacterial population ratio of Saccharomyces cerevisiae and Brettanomyces, and target aroma spectrum parameters;

[0086] Wort fermentable sugar concentration indicates IPA beer fermentation batch number Corresponding to the fermentable sugar concentration in wort, unit is The pH value indicates the initial pH of the wort; the Saccharomyces cerevisiae population ratio at the beginning of fermentation is expressed in percentage (%), and is defined as the ratio of the number of Saccharomyces cerevisiae cells to the total number of cells; the initial Brettanomyces population ratio indicates the population ratio of Brettanomyces cerevisiae, and is expressed in percentage (%). The target aroma spectrum parameters are used to characterize the sensory distribution structure of the target volatiles at the end of fermentation.

[0087] In this embodiment, brewer's yeast and Brettanomyces yeast are just two examples of yeasts. In actual applications, other IPA beer yeasts can be cached.

[0088] S12. Constructing the initial state space for dynamic optimization of IPA beer yeast population proportions , the initial state vector in the initial state space records the five-dimensional features of the initial data of IPA beer fermentation.

[0089] In this embodiment, S2 includes the following steps:

[0090] S21. Number each IPA beer fermentation batch Corresponding IPA beer fermentation initial data Each feature data in is normalized to obtain the normalized feature value;

[0091] S22. Recombining all normalized eigenvalues ​​to form an IPA beer fermentation batch number The normalized state vector of ;

[0092] S23. Perform wavelet reconstruction and noise reduction on each eigenvalue in the standardized state vector to obtain the denoised eigenvalue;

[0093] S24. Recombine all the denoised eigenvalues ​​to form the IPA beer fermentation batch number The denoised normalized state vector , and use the noise-reduced normalized state vector as the normalized real-time process data set .

[0094] In this embodiment, S3 includes the following steps:

[0095] S31. Denoise the normalized state vector With the initial state space Perform fusion and construct IPA beer fermentation batch number The comprehensive state vector :

[0096] ;

[0097] in, IPA beer fermentation batch number The denoised normalized state vector, is the corresponding original initial state vector, is the fusion weight coefficient;

[0098] S32 sets the total fermentation time interval, the total fermentation time interval is composed of the fermentation start time and fermentation end time of the batch of IPA beer, the fermentation start time indicates the time when the IPA beer wort enters the fermentation tank, and the fermentation end time indicates the time when the beer enters the maturation stage;

[0099] S33. A threshold set for stage division is set based on the changing trends of the fermentable sugar concentration rate, bacterial cell concentration curve, and dissolved oxygen curve in the real-time process data. The threshold set for stage division includes the end time of the sugar harvesting stage and the end time of the main fermentation stage. The end time of the sugar harvesting stage is defined as the time when the rate of decrease of the fermentable sugar concentration first increases significantly. The end time of the main fermentation stage is defined as the time when the dissolved oxygen concentration stabilizes and the ethyl acetate production rate reaches a plateau.

[0100] S34. Based on the stage division threshold set, the total fermentation time interval is divided into three consecutive stages: a sugar collection stage, a main fermentation stage, and a post-ripening stage. The start and end time of the sugar collection stage is from the fermentation start time to the sugar collection stage end time, the start and end time of the main fermentation stage is from the sugar collection stage end time to the main fermentation stage end time, and the start and end time of the post-ripening stage is from the main fermentation stage end time to the fermentation end time.

[0101] S35. Record the start and end time of each fermentation stage, and combine the start and end times of three consecutive stages in sequence to form a stage time label vector The stage time label vector includes the start and end time of the sugar harvesting stage, the start and end time of the main fermentation stage, and the start and end time of the ripening stage:

[0102] ;

[0103] in, and Respectively represent The start and end time of each phase.

[0104] In this embodiment, S4 includes the following steps:

[0105] S41. Based on the integrated state vector and stage time label vector , extracting key indicators in the sugar collection stage, main fermentation stage and after-ripening stage, wherein the key indicators include yeast growth rate, ethyl acetate concentration, isovaleric acid concentration and fermentation time, respectively constituting a stage optimization indicator set for each fermentation stage;

[0106] S42. For each stage of fermentation, the stage optimization index is concentrated on each index, and the real-time fluctuation intensity factor of the index in that stage is calculated. The real-time volatility intensity factor is the ratio of the standard deviation of the real-time measurement value of the indicator in the current period to the average value of the period:

[0107] ;

[0108] in, For the In the stage The real-time fluctuation intensity factor of each optimization index reflects the dynamic fluctuation characteristics of each optimization index at different fermentation stages;

[0109] S43. Calculate the characteristic dynamic weight factor for the optimization index of each fermentation stage The characteristic dynamic weight factor is the real-time fluctuation intensity factor and the average absolute gradient change rate of the indicator in the current stage. The product of , where the mean absolute gradient rate of change is:

[0110] ;

[0111] Among them, among them, For the first stage The real-time data of an indicator measured at the tth time, For the first stage The real-time data of the indicator measured at the t+1th time, For the The total number of real-time measurement data in the stage;

[0112] The dynamic weighting factor of the feature proposed in S43 , which is determined by the real-time fluctuation intensity factor Mean absolute gradient rate The product reflects the uncertainty of the change of a certain indicator in a certain fermentation stage × the intensity of fluctuation. It is an effective parameter to measure the sensitivity of this feature in the yeast fermentation system to the disturbance of the control system.

[0113] Different from the conventional information entropy calculation method of equal weight or fixed weight processing of indicators, the present invention directly encodes the data-driven dynamic process characteristics into the weight modeling process, and proposes the concept that the entropy weight factor is no longer determined solely by the distribution, but should be strengthened by dynamic performance, which significantly enhances the responsiveness and adaptability of the information entropy model under actual industrial disturbances.

[0114] S44. Calculate the improved information entropy value for each optimization index in each fermentation stage :

[0115] ;

[0116] in, For the Phase The improved information entropy value of the optimization index, In order to prevent the occurrence of extremely small positive constants with logarithmic zero values, the dynamic change sensitivity of key indicators at each stage of IPA beer fermentation and the importance of stage characteristics are reflected;

[0117] S44 introduces a feature dynamic weight factor based on the traditional information entropy formula ,By linking the entropy calculation of each indicator with its dynamic response ,characteristics, an adaptive entropy expression is formed, so that the ,calculation of entropy not only reflects the numerical distribution ,characteristics, but also maps its contribution to the system output and ,stability risk.

[0118] The formula formally introduces control sensitivity into information theory modeling, realizing the paradigm shift from statistical information entropy to dynamic feature sensitive entropy. For the first time, a high-order coupling relationship between information entropy theory and real-time adjustable variables of the fermentation process is established in the fermentation optimization scenario, breaking the limitation of previous information entropy algorithms on the lack of ability to capture time dynamics.

[0119] S45. Based on the improved information entropy value, the entropy weight of the optimization index in each fermentation stage is calculated:

[0120] ;

[0121] in, For the Phase The staged entropy weight of each optimization indicator reflects the dynamic optimization priority of the indicator in each fermentation stage;

[0122] S46. Combine the staged entropy weights of the optimization indicators of each stage into a staged entropy weight vector ,in, , IPA beer fermentation batch number The staged entropy weight vector is used as the dynamic basis for the stage indicator weights in the multi-objective fitness function during the dynamic optimization of the ratio of brewer's yeast and Brettanomyces yeast populations in IPA beer.

[0123] The present invention introduces a multi-objective weight adaptive mechanism driven by staged information entropy to solve the problems of stage objective conflict and dynamic weight imbalance, constructs a three-stage (sugar collection, main fermentation, and post-ripening) dynamic segmentation mechanism based on the actual IPA beer fermentation process, and integrates the real-time fluctuation intensity factor and the average gradient change rate in each stage to calculate the improved information entropy value, and then generates a staged entropy weight vector based on this. It can accurately capture the optimization sensitivity of key indicators in different fermentation stages in different time periods, realize the temporal reconstruction of the objective function structure in the optimization process, show higher multi-objective coordination ability in complex dynamic systems, and significantly improve the stability of microbial metabolic interaction regulation and the accuracy of flavor control.

[0124] In this embodiment, S5 includes the following steps:

[0125] S51. Set IPA beer fermentation batch number The optimized decision variable vector of :

[0126] ;

[0127] in, Indicates the optimized bacterial population ratio of this batch of brewer's yeast, Indicates the optimized bacterial population ratio of Brettanomyces, and Respectively represent the sugar supplementation time point and nitrogen supplementation time point of this batch, Indicates the optimized fermentation temperature;

[0128] S52. Combining the comprehensive state vector and the staged entropy weight vector , define the set of phased optimization objective functions:

[0129] ;

[0130] in, For fermentation batch In the The stage-by-stage multi-objective loss function, Indicates the The optimization index is in the optimization variable vector The predicted value under is the target value of the indicator at this stage;

[0131] S53. The optimization objective functions of the sugar harvesting stage, the main fermentation stage, and the post-ripening stage are combined to construct a multi-objective comprehensive optimization function for the entire IPA beer fermentation process:

[0132] ;

[0133] in, Number the fermentation batch The multi-objective comprehensive optimization function, For the The importance coefficient of the stage is used to express the weight adjustment requirements for the flavor control of the IPA fermentation stage;

[0134] S54. Constructing a multi-objective optimization constraint set , including bacterial population ratio constraints, fermentation temperature adjustable range constraints and feeding window range constraints, which are specifically defined as follows:

[0135] ;

[0136] The bacterial population ratio is limited to ensure a balanced and coordinated fermentation process. The temperature control range is within the growth range of the IPA-compatible yeast population. The sugar and nitrogen supplementation time window is between the end of the sugar collection stage and the beginning of the main fermentation stage.

[0137] S55. In the multi-objective integrated optimization function and constraints Based on this, a mathematical model for dynamic optimization of the ratio of brewer's yeast and Brettanomyces yeast populations in IPA beer was constructed. The goal is to minimize the weighted sum of squared multi-objective deviations. The optimization variables include the proportion of bacterial populations, the timing of sugar / nitrogen supplementation, and the fermentation temperature, which constitutes a multivariable coupling optimization problem for the IPA beer fermentation system.

[0138] In this embodiment, S6 includes the following steps:

[0139] S61. Initialize IPA beer fermentation batch number The corresponding bat population, the first The position vector of each bat is expressed as the optimization decision variable vector:

[0140] ;

[0141] in, 、 Respectively represent In the iteration The optimized bacterial population ratio of brewer's yeast and brewer's yeast of bats, 、 Represent the time points of sugar and nitrogen supplementation, Indicates the corresponding optimized fermentation temperature;

[0142] S62. Based on the physiological characteristics of the decay of the activities of brewer's yeast and Brettanomyces yeast during IPA beer fermentation, a dynamic loudness attenuation strategy update formula is proposed:

[0143] ;

[0144] in, For the The bat in the The updated dynamic loudness value in the iteration, For the The bat in the The dynamic loudness value in the iteration, is the basic loudness attenuation coefficient, is the yeast activity sensitivity coefficient in the IPA brewing process, The real-time fluctuation intensity factor of yeast growth rate reflects the dynamic constraint effect of yeast activity fluctuation on search step length during the fermentation stage;

[0145] S62 real-time fluctuation intensity factor based on yeast growth rate Loudness value By adjusting the ratio, the system converges slowly when the fermentation activity fluctuates violently and converges quickly when the activity stabilizes, simulating the biological learning mechanism of maintaining diversity when the system is uncertain and pursuing efficiency when the system is stable.

[0146] This formula breaks through the classic limitation of the bat algorithm that loudness only decays exponentially at a fixed ratio. It proposes a dynamic attenuation mechanism driven by the response of physiological parameters of the brewing process, and embeds real-time feedback of the fermentation system status into the core of the optimization strategy control. It is a paradigm for the integration of swarm intelligence algorithms and the dynamic physiological processes of industrial yeast, and is practical in the field of bioprocess control optimization.

[0147] S63. The staged entropy weight vector Introducing bat local search position update, a phased entropy weight-guided local search strategy is constructed for optimizing the proportion of IPA beer yeast population:

[0148] ;

[0149] in, is the local search step factor, Indicates the The importance coefficient of the stage, is the predicted value of the jth optimization indicator in the nth stage corresponding to the position of the i-th bat in the t-1th iteration, is the predicted value of the jth optimization indicator in the nth stage corresponding to the global optimal bat position in the t-1th iteration;

[0150] S63 implements a stage-by-index-weighted ontology-guided search mechanism. The traditional bat algorithm relies only on global optimization and random perturbations to update particle positions, ignoring the physiological stage characteristics and coupling weights of the optimization target. Through stage decomposition + entropy weight embedding, the algorithm update path has biological behavior-driven characteristics, giving each optimization action a clear metabolic background meaning, which is highly consistent with the actual flavor generation rules in IPA fermentation.

[0151] Not only does it structurally achieve deep integration and embedded weighted modeling of the traditional bat algorithm based on yeast physiological mechanisms and flavor evolution logic, but it also demonstrates algorithm adaptability, optimization path controllability and multi-objective harmony for complex brewing environments at the functional level.

[0152] S65. Set the convergence judgment condition of the IPA beer yeast population ratio optimization problem as follows: the change of the multi-objective comprehensive optimization function corresponding to the optimal position of the population for Q consecutive iterations is lower than the threshold. If this condition is met or the preset maximum number of iterations is reached, stop the iteration and obtain the IPA beer fermentation batch number. The optimal bacterial population ratio parameter set and the optimal fermentation control parameter set:

[0153] ;

[0154] in, 、 、 、 and Respectively represent IPA beer fermentation batch numbers The optimized bacterial population ratio, the best time to supplement sugar and nitrogen, and the optimal fermentation temperature.

[0155] In this embodiment, the bat algorithm is improved to search for flavor fluctuations during the IPA beer brewing phase and dynamically adjust the adaptive pulse frequency of the bat population. The adaptive pulse frequency parameter is composed of the minimum pulse frequency parameter and the maximum pulse frequency parameter, and changes in an exponential function manner. The exponential function attenuation of the pulse frequency is based on the function value change amplitude of the multi-objective comprehensive optimization function between two adjacent iterations. control:

[0156] ;

[0157] in, It represents the variation of the multi-objective comprehensive optimization function value between two adjacent iterations, reflecting the flavor fluctuation trend during the IPA beer brewing stage. 、 Respectively represent the lower and upper limits of the pulse frequency, Frequency adjustment sensitivity factor optimized for IPA brewing.

[0158] The exponential decay of the pulse frequency parameter is based on the change in the function value of the multi-objective comprehensive optimization function between two adjacent iterations. , that is, the difference between the optimization function value of the current bat and the optimization function value of the bat in the previous iteration. The larger the amplitude, the more unstable the convergence trend of the objective function, and the bat should maintain a higher frequency to enhance the global search ability; the smaller the amplitude, the more stable the convergence trend of the objective function, and the bat should reduce the frequency to enhance the local search accuracy. The adaptive pulse frequency parameter It is between the minimum pulse frequency parameter and the maximum pulse frequency parameter, with the current multi-objective comprehensive optimization function variation range The result of exponential regulation of the variables reflects the influence of the convergence trend of flavor characteristics on the dynamic control of search frequency during the fermentation process of IPA beer.

[0159] This embodiment constructs an exponential decay model based on the change range of the optimization function value. Real-time adjustment of pulse frequency That is, the greater the fluctuation of the optimization path, the lower the jumping frequency of the bat algorithm and the more cautious the search; on the contrary, when the convergence trend is obvious, the frequency increases and the jumping search is accelerated. This is consistent with the control characteristics of the IPA brewing process, which requires exploration in the early stage of fermentation and stability in the later stage.

[0160] By directly using the convergence characteristics of the objective function as the main driving factor for adjusting the search frequency, the bat algorithm is transformed from a passive parameter control mode to a behavioral intelligent agent structure that is feedback-driven and responds to target changes. In the yeast colony control task with multiple targets and coupled disturbances, the algorithm's responsiveness and process robustness are significantly improved.

[0161] The present invention proposes a pulse frequency and loudness control strategy for brewing disturbance adaptation, which enhances the robustness and directional guidance capability of optimization search. Aiming at the common problems of raw material fluctuation, temperature offset and uneven yeast metabolic activity in the IPA beer fermentation process, the flavor change trend is introduced into the pulse frequency function, and a frequency control formula based on the amplitude of the difference change of the optimization function is designed to give the bat algorithm the ability to respond dynamically to local disturbances. At the same time, a loudness attenuation mechanism is designed in combination with the real-time fluctuation factor of the yeast growth rate to suppress excessive search behavior when the yeast is unbalanced. The dual control structure for perturbation scenarios avoids the problems of frequent oscillation or premature convergence in the standard bat algorithm, and significantly enhances the convergence stability and global optimal search capability of the algorithm in nonlinear complex process space.

[0162] In this embodiment, S7 includes the following steps:

[0163] S71, the optimal bacterial population ratio parameter set and the optimal fermentation control parameter set Transfer to the microfluidic dosing device, which optimizes the bacterial population ratio according to Saccharomyces cerevisiae Optimize the bacterial flora ratio with Brettanomyces The target injection rate and duration of the two independent injection channels are controlled separately, so that the actual injection ratio reaches the optimization target;

[0164] S72, the time point of sugar supplementation and nitrogen supplementation time point They are respectively sent as instructions to the feeding control unit, and the feeding control unit starts the precision peristaltic pump at the corresponding time point to perform the nutrient material injection operation;

[0165] S73, the optimal fermentation temperature The input is sent to the variable frequency temperature control unit, which dynamically adjusts the water temperature of the fermentation tank cooling jacket and the stirring frequency at a response rate of seconds to keep the temperature inside the tank stable. ±0.3°C error range;

[0166] S74. Three types of control instructions are issued simultaneously through the central process execution platform to form a multi-parameter collaborative control closed loop of "microbial flora delivery - nutrient supplementation - temperature control adjustment", and the dynamic microbial flora optimization execution process of the IPA beer fermentation process is completed in real time to ensure the real-time consistency between the actual execution parameters and the optimal decision parameters, thereby improving the stability and control accuracy of the fermentation flavor expression.

[0167] In this embodiment, the nutrient material injection operation specifically includes:

[0168] Sugar supplementation operation: When the fermentable sugar liquid with preset mass concentration is injected into the fermentation tank, the sugar replenishment volume is determined by the current wort residual sugar concentration and fermentation trend;

[0169] Nitrogen supplementation operation: When fermenting, inject amino acid nitrogen source slow-release liquid into the fermentation tank to activate the metabolic potential of Brettanomyces yeast and enhance the ester production capacity in the later stage of fermentation.

[0170] Example 1: A craft brewery located in Area A with an annual production capacity of approximately 3,200 tons applied the present invention as the main fermentation control system for the first time during the execution of the 125th batch of IPA fermentation task (batch number B125), and performed dynamic decision-making and closed-loop regulation for the entire process of the batch.

[0171] During the fermentation preparation phase, on-site personnel used a wort flow meter and an automatic sampling unit to collect the following initial data:

[0172] Fermentable sugar concentration: 167.4 g / L; initial pH: 5.06; initial proportion of Saccharomyces cerevisiae: 74.0%; proportion of Brettanomyces: 26.0%; main components of the target aroma spectrum (first three GC-MS peaks): limonene, ethyl acetate, and β-myrcene;

[0173] The data constitutes the initial state vector , and mapped to the initial five-dimensional feature space .

[0174] After fermentation starts, the system collects and inputs raw data from the fiber optic DO probe, pH electrode, and infrared yeast counter every 5 minutes. The sugar concentration, dissolved oxygen, and conductivity are standardized using range normalization. The one-dimensional Daubechies-4 wavelet filter is used for noise reduction to obtain a smooth curve, and finally the standardized state vector after noise reduction is formed. , building a standardized real-time process data set .

[0175] Identify three physiological stages through real-time specific gravity curve and yeast concentration growth rate change rate:

[0176] Sugar harvesting stage: 0–8.3h; main fermentation stage: 8.3–51.5h; ripening stage: 51.5–80.2h; the start and end time combination of each stage is a vector .

[0177] The system dynamically calculates the fluctuation intensity and change rate of the optimization indicators (yeast growth rate, ethyl acetate concentration, isovaleric acid concentration, and stage duration) based on the data within each stage, and constructs a dynamic entropy weight model:

[0178] The fluctuation intensity of ethyl acetate in the main fermentation stage is 0.36; the fluctuation intensity of isovaleric acid in the ripening stage is 0.41; the final stage entropy weight vector is:

[0179] ;

[0180] Combining the staged entropy weight vector and the comprehensive state vector, the staged objective function and the overall optimization function are constructed:

[0181] Optimization variables: bacterial population ratio, sugar supplementation time, nitrogen supplementation time, temperature; optimization goal: minimize the sum of square deviations of indicators at each stage; constraints: , ; The final optimization model constructed:

[0182] ;

[0183] Improved Bat Algorithm Search Algorithm Startup adopts,adaptive adjustment of pulse frequency according to flavor function jitter amplitude,loudness control combined with dynamic attenuation of yeast growth activity,and staged entropy weight for local search direction offset;

[0184] During the 17th iteration, a bat numbered BA-19 output the optimal solution. The optimal bacterial community ratio was: 68.4% for brewer's yeast and 31.6% for brewer's yeast. The optimal temperature was 20.9°C, the sugar replenishment time was 9.7h, and the nitrogen replenishment time was 24.2h. The system recorded the function convergence trajectory, and the iteration took 21.3 seconds.

[0185] The execution module adjusts the following actions: the microfluidic pump increases the amount of Brettanomyces from 26 mL to 48 mL; the temperature control device adjusts the temperature at a rate of 0.25°C / min to 20.9°C; the precision sugar feeding device automatically injects 62 g of glucose solution at 9.7 hours; and the nitrogen feeding device injects 23 mL of amino nitrogen complex solution at 24.2 hours.

[0186] At the same time, the system records the execution log (number LOG-B125) and feedbacks the error of each operation (the control accuracy is better than ±0.4%).

[0187] Table 1 Comparative data of the effects of the method of the present invention and the traditional fixed ratio

[0188]

[0189] This Example 1 demonstrates that the full-link closed-loop process from fermentation data collection, state space construction, entropy weight modeling, algorithm optimization, parameter update to industrial execution forms a highly structured fermentation optimization application system, verifying the comprehensive advantages of the present invention in flavor consistency control, fermentation efficiency improvement and actual workshop adaptability.

[0190] The present invention integrates a local search update strategy guided by multi-stage indicator deviations to achieve distributed convergence optimization of flavor targets. It introduces an embedded structure of staged entropy weight-target deviation into the local search update formula to dynamically adjust the direction and amplitude of each local search. The update strategy no longer relies on traditional position differences or historical optimal paths, but instead evaluates the normalized deviation between the key indicators of each stage and their target values ​​under the current decision variables, and then multiplies it by the entropy weight coefficient of each stage to form a comprehensive directional factor, which directly drives variable adjustment. This can effectively guide the optimization path to approach the flavor synergistic expression area, greatly improving the spatial consistency of yeast metabolism and aroma generation.

[0191] In summary, the present invention not only makes structural improvements to swarm intelligence optimization from the algorithm level, but also closely combines the physical properties and execution characteristics of the IPA brewing process, achieving a balance between optimization accuracy and industrial feasibility in a dynamic environment, demonstrating its wide promotion and application value.

[0192] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for optimizing the microbial flora of craft IPA beer based on an improved bat algorithm, characterized in that: The steps include: S1. Collect initial data on IPA beer fermentation and construct an initial state space for dynamic optimization of yeast population proportions; S2. Standardize the initial data of IPA beer fermentation to form standardized real-time process data; S3. Fusing the standardized real-time process data with the initial state space to obtain an updated comprehensive state vector, and dividing the entire fermentation process into the sugar collection stage, the main fermentation stage, and the post-ripening stage in chronological order; S4. Based on the comprehensive state vector, yeast growth rate, ethyl acetate concentration, isovaleric acid concentration, and fermentation time are selected as multi-objective optimization indicators. The information entropy of the indicators is used to calculate the stage-by-stage entropy weight for each fermentation stage to obtain the stage-by-stage entropy weight vector. S5. Combine the staged entropy weight vector and the comprehensive state vector to construct a staged entropy weight embedded multi-objective fitness function. The multi-objective optimization problem is formulated using the Saccharomyces cerevisiae and Brettanomyces yeast population ratio, the sugar and nitrogen supplementation window, and the fermentation temperature as optimization variables. S6. Apply the improved bat algorithm to the multi-objective optimization problem, search within the integrated state vector, and obtain the optimal bacterial population ratio parameter set and the optimal fermentation control parameter set; S7. Transmit the optimal bacterial flora ratio parameter set and the optimal fermentation control parameter set to the microfluidic dosing device, feeding control unit, and variable frequency temperature control unit to adjust the ratio of brewer's yeast and wine yeast, and simultaneously adjust the sugar and nitrogen supplementation amount and fermentation temperature to achieve the optimization of the craft IPA beer flora.

2. The method for optimizing the microbial flora of craft IPA beer based on the improved bat algorithm according to claim 1, characterized in that: Said S1 comprises the following steps: S11. Set IPA beer fermentation batch number , collect initial data of IPA beer fermentation for each fermentation batch , including wort fermentable sugar concentration, pH value, initial bacterial population ratio of Saccharomyces cerevisiae and Brettanomyces, and target aroma spectrum parameters; S12. Constructing the initial state space for dynamic optimization of IPA beer yeast population proportions , the initial state vector in the initial state space records the five-dimensional features of the initial data of IPA beer fermentation.

3. The method for optimizing the craft IPA beer flora based on the improved bat algorithm according to claim 2, characterized in that: The S2 comprises the following steps: S21. Number each IPA beer fermentation batch Corresponding IPA beer fermentation initial data Each feature data in is normalized to obtain the normalized feature value; S22. Recombining all normalized eigenvalues ​​to form an IPA beer fermentation batch number The normalized state vector of ; S23. Perform wavelet reconstruction and noise reduction on each eigenvalue in the standardized state vector to obtain the denoised eigenvalue; S24. Recombine all the denoised eigenvalues ​​to form the IPA beer fermentation batch number The denoised normalized state vector , and the noise-reduced normalized state vector is used as the normalized real-time process data set.

4. The method for optimizing the craft IPA beer flora based on the improved bat algorithm according to claim 3, characterized in that: The S3 includes the following steps: S31. Denoise the normalized state vector With the initial state space Perform fusion and construct IPA beer fermentation batch number The comprehensive state vector ; S32 sets the total fermentation time interval, the total fermentation time interval is composed of the fermentation start time and fermentation end time of the batch of IPA beer; S33. A threshold set for stage division is set based on the changing trend of the fermentable sugar concentration rate, bacterial concentration curve, and dissolved oxygen curve in the real-time process data. The threshold set includes the end time of the sugar harvesting stage and the end time of the main fermentation stage. The end time of the sugar harvesting stage is defined as the time when the rate of decline of the fermentable sugar concentration first increases significantly. The end time of the main fermentation stage is defined as the time when the dissolved oxygen concentration stabilizes and the ethyl acetate production rate reaches a plateau. S34. The total fermentation time interval is divided into three consecutive stages based on the stage division threshold set, namely, the sugar collection stage, the main fermentation stage, and the ripening stage; S35. Record the start and end time of each fermentation stage, and combine the start and end times of three consecutive stages in sequence to form a stage time label vector The stage time labeling vector includes the start time and end time of the sugar collection stage, the start time and end time of the main fermentation stage, and the start time and end time of the ripening stage.

5. The method for optimizing the microbial flora of craft IPA beer based on the improved bat algorithm according to claim 4, characterized in that: The S4 comprises the following steps: S41. Based on the integrated state vector and stage time label vector , extracting key indicators in the sugar collection stage, main fermentation stage and after-ripening stage, wherein the key indicators include yeast growth rate, ethyl acetate concentration, isovaleric acid concentration and fermentation time, respectively constituting a stage optimization indicator set for each fermentation stage; S42. For each stage of fermentation, the stage optimization index is concentrated on each index, and the real-time fluctuation intensity factor of the index in that stage is calculated. , the real-time volatility intensity factor is the ratio of the standard deviation of the real-time measurement value of the indicator in the current stage to the stage average value; S43. Calculate the characteristic dynamic weight factor for the optimization index of each fermentation stage The characteristic dynamic weight factor is the real-time fluctuation intensity factor and the average absolute gradient change rate of the indicator in the current stage. The product of S44. Calculate the improved information entropy value for each optimization index in each fermentation stage : ； in, To prevent the occurrence of extremely small positive constants with zero logarithmic values; S45. Based on the improved information entropy value, calculate the stage entropy weight of the optimization index in each fermentation stage , reflecting the dynamic optimization priority of optimization indicators in each fermentation stage; S46. Combine the staged entropy weights of the optimization indicators of each stage into a staged entropy weight vector ,in, .

6. The method for optimizing the craft IPA beer flora based on the improved bat algorithm according to claim 5, characterized in that: The S5 comprises the following steps: S51. Set IPA beer fermentation batch number The optimized decision variable vector of ,in, Indicates the optimized bacterial population ratio of this batch of brewer's yeast, Indicates the optimized bacterial population ratio of Brettanomyces, and Respectively represent the sugar supplementation time point and nitrogen supplementation time point of this batch, Indicates the optimized fermentation temperature; S52. Combining the comprehensive state vector and the staged entropy weight vector , define the stage optimization objective function ; S53. Combine the optimization objective functions of the sugar harvesting stage, main fermentation stage and post-ripening stage to construct a multi-objective comprehensive optimization function for the entire IPA beer fermentation process ; S54. Constructing a multi-objective optimization constraint set , including bacterial population ratio constraints, fermentation temperature adjustable range constraints and feeding window range constraints; S55. In the multi-objective integrated optimization function and constraints Based on this, a mathematical model for dynamic optimization of the ratio of brewer's yeast and Brettanomyces yeast populations in IPA beer was constructed. .

7. The method for optimizing the craft IPA beer flora based on the improved bat algorithm according to claim 6, characterized in that: The S6 comprises the following steps: S61. Initialize IPA beer fermentation batch number The corresponding bat population, the first The position vector of each bat is represented as the optimization decision variable vector ; S62. Based on the physiological characteristics of the decay of the activities of brewer's yeast and Brettanomyces yeast during IPA beer fermentation, a dynamic loudness attenuation strategy update formula is proposed: ； in, For the The bat in the The updated dynamic loudness value in the iteration, For the The bat in the The dynamic loudness value in the iteration, is the basic loudness attenuation coefficient, is the yeast activity sensitivity coefficient in the IPA brewing process, The real-time fluctuation intensity factor of yeast growth rate reflects the dynamic constraint effect of yeast activity fluctuation on search step length during the fermentation stage; S63. The staged entropy weight vector By introducing the bat local search position update, a phased entropy weight guided local search strategy is constructed for optimizing the proportion of IPA beer yeast population. S65. Set the convergence judgment condition of the IPA beer yeast population ratio optimization problem as follows: the change of the multi-objective comprehensive optimization function corresponding to the optimal position of the population for Q consecutive iterations is lower than the threshold. If this condition is met or the preset maximum number of iterations is reached, stop the iteration and obtain the IPA beer fermentation batch number. The optimal bacterial population ratio parameter set and the optimal fermentation control parameter set .

8. The method for optimizing the microbial flora of craft IPA beer based on the improved bat algorithm according to claim 7, characterized in that: During the search process of the improved bat algorithm, the flavor fluctuation trend of IPA beer brewing stage is used to dynamically adjust the adaptive pulse frequency of the bat population. The adaptive pulse frequency parameter is composed of the minimum pulse frequency parameter and the maximum pulse frequency parameter, and changes in an exponential function manner. The exponential function attenuation of the pulse frequency is based on the function value change amplitude of the multi-objective comprehensive optimization function between two adjacent iterations. control.

9. The method for optimizing the microbial flora of craft IPA beer based on the improved bat algorithm according to claim 7, characterized in that: The S7 comprises the following steps: S71, the optimal bacterial population ratio parameter set and the optimal fermentation control parameter set Transfer to the microfluidic dosing device, which optimizes the bacterial population ratio according to Saccharomyces cerevisiae Optimize the bacterial flora ratio with Brettanomyces The target injection rate and duration of the two independent injection channels are controlled separately, so that the actual injection ratio reaches the optimization target; S72, the time point of sugar supplementation and nitrogen supplementation time point They are respectively sent as instructions to the feeding control unit, and the feeding control unit starts the precision peristaltic pump at the corresponding time point to perform the nutrient material injection operation; S73, the optimal fermentation temperature The input is sent to the variable frequency temperature control unit, which dynamically adjusts the water temperature of the fermentation tank cooling jacket and the stirring frequency at a response rate of seconds to keep the temperature inside the tank stable. ±0.3°C error margin.

10. The method for optimizing the microbial flora of craft IPA beer based on the improved bat algorithm according to claim 7, characterized in that: The nutrient material injection operation specifically includes: Sugar supplementation operation: When the fermentable sugar liquid with preset mass concentration is injected into the fermentation tank, the sugar replenishment volume is determined by the current wort residual sugar concentration and fermentation trend; Nitrogen supplementation operation: When fermenting, inject amino acid nitrogen source slow-release liquid into the fermentation tank to activate the metabolic potential of Brettanomyces yeast.

Images