Method for judging that saccharification heap reaches high-yield wine heap turning standard

Abstract

The invention belongs to the technical field of wine brewing, and relates to a method for judging that a saccharification pile reaches a high-yield wine pile turning standard. According to the distinguishing method, the saccharification heap is distinguished to reach the high-yield wine turning standard according to the stacking fermentation indexes of the saccharification heap, and the stacking fermentation indexes comprise the saccharification heap top temperature, the saccharification heap bottom temperature, the saccharification heap medium temperature and the like. The method comprises the following steps: (1) obtaining modeling characteristic data; (2) sorting model data feature importance; (3) screening and combining modeling data; and (4) constructing a saccharification pile high-yield wine turning standard judgment model and predicting. The method provided by the invention has more comprehensive stacking fermentation indexes for saccharification piles in the high-temperature stacking process, and is also more scientific and reasonable. According to the method, the pile turning condition of the saccharification pile in the pile fermentation process is evaluated through the model, and whether the state of the saccharification pile meets the pile turning standard of high-yield wine or not is judged based on the model, so that the stable and high-yield wine yield is achieved.

Description

Technical Field

[0001] The invention belongs to the technical field of wine making, and relates to a method for judging whether a saccharification pile has reached a high-yield wine turning standard. Background Art

[0002] High-temperature stacking is an important fermentation method in the brewing process of Maotai-flavor liquor. During the stacking process, it can grow and reproduce through microorganisms in the network environment, and promote microbial metabolism of Maotai-flavor liquor flavor precursors, preparing for the subsequent anaerobic fermentation of the cellar. During the stacking fermentation process, the turning process will be used to make the temperature inside the saccharification pile rise evenly, improve the inoculation area of ​​microorganisms and saccharification piles, and then facilitate the fermentation of the saccharification pile and promote the wine production rate. Prior art CN117821188A discloses a method for stacking fermentation of Maotai-flavor liquor saccharification piles. In this technical scheme, the stacking fermentation process is optimized, and the turning node is the first turning when the saccharification pile is stacked for 48-72h and the top temperature of the pile reaches 35°C. The turning node is based only on the stacking time and the top temperature of the saccharification pile. The reference basis is relatively single and has not been linked to the wine production rate. In addition, at present, whether to turn the pile during the fermentation of Maotai-flavor liquor is mainly determined by the senses of the technicians. However, due to differences in skills and experience among technicians, the turning nodes are inconsistent, which affects the subsequent cellar fermentation and may lead to unstable wine production. Therefore, based on a machine learning model, the present invention combines turning the pile with the wine production rate under comprehensive consideration of multiple factors, and judges whether the saccharification pile meets the high-yield wine turning standard during the high-temperature stacking process, so as to promote a better and more stable wine production rate. Summary of the invention

[0003] Based on the fact that there is currently no more comprehensive reference basis for when to turn the saccharification pile during the high-temperature stacking fermentation of Maotai-flavor liquor, and at the same time, in order to solve the problem of different wine yields caused by different saccharification pile turning nodes between work teams due to human factors, the present invention uses the process parameters of the high-temperature stacking saccharification pile and the physical and chemical indicators of the saccharification pile as the basis for judging whether the saccharification pile meets the turning standard, and uses the relevant indicators corresponding to the high-yield saccharification pile to construct a model based on machine learning to judge whether the high-yield wine turning standard is met, so as to optimize the stacking fermentation process and improve the wine yield.

[0004] To achieve the above objectives, in some embodiments, the present invention provides a method for determining whether a saccharification pile has reached the high-yield wine turning standard, wherein the saccharification pile is determined to have reached the high-yield wine turning standard based on the stacking fermentation index of the saccharification pile, and the stacking fermentation index includes the top temperature of the saccharification pile, the bottom temperature of the saccharification pile and the middle temperature of the saccharification pile.

[0005] In some embodiments, the pile fermentation index in the discrimination method further includes at least one of the content of lactic acid in the saccharification pile, the content of reducing sugar in the saccharification pile, the content of water in the saccharification pile, the sampling room temperature, the acidity of the saccharification pile, the content of starch in the saccharification pile, and the content of acetic acid in the saccharification pile.

[0006] In some embodiments, the pile fermentation index further comprises a combination of the saccharification pile top temperature, the saccharification pile bottom temperature, the saccharification pile middle temperature, the saccharification pile lactic acid content, the saccharification pile reducing sugar content, the saccharification pile water content, the sampling room temperature, the saccharification pile acidity and the saccharification pile starch content.

[0007] In some embodiments, the stack fermentation index further comprises a combination of saccharification pile top temperature, saccharification pile bottom temperature, saccharification pile middle temperature, saccharification pile lactic acid content, saccharification pile reducing sugar content, saccharification pile water content and sampling room temperature.

[0008] In some embodiments, the pile fermentation index further comprises a combination of the saccharification pile top temperature, the saccharification pile bottom temperature, the saccharification pile middle temperature, the saccharification pile lactic acid content and the saccharification pile reducing sugar content.

[0009] In some embodiments, the present invention also provides a method for determining whether a saccharification pile has reached the high-yield wine turning standard, characterized in that the method comprises: obtaining a stacking fermentation index of the saccharification pile; inputting the stacking fermentation index into a discrimination model to obtain a discrimination result of reaching the high-yield wine turning standard; the stacking fermentation index comprises the top temperature of the saccharification pile, the bottom temperature of the saccharification pile and the middle temperature of the saccharification pile.

[0010] In some embodiments, the stack fermentation index further includes at least one of the content of lactic acid in the saccharification pile, the content of reducing sugar in the saccharification pile, the content of water in the saccharification pile, the sampling room temperature, the acidity of the saccharification pile, the content of starch in the saccharification pile, and the content of acetic acid in the saccharification pile.

[0011] In some embodiments, the pile fermentation index further comprises a combination of the saccharification pile top temperature, the saccharification pile bottom temperature, the saccharification pile middle temperature, the saccharification pile lactic acid content, the saccharification pile reducing sugar content, the saccharification pile water content, the sampling room temperature, the saccharification pile acidity and the saccharification pile starch content.

[0012] In some embodiments, the stack fermentation index further comprises a combination of saccharification pile top temperature, saccharification pile bottom temperature, saccharification pile middle temperature, saccharification pile lactic acid content, saccharification pile reducing sugar content, saccharification pile water content and sampling room temperature.

[0013] In some embodiments, the pile fermentation index further comprises a combination of the saccharification pile top temperature, the saccharification pile bottom temperature, the saccharification pile middle temperature, the saccharification pile lactic acid content and the saccharification pile reducing sugar content.

[0014] In some embodiments, the discriminant model includes a support vector machine binary classification model, an extreme gradient boosting binary classification model, a random forest binary classification model, a decision tree binary classification model, a K nearest neighbor binary classification model, or a neural network binary classification model.

[0015] In some embodiments, the discriminant model includes a vector machine binary classification model, an extreme gradient boosting binary classification model, or a random forest binary classification model.

[0016] In some embodiments, the discriminant model is a random forest binary classification model.

[0017] In some embodiments, the method for constructing the discriminant model is as follows:

[0018] The pile fermentation index is used as an independent variable as an input parameter, and the dependent variable whether to turn the pile is used as an output parameter. The input parameters and output parameters are modeling data. The modeling data is imported into a random forest binary classification model, and the importance of data features is sorted and screened to obtain a judgment model.

[0019] In some embodiments, the random forest binary classification model is a random forest binary classification model in R language.

[0020] In some embodiments, the modeling data is divided into a training set and a test set; the training set data is input into a random forest binary classification algorithm to establish a training model, and the test set is input into the training model to obtain a prediction result.

[0021] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0022] ntree=100~1000, mtry=2~10, importance=T.

[0023] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0024] ntree=300~1000, mtry=2~8, importance=T.

[0025] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0026] ntree=300~800, mtry=2~6, importance=T.

[0027] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0028] ntree=500, mtry=6, importance=T.

[0029] In some embodiments, the modeling data is divided into a training set and a test set in a ratio of 3:1.

[0030] In some embodiments, the saccharification pile top temperature, saccharification pile middle temperature and saccharification pile bottom temperature in the identification method are measured by a temperature measuring rod.

[0031] In some embodiments, the saccharification pile top temperature, saccharification pile middle temperature and saccharification pile bottom temperature are the temperatures 10 to 20 minutes after the temperature measuring rod is inserted into the saccharification pile.

[0032] In some embodiments, the top temperature of the saccharification pile is measured at a distance of 10 to 30 cm from the top of the saccharification pile.

[0033] In some embodiments, the temperature in the saccharification pile is measured at a distance of 80 to 100 cm from the bottom of the saccharification pile.

[0034] In some embodiments, the bottom temperature of the saccharification pile is measured at a distance of 10 to 30 cm from the bottom of the saccharification pile.

[0035] In some embodiments, the acidity of the saccharification pile, the water content of the saccharification pile, the starch content of the saccharification pile, the reducing sugar content of the saccharification pile, the lactic acid content of the saccharification pile and the acetic acid content of the saccharification pile in the discrimination method are near-infrared spectral information obtained by detecting with a near-infrared analyzer.

[0036] In some embodiments, the near-infrared spectrum is a continuous spectrum with a wavelength range of 800 to 2500 nm.

[0037] In some embodiments, the present invention provides a method for constructing a standard discrimination model for turning a saccharification pile to achieve high-yield wine, which specifically comprises the following steps:

[0038] S1 obtains modeling data: obtains the pile temperature, physical and chemical indicators, sampling room temperature, etc. from the completion of the saccharification pile to the storage, a total of 290 groups, 2900 modeling data, and obtains the modeling database, which specifically includes: the independent variables are the top temperature, middle temperature, bottom temperature, physical and chemical indicators (water content, acidity, reducing sugar, starch, lactic acid, acetic acid content) and sampling room temperature of the saccharification pile, and the dependent variable is whether to turn the pile, as categorical data, where turning the pile is "1" and not turning the pile is "0". Finally, according to the obtained wine yield data, the stacking fermentation index of the saccharification pile corresponding to the high-yield wine is traced back, and it is judged as turning the pile "1" (in order to establish the connection between high-yield wine and the turning process, and form a high-yield wine turning model), and the remaining data (the stacking fermentation index of the saccharification pile at each stage of non-high-yield wine, and the stacking fermentation index of the saccharification pile at the completion and storage stages of high-yield wine) is judged as not turning the pile "0".

[0039] S2: ranking the importance of modeling data features: bringing the modeling data obtained in step S1 into the algorithm to rank the importance of data features;

[0040] S3: screening and combining modeling data: screening according to the importance ranking of data features in step S2, and obtaining the selected preferred indicators as modeling feature data;

[0041] S4 constructs a model for judging the standards for turning over saccharification piles for high-yield wine: Based on the modeling feature data, machine learning is used to construct a model for judging the standards for turning over high-yield wine.

[0042] In some embodiments, the modeling data includes a pile fermentation index as an independent variable as an input parameter, and a dependent variable of whether to turn the pile as an output parameter.

[0043] In some embodiments, the pile fermentation indexes include saccharification pile top temperature, saccharification pile bottom temperature and saccharification pile middle temperature.

[0044] In some embodiments, the stack fermentation index further includes at least one of the content of lactic acid in the saccharification pile, the content of reducing sugar in the saccharification pile, the content of water in the saccharification pile, the sampling room temperature, the acidity of the saccharification pile, the content of starch in the saccharification pile, and the content of acetic acid in the saccharification pile.

[0045] In some embodiments, the pile fermentation index further comprises a combination of the saccharification pile top temperature, the saccharification pile bottom temperature, the saccharification pile middle temperature, the saccharification pile lactic acid content, the saccharification pile reducing sugar content, the saccharification pile water content, the sampling room temperature, the saccharification pile acidity and the saccharification pile starch content.

[0046] In some embodiments, the stack fermentation index further comprises a combination of saccharification pile top temperature, saccharification pile bottom temperature, saccharification pile middle temperature, saccharification pile lactic acid content, saccharification pile reducing sugar content, saccharification pile water content and sampling room temperature.

[0047] In some embodiments, the pile fermentation index further comprises a combination of the saccharification pile top temperature, the saccharification pile bottom temperature, the saccharification pile middle temperature, the saccharification pile lactic acid content and the saccharification pile reducing sugar content.

[0048] In some embodiments, the algorithm in step S2 and the algorithm used in the machine learning in step S4 include a support vector machine binary classification model, an extreme gradient boosting binary classification model, a random forest binary classification model, a decision tree binary classification model, a K nearest neighbor binary classification model or a neural network binary classification model.

[0049] In some embodiments, the algorithm in step S2 and the algorithm used in the machine learning in step S4 include a vector machine binary classification model, an extreme gradient boosting binary classification model or a random forest binary classification model.

[0050] In some embodiments, the algorithm in step S2 and the algorithm used in the machine learning in step S4 are random forest binary classification models.

[0051] In some embodiments, the random forest binary classification model is a random forest binary classification model in R language.

[0052] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0053] ntree=100~1000, mtry=2~10, importance=T.

[0054] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0055] ntree=300~1000, mtry=2~8, importance=T.

[0056] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0057] ntree=300~800, mtry=2~6, importance=T.

[0058] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0059] ntree=500, mtry=6, importance=T.

[0060] In some embodiments, in step S1 of the construction method, the physical and chemical index data of the acidity, moisture, starch, reducing sugar, lactic acid and acetic acid content of the saccharification pile are obtained by quick detection using a FOSS near-infrared multifunctional quality analyzer (the detection method is to fill the sample to be tested into the sample cup of the device, ensure that the bottom of the cup is opaque, place it on the measuring pool after filling, cover it with a lid, select the measurement template and click OK to perform the detection).

[0061] In some embodiments, in step S2 and step S4 of the method, the algorithm parameters for ranking the importance of random forest data features are set to:

[0062] ntree=100~1000, mtry=2~10, importance=T.

[0063] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0064] ntree=300~1000, mtry=2~8, importance=T.

[0065] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0066] ntree=300~800, mtry=2~6, importance=T.

[0067] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0068] ntree=500, mtry=6, importance=T.

[0069] In some embodiments, in step S2 of the method, the random forest algorithm ranks the modeling features in descending order of importance as follows: top temperature, bottom temperature, middle temperature, lactic acid, reducing sugar, moisture, sampling room temperature, acidity, starch, and acetic acid of the saccharification pile.

[0070] In some embodiments, the construction of the high-yield wine compost turning standard discrimination model comprises the following steps:

[0071] Step 1: Screen the combination according to the modeling data and build a modeling feature database;

[0072] Step 2: Divide the modeling data into a training set (75%) and a test set (25%);

[0073] Step 3: Input the training set data into the random forest binary classification algorithm to establish a training model;

[0074] Step 4: Input the test set into the training model to obtain the prediction results, and complete the judgment on whether the saccharification pile meets the high-yield wine turning standard.

[0075] In some embodiments, the present invention provides a more comprehensive basis for turning the saccharification pile during high-temperature stacking, including the top temperature, middle temperature, bottom temperature of the pile during turning, as well as the acidity, reducing sugar content, starch content, etc. of the pile. The prior art uses a relatively simple basis for judging whether to turn the pile, including the top temperature of the saccharification pile and the stacking time. In comparison, the present invention provides a more comprehensive index and is more scientific and reasonable.

[0076] In some embodiments, the present invention establishes a connection between the wine yield and the compost turning process. Based on the present invention, the compost fermentation conditions can be reasonably adjusted to improve the wine yield. The prior art does not combine the compost turning process with the wine yield, and has limited guidance for liquor brewing production.

[0077] In some embodiments, the present invention further optimizes the pile fermentation process for brewing Maotai-flavor liquor, and solves the problem of differences in pile turning standards due to human factors among different teams.

[0078] In some embodiments, the present invention uses a model to judge the turning standard of the saccharification pile during the stack fermentation process, and based on this model, judges whether the state of the saccharification pile meets the turning standard for high-yield wine, so as to achieve a stable and high-yield wine production rate.

[0079] In some embodiments, the present invention provides a device for determining whether a saccharification pile has reached a high-yield wine turning standard, the device comprising:

[0080] A data management module is used to obtain the pile fermentation index of the saccharification pile;

[0081] A data sorting and screening module is used to sort and screen the pile fermentation indicators according to their importance, and obtain the selected preferred indicators as modeling feature numbers;

[0082] A model building module, used for discriminating a model based on the modeling feature data;

[0083] The data discrimination module is used to input the pile fermentation index into the discrimination model and output the discrimination result of whether the saccharification pile reaches the high-yield wine turning standard.

[0084] In some embodiments, the stack fermentation indexes include the saccharification pile top temperature, the saccharification pile bottom temperature and the saccharification pile middle temperature.

[0085] In some embodiments, the stack fermentation index further includes at least one of the content of lactic acid in the saccharification pile, the content of reducing sugar in the saccharification pile, the content of water in the saccharification pile, the sampling room temperature, the acidity of the saccharification pile, the content of starch in the saccharification pile, or the content of acetic acid in the saccharification pile.

[0086] In some embodiments, the pile fermentation index further comprises a combination of the saccharification pile top temperature, the saccharification pile bottom temperature, the saccharification pile middle temperature, the saccharification pile lactic acid content, the saccharification pile reducing sugar content, the saccharification pile water content, the sampling room temperature, the saccharification pile acidity and the saccharification pile starch content.

[0087] In some embodiments, the stack fermentation index further comprises a combination of saccharification pile top temperature, saccharification pile bottom temperature, saccharification pile middle temperature, saccharification pile lactic acid content, saccharification pile reducing sugar content, saccharification pile water content and sampling room temperature.

[0088] In some embodiments, the pile fermentation index further comprises a combination of the saccharification pile top temperature, the saccharification pile bottom temperature, the saccharification pile middle temperature, the saccharification pile lactic acid content and the saccharification pile reducing sugar content.

[0089] In some embodiments, the discriminant model includes a support vector machine binary classification model, an extreme gradient boosting binary classification model, a random forest binary classification model, a decision tree binary classification model, a K nearest neighbor binary classification model, or a neural network binary classification model.

[0090] In some embodiments, the discriminant model includes a vector machine binary classification model, an extreme gradient boosting binary classification model, or a random forest binary classification model.

[0091] In some embodiments, the discriminant model is a random forest binary classification model.

[0092] In some embodiments, the random forest binary classification model is a random forest binary classification model in R language.

[0093] In some embodiments, the modeling data is divided into a training set and a test set; the training set data is input into a random forest binary classification algorithm to establish a training model, and the test set is input into the training model to obtain a prediction result.

[0094] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0095] ntree=100~1000, mtry=2~10, importance=T.

[0096] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0097] ntree=300~1000, mtry=2~8, importance=T.

[0098] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0099] ntree=300~800, mtry=2~6, importance=T.

[0100] In some embodiments, the parameters of the random forest binary classification model are set as follows:

[0101] ntree=500, mtry=6, importance=T.

[0102] In some embodiments, the modeling data is divided into a training set and a test set in a ratio of 3:1.

[0103] In some embodiments, the pile fermentation index in the identification device is detected and collected by a near-infrared analyzer, a temperature measuring rod and a thermometer.

[0104] In some embodiments, the method or device for judging whether a saccharification pile has reached a high wine-yielding pile turning standard provided by the present invention is used to improve the wine yield. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] Figure 1 Schematic diagram of the saccharification pile.

[0106] Figure 2This is a ranking diagram of the importance of random forest data features. DW refers to the top temperature of the saccharification pile, dw refers to the bottom temperature of the saccharification pile, ZW refers to the middle temperature of the saccharification pile, RS refers to the lactic acid content of the saccharification pile, HYT refers to the reducing sugar content of the saccharification pile, SF refers to the water content of the saccharification pile, SW refers to the sampling room temperature, SD refers to the acidity of the saccharification pile, DF refers to the starch content of the saccharification pile, and YS refers to the acetic acid content of the saccharification pile. DETAILED DESCRIPTION

[0107] The technical solution of the present invention is further described below by specific embodiments, which do not limit the protection scope of the present invention. Some non-essential modifications and adjustments made by others based on the concept of the present invention are still protected by the present invention.

[0108] The "saccharification pile" mentioned in the embodiment of the present invention refers to the pile of grain (sorghum) after steaming and gelatinization mixed with Daqu. The saccharification pile fermentation includes three stages: finishing the pile, turning the pile, and putting it into the cellar. When piling, it is piled from all sides, with one layer covering another, so that it is loose and breathable. When the pile is finished, the height of the saccharification pile is 1.8-2.0m, the bottom diameter is 6.0-6.5m, and the weight is 13-14 tons.

[0109] The "turning pile process" specifically refers to: during the saccharification pile stacking process, a common method to help the fermentation of the mash is to transfer the saccharification pile at the top of the original saccharification pile to the target area as the new saccharification pile center; peel off the remaining saccharification piles around the initial saccharification pile and surround them at the foot of the pile; transfer the saccharification piles at the foot of the pile from the outside to the inside in batches to the center of the saccharification pile covering the target area, and after multiple cycles, a new saccharification pile is formed, and the turning pile is completed. The height of the saccharification pile formed after turning the pile is greater than 2m, and the bottom diameter is less than 6m.

[0110] "Down to the cellar" specifically means: when the saccharification pile is heated up to the temperature required for cellaring during the stacking fermentation process, the saccharification pile is transferred to the fermentation cellar for further fermentation.

[0111] "High-yield wine" specifically refers to: the base wine yield rate in the round that is higher than the average base wine yield rate is a high-yield wine, where "base wine yield rate" is the proportion of the base wine output of the corresponding cellar of each saccharification pile in the round to the total amount of grain input; "average base wine yield rate" is the average wine yield rate of all 120 cellars in a round.

[0112] "Saccharification pile top temperature" (DW) specifically refers to: the temperature at the top of the saccharification pile during pile fermentation (℃);

[0113] "Saccharification pile middle temperature" (ZW) specifically refers to: the center temperature of the saccharification pile during pile fermentation (℃);

[0114] "Saccharification pile bottom temperature" (dw) specifically refers to: the temperature at the bottom of the saccharification pile during pile fermentation (℃);

[0115] "Saccharification pile acidity" (SD) specifically refers to: the amount of acid contained in the saccharification pile during pile fermentation (mmol / 10g);

[0116] "Saccharification pile moisture" (SF) specifically refers to: the moisture content (%) in the saccharification pile (mixture of sorghum and Daqu) during pile fermentation;

[0117] "Saccharification pile starch" (DF) specifically refers to: the starch content (%) of the saccharification pile during pile fermentation;

[0118] "Hydrogenation pile reducing sugar" (HYT) specifically refers to: the reducing sugar content (%) of the saccharification pile during pile fermentation;

[0119] "Saccharification pile lactic acid" (RS) specifically refers to: the lactic acid content (%) of the saccharification pile during pile fermentation;

[0120] "Acetic acid in saccharification pile" (YS) specifically refers to: the acetic acid content (%) of the saccharification pile during pile fermentation;

[0121] "Sampling room temperature" (SW) specifically refers to: the temperature (°C) inside the saccharification pile stacking plant during stack fermentation.

[0122] Note: The top temperature, middle temperature and bottom temperature of the saccharification pile are measured by inserting temperature measuring rods into the saccharification pile, and the temperature data are recorded after waiting for 10 to 20 minutes. The top temperature is measured at 10 to 30 cm from the top of the saccharification pile, the bottom temperature is measured at 10 to 30 cm from the bottom of the saccharification pile, and the middle temperature of the saccharification pile is measured in the middle of the saccharification pile (80 to 100 cm from the bottom of the saccharification pile); the sampling room temperature data comes from the real-time reading of the thermometer in the middle of the plant; the other physical and chemical indicators come from the near-infrared analyzer detection data of the saccharification pile in the middle layer of the saccharification pile; the moisture, starch, reducing sugar, lactic acid and acetic acid content (%) of the saccharification pile refers to the mass percentage (W / W).

[0123] Example 1: Establishment of a model for judging the standard of turning over saccharification piles for high-yield wine

[0124] 1. Obtain modeling data

[0125] Obtain relevant indicators at each stage during the saccharification pile fermentation, the room temperature during sampling during the saccharification pile, the top temperature, middle temperature, bottom temperature of the saccharification pile, and the physical and chemical indicators of the saccharification pile. The details are as follows:

[0126] The temperature of the saccharification pile and the daily sampling room temperature data during the stacking fermentation process rely on manual recording. The physical and chemical indicators of the saccharification pile, including the acidity, moisture, starch, reducing sugar, lactic acid and acetic acid of the saccharification pile, are obtained by detection using a FOSS near-infrared multifunctional quality analyzer (instrument model NIR DS2500 F, detection range 800-2500nm, data sampling interval 10S, and about 60S to complete the detection of one sample). The stacking fermentation indicators are specifically: saccharification pile top temperature (DW); saccharification pile middle temperature (ZW); saccharification pile bottom temperature (dw); sampling room temperature (SW); saccharification pile moisture content (SF); saccharification pile acidity (SD); saccharification pile starch content (DF); saccharification pile reducing sugar content (HYT); saccharification pile lactic acid content (RS); saccharification pile acetic acid content (YS). Finally, according to the obtained wine yield data, the pile fermentation index of high-yield wine corresponding to the saccharification pile when turning the pile was traced, and it was judged as turning the pile "1", and the rest of the data (saccharification pile fermentation index of each stage of non-high-yield wine, saccharification pile fermentation index of high-yield wine in the stage of finishing and putting into the cellar) was judged as no turning the pile "0". There were 290 groups of relevant data and 2,900 modeling data to obtain the modeling database.

[0127] 2. Construction of a standard model for high-yield wine compost turning

[0128] The independent variables of saccharification pile top temperature (DW), middle temperature (ZW), bottom temperature (dw), sampling room temperature (SW), moisture content (SF), acidity (SD), starch content (DF), reducing sugar content (HYT), lactic acid content (RS) and acetic acid content (YS) were taken as input parameters, and the dependent variable whether the pile was turned was the output parameter (the output was "1" for turning the pile and "0" for not turning the pile). The input parameters and output parameters were respectively imported into the random forest binary classification model, XGBoost binary classification model and SVM binary classification model in R language to construct a judgment model for whether the corresponding saccharification pile met the high-yield wine turning standard.

[0129] Construction of random forest binary classification model:

[0130] 1. 75% of the data in the modeling database is divided into a training set, and 25% of the data is divided into a test set for modeling.

[0131] 2. Input the training set data into the random forest binary classification algorithm to establish a training model.

[0132] 3. Input the test set into the training model to obtain the prediction results, and complete the judgment on whether the saccharification pile meets the high-yield wine turning standard.

[0133] The optimal parameter settings for the random forest binary classification model are as follows: ntree=500, mtry=6, importance=T.

[0134] Construction of the Extreme Gradient Boosting (XGBoost) binary classification model:

[0135] 1. 75% of the data in the modeling database is divided into a training set, and 25% of the data is divided into a test set for modeling.

[0136] 2. Input the training set data into the extreme gradient boosting (XGBoost) binary classification algorithm to establish a training model.

[0137] 3. Input the test set into the training model to obtain the prediction results, and complete the judgment on whether the saccharification pile meets the high-yield wine turning standard.

[0138] The optimal parameter settings for the extreme gradient boosting (XGBoost) binary classification model are as follows: eta = 0.3, gamma = 0.001, max_depth = 2, subsample = 0.70, colsample_bytree = 0.4, objective = "binary:logistic", nround = 1000, watchlist = watchlist, verbose = 1, print_every_n = 100, early_stopping_rounds = 200.

[0139] Construction of Support Vector Machine (SVM) binary classification model:

[0140] 1. 75% of the data in the modeling database is divided into a training set, and 25% of the data is divided into a test set for modeling.

[0141] 2. Input the training set data into the support vector machine (SVM) binary classification algorithm to establish a training model.

[0142] 3. Input the test set into the training model to obtain the prediction results, and complete the judgment on whether the saccharification pile meets the high-yield wine turning standard.

[0143] The optimal parameter settings for the support vector machine (SVM) binary classification model are as follows: kernel = "radial", cost = 1, gamma = 0.1, probability = T.

[0144] 3. Model Accuracy Evaluation

[0145] Accuracy, F1 value and AUC value were used as evaluation indicators of the binary classification model to evaluate the accuracy of the high-yield wine compost turning model.

[0146] Table 1 Evaluation results of different models

[0147]

[0148] Accuracy refers to the ratio of the number of samples correctly predicted by the model to the total number of samples. The calculation formula for accuracy is as follows:

[0149] Accuracy=(TP+TN) / (TP+TN+FP+FN)

[0150] Among them, TP (True Positive) indicates the number of samples predicted by the model as "1" and actually "1"; TN (True Negative) indicates the number of samples predicted by the model as "0" and actually "0"; FP (False Positive) indicates the number of samples predicted by the model as "1" but actually "0"; FN (False Negative) indicates the number of samples predicted by the model as "0" but actually "1". ("1" means turning the compost, "0" means not turning the compost)

[0151] The F1 value represents the harmonic mean of precision and recall, which takes into account classification accuracy and classification completeness. The higher the F1 value, the better the model performance. The calculation formula is as follows:

[0152] F1 = 2 × (precision × recall) / (precision + recall)

[0153] The precision rate indicates the proportion of samples predicted as "1" that are actually "1", and the recall rate indicates the proportion of samples actually "1" that are predicted as "1".

[0154] The AUC value represents the area under the ROC curve and the coordinate axis, and the value range of AUC is between 0.5 and 1. The closer the AUC is to 1.0, the better the performance of the model.

[0155] The three evaluation indicators of accuracy, F1 value, and AUC value are directly output by the binary classification model, and the results are shown in Table 1. According to the results in Table 1, the accuracy, F1 value, and AUC value of the training sets of the three models are all above 0.9, while in the test set, the accuracy and F1 value of the XGBoost binary classification model are relatively low, indicating that the accuracy and classification integrity of the model are relatively poor, and the three evaluation parameters of the test sets of the random forest binary classification model and the SVM binary classification model are all maintained above 0.9. Combined with the evaluation index of the training set, in the judgment of whether the saccharification pile of the present invention reaches the high-yield wine turning standard, the random forest binary classification model and the SVM binary classification model have better fitting effects than the XGBoost binary classification model, and can better realize the prediction of the turning node of the high-yield wine of the sauce-flavor liquor.

[0156] Example 2: Ranking of modeling data feature importance

[0157] 1. Obtain modeling data

[0158] Obtain relevant indicators during the fermentation of the saccharification pile: room temperature when sampling the saccharification pile, top temperature, middle temperature, bottom temperature of the pile, physical and chemical indicators of the saccharification pile, etc. The details are as follows:

[0159] The temperature of the saccharification pile and the room temperature data information during sampling are recorded manually. The physical and chemical indicators of the saccharification pile include acidity, moisture, starch, reducing sugar, lactic acid, and acetic acid. The contents are obtained by using the FOSS near-infrared multifunctional quality analyzer. The specific indicators of pile fermentation are: top temperature of the saccharification pile (DW); middle temperature of the saccharification pile (ZW); bottom temperature of the saccharification pile (dw); sampling room temperature (SW); moisture of the saccharification pile (SF); acidity of the saccharification pile (SD); starch content of the saccharification pile (DF); reducing sugar content of the saccharification pile (HYT); lactic acid content of the saccharification pile (RS); acetic acid content of the saccharification pile (YS). The above modeling data also traces the pile fermentation indicators of the high-yield wine when the saccharification pile is turned according to the obtained wine yield data, and judges it as "1" for turning the pile. The rest of the data (the pile fermentation indicators of the saccharification pile at each stage of non-high-yield wine, and the pile fermentation indicators of the saccharification pile at the completion and storage stages of high-yield wine) are judged as "0" without turning the pile.

[0160] 2. Ranking of modeling data feature importance

[0161] The independent variables of saccharification pile top temperature (DW), middle temperature (ZW), bottom temperature (dw), sampling room temperature (SW), moisture content (SF), acidity (SD), starch content (DF), reducing sugar content (HYT), lactic acid content (RS), and acetic acid content (YS) were used as input parameters, and the dependent variable whether to turn the pile was used as the output parameter (the output of turning the pile was "1", and the output of not turning the pile was "0"). The input parameters and output parameters were imported into the independent variable importance ranking algorithm of random forest in R language to generate the modeling data feature importance ranking results. The algorithm parameters were: ntree=500, mtry=6, importance=T;

[0162] The ranking results of the importance of the modeling data features of random forest are as follows Figure 2 , from large to small, specifically: saccharification pile top temperature (DW) > saccharification pile bottom temperature (dw) > saccharification pile middle temperature (ZW) > saccharification pile lactic acid content (RS) > saccharification pile reducing sugar content (HYT) > saccharification pile water content (SF) > sampling room temperature (SW) > saccharification pile acidity content (SD) > saccharification pile starch content (DF) > saccharification pile acetic acid content (YS).

[0163] Example 3: Determination of the optimal model for judging the turning standard of saccharification piles for high-yield wine

[0164] 1. Modeling data screening

[0165] Combined with the importance ranking of the screening data features in Example 2, and taking into account the difficulty of obtaining subsequent modeling data and the model effect, the stacking fermentation indicators of the top 9, 7, 5, and 3 of the data feature rankings are selected as modeling data. Among them:

[0166] Top 9: saccharification pile top temperature (DW) > saccharification pile bottom temperature (dw) > saccharification pile middle temperature (ZW) > saccharification pile lactic acid content (RS) > saccharification pile reducing sugar content (HYT) > saccharification pile water content (SF) > sampling room temperature (SW) > saccharification pile acidity (SD) > saccharification pile starch content (DF), a total of 290 groups, 2610 modeling data.

[0167] Top 7: saccharification pile top temperature (DW) > saccharification pile bottom temperature (dw) > saccharification pile middle temperature (ZW) > saccharification pile lactic acid content (RS) > saccharification pile reducing sugar content (HYT) > saccharification pile water content (SF) > sampling room temperature (SW), a total of 290 groups, 2030 modeling data.

[0168] Top 5: saccharification pile top temperature (DW) > saccharification pile bottom temperature (dw) > saccharification pile middle temperature (ZW) > saccharification pile lactic acid content (RS) > saccharification pile reducing sugar content (HYT), a total of 290 groups, 1450 modeling data.

[0169] The first 3: saccharification pile top temperature (DW) > saccharification pile bottom temperature (dw) > saccharification pile middle temperature (ZW), a total of 290 groups, 870 modeling data.

[0170] The above modeling data is also based on the wine yield data obtained, tracing the stacking fermentation indicators of the corresponding saccharification pile when the high-yield wine is turned over, and judging it as "1" for turning over, and the rest of the data (the stacking fermentation indicators of the saccharification pile at each stage of non-high-yield wine, and the stacking fermentation indicators of the saccharification pile at the completion and cellaring stages of high-yield wine) are judged as "0" for not turning over the pile.

[0171] 2. Screening of the optimal model for judging the high-yield wine compost turning standard

[0172] The pile fermentation indicators ranked first 9, 7, 5, and 3 in the selected data features were imported into the random forest binary classification model, XGBoost binary classification model, and SVM binary classification model in R language respectively, and the optimal model was further screened to construct the optimal judgment model for the high-yield wine pile turning standard. 75% of the data in the modeling database was divided into a training set, and 25% of the data was divided into a test set for modeling.

[0173] The specific method for constructing the model is shown in Example 1.

[0174] The optimal parameter settings for the random forest binary classification model are as follows: ntree=500, mtry=6, importance=T.

[0175] The optimal parameter settings for the extreme gradient boosting (XGBoost) binary classification model are as follows: eta = 0.3, gamma = 0.001, max_depth = 2, subsample = 0.70, colsample_bytree = 0.4, objective = "binary:logistic", nround = 1000, watchlist = watchlist, verbose = 1, print_every_n = 100, early_stopping_rounds = 200.

[0176] The optimal parameter settings for the support vector machine (SVM) binary classification model are as follows: kernel = "radial", cost = 1, gamma = 0.1, probability = T.

[0177] 3. Model Accuracy Evaluation

[0178] Similarly, accuracy, F1 value, and AUC value are used as evaluation indicators of the binary classification model to evaluate the accuracy of each established model.

[0179] Table 2 Evaluation results of different models

[0180]

[0181]

[0182] The three evaluation indicators, accuracy, F1 value, and AUC value, are directly output by the binary classification model, and the results are shown in Table 2. The results in Table 2 show that after the pile fermentation indicators ranked before 9, 7, 5, and 3 in the data feature ranking were imported into the random forest binary classification model, the XGBoost binary classification model, and the SVM binary classification model, the accuracy, F1 value, and AUC value of the training set and test set of each model were compared. It was found that when the model parameters were set to the indicators ranked before 7 in importance, the accuracy, F1 value, and AUC value of the training set and test set of the random forest binary classification model were the best in all test groups, with the training set being 1.0000 and the test set being greater than 0.96, indicating that the model has high accuracy and completeness. Therefore, under the premise of comprehensively considering the model effect and the convenience of obtaining modeling data, the random forest binary classification model was selected as the model for judging whether the saccharification pile meets the high-yield wine turning standard during the high-temperature stacking process.

[0183] Example 4: Verification of the model for judging the standard of turning the saccharification pile for high-yield wine

[0184] In combination with the optimal results in Example 3, the saccharification pile top temperature (DW), saccharification pile bottom temperature (dw), saccharification pile middle temperature (ZW), saccharification pile lactic acid content (RS), saccharification pile reducing sugar content (HYT), saccharification pile water content (SF), and sampling room temperature (SW) of 10 saccharification piles in the stacking fermentation stage were collected and imported into a random forest binary classification model. The model was used to determine whether each saccharification pile met the high-yield wine turning standard during the stacking process. If the model result output is "1", that is, it meets the high-yield wine turning standard, the turning step can be performed. After further stacking and fermentation, the cellar fermentation link is entered to obtain the base wine yield corresponding to each saccharification pile, and the accuracy of the saccharification pile high-yield wine turning standard judgment model is verified; if the model result output is "0", it does not meet the high-yield wine turning standard. After a period of time, the stacking fermentation indicators are collected again, and the random forest binary classification model is imported for judgment. Until the turning standard is met, the turning and cellar fermentation are performed.

[0185] Table 3 Base wine yield rate corresponding to the predicted compost turning and saccharification compost under the optimal model

[0186]

[0187] Note: The average base wine yield is 12.03%.

[0188] The results in Table 3 show that the random forest binary classification model can judge whether the saccharification pile has reached the high-yield wine turning standard during the fermentation process, as well as the actual wine yield corresponding to each saccharification pile. The wine yields corresponding to the 10 saccharification piles are relatively high, and the model is relatively stable.

[0189] Comparative Example 1: The actual wine yield of the compost turning predicted by traditional experience and its corresponding saccharification pile

[0190] In order to further demonstrate the accuracy and stability of the preferred high-yield wine compost turning standard judgment model of the present invention, the saccharification pile top temperature (DW), saccharification pile bottom temperature (dw), saccharification pile middle temperature (ZW), saccharification pile lactic acid (RS), saccharification pile reducing sugar (HYT), saccharification pile moisture (SF) and sampling room temperature (SW) of 10 saccharification piles that were predicted to be able to be turned according to traditional experience during the stacking and fermentation stage were also collected. At the same time, the actual wine yield corresponding to each saccharification pile was obtained to further verify the accuracy and stability of the saccharification pile high-yield wine compost turning standard judgment model.

[0191] Table 4 Traditional experience predicts the base wine yield rate corresponding to the compost turning and its saccharification pile

[0192]

[0193]

[0194] Note: The average base wine yield is 12.03%.

[0195] The results in Table 4 show that after turning the compost according to traditional experience, the wine yields of 6 out of 10 saccharification piles were higher than the average wine yield, and the wine yields of 4 saccharification piles were lower than the average wine yield, and their stability was relatively poor compared with the method of the present invention, indicating that the standard judgment model for turning the compost for high wine yield of the present invention was successfully constructed, with good effects and great application value.

[0196] The above-mentioned implementation modes are only preferred implementation modes of the present application and cannot be used to limit the scope of protection of the present application. Any non-substantial changes and substitutions made by technicians in this field on the basis of the present application shall fall within the scope of protection required by the present application.

Claims

1. A method for determining whether a saccharification pile has reached the high-yield wine turning standard, characterized in that: The saccharification pile is judged to have reached the high-yield wine turning standard according to the pile fermentation index of the saccharification pile, and the pile fermentation index includes the top temperature of the saccharification pile, the bottom temperature of the saccharification pile and the middle temperature of the saccharification pile.

2. The identification method according to claim 1, characterized in that: The pile fermentation index further includes at least one of the content of lactic acid in the saccharification pile, the content of reducing sugar in the saccharification pile, the content of water in the saccharification pile, the sampling room temperature, the acidity of the saccharification pile, the content of starch in the saccharification pile, and the content of acetic acid in the saccharification pile; Preferably, the pile fermentation index further comprises a combination of saccharification pile top temperature, saccharification pile bottom temperature, saccharification pile middle temperature, saccharification pile lactic acid content, saccharification pile reducing sugar content, saccharification pile water content, sampling room temperature, saccharification pile acidity and saccharification pile starch content; Preferably, the pile fermentation index further comprises a combination of saccharification pile top temperature, saccharification pile bottom temperature, saccharification pile middle temperature, saccharification pile lactic acid content, saccharification pile reducing sugar content, saccharification pile water content and sampling room temperature; Preferably, the pile fermentation index further comprises a combination of the saccharification pile top temperature, the saccharification pile bottom temperature, the saccharification pile middle temperature, the saccharification pile lactic acid content and the saccharification pile reducing sugar content.

3. A method for determining whether a saccharification pile has reached the high-yield wine turning standard, characterized in that: The method comprises: Obtaining a pile fermentation index of the saccharification pile; inputting the pile fermentation index into a discrimination model to obtain a discrimination result of reaching the high-yield wine turning standard; the pile fermentation index includes the top temperature of the saccharification pile, the bottom temperature of the saccharification pile and the middle temperature of the saccharification pile; Preferably, the stack fermentation index further comprises at least one of the content of lactic acid in the saccharification pile, the content of reducing sugar in the saccharification pile, the content of water in the saccharification pile, the sampling room temperature, the acidity of the saccharification pile, the content of starch in the saccharification pile, and the content of acetic acid in the saccharification pile; Preferably, the pile fermentation index further comprises a combination of saccharification pile top temperature, saccharification pile bottom temperature, saccharification pile middle temperature, saccharification pile lactic acid content, saccharification pile reducing sugar content, saccharification pile water content, sampling room temperature, saccharification pile acidity and saccharification pile starch content; Preferably, the pile fermentation index further comprises a combination of saccharification pile top temperature, saccharification pile bottom temperature, saccharification pile middle temperature, saccharification pile lactic acid content, saccharification pile reducing sugar content, saccharification pile water content and sampling room temperature; Preferably, the pile fermentation index further comprises a combination of the saccharification pile top temperature, the saccharification pile bottom temperature, the saccharification pile middle temperature, the saccharification pile lactic acid content and the saccharification pile reducing sugar content; Preferably, the discriminant model includes a support vector machine binary classification model, an extreme gradient boosting binary classification model, a random forest binary classification model, a decision tree binary classification model, a K nearest neighbor binary classification model or a neural network binary classification model; Preferably, the discriminant model includes a vector machine binary classification model, an extreme gradient enhancement binary classification model or a random forest binary classification model; Preferably, the discriminant model is a random forest binary classification model; Preferably, the method for constructing the discriminant model is as follows: The pile fermentation index is used as an independent variable as an input parameter, and the dependent variable whether the pile is turned is used as an output parameter. The input parameters and the output parameters are modeling data. The modeling data is imported into a random forest binary classification model, and the importance of data features is sorted and screened to obtain a judgment model; Preferably, the random forest binary classification model is a random forest binary classification model in R language; Preferably, the modeling data is divided into a training set and a test set; the training set data is input into a random forest binary classification algorithm to establish a training model, and the test set is input into the training model to obtain a prediction result; Preferably, the parameters of the random forest binary classification model are set as follows: ntree=100~1000, mtry=2~10, importance=T; Preferably, the parameters of the random forest binary classification model are set as follows: ntree=300~1000, mtry=2~8, importance=T; Preferably, the parameters of the random forest binary classification model are set as follows: ntree=300~800, mtry=2~6, importance=T; More preferably, the parameters of the random forest binary classification model are set as follows: ntree=500, mtry=6, importance=T; Preferably, the modeling data is divided into a training set and a test set in a ratio of 3:

1.

4. The identification method according to any one of claims 1 to 3, characterized in that: The saccharification pile top temperature, saccharification pile middle temperature and saccharification pile bottom temperature are measured by a temperature measuring rod; Preferably, the saccharification pile top temperature, saccharification pile middle temperature and saccharification pile bottom temperature are the temperatures 10 to 20 minutes after the temperature measuring rod is inserted into the saccharification pile; Preferably, the saccharification pile top temperature is measured at a distance of 10 to 30 cm from the top of the saccharification pile; Preferably, the temperature in the saccharification pile is measured at a distance of 80 to 100 cm from the bottom of the saccharification pile; Preferably, the bottom temperature of the saccharification pile is measured at a distance of 10 to 30 cm from the bottom of the saccharification pile.

5. The identification method according to any one of claims 2 to 3, characterized in that: The acidity of the saccharification pile, the content of water in the saccharification pile, the content of starch in the saccharification pile, the content of reducing sugar in the saccharification pile, the content of lactic acid in the saccharification pile and the content of acetic acid in the saccharification pile are near-infrared spectral information obtained by detecting with a near-infrared analyzer; Preferably, the near-infrared spectrum is a continuous spectrum with a wavelength range of 800-2500nm.

6. A method for constructing a model for determining the standard for turning a saccharification pile to achieve high-yield wine, characterized in that: The following steps are involved: Obtain modeling data; Importance ranking of modeling data features: Bring the obtained modeling data into the algorithm to sort the importance of data features; Modeling data screening combination: screening is performed according to the importance ranking of the data features, and the selected preferred indicators are obtained as modeling feature data; Constructing a model for distinguishing high-yield wine compost turning standards in saccharification piles: Based on the modeling feature data, a model for distinguishing high-yield wine compost turning standards is constructed using machine learning; Preferably, the modeling data includes a pile fermentation index as an independent variable as an input parameter, and a dependent variable of whether to turn the pile as an output parameter; Preferably, the pile fermentation indexes include saccharification pile top temperature, saccharification pile bottom temperature and saccharification pile middle temperature; Preferably, the stack fermentation index further comprises at least one of the content of lactic acid in the saccharification pile, the content of reducing sugar in the saccharification pile, the content of water in the saccharification pile, the sampling room temperature, the acidity of the saccharification pile, the content of starch in the saccharification pile, and the content of acetic acid in the saccharification pile; Preferably, the pile fermentation index further comprises a combination of saccharification pile top temperature, saccharification pile bottom temperature, saccharification pile middle temperature, saccharification pile lactic acid content, saccharification pile reducing sugar content, saccharification pile water content, sampling room temperature, saccharification pile acidity and saccharification pile starch content; Preferably, the pile fermentation index further comprises a combination of saccharification pile top temperature, saccharification pile bottom temperature, saccharification pile middle temperature, saccharification pile lactic acid content, saccharification pile reducing sugar content, saccharification pile water content and sampling room temperature; Preferably, the pile fermentation index further comprises a combination of the saccharification pile top temperature, the saccharification pile bottom temperature, the saccharification pile middle temperature, the saccharification pile lactic acid content and the saccharification pile reducing sugar content; Preferably, the algorithm and the algorithm used in machine learning include a support vector machine binary classification model, an extreme gradient boosting binary classification model, a random forest binary classification model, a decision tree binary classification model, a K nearest neighbor binary classification model or a neural network binary classification model; Preferably, the algorithm and the algorithm used in machine learning include a vector machine binary classification model, an extreme gradient enhancement binary classification model or a random forest binary classification model; Preferably, the algorithm and the algorithm used in machine learning are random forest binary classification models; Preferably, the random forest binary classification model is a random forest binary classification model in R language; Preferably, the parameters of the random forest binary classification model are set as follows: ntree=100~1000, mtry=2~10, importance=T; Preferably, the parameters of the random forest binary classification model are set as follows: ntree=300~1000, mtry=2~8, importance=T; Preferably, the parameters of the random forest binary classification model are set as follows: ntree=300~800, mtry=2~6, importance=T; More preferably, the parameters of the random forest binary classification model are set as follows: ntree=500, mtry=6, importance=T.

7. The construction method according to claim 6, characterized in that: The construction of the high-yield wine compost turning standard discrimination model comprises the following steps: (1) constructing a modeling feature data set according to the modeling data screening combination of claim 6; (2) The modeling data is divided into a training set and a test set in a ratio of 3:1; (3) Input the training set data into the random forest binary classification algorithm to establish a training model; (4) The test set is input into the training model to obtain the prediction results, thereby completing the judgment on whether the saccharification pile has reached the high-yield wine turning standard.

8. A device for determining whether a saccharification pile has reached the high-yield wine turning standard, characterized in that: The device comprises: A data management module is used to obtain the pile fermentation index of the saccharification pile; A data sorting and screening module is used to sort and screen the pile fermentation indicators according to their importance, and obtain the selected preferred indicators as modeling feature data; A model building module, used to build a discriminant model based on the modeling feature data; A data discrimination module, used for inputting the pile fermentation index into the discrimination model, and outputting a discrimination result of whether the saccharification pile reaches the high-yield wine turning standard; Preferably, the pile fermentation indexes include saccharification pile top temperature, saccharification pile bottom temperature and saccharification pile middle temperature; Preferably, the stack fermentation index further comprises at least one of the content of lactic acid in the saccharification pile, the content of reducing sugar in the saccharification pile, the content of water in the saccharification pile, the sampling room temperature, the acidity of the saccharification pile, the content of starch in the saccharification pile, or the content of acetic acid in the saccharification pile; Preferably, the pile fermentation index further comprises a combination of saccharification pile top temperature, saccharification pile bottom temperature, saccharification pile middle temperature, saccharification pile lactic acid content, saccharification pile reducing sugar content, saccharification pile water content, sampling room temperature, saccharification pile acidity and saccharification pile starch content; Preferably, the pile fermentation index further comprises a combination of saccharification pile top temperature, saccharification pile bottom temperature, saccharification pile middle temperature, saccharification pile lactic acid content, saccharification pile reducing sugar content, saccharification pile water content and sampling room temperature; Preferably, the pile fermentation index further comprises a combination of the saccharification pile top temperature, the saccharification pile bottom temperature, the saccharification pile middle temperature, the saccharification pile lactic acid content and the saccharification pile reducing sugar content; Preferably, the discriminant model includes a support vector machine binary classification model, an extreme gradient boosting binary classification model, a random forest binary classification model, a decision tree binary classification model, a K nearest neighbor binary classification model or a neural network binary classification model; Preferably, the discriminant model includes a vector machine binary classification model, an extreme gradient enhancement binary classification model or a random forest binary classification model; Preferably, the discriminant model is a random forest binary classification model; Preferably, the random forest binary classification model is a random forest binary classification model in R language; Preferably, the modeling data is divided into a training set and a test set; the training set data is input into a random forest binary classification algorithm to establish a training model, and the test set is input into the training model to obtain a prediction result; Preferably, the parameters of the random forest binary classification model are set as follows: ntree=100~1000, mtry=2~10, importance=T; Preferably, the parameters of the random forest binary classification model are set as follows: ntree=300~1000, mtry=2~8, importance=T; Preferably, the parameters of the random forest binary classification model are set as follows: ntree=300~800, mtry=2~6, importance=T; More preferably, the parameters of the random forest binary classification model are set as follows: ntree=500, mtry=6, importance=T; Preferably, the modeling data is divided into a training set and a test set in a ratio of 3:

1.

9. The identification device according to claim 8, characterized in that: The pile fermentation index is detected and collected by a near infrared analyzer, a temperature measuring rod and a thermometer.

10. Use of the identification method according to any one of claims 1 to 5 or the identification device according to any one of claims 8 to 9 in improving wine production rate.

Images