System, method and equipment for automatically predicting retort feeding height in white spirit brewing

Abstract

The invention discloses a system, a method and equipment for automatically predicting the retort feeding height of white spirit brewing, and the system comprises a data collection module which is in communication connection with a field DCS and a production report system of a white spirit brewing production line, and can obtain historical production report data from the field DCS and the production report system; the data cleaning module is in communication connection with the data collection module and can automatically remove abnormal data entries from the historical production report data obtained from the data collection module according to a preset rule; and the retort feeding height prediction model can be used for distributing an independent prediction model for each retort pot and predicting the optimal retort feeding height of the corresponding retort pot according to the historical production report data cleaned by the data cleaning module. The method can accurately predict the retort feeding height, is beneficial to control the optimal cover closing time, and reduces the waste of cover closing raw wine; meanwhile, the steaming time is shortened; the overall brewing efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Baijiu brewing, and particularly relates to a system and method for automatically predicting the height of a distillation head in Baijiu brewing. BACKGROUND

[0002] In the production of Baijiu, the distillation step is one of the key steps that determine the quality of the Baijiu, and the height of the distillation head, as an important parameter in the distillation process, directly affects the distillation efficiency and the extraction of flavoring substances. Accurate prediction and control of the height of the distillation head is of great significance to ensuring the quality of Baijiu and improving production efficiency. Measurement based on physical means involves adjustment of the production line, which is difficult, and there is a situation of measurement uncertainty, which can easily lead to waste of raw liquor. In addition, accurate determination of the time to close the lid is also a key to improving the distillation efficiency. With the complication of the brewing process, it is particularly necessary to use data analysis and machine learning methods to predict key parameters in the brewing process. The prior art has not fully utilized historical data and machine learning techniques to automatically predict the height of the distillation head and optimize the time to close the lid, in order to reduce the loss of raw liquor in the distillation process and improve the efficiency of the distillation head.

[0003] In view of this, the present application is proposed SUMMARY

[0004] The present application provides a system and method for automatically predicting the height of a distillation head in Baijiu brewing, which can accurately predict the height of the distillation head based on multiple historical process parameters in the Baijiu brewing process, thereby optimizing the distillation process and improving the quality and production efficiency of Baijiu, and effectively solving the above technical problems in the prior art.

[0005] The present application is achieved by the following technical solutions: A system for automatically predicting the height of a distillation head in Baijiu brewing, comprising: a data collection module, a data cleaning module, and a distillation head height prediction model, wherein the data collection module is communicatively connected to a field distributed control system and a production report system of a Baijiu brewing production line, and can obtain historical production report data from the field distributed control system and the production report system; the data cleaning module is communicatively connected to the data collection module, and can automatically exclude abnormal data entries from the historical production report data obtained from the data collection module according to a preset rule; the distillation head height prediction model can assign an independent prediction model to each distillation pot, and predict the optimal height of the distillation head of the corresponding distillation pot based on the historical production report data cleaned by the data cleaning module.

[0006] A method for automatically predicting the height of a distillation head in Baijiu brewing for the system of the present application, comprising: acquire historical production report data from the field distributed control system and the production report system through a data collection module of the system; automatically eliminate abnormal data entries from the historical production report data acquired from the data collection module according to a preset rule through a data cleaning module of the system; assign an independent prediction model to each pot through a distillation height prediction model of the system, and predict the optimal distillation height of the corresponding pot according to the historical production report data cleaned by the data cleaning module.

[0007] A processing device comprises: at least one memory for storing one or more programs; at least one processor capable of executing one or more programs stored in the memory, so that the processor can implement the method described in the present application when the one or more programs are executed by the processor.

[0008] Compared with the prior art, the white liquor brewing distillation height automatic prediction system, method and device provided by the present application have the following beneficial effects: By using the distillation height prediction model, the optimal distillation height for a new brewing batch can be predicted through learning of historical production process data related to the distillation height, and real-time adjustment can be made according to the dynamic conditions in the distillation process. After the distillation height is obtained, the steam pressure is controlled more finely based on the "two small and one large method", the waste of raw liquor is reduced, the time used in the distillation process is reduced, and the cost is reduced and the efficiency is increased for the intelligent brewing production of white liquor. BRIEF DESCRIPTION OF DRAWINGS

[0009] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0010] Figure 1 The white liquor brewing distillation height automatic prediction system provided by the embodiment of the present application is shown in the structural diagram.

[0011] Figure 2 The white liquor brewing distillation height automatic prediction method provided by the embodiment of the present application is shown in the flowchart. DETAILED DESCRIPTION

[0012] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the specific contents of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments of the present application, which do not constitute a limitation of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the protection scope of the present application.

[0013] Firstly, the terms possibly used in the present application are described as follows: The term "and / or" means either of the two or both, for example, X and / or Y means three cases including "X" or "Y" or "X and Y".

[0014] The terms "include", "contain", "have", "possess" or other similar semantic descriptions should be interpreted as non-exclusive inclusion. For example, including a technical feature element (such as raw materials, components, ingredients, carriers, dosage forms, materials, sizes, parts, components, mechanisms, devices, steps, processes, methods, reaction conditions, processing conditions, parameters, algorithms, signals, data, products or articles, etc.) should be interpreted as not only including the explicitly listed technical feature element, but also including other technical feature elements not explicitly listed in the art.

[0015] The term "consisting of" means excluding any technical feature element not explicitly listed. If this term is used in the claims, the term will make the claim closed, so that it does not contain technical feature elements other than the explicitly listed technical feature elements, except for conventional impurities related thereto. If the term only appears in a certain clause of the claim, it is only limited to the elements explicitly listed in the clause, and the elements described in other clauses are not excluded from the overall claim.

[0016] Unless otherwise specifically specified or limited, the terms "mount", "connect", "connect", "fix", and the like should be understood broadly, for example: it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium, or it can be connected inside two elements. For those of ordinary skill in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0017] When concentration, temperature, pressure, size or other parameters are expressed in the form of a numerical range, the numerical range should be understood to specifically disclose all ranges formed by the pairing of any upper limit, lower limit, or preferred value within the numerical range, regardless of whether the range is explicitly stated. For example, if a numerical range of "2 to 8" is stated, the numerical range should be interpreted as including ranges of "2 to 7," "2 to 6," "5 to 7," "3 to 4 and 6 to 7," "3 to 5 and 7," "2 and 5 to 7," etc. Unless otherwise specified, the numerical ranges stated herein include both their endpoints and all integers and fractions within the numerical range.

[0018] The terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings and are only for the convenience and simplification of description, and do not explicitly or implicitly indicate that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore should not be understood as a limitation to this document.

[0019] The scheme provided by the present invention is described in detail below. The contents not described in detail in the examples of the present invention belong to the prior art known to professionals in this field. If specific conditions are not specified in the examples of the present invention, they are carried out according to conventional conditions in the field or conditions recommended by the manufacturer. If the manufacturer of the reagents or instruments used in the examples of the present invention is not specified, they are all conventional products that can be purchased commercially.

[0020] like Figure 1 As shown, the embodiment of the present invention provides a highly automated prediction system for the upper steamer of liquor brewing, comprising: Data collection module, data cleaning module and upper steamer height prediction model; among them, The data collection module is connected to the on-site distributed control system (DCS) and production reporting system of the liquor brewing production line, and can obtain historical production report data from the on-site distributed control system and production reporting system; The data cleaning module is in communication with the data collection module and can automatically remove abnormal data entries from the historical production report data obtained from the data collection module according to preset rules; The upper steamer height prediction model can assign an independent prediction model to each steamer pot, and predict the optimal upper steamer height of the corresponding steamer pot based on the historical production report data cleaned by the data cleaning module.

[0021] Preferably, the system further comprises a pressure control module, which is communicatively connected with the still height prediction model and is capable of controlling the steam pressure of the distillation process by the two-small-one-big method according to the optimal still height predicted by the still height prediction model. The two-small-one-big method refers to a segmented steam pressure control method in which small pressures of 3-12 kPa are used in the front and tail sections, and a large pressure of 26-30 kPa is used in the middle section during the still process.

[0022] Preferably, in the system, the historical production report data includes: source type, water weight for moistening grains, weight of fermented grains, weight of bran, weight of added grains, weight of mixing water, and still number.

[0023] Preferably, in the system, the still height prediction model is a machine learning model based on an XGboost algorithm.

[0024] Preferably, in the system, the machine learning model based on the XGboost algorithm is an XGBoost (eXtreme Gradient Boosting) algorithm core, which is an ensemble of gradient boosted trees. The model is composed of multiple CART (Classification and Regression Tree) regression trees, and each sub-model (i.e., decision tree) is trained in series. Each tree is fitted based on the residual error of the previous round of model, thereby gradually optimizing the overall prediction performance. The structure of each regression tree is generated by minimizing the objective function (the sum of the loss term and the regularization term), and supports automatic feature selection, pruning, and underfitting or overfitting control. The loss function used is the mean squared error (MSE).

[0025] The objective function of the machine learning model based on the XGboost algorithm is: ; Wherein, is the value of the overall objective function in the tth round; n is the total number of training samples; i is the sample index; is the actual value of the still height of the ith sample; is the prediction value of the model for the ith sample in the t-1th round; is the prediction increment of the tth new tree for the ith sample; is the training loss function, which is used to measure the difference between the prediction value of the current model and the true value, and is calculated using the squared error, i.e., ; is the prediction value of the previous tree for the sample ; ​The predicted cumulative value of It is a regularization term used to penalize the complexity of the model, including the number of leaf nodes in the tree and the L2 norm of the leaf node weights; The training method of the machine learning model based on the XGboost algorithm is as follows: Initialization: First, the machine learning model starts with an initial prediction value Start with the initial prediction value , in this method The value is the average value of the upper steamer height; Iterative training: (1) In each iteration t, the goal of XGBoost is to train a new decision tree , so that it can fit the residuals of the previous round of model predictions. The residuals here are gradient information, specifically the first-order and second-order gradients of the loss function with respect to the current prediction value.

[0026] (2) For each sample , calculate its first-order gradient under the current model and the second-order gradient ; (3) Using this gradient information, a new decision tree is constructed. During the construction process, a greedy algorithm is used to find the best split point, that is, the split point where the objective function drops the most.

[0027] (4) When a tree is built, an optimal weight is calculated for each leaf node ,in is the sum of the first-order gradients of all samples in leaf node j, is the sum of the second-order gradients of all samples in leaf node j.

[0028] (5) Weight this newly trained tree into the current model.

[0029] (6) Stop condition: The training process continues until the preset maximum number of iterations is reached or the performance of the validation set no longer improves.

[0030] The embodiment of the present invention further provides a method for automatically predicting the height of the upper retort of liquor brewing for the above system, comprising: Obtain historical production report data from the on-site distributed control system and production report system through the data collection module of the system; Automatically remove abnormal data entries from the historical production report data obtained from the data collection module according to preset rules by the data cleaning module of the system; The system is provided with an upper distillation height prediction model, which assigns an independent prediction model to each distillation pot, and predicts the optimal upper distillation height of the corresponding distillation pot according to the historical production report data cleaned by the data cleaning module.

[0031] Preferably, in the system, the pressure control module controls the steam pressure of the distillation process according to the optimal upper distillation height predicted by the upper distillation height prediction model by using a two-small-one-big method. The two-small-one-big method refers to a segmented steam pressure control method in which small pressures of 3-12 kPa are used in the front and tail sections, and a large pressure of 26-30 kPa is used in the middle section during the upper distillation process.

[0032] Preferably, in the system, the historical production report data includes: source type, grain moistening water weight, fermented grains weight, bran weight, added grain weight, mixing water weight, and upper distillation pot number.

[0033] Preferably, in the system, the upper distillation height prediction model is a machine learning model based on an XGboost algorithm. The machine learning model based on the XGboost algorithm is an XGBoost (eXtreme Gradient Boosting) algorithm, and the core of the algorithm is an Ensemble of Gradient Boosted Trees. The model is composed of multiple CART (Classification and Regression Tree) regression trees, and each sub-model (i.e., decision tree) is trained in series. Each tree is fitted based on the residual error of the previous round of model, thereby gradually optimizing the overall prediction performance. The structure of each regression tree is generated by minimizing an objective function (the sum of a loss term and a regularization term), and supports automatic feature selection, pruning, and underfitting or overfitting control.

[0034] The objective function of the machine learning model based on the XGboost algorithm is: ; wherein, is the value of the overall objective function in the tth round; n is the total number of training samples; i is the sample index; is the actual value of the upper distillation height of the ith sample; is the prediction value of the model for the ith sample in the (t-1)th round; is the prediction increment of the tth new tree for the ith sample; is a training loss function, which is used to measure the difference between the prediction value of the current model and the true value, and is calculated using squared error, i.e., ; is the prediction value of the (t-1)th round of model for the ith sample; ; is the prediction value of the tth new tree for the ith sample. a tree for a sample a predicted cumulative value of the sample; is a regularization term, used to penalize the complexity of the model, including the number of leaf nodes of the tree and the L2 norm of the leaf node weights; The training method of the machine learning model based on the XGboost algorithm is as follows: Initialization: First, the machine learning model starts from an initial prediction value , which is the initial prediction value in this method , which is the average value of the height of the upper distillation; Iterative training: (1) In each iteration t, the goal of XGBoost is to train a new decision tree that can fit the residual of the prediction of the previous round model. The residual here is the gradient information, specifically the first and second order gradients of the loss function with respect to the current prediction value.

[0035] (2) For each sample , calculate its first order gradient and second order gradient under the current model; (3) Use these gradient information to build a new decision tree. In the construction process, a greedy algorithm is used to find the best split point, i.e. the split point with the maximum decrease in the objective function.

[0036] (4) When a tree is constructed, an optimal weight is calculated for each leaf node , where is the sum of the first order gradients of all samples in the leaf node , and is the sum of the second order gradients of all samples in the leaf node .

[0037] (5) Add this newly trained tree to the current model with a weight.

[0038] (6) Stopping condition: the training process continues until the maximum number of iterations is reached or the performance on the validation set no longer improves.

[0039] The embodiment of the application further provides a processing device, comprising: at least one memory for storing one or more programs; at least one processor capable of executing one or more programs stored in the memory, so that the processor can implement the above method when the one or more programs are executed by the processor.

[0040] In summary, the system and method provided by the embodiments of the present invention, by adopting an XGBoost-based upper steamer height prediction model and utilizing the analysis and learning of historical production process data related to the upper steamer height, analyzes the relationship between key parameters in the upper steamer process and the upper steamer height, and achieves accurate prediction of the upper steamer height. It can accurately predict the optimal upper steamer height for a new brewing batch and make real-time adjustments based on the dynamic conditions in the upper steamer process. After obtaining the upper steamer height, the steam pressure is more precisely controlled based on the "two small and one large method", thereby optimizing the distillation process, reducing the waste of raw liquor, reducing the time required for the upper steamer process, and improving the quality and production efficiency of liquor.

[0041] In order to more clearly demonstrate the technical solution and technical effects provided by the present invention, the solution provided by the embodiment of the present invention is described in detail with reference to specific embodiments below. Example

[0042] like Figure 1 As shown, this embodiment provides a system for automatically predicting the height of the upper retort in liquor brewing, comprising: a data collection module, a data cleaning module and a prediction model for the height of the upper retort; wherein, The data collection module is responsible for connecting to the on-site DCS and production reporting system, and obtaining historical production process data through scripts; The data cleaning module is in communication with the data collection module and is responsible for automatically removing abnormal data entries in the historical production process data according to preset rules; The upper steamer height prediction model is in communication with the data cleaning module and can assign an independent prediction model to each steamer pot, and predict the optimal upper steamer height of the corresponding steamer pot based on the historical production process data after cleaning.

[0043] Specifically, based on XGBoost and historical retort data, a model parameter space was established, encompassing key parameters such as mash source type, grain water weight, mash weight, husk weight, added grain weight, and mixing water weight. This model was then constructed as an XGBoost model to predict the upper retort height. XGBoost is a machine learning method based on gradient boosting, providing a scalable, portable, and distributed gradient boosting (GBM, GBRT, GBDT) library.

[0044] The system of this embodiment establishes a machine learning model based on XGboost by collecting historical production process data related to the upper steamer during the liquor brewing process (including the type of mash source, the weight of the water used to moisten the grain, the weight of the mash, the weight of the husk, the weight of the added grain, the weight of the mixing water, etc.). By learning from this data, the model can predict the optimal upper steamer height for each new brewing batch and make real-time adjustments based on the dynamic conditions during the upper steamer process. Once the upper steamer height is determined, the steam pressure is more precisely controlled based on the "two small and one large method", reducing the waste of raw liquor when the lid is closed. At the same time, the upper steamer time is shortened, reducing the time required for the upper steamer process, improving overall brewing efficiency, and reducing costs and increasing efficiency for intelligent liquor brewing production.

[0045] In the above system, the retorting process refers to the process of layering the mixed mash into the retort pot. The alcohol in the mash is distilled by adding steam at a certain pressure to the bottom of the pot. The "two-low, one-large" method refers to the steam pressure control method during the retorting process. In the early stages of retorting, to prevent excessive steam pressure from splashing the mash, a lower steam pressure, such as 3-12 kPa, is used. In the middle stages of retorting, to ensure production efficiency, the steam pressure is adjusted to a higher range, such as 26-30 kPa. Towards the end of retorting, excessive steam pressure can cause the wine to escape into the air as vapor when the lid is closed, resulting in waste of the original wine. Therefore, a lower steam pressure, still 3-12 kPa, is used near the end of retorting. Example

[0046] This embodiment provides a method for automatically predicting the height of the upper retort during liquor brewing, which is applied to the system of embodiment 1 and includes: Obtain historical production report data from the on-site DCS and production report system through the data collection module of the system; Automatically remove abnormal data entries from the historical production report data obtained from the data collection module according to preset rules by the data cleaning module of the system; An independent prediction model is allocated to each steamer pot through the upper steamer height prediction model of the system, and the optimal upper steamer height of the corresponding steamer pot is predicted based on the historical production report data cleaned by the data cleaning module.

[0047] Table 1 shows the data verification effect of the upper steamer height prediction method of the present invention in a brewing workshop: .

[0048] The method of the embodiment is verified by multiple production line data, and the average absolute error of the predicted liquor brewing distillation height can be controlled below 23 mm based on the XGBoost model and historical distillation data. It is helpful to control the optimal covering time, reduce the waste of covering base liquor, shorten the distillation time, and improve the overall brewing efficiency. A1, A2, A3, A4 and A5 in Table 1 respectively represent the data sets containing historical distillation height of five different production lines in the same workshop within seven days.

[0049] The automatic prediction system and method of the present application can be deployed in the corresponding production line of the intelligent brewing workshop of the liquor factory to build an intelligent prediction platform for the distillation height, which is convenient for the on-site staff to use.

[0050] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by programs instructing related hardware. The programs can be stored in a computer-readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. The storage medium can be a magnetic disc, an optical disc, a read-only memory (ROM) or a random access memory (RAM), etc.

[0051] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited to this. Any changes or replacements within the technical scope disclosed in the present application can be easily thought of by those skilled in the art, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims. The information disclosed in the background section of the present application is only intended to deepen the understanding of the general background of the present application, and should not be regarded as acknowledging or implying in any form that the information constitutes prior art known to those skilled in the art.

Claims

1. A highly automated prediction system for the steamer in liquor brewing, characterized in that: include: Data collection module, data cleaning module and upper steamer height prediction model; among them, The data collection module is connected to the on-site distributed control system and production reporting system of the liquor brewing production line, and can obtain historical production report data from the on-site DCS and production reporting system; The data cleaning module is in communication with the data collection module and can automatically remove abnormal data entries from the historical production report data obtained from the data collection module according to preset rules; The upper steamer height prediction model can assign an independent prediction model to each steamer pot, and predict the optimal upper steamer height of the corresponding steamer pot based on the historical production report data cleaned by the data cleaning module.

2. The highly automated prediction system for the upper steamer of liquor brewing according to claim 1, characterized in that: Also includes: The pressure control module is communicated with the upper steamer height prediction model and can control the steam pressure of the distillation process through the "two small and one large" method according to the optimal upper steamer height predicted by the upper steamer height prediction model; the "two small and one large" method refers to a segmented steam pressure control method in which a small pressure of 3 to 12 kPa is used in the front and last sections and a large pressure of 26 to 30 kPa is used in the middle section during the upper steamer process.

3. The highly automated prediction system for the upper steamer of liquor brewing according to claim 1 or 2, characterized in that: The historical production report data includes: Type of mash source, weight of moistening water, weight of mash, weight of bran, weight of added grain, weight of mixing water and number of steamer.

4. The highly automated prediction system for the upper steamer of liquor brewing according to claim 3 is characterized in that: The upper steamer height prediction model is a machine learning model based on the XGboost algorithm.

5. The highly automated prediction system for the upper steamer of liquor brewing according to claim 4 is characterized in that: The machine learning model based on the XGboost algorithm is: A weighted gradient boosting decision tree ensemble model based on the XGBoost algorithm. This model consists of multiple classification and regression trees. Each sub-model, acting as a decision tree, is trained sequentially in series. Each decision tree is fitted based on the residuals of the previous model, gradually optimizing the overall prediction performance. The structure of each regression tree is generated by minimizing the objective function and supports automatic feature selection, pruning, and underfitting and overfitting control. The objective function of the machine learning model based on the XGboost algorithm for: ； in, is the overall objective function value of the tth round; n is the total number of training samples; i is the sample index; is the actual value of the upper steamer height of the i-th sample; is the model's predicted value for sample i in round t-1; is the prediction increment of the t-th new tree for sample i; Is the training loss function, used to measure the prediction value of the current model and the true value The difference between them is calculated using the squared error, i.e. ; It is before Tree pairs sample The predicted cumulative value of ; is the predicted value of the sample by the th tree; It is a regularization term used to penalize the complexity of the model, including the number of leaf nodes in the tree and the L2 norm of the leaf node weights; The training method of the machine learning model based on the XGboost algorithm is as follows: Initialization: The machine learning model starts with an initial prediction value Initially, the initial prediction value The value is the average value of the upper steamer height; Iterative training: In each iteration t, the goal of the XGBoost algorithm is to train a new decision tree , so that the decision tree It can fit the residual of the previous round of model prediction. The residual is the gradient information, which is the first-order and second-order gradient of the loss function to the current prediction value. For each sample , calculate the sample The first-order gradient under the current model and the second-order gradient ; Using this gradient information, a new decision tree is constructed. During the construction process, a greedy algorithm is used to find the best split point, that is, the split point where the objective function drops the most; When a tree is built, an optimal weight is calculated for each leaf node ,in Is a leaf node The sum of the first-order gradients of all samples in , Is a leaf node The sum of the second-order gradients of all samples in ; Add this newly trained tree to the current machine learning model; Stop condition: The training process continues until the preset maximum number of iterations is reached or the performance of the validation set no longer improves.

6. A method for highly automated prediction of the upper retort of liquor brewing for use in the system according to any one of claims 1 to 5, characterized in that: include: Obtain historical production report data from the on-site distributed control system and production report system through the data collection module of the system; Automatically remove abnormal data entries from the historical production report data obtained from the data collection module according to preset rules by the data cleaning module of the system; An independent prediction model is allocated to each steamer pot through the upper steamer height prediction model of the system, and the optimal upper steamer height of the corresponding steamer pot is predicted based on the historical production report data cleaned by the data cleaning module.

7. The method for highly automated prediction of the upper steamer for liquor brewing according to claim 6, characterized in that: The system's pressure control module uses a two-small-one-large method to control the steam pressure of the distillation process based on the optimal upper steamer height predicted by the upper steamer height prediction model; the two-small-one-large method refers to a segmented steam pressure control method in which a small pressure of 3 to 12 kPa is used in the front and last sections, and a large pressure of 26 to 30 kPa is used in the middle section during the upper steamer process.

8. The method for highly automated prediction of the steamer level in liquor brewing according to claim 6 or 7, characterized in that: The historical production report data includes: Type of mash source, weight of moistening water, weight of mash, weight of bran, weight of added grain, weight of mixing water and number of steamer.

9. The method for highly automated prediction of the upper steamer for liquor brewing according to claim 3, characterized in that: The upper steamer height prediction model is a machine learning model based on XGboost. The machine learning model based on the XGboost algorithm is: A weighted gradient boosting decision tree ensemble model based on the XGBoost algorithm. This model consists of multiple classification and regression trees. Each sub-model, acting as a decision tree, is trained sequentially in series. Each decision tree is fitted based on the residuals of the previous model, gradually optimizing the overall prediction performance. The structure of each regression tree is generated by minimizing the objective function and supports automatic feature selection, pruning, and underfitting or overfitting control. The objective function of the machine learning model based on the XGboost algorithm is: ； in, is the overall objective function value of the tth round; n is the total number of training samples; i is the sample index; is the actual value of the upper steamer height of the i-th sample; is the model's predicted value for sample i in round t-1; is the prediction increment of the t-th new tree for sample i; Is the training loss function, used to measure the prediction value of the current model and the true value The difference between them is calculated using the squared error, i.e. ; It is before Tree pairs sample The predicted cumulative value of It is a regularization term used to penalize the complexity of the model, including the number of leaf nodes in the tree and the L2 norm of the leaf node weights; The training method of the machine learning model based on the XGboost algorithm is as follows: Initialization: The machine learning model starts with an initial prediction value Initially, the initial prediction value The value is the average value of the upper steamer height; Iterative training: In each iteration t, the goal of the XGBoost algorithm is to train a new decision tree , so that the decision tree It can fit the residual of the previous round of model prediction. The residual is the gradient information, which is the first-order and second-order gradient of the loss function to the current prediction value. For each sample , calculate the sample The first-order gradient under the current model and the second-order gradient ; Using this gradient information, a new decision tree is constructed. During the construction process, a greedy algorithm is used to find the best split point, that is, the split point where the objective function drops the most; When a tree is built, an optimal weight is calculated for each leaf node ,in Is a leaf node The sum of the first-order gradients of all samples in , Is a leaf node The sum of the second-order gradients of all samples in ; Add this newly trained tree to the current machine learning model; Stop condition: The training process continues until the preset maximum number of iterations is reached or the performance of the validation set no longer improves.

10. A processing device, characterized in that include: at least one memory for storing one or more programs; At least one processor is capable of executing one or more programs stored in the memory, and when the one or more programs are executed by the processor, the processor is capable of implementing the method according to any one of claims 6 to 9.

Images