Intelligent temperature control method and system for spreading and cooling fermented grains

Abstract

The invention discloses an intelligent temperature control method for spreading and cooling fermented grains, which comprises the following steps of: acquiring key parameters such as environment temperature, temperature and humidity of the fermented grains, addition amount of starch and water, frequency of a fan and the like in real time by deploying a plurality of sensors, constructing a data set and screening out steady-state data; using the data to construct a multilayer perceptron (MLP) model, taking the temperature of the fermented grains at the outlet as output, and taking the rest data as input; a genetic algorithm is used as an intelligent optimization strategy, an MLP model is used as a fitness function, a difference value between a predicted temperature and a target temperature is calculated to serve as a loss value, and an optimal fan frequency combination is obtained through iterative optimization; sensor data are collected in real time, when the temperature of fermented grains at an outlet exceeds a set threshold value, the fan frequency is adjusted to the optimal state through a genetic algorithm, and accurate control over the temperature is achieved; according to the invention, real-time and automatic regulation and control of the temperature in the spreading and cooling process of the fermented grains are realized, and an efficient and intelligent solution is provided for fermented grain production.

Description

Technical Field

[0001] The invention relates to the technical field of liquor production and brewing, and in particular to an intelligent temperature control method and system for cooling mash. Background Art

[0002] In the liquor brewing process, the cooling of the mash is a crucial link, and its temperature control directly affects the subsequent fermentation process and the quality of the final liquor. The traditional cooling process of the mash mainly relies on manual experience and simple mechanical equipment for temperature control, which has problems such as large temperature fluctuations, low control accuracy, high energy consumption and low production efficiency. Especially in large-scale production, due to the complexity of environmental conditions and process parameters, manual control often cannot achieve the ideal temperature control effect, resulting in unstable mash quality and even affecting the efficiency and economic benefits of the entire brewing production line.

[0003] At present, although some automated control systems are used in the process of cooling the mash, most of these systems use fixed control strategies and cannot make intelligent adjustments according to real-time working conditions, resulting in limited temperature control effects. In addition, these systems often lack accurate temperature prediction models, making it difficult to accurately predict the temperature of the exported mash, resulting in large lags and errors in temperature control.

[0004] In addition, most existing temperature control methods do not consider the impact of key parameters such as fan frequency on temperature control, which makes it difficult to achieve accurate temperature control during the temperature regulation process. At the same time, due to the lack of comprehensive collection and effective use of sensor data, the formulation of temperature control strategies lacks comprehensive and accurate data support, which further affects the effect of temperature control. Summary of the invention

[0005] The purpose of the present invention is to provide an intelligent temperature control method and system for cooling fermented mash in view of the problems existing in the prior art. By deploying multiple sensors to monitor key parameters in real time, building an accurate temperature prediction model, and using an intelligent optimization strategy to automatically adjust the fan frequency combination, the temperature of the exported fermented mash can be accurately controlled, energy consumption and manual intervention can be reduced, and the quality and production efficiency of the fermented mash can be improved.

[0006] To achieve the above object, the technical solution adopted by the present invention is:

[0007] An intelligent temperature control method for cooling fermented grains comprises the following steps:

[0008] S1. Collect data on ambient temperature, temperature and humidity of inlet mash, amount of pulp water added, fan frequency of each wind chamber, temperature and humidity of mash in each wind chamber, plate chain speed, amount of koji powder added, and temperature and humidity of outlet mash during mash cooling process through sensors;

[0009] S2, cleaning the sensor data collected in step S1, constructing a data set, and filtering out steady-state data;

[0010] S3, using the export mash temperature data as output and the remaining sensor data as input to construct the MLP model;

[0011] S4. Use genetic algorithm as intelligent optimization strategy, MLP model as fitness function, the difference between predicted temperature and target temperature as loss value, calculate the fitness value corresponding to each fan frequency combination, and obtain the best fan frequency combination through continuous iterative optimization;

[0012] S5. Collect sensor data in real time. When the outlet mash temperature is higher than the set threshold, adjust the fan frequency combination to the optimal state through genetic algorithm to achieve real-time temperature control during the mash cooling process.

[0013] Step S1 includes:

[0014] S1.1. A first telescopic temperature probe is provided at the inlet trough of the fermented grains cooling machine to measure the inlet fermented grains temperature, and a first fermented grains moisture detection device is provided to measure the inlet fermented grains humidity;

[0015] A second telescopic temperature probe is provided at the outlet trough of the fermented mash cooling machine for measuring the outlet fermented mash temperature, and a second fermented mash moisture detection device is provided for measuring the outlet fermented mash humidity;

[0016] In each wind chamber, an infrared thermal imager is respectively arranged to measure the wind chamber temperature, an infrared point thermometer is arranged to measure the plate chain temperature, a third telescopic temperature probe is arranged to measure the wind chamber fermented grain temperature, and a third fermented grain moisture detection device is arranged to measure the fermented grain humidity in the wind chamber;

[0017] A slurry water weighing sensor is set in the slurry water bin to measure the amount of slurry water added, and a koji powder weighing sensor is set in the koji powder bin to measure the amount of koji powder added;

[0018] S1.2. When the mash cooling machine is operating normally, all sensor data are collected regularly through the control cabinet.

[0019] 3. The intelligent temperature control method for cooling fermented grains according to claim 1, characterized in that step S2 comprises:

[0020] S2.2. Delete records with a large amount of missing data;

[0021] For individual records with missing data, missing values ​​are filled using statistical value filling, local linear interpolation, regression prediction, or forward and backward filling methods;

[0022] For outliers exceeding the threshold, statistical value filling, local linear interpolation, regression prediction or forward and backward filling methods are used to process them;

[0023] S2.3, for the cleaned data, by calculating the time delay of the fermented grains raw materials during the transmission process and the distance between each detection point, the data of each detection point is aligned according to the actual transmission order to construct a data set;

[0024] S2.4. According to the on-site process and business logic, calculate the change of a main measurement data within a preset period of time. If the change of the main measurement data within this period of time does not exceed the set threshold, the data within this period of time is considered to be steady-state data.

[0025] Step S3 includes:

[0026] S3.1. Use a multi-layer perceptron MLP as a regression model, where the MLP model includes an input layer, at least one hidden layer and an output layer, and each hidden layer uses a ReLU activation function;

[0027] S3.2, divide the data set into a training set and a validation set, pass the input data to the model through forward propagation, and output the predicted outlet mash temperature;

[0028] S3.3, set the mean square error as the loss function and calculate the loss value between the predicted value and the true value;

[0029] S3.4. Use backpropagation to update the model weights and biases according to the gradient of the loss function to minimize the loss.

[0030] S3.5. Repeat the forward propagation and back propagation process until the preset training round is reached or the loss value drops below a certain threshold.

[0031] In step S3, after the MLP model is constructed, the MLP model is calibrated to achieve continuous optimization of the model accuracy, wherein the calibration process specifically includes:

[0032] S3.6. Continuously monitor the difference between the actual outlet mash temperature and the outlet temperature predicted by the MLP model;

[0033] S3.7. When the temperature control effect monitoring finds that the deviation exceeds the preset range, all sensor data at the current time point are immediately recorded, and the data are cleaned, aligned and screened;

[0034] S3.8. After determining that the prediction deviation of the newly collected data samples is not caused by abnormal data, the processed new data samples are used as training samples, and the MLP model is retrained using transfer learning or incremental learning methods to obtain a new model;

[0035] S3.9. Design a validation set that includes samples from the original data validation set and samples from the new data validation set to evaluate the performance of the new model and verify whether its prediction accuracy, loss value, and robustness indicators have improved;

[0036] S3.10. After confirming that the overall performance of the new model has not decreased but has improved, replace the original model with the new model and complete the model calibration.

[0037] Batch normalization and regularization modules are added during the training process of the MLP model; among them,

[0038] The calculation process of batch normalization includes:

[0039] Calculate the mean and variance of the output of each layer;

[0040] Use the calculated mean and variance to standardize the input so that the mean of the input data is close to zero and the standard deviation is close to one;

[0041] Perform linear transformation on the standardized results;

[0042] The calculation formula of the L2 regularization is:

[0043]

[0044] Among them, L(θ) is the original loss function, λ is the hyperparameter of regularization strength, is the sum of squares of all weight parameters.

[0045] Step S4 includes:

[0046] S4.1, encoding the fan frequency combination as a chromosome of the genetic algorithm;

[0047] S4.2, randomly generate a certain number of fan frequency combinations as the initial population;

[0048] S4.3, using the MLP model to predict the outlet temperature corresponding to each group of fan frequencies, and calculating the difference between the predicted outlet temperature and the set outlet temperature as the fitness value;

[0049] S4.4, according to the fitness value, use the roulette wheel selection or tournament selection method to select an excellent fan frequency combination from the current population as the parent generation;

[0050] S4.5, performing crossover and mutation operations on the selected parent chromosome to generate a new fan frequency combination as the offspring;

[0051] S4.6. Repeat the above steps S4.3-S4.5, iteratively approaching the optimal solution step by step until the predetermined number of rounds is reached.

[0052] In step S4.5, the crossover and mutation operations on the selected parent chromosome include:

[0053] In the early stages of evolution, a lower crossover probability and a higher mutation probability are used; in the later stages, a higher crossover probability and a lower mutation probability are used.

[0054] In the early stage of evolution, multiple fan frequency combinations are used to increase search diversity, and in the later stage, the number of fan frequency combinations is reduced to improve search efficiency.

[0055] An intelligent temperature control system for cooling fermented grains, comprising:

[0056] The data acquisition module collects data on the ambient temperature during the cooling process of the mash, the temperature and humidity of the inlet mash, the amount of pulp water added, the fan frequency of each wind chamber, the temperature and humidity of the mash in each wind chamber, the speed of the plate chain, the amount of koji added, and the temperature and humidity of the outlet mash through sensors;

[0057] The data processing module cleans the sensor collected data, builds the data set, and filters out the steady-state data;

[0058] The MLP model module uses the temperature data of the exiting mash as output and the rest of the sensor data as input;

[0059] The intelligent optimization module uses a genetic algorithm, takes the MLP model as the fitness function, and takes the difference between the predicted temperature and the target temperature as the loss value. It calculates the fitness value corresponding to each fan frequency combination and obtains the optimal fan frequency combination through continuous iterative optimization.

[0060] The temperature control module collects sensor data in real time. When the outlet mash temperature is higher than the set threshold, the fan frequency combination is adjusted to the optimal state through the genetic algorithm, thereby realizing real-time temperature control during the mash cooling process.

[0061] A computer-readable storage medium stores a computer program, wherein the computer program implements the above method steps when executed by a processor.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] 1. Realize real-time and accurate temperature control during the cooling process of the fermented grains, avoiding the problem of reduced fermented grain quality and production efficiency due to excessive temperature fluctuations. At the same time, the fan frequency combination is automatically adjusted through intelligent optimization strategies, reducing manual intervention and energy waste, and improving fermented grain quality and production efficiency;

[0064] 2. The multi-layer perceptron (MLP) is used as the regression model to accurately predict the temperature of the exported mash. By continuously monitoring the difference between the actual outlet temperature and the predicted temperature and calibrating the model, it is ensured that the model always maintains high prediction accuracy and robustness in practical applications.

[0065] 3. Adding batch normalization and regularization modules during the MLP model training process further improves the training efficiency and stability of the model and prevents the model from overfitting;

[0066] 4. Genetic algorithm is used as an intelligent optimization strategy to automatically adjust the fan frequency combination to the optimal state, realizing precise control of the temperature of the exported mash;

[0067] 5. Through the adaptive adjustment of crossover and mutation probabilities and the dynamic adjustment of population size, the optimization efficiency of the genetic algorithm is significantly improved, so that the optimal fan frequency combination can be found more quickly and accurately during the cooling process of the mash;

[0068] 6. By deploying a variety of sensors at key locations of the mash cooler, key parameters such as ambient temperature, temperature and humidity of the inlet and outlet mash, amount of slurry added, fan frequency, etc. can be monitored in real time, providing comprehensive and accurate data support for subsequent temperature control strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0070] Figure 1 This is a schematic diagram of the sensor deployment position of the intelligent cooling machine in the embodiment of the present application;

[0071] Figure 2 This is a flow chart of the intelligent optimization strategy in the embodiment of this application;

[0072] In the figure: 1. first telescopic temperature probe rod; 2. first fermented mash moisture detection device; 3. second telescopic temperature probe rod; 4. second fermented mash moisture detection device; 5. infrared thermal imager; 6. infrared point thermometer; 7. third telescopic temperature probe rod; 8. third fermented mash moisture detection device; 9. slurry water weighing sensor; 10. koji powder weighing sensor. DETAILED DESCRIPTION

[0073] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0074] In a first aspect of the present application, a method for intelligent temperature control of fermented grains cooling is proposed, comprising the following steps:

[0075] S1. The ambient temperature during the cooling process of the mash, the temperature and humidity of the inlet mash, the amount of pulp water added, the fan frequency of each wind chamber, the temperature and humidity of the mash in each wind chamber, the speed of the plate chain, the amount of koji added, and the temperature and humidity of the outlet mash are collected through sensors.

[0076] S2. Clean the data collected by the sensor in step S1, construct a data set, and filter out steady-state data.

[0077] S3. Use the export mash temperature data as output and the remaining sensor data as input to build the MLP model.

[0078] S4. Use genetic algorithm as the intelligent optimization strategy, MLP model as the fitness function, the difference between the predicted temperature and the target temperature as the loss value, calculate the fitness value corresponding to each fan frequency combination, and obtain the optimal fan frequency combination through continuous iterative optimization.

[0079] S5. Collect sensor data in real time. When the outlet mash temperature is higher than the set threshold, adjust the fan frequency combination to the optimal state through genetic algorithm to achieve real-time temperature control during the mash cooling process.

[0080] Through the above method and steps, real-time and precise control of the temperature during the cooling process of the mash is achieved, avoiding the problems of mash quality degradation and reduced production efficiency due to excessive temperature fluctuations. At the same time, the fan frequency combination is automatically adjusted through intelligent optimization strategies, reducing manual intervention and energy waste, and improving mash quality and production efficiency.

[0081] Each step is described in detail below with reference to the accompanying drawings.

[0082] See also Figure 1 , step S1 comprises:

[0083] S1.1. A first telescopic temperature probe rod 1 is provided at the inlet trough of the fermented grains cooling machine to measure the inlet fermented grains temperature, and a first fermented grains moisture detection device 2 is provided to measure the inlet fermented grains humidity.

[0084] A second telescopic temperature probe rod 3 is provided at the outlet trough of the mash cooling machine for measuring the outlet mash temperature, and a second mash moisture detection device 4 is provided for measuring the outlet mash humidity.

[0085] In each wind chamber, an infrared thermal imager 5 is provided to measure the wind chamber temperature, an infrared point thermometer 6 is provided to measure the plate chain temperature, a third telescopic temperature probe rod 7 is provided to measure the wind chamber mash temperature, and a third mash moisture detection device 8 is provided to measure the wind chamber mash humidity.

[0086] A slurry water weighing sensor 9 is arranged in the slurry water bin to measure the amount of slurry water added, and a koji powder weighing sensor 10 is arranged in the koji powder bin to measure the amount of koji powder added.

[0087] S1.2. When the mash cooling machine is operating normally, all sensor data are collected regularly through the control cabinet.

[0088] Specifically, the data collection time interval is fixed to ensure the continuity and comparability of the data. As the center of data collection and processing, the control cabinet is responsible for receiving the data from each sensor and storing it in the data set.

[0089] Through the sensor deployment and data collection methods in the above steps, real-time monitoring and recording of key parameters in the mash cooling process can be achieved, providing data support for the subsequent formulation of temperature control strategies.

[0090] Step S2 includes:

[0091] S2.2. Delete records with a large amount of missing data.

[0092] For individual records with missing data, missing values ​​are filled using statistical value filling, local linear interpolation, regression prediction, or forward and backward filling methods.

[0093] Outliers exceeding the threshold are processed by statistical value filling, local linear interpolation, regression prediction or forward and backward filling.

[0094] S2.3. For the cleaned data, by calculating the time delay of the mash raw material in the transmission process and the distance between each detection point, the time difference between the data is calculated, and the data of each detection point are aligned according to the actual transmission order to ensure that the data at the same time can accurately reflect the status of the mash at each stage. The aligned data are constructed into a data set.

[0095] S2.4. According to the on-site process and business logic, calculate the change of a main measurement data (such as the inlet mash temperature) within a preset time period. If the change of the main measurement data within this period does not exceed the set threshold, the data within this period is considered to be steady-state data.

[0096] Step S3 includes:

[0097] S3.1. Use a multi-layer perceptron (MLP) as a regression model, where the MLP model includes an input layer, at least one hidden layer, and an output layer, and each hidden layer uses a ReLU activation function.

[0098] Specifically, the ambient temperature, inlet mash temperature and humidity, fan frequency of each wind chamber, mash temperature and humidity of each wind chamber, plate chain temperature of each wind chamber, plate chain speed, outlet mash temperature and humidity, etc. are used as input parameters of the input layer to reflect the state changes of the mash during the cooling process.

[0099] In this embodiment, a three-layer hidden layer structure is adopted, and the number of neurons in each hidden layer is 128, 64 and 32 respectively. Each hidden layer uses a ReLU activation function. The output layer uses a single neuron to output the predicted outlet mash temperature.

[0100] S3.2. Divide the data set into a training set and a validation set. Pass the input data to the model through forward propagation and output the predicted export mash temperature.

[0101] S3.3. Set the mean square error as the loss function and calculate the loss value between the predicted value and the true value.

[0102] S3.4. Use backpropagation to update the model’s weights and biases according to the gradient of the loss function to minimize the loss.

[0103] S3.5. Repeat the forward propagation and back propagation process until the preset training round is reached or the loss value drops below a certain threshold.

[0104] Through the above steps, a nonlinear regression model based on multi-layer perceptron was constructed to achieve accurate prediction of the outlet temperature during the cooling process of the mash.

[0105] In step S3, after the MLP model is constructed, the MLP model is calibrated to achieve continuous optimization of the model accuracy, wherein the calibration process specifically includes:

[0106] S3.6. Continuously monitor the difference between the actual outlet mash temperature and the outlet temperature predicted by the MLP model.

[0107] S3.7. When the temperature control effect monitoring finds that the deviation exceeds the preset range, all sensor data at the current time point are immediately recorded, and the data are cleaned, aligned and screened.

[0108] S3.8. After determining that the prediction deviation is caused by the non-abnormal data of the newly collected data samples, the processed new data samples are used as training samples, and the MLP model is retrained using transfer learning or incremental learning methods to obtain a new model; among them, transfer learning can transfer part of the knowledge of the trained model to the new task to accelerate the training process of the new model; incremental learning can enable the model to learn new knowledge while retaining the old knowledge, so as to achieve continuous updating and optimization of the model.

[0109] S3.9. Design a validation set that includes samples from the original data validation set and samples from the new data validation set to evaluate the performance of the new model and verify whether its prediction accuracy, loss value, and robustness indicators have improved; the evaluation indicators include prediction accuracy, loss value, and robustness indicators. By comparing the performance of the new and old models on the validation set, determine whether the overall performance of the new model has improved and has not declined.

[0110] S3.10. After confirming that the overall performance of the new model has not decreased but has improved, replace the original model with the new model and complete the model calibration.

[0111] Through the above model calibration process, it can be ensured that the MLP model always maintains high prediction accuracy and robustness in practical applications, thereby achieving real-time and precise control of the temperature during the cooling process of the mash.

[0112] Batch normalization and regularization modules are added during the training process of the MLP model; among them,

[0113] The calculation process of batch normalization includes:

[0114] The mean and variance of the output of each layer are calculated.

[0115] The input is normalized using the calculated mean and variance so that the mean of the input data is close to zero and the standard deviation is close to one.

[0116] Perform a linear transformation on the standardized results.

[0117] The calculation formula of the L2 regularization is:

[0118]

[0119] Among them, L(θ) is the original loss function, λ is the hyperparameter of regularization strength, is the sum of squares of all weight parameters.

[0120] Through the above steps, during the training process of the MLP model, the input data of each layer is batch normalized to improve the training efficiency and stability of the model. When calculating the loss function, the L2 regularization term is added to the original loss function, and the strength of regularization is controlled by adjusting the value of λ to prevent the model from overfitting.

[0121] See also Figure 2 , step S4 comprises:

[0122] S4.1. Encode the fan frequency combination as chromosomes of the genetic algorithm.

[0123] S4.2. Randomly generate a certain number of fan frequency combinations as the initial population.

[0124] S4.3. Use the MLP model to predict the outlet temperature corresponding to each group of fan frequencies, and calculate the difference between the predicted outlet temperature and the set outlet temperature as the fitness value; wherein, the smaller the fitness value, the closer the group frequency combination is to the optimal solution.

[0125] S4.4. According to the fitness value, use the roulette wheel selection or tournament selection method to select an excellent fan frequency combination from the current population as the parent generation.

[0126] S4.5. Perform crossover and mutation operations on the selected parent chromosome to generate a new fan frequency combination as the offspring.

[0127] Specifically, a crossover operation is performed on the selected parent chromosomes to simulate the gene recombination process in biological evolution. New daughter chromosomes (i.e., new fan frequency combinations) are generated by exchanging some genes (i.e., some fan frequency values) of the parent chromosomes.

[0128] The newly generated offspring chromosomes (i.e., new fan frequency combinations) are mutated to randomly change certain gene values ​​in the chromosomes (i.e., randomly adjust some fan frequencies) to increase the diversity of the population and prevent the algorithm from maturing prematurely.

[0129] S4.6, repeat the above steps S4.3-S4.5, iteratively approaching the optimal solution step by step until the predetermined round is reached. In each iteration, the pros and cons of all individuals in the population are re-evaluated according to the new fitness value, and a new population is generated through selection, crossover and mutation operations. The iterative process will continue until the predetermined round is reached or the fitness value reaches the preset threshold.

[0130] Through the above-mentioned intelligent optimization steps, the fan frequency can be automatically adjusted during the cooling process of the mash, so as to achieve precise control of the temperature of the outlet mash.

[0131] In step S4.5, the crossover and mutation operations on the selected parent chromosome include:

[0132] In the early stages of evolution, the population diversity is high. In order to explore a wider solution space, a lower crossover probability and a higher mutation probability are used. A lower crossover probability helps maintain the diversity of the population and avoids premature convergence, while a higher mutation probability can increase the degree of population variation and promote the generation of new solutions. In the later stages, as the evolutionary process proceeds, the population gradually converges. In order to strengthen the local search for existing excellent solutions, a higher crossover probability and a lower mutation probability are used. A higher crossover probability helps the combination of excellent genes and accelerates the approach to the optimal solution, while a lower mutation probability can reduce unnecessary variation and maintain the stability of the population.

[0133] The specific formula is as follows: Let the crossover probability be P c , the mutation probability is P m :

[0134]

[0135] Among them, D is the population diversity index, t is the evolutionary generation, T is the total evolutionary generation, and a, b, c, and d are constants.

[0136] In order to increase the diversity of the search, in the early stage of evolution, a variety of fan frequency combinations are used to increase the search diversity. A larger population can cover a wider solution space and increase the possibility of finding the global optimal solution. In the later stage, as the population gradually converges, the number of fan frequency combinations is reduced to improve the search efficiency. A smaller population can reduce the amount of calculation and speed up the search while maintaining the ability to search locally for excellent solutions.

[0137] The specific formula is as follows: Assume that the initial population size is N0 and define the population size N t is the evolutionary generation t, the average fitness F avg And the function of diversity index D:

[0138]

[0139] Among them, α and β are adjustment parameters, F max is the theoretical maximum fitness value.

[0140] Through the above-mentioned adaptive adjustment of the crossover and mutation probabilities and the dynamic adjustment of the population size, the optimization efficiency of the genetic algorithm can be significantly improved, so that the optimal fan frequency combination can be found more quickly and accurately during the cooling process of the mash, thereby achieving precise control of the outlet mash temperature.

[0141] In a second aspect of the present application, an intelligent temperature control system for cooling fermented grains is provided, comprising:

[0142] The data acquisition module collects data on the ambient temperature during the cooling process of the mash, the temperature and humidity of the inlet mash, the amount of slurry added, the fan frequency of each wind chamber, the temperature and humidity of the mash in each wind chamber, the speed of the plate chain, the amount of koji added, and the temperature and humidity of the outlet mash through sensors.

[0143] The data processing module cleans the data collected by the sensor, builds the data set, and filters out the steady-state data.

[0144] The MLP model module takes the export mash temperature data as output and the rest of the sensor data as input.

[0145] The intelligent optimization module adopts a genetic algorithm, uses the MLP model as the fitness function, and takes the difference between the predicted temperature and the target temperature as the loss value. It calculates the fitness value corresponding to each fan frequency combination and obtains the optimal fan frequency combination through continuous iterative optimization.

[0146] The temperature control module collects sensor data in real time. When the outlet mash temperature is higher than the set threshold, the fan frequency combination is adjusted to the optimal state through the genetic algorithm, thereby realizing real-time temperature control during the mash cooling process.

[0147] According to a third aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, wherein the computer program implements the above-mentioned method steps when executed by a processor.

[0148] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0149] In the description of the present specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish descriptions and cannot be understood as indicating or suggesting relative importance. It should also be understood that although the terms "first", "second", etc. are used to describe various elements in some embodiments of the present application in the text, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element. For example, the first table can be named as the second table, and similarly, the second table can be named as the first table without departing from the scope of the various described embodiments. The first table and the second table are both tables, but they are not the same table.

[0150] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0151] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent temperature control method for cooling fermented grains, It is characterized in that The following steps are involved: S1. Collect data on ambient temperature, temperature and humidity of inlet mash, amount of pulp water added, fan frequency of each wind chamber, temperature and humidity of mash in each wind chamber, plate chain speed, amount of koji powder added, and temperature and humidity of outlet mash during mash cooling process through sensors; S2, cleaning the sensor data collected in step S1, constructing a data set, and filtering out steady-state data; S3, using the export mash temperature data as output and the remaining sensor data as input to construct the MLP model; S4. Use genetic algorithm as intelligent optimization strategy, MLP model as fitness function, the difference between predicted temperature and target temperature as loss value, calculate the fitness value corresponding to each fan frequency combination, and obtain the best fan frequency combination through continuous iterative optimization; S5. Collect sensor data in real time. When the outlet mash temperature is higher than the set threshold, adjust the fan frequency combination to the optimal state through genetic algorithm to achieve real-time temperature control during the mash cooling process.

2. The intelligent temperature control method for cooling fermented grains according to claim 1, It is characterized in that Step S1 includes: S1.

1. A first telescopic temperature probe is provided at the inlet trough of the fermented grains cooling machine to measure the inlet fermented grains temperature, and a first fermented grains moisture detection device is provided to measure the inlet fermented grains humidity; A second telescopic temperature probe is provided at the outlet trough of the fermented mash cooling machine for measuring the outlet fermented mash temperature, and a second fermented mash moisture detection device is provided for measuring the outlet fermented mash humidity; In each wind chamber, an infrared thermal imager is respectively arranged to measure the wind chamber temperature, an infrared point thermometer is arranged to measure the plate chain temperature, a third telescopic temperature probe is arranged to measure the wind chamber fermented grain temperature, and a third fermented grain moisture detection device is arranged to measure the fermented grain humidity in the wind chamber; A slurry water weighing sensor is set in the slurry water bin to measure the amount of slurry water added, and a koji powder weighing sensor is set in the koji powder bin to measure the amount of koji powder added; S1.

2. When the mash cooling machine is operating normally, all sensor data are collected regularly through the control cabinet.

3. The intelligent temperature control method for cooling fermented grains according to claim 1, It is characterized in that Step S2 includes: S2.

2. Delete records with a large amount of missing data; For individual records with missing data, missing values ​​are filled using statistical value filling, local linear interpolation, regression prediction, or forward and backward filling methods; For outliers exceeding the threshold, statistical value filling, local linear interpolation, regression prediction or forward and backward filling methods are used to process them; S2.3, for the cleaned data, by calculating the time delay of the fermented grains raw materials during the transmission process and the distance between each detection point, the data of each detection point is aligned according to the actual transmission order to construct a data set; S2.

4. According to the on-site process and business logic, calculate the change of a main measurement data within a preset period of time. If the change of the main measurement data within this period of time does not exceed the set threshold, the data within this period of time is considered to be steady-state data.

4. The intelligent temperature control method for cooling fermented grains according to claim 1, It is characterized in that Step S3 includes: S3.

1. Use a multi-layer perceptron MLP as a regression model, where the MLP model includes an input layer, at least one hidden layer and an output layer, and each hidden layer uses a ReLU activation function; S3.2, divide the data set into a training set and a validation set, pass the input data to the model through forward propagation, and output the predicted outlet mash temperature; S3.3, set the mean square error as the loss function and calculate the loss value between the predicted value and the true value; S3.

4. Use backpropagation to update the model weights and biases according to the gradient of the loss function to minimize the loss. S3.

5. Repeat the forward propagation and back propagation process until the preset training round is reached or the loss value drops below a certain threshold.

5. The intelligent temperature control method for cooling fermented grains according to claim 1, It is characterized in that In step S3, after the MLP model is constructed, the MLP model is calibrated to achieve continuous optimization of the model accuracy, wherein the calibration process specifically includes: S3.

6. Continuously monitor the difference between the actual outlet mash temperature and the outlet temperature predicted by the MLP model; S3.

7. When the temperature control effect monitoring finds that the deviation exceeds the preset range, all sensor data at the current time point are immediately recorded, and the data are cleaned, aligned and screened; S3.

8. After determining that the prediction deviation of the newly collected data samples is not caused by abnormal data, the processed new data samples are used as training samples, and the MLP model is retrained using transfer learning or incremental learning methods to obtain a new model; S3.

9. Design a validation set that includes samples from the original data validation set and samples from the new data validation set to evaluate the performance of the new model and verify whether its prediction accuracy, loss value, and robustness indicators have improved; S3.

10. After confirming that the overall performance of the new model has not decreased but has improved, replace the original model with the new model and complete the model calibration.

6. The intelligent temperature control method for cooling fermented grains according to claim 4, It is characterized in that Batch normalization and regularization modules are added during the training process of the MLP model; among them, The calculation process of batch normalization includes: Calculate the mean and variance of the output of each layer; Use the calculated mean and variance to standardize the input so that the mean of the input data is close to zero and the standard deviation is close to one; Perform linear transformation on the standardized results; The calculation formula of the L2 regularization is: Among them, L(θ) is the original loss function, λ is the hyperparameter of regularization strength, is the sum of squares of all weight parameters.

7. The intelligent temperature control method for cooling fermented grains according to claim 1, It is characterized in that Step S4 includes: S4.1, encoding the fan frequency combination as a chromosome of the genetic algorithm; S4.2, randomly generate a certain number of fan frequency combinations as the initial population; S4.3, using the MLP model to predict the outlet temperature corresponding to each group of fan frequencies, and calculating the difference between the predicted outlet temperature and the set outlet temperature as the fitness value; S4.4, according to the fitness value, use the roulette wheel selection or tournament selection method to select an excellent fan frequency combination from the current population as the parent generation; S4.5, performing crossover and mutation operations on the selected parent chromosome to generate a new fan frequency combination as the offspring; S4.

6. Repeat the above steps S4.3-S4.5, iteratively approaching the optimal solution step by step until the predetermined number of rounds is reached.

8. The intelligent temperature control method for cooling fermented grains according to claim 7, It is characterized in that In step S4.5, the crossover and mutation operations on the selected parent chromosome include: In the early stages of evolution, a lower crossover probability and a higher mutation probability are used; in the later stages, a higher crossover probability and a lower mutation probability are used. In the early stage of evolution, multiple fan frequency combinations are used to increase search diversity, and in the later stage, the number of fan frequency combinations is reduced to improve search efficiency.

9. An intelligent temperature control system for cooling fermented grains. It is characterized in that include: The data acquisition module collects data on the ambient temperature during the cooling process of the mash, the temperature and humidity of the inlet mash, the amount of pulp water added, the fan frequency of each wind chamber, the temperature and humidity of the mash in each wind chamber, the speed of the plate chain, the amount of koji added, and the temperature and humidity of the outlet mash through sensors; The data processing module cleans the sensor collected data, builds the data set, and filters out the steady-state data; The MLP model module uses the temperature data of the exiting mash as output and the rest of the sensor data as input; The intelligent optimization module uses a genetic algorithm, takes the MLP model as the fitness function, and takes the difference between the predicted temperature and the target temperature as the loss value. It calculates the fitness value corresponding to each fan frequency combination and obtains the optimal fan frequency combination through continuous iterative optimization. The temperature control module collects sensor data in real time. When the outlet mash temperature is higher than the set threshold, the fan frequency combination is adjusted to the optimal state through the genetic algorithm, thereby realizing real-time temperature control during the mash cooling process.

10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, wherein the computer program implements the method steps of any one of claims 1 to 8 when executed by a processor.

Images