Daqu fermentation control method and Daqu fermentation system

Abstract

The invention provides a yeast fermentation control method and a yeast fermentation system, and belongs to the technical field of brewing and automatic control. The yeast fermentation system comprises a yeast room, an execution mechanism, a temperature sensor, a humidity sensor, a yeast fermentation controller and an upper computer. The yeast fermentation control method comprises the following steps: constructing and optimizing a yeast fermentation controller; fusing the temperature detection data and the humidity detection data; the yeast fermentation controller carries out learning and training, and outputs the starting time of each execution mechanism; and the upper computer controls the execution starting or stopping of the corresponding execution mechanism. According to the method, the adaptive weighted fusion algorithm based on deep learning is adopted to perform data fusion, and the group intelligent optimization strategy is adopted to optimize the yeast fermentation controller, so that the control requirements of the yeast fermentation system on temperature and humidity are met, the accuracy of yeast room environment temperature and humidity control and good follow-up performance of the environment temperature and humidity are ensured, and the fermentation quality is improved; the energy consumption is reduced and the fermentation quality is ensured.

Description

Technical Field

[0001] The invention belongs to the technical field of brewing and automatic control, and in particular relates to a Daqu fermentation control method and a Daqu fermentation system. Background Art

[0002] As a fermentation agent for solid-state brewing, Daqu plays a key role in the winemaking process. In winemaking production, there is a saying that "flavor is produced by distillation, and aroma is produced by fermentation." The so-called aroma production refers to the rich active microorganisms and enzymes in the qu being mixed into the starch-containing grain dregs that have been cooled after high-temperature steaming and gelatinization, participating in the fermentation process and producing complex aroma and flavor substances. A large amount of heat will be generated during the fermentation of Daqu, and the rise in temperature will also lead to the evaporation of the water in the Daqu, and the concentrations of O2 and CO2 will also change. Daqu plays the role of the soul in the fermentation process of winemaking, and the environment of the qu room, as the fermentation place of Daqu, has a great influence on the quality of Daqu.

[0003] At present, some domestic wineries still use the traditional fermentation method in the koji room for Daqu fermentation, that is, through artificial timed humidification, opening and closing doors and windows, turning over koji and other operations, the temperature and humidity of the Daqu fermentation process in the koji room and environmental parameters such as O2 and CO2 are regulated. Therefore, operators are required to observe the Daqu and the environment in the koji room, and then according to the company's existing experience, open and close doors and windows (adjust temperature, humidity, and dehumidification), turn over koji multiple times, pile burn, gather and cover with straw curtains, etc., to regulate the environmental parameters during Daqu fermentation to achieve the optimal environmental state required for Daqu fermentation. The traditional Daqu fermentation production process not only relies on the subjective experience of the operators in the koji workshop to make judgments and decisions, but is prone to poor control of the Daqu fermentation process. In addition, the fermentation quality of Daqu in different batches or different koji rooms of the same batch and different locations in the same koji room is very different, so that uniformity is difficult to guarantee. In addition, the production environment during Daqu fermentation is poor, and the high temperature and high humidity environment is not conducive to the health of workshop operators.

[0004] In response to the above problems, many researchers have built intelligent koji rooms, such as the temperature and humidity monitoring system of a winery's koji fermentation workshop. The system uses DS18B20 and DHIH10N as temperature and humidity detection devices to form a temperature and humidity detection and adjustment system based on the CAN bus network, and adopts PID control algorithm for system control, which improves the level of automation and reduces labor intensity. For another example, a company proposed the digitalization of the production process of Daqu, applied the digital control system to its own winery, collected, analyzed and processed data on environmental parameters such as temperature, humidity, and carbon dioxide, and performed digital and automatic opening and closing operations on the windows of the koji room based on the results, changing the traditional management mode of koji room cultivation and making a leap to digital management. For another example, a group successfully applied the automatic opening and closing door and window technology to the Daqu fermentation koji room for the first time. By detecting parameters such as the koji core temperature, environmental temperature and humidity in the koji room, when the data exceeds the limit value, the doors and windows are automatically opened for ventilation, the koji room is cooled down, the Daqu is kept fermented stably, and the fermentation quality of Daqu is guaranteed.

[0005] The above research results have promoted the industrial intelligence of Daqu fermentation through modern science and technology, but they all have the problem of not being combined with the actual characteristics of the qu room itself and the temperature and humidity characteristics of the Daqu fermentation environment. The intelligent control of temperature and humidity in the Daqu fermentation process has not been achieved, and the quality and uniformity of Daqu fermentation cannot be guaranteed. Summary of the invention

[0006] The present invention aims to provide a Daqu fermentation control method and a Daqu fermentation system, which adopts an adaptive weighted fusion algorithm based on deep learning to fuse temperature detection data and detection data, and then proposes a Daqu fermentation controller based on a swarm intelligence optimization strategy by fusing a fuzzy logic control algorithm and a neural network. The optimized Daqu fermentation controller can meet the temperature and humidity control requirements of the Daqu fermentation system, achieve more accurate start and stop time of the actuator, ensure the accuracy of the temperature and humidity control of the qu room environment and the good followability of the ambient temperature and humidity, improve the fermentation quality, reduce energy consumption, and ensure the fermentation quality.

[0007] The technical solution adopted by the present invention is: A Daqu fermentation system, comprising: Qufang; Actuators, the actuators are dispersedly arranged in the curved room; wherein the actuators include a heater, a humidifier, an electric window, an inner circulation fan and an outer circulation fan; Temperature sensors are dispersedly arranged in the qu room to collect temperature detection data at different monitoring points in the qu room; Humidity sensors are dispersedly arranged in the qu room to collect humidity detection data at different monitoring points in the qu room; A Daqu fermentation controller, the Daqu fermentation controller is constructed based on fuzzy logic control algorithm and neural network fusion, and is optimized by swarm intelligence optimization strategy; the optimized Daqu fermentation controller uses an adaptive weighted fusion algorithm based on deep learning to perform data fusion processing on the temperature detection data and humidity detection data, and then outputs the temperature deviation and temperature deviation change rate between the actual temperature value of the qu room and the target temperature value, as well as the humidity deviation and humidity deviation change rate between the actual humidity value of the qu room and the target humidity value as input signals, and outputs the opening time of each of the actuators as output signals; wherein the temperature deviation change rate is the ratio of the temperature deviation to the target temperature value, and the humidity deviation change rate is the ratio of the humidity deviation to the target humidity value; The host computer receives all the output signals and converts the output signals into corresponding Boolean signals one by one, and then the host computer controls the execution start or stop operation of the corresponding actuator according to the Boolean signals.

[0008] Furthermore, the Daqu fermentation controller includes a front-end network and a back-end network; The first layer of the antecedent network is composed of an input layer, and each node in the input layer is connected to each component of an input vector and transmitted to the next layer; the input vector is the input signal; The second layer of the antecedent network is composed of a fuzzification layer, and the fuzzification layer fuzzifies the data of the input layer through a Gaussian membership function; The third layer of the antecedent network is composed of a rule layer, and the rule layer is composed of 49*49 neuron nodes, and each neuron node of the rule layer represents a fuzzy rule; The fourth layer of the antecedent network is composed of a normalization layer, which is used to perform normalization calculations and serves as the connection weight of the third layer of the consequent network; the number of neuron nodes in the normalization layer is 49*49; The first layer of the subsequent network consists of the input layer and transmits the input vector to the next layer; The second layer of the consequent network is composed of a rule calculation layer, which is composed of 49*49 neuron nodes, each of which represents a calculation rule; the rule calculation layer is used to calculate the consequent of the fuzzy rule; The third layer of the subsequent network consists of an output layer, which performs weighted summation of the output of the normalization layer and the output of the rule calculation layer.

[0009] Furthermore, the host computer also has a visualization module; the visualization module is used for visual display of chart data.

[0010] Based on the same inventive concept, the present invention also provides a method for controlling Daqu fermentation, comprising the following steps: Step S1, constructing a Daqu fermentation controller based on fuzzy logic control algorithm and neural network fusion, and optimizing the Daqu fermentation controller using swarm intelligence optimization strategy; Step S2, collecting temperature detection data and humidity detection data from different monitoring points in the Qu room, and using an adaptive weighted fusion algorithm based on deep learning to perform data fusion processing on the temperature detection data and the humidity detection data, to obtain the fused actual temperature value of the Qu room and the actual humidity value of the Qu room, and then outputting the temperature deviation and the temperature deviation change rate between the actual temperature value of the Qu room and the target temperature value, and outputting the humidity deviation and the humidity deviation change rate between the actual humidity value of the Qu room and the target humidity value; wherein the temperature deviation change rate is the ratio of the temperature deviation to the target temperature value, and the humidity deviation change rate is the ratio of the humidity deviation to the target humidity value; Step S3, using the temperature deviation and the temperature deviation change rate, as well as the humidity deviation and the humidity deviation change rate as input signals to be introduced into the optimized Daqu fermentation controller for learning and training, and outputting the opening time of each actuator as an output signal; wherein the actuator includes a heater, a humidifier, an electric window, an internal circulation fan and an external circulation fan; Step S4, the host computer receives all the output signals and converts the output signals into corresponding Boolean signals one by one, and then the host computer controls the corresponding actuator to start or stop according to the Boolean signals.

[0011] Furthermore, the specific process of optimizing the Daqu fermentation controller by using the swarm intelligence optimization strategy in step S1 is as follows: Step S11, initializing the Daqu fermentation controller; wherein the Daqu fermentation controller includes a front-end network and a back-end network; Assume that there are m individuals in the D-dimensional search space, and the specific meaning of each individual is the parameter combination of the antecedent network and the consequent network; the fitness of each individual in the swarm intelligence optimization strategy is the difference between the start and stop time of the actuator actually output and the expected start and stop time, and the position and speed of the individual are updated according to the fitness of each individual; wherein, the speed and position of the individual are updated by the individual extreme value and the group extreme value using the following formula: ; ; In the formula, It represents the speed of the i-th individual in the k+1th iteration of the swarm intelligence optimization strategy in the j-th dimension in the D-dimensional search space; It represents the speed of the kth iteration of the swarm intelligence optimization strategy of the i-th individual in the j-th dimension in the D-dimensional search space; represents the inertia weight; c1 and c2 represent acceleration constants; r1 and r2 represent random numbers in the interval [0, 1]; It represents the optimal position of the i-th individual in the k-th iteration of the swarm intelligence optimization strategy on the j-th dimension in the D-dimensional search space; Represents the position of the i-th individual at the k-th iteration of the swarm intelligence optimization strategy on the j-th dimension in the D-dimensional search space; It represents the position of the i-th individual at the k+1th iteration of the swarm intelligence optimization strategy on the j-th dimension in the D-dimensional search space; It represents the global optimal position of the i-th individual in the k-th iteration of the swarm intelligence optimization strategy in the D-dimensional search space; Step S12, presetting the maximum number of iterations of the swarm intelligence optimization strategy, and evaluating the fitness of each individual after each iteration; after reaching the preset maximum number of iterations, terminating the iteration and executing the next step; Step S13, presetting the maximum number of iterations of the Daqu fermentation controller, selecting the optimal value of the individual as the combined parameter weight of the antecedent network and the consequent network and importing it into the Daqu fermentation controller; after reaching the preset maximum number of iterations, terminating the iteration to obtain the optimized Daqu fermentation controller.

[0012] Furthermore, before performing the fusion process in step S2, the collected temperature detection data or the humidity detection data is firstly subjected to consistency detection.

[0013] Furthermore, the specific process of performing consistency detection on the temperature detection data in step S2 is: Determine one by one whether the difference between the temperature detection data currently collected by the same temperature sensor and the temperature detection data collected at the last moment is less than or equal to the corresponding temperature monitoring threshold; If it is less than or equal to the corresponding temperature monitoring threshold, performing subsequent fusion processing on the temperature detection data currently collected by the same temperature sensor; If it is greater than the corresponding temperature monitoring threshold, taking the temperature detection average value of all the temperature detection data collected by the same temperature sensor to perform subsequent fusion processing; The specific process of performing consistency detection on the humidity detection data in step S2 is: Determine one by one whether the difference between the humidity detection data currently collected by the same humidity sensor and the humidity detection data collected at the last moment is less than or equal to the corresponding humidity monitoring threshold; If it is less than or equal to the corresponding humidity monitoring threshold, performing subsequent fusion processing on the humidity detection data currently collected by the same humidity sensor; If it is greater than the corresponding humidity monitoring threshold, the humidity detection average value of all the humidity detection data collected by the same humidity sensor is taken to perform subsequent fusion processing.

[0014] Furthermore, in step S2, the temperature detection data is subjected to data fusion processing, and the fused actual temperature value of the room is Obtained by the following formula: ; ; ; In the formula, Indicates the actual temperature value of the fused qufang; Indicates the latest temperature detection data of the p-th temperature sensor among the n temperature sensors or the temperature detection average value of all temperature detection data of the p-th temperature sensor; represents the variance of the temperature detection data of the pth temperature sensor among the n temperature sensors; Represents the average value of the temperature detection data of n temperature sensors; Similarly, when the fusion processing of the humidity detection data is performed in step S2, reference is made to the fusion processing process of the temperature detection data.

[0015] Furthermore, when learning and training are performed in the Daqu fermentation controller in step S3, a back propagation algorithm is used to change the weights of the second layer of the successor network in the Daqu fermentation controller, and the center value and width of the Gaussian membership function used in the second layer of the predecessor network in the Daqu fermentation controller are adjusted, so that the Daqu fermentation controller can stably track the optimization target value.

[0016] Furthermore, the specific process of learning and training in the Daqu fermentation controller in step S3 is as follows: The learning algorithm for the weights of the second layer of the subsequent network is: ; ; Where E is the loss function; is the weight change between the kth output and the ith input in the pth calculation rule in the consequent network; is the output of the pth calculation rule in the consequent network; is the output of the kth neuron node in the pth calculation rule in the consequent network; is the expected output of the pth calculation rule in the consequent network; is the membership degree of the jth fuzzy set; is the component of the i-th input vector; r is the learning rate; l is the iteration number of the Daqu fermentation controller; In parameter value After determination, the influence of the error on each layer in the antecedent network is checked, and the derived formula is as follows; ; ; ; In the formula, represents the error term of the i-th neuron node in the fifth layer of the antecedent network; the fifth layer of the antecedent network consists of the output layer; Represents the error term of the kth neuron node in the fourth layer of the antecedent network; Represents the error term of the kth neuron node in the third layer of the antecedent network; t i represents the expected output of the i-th output node; y i Represents the actual output of the i-th output node; y ik Represents the connection weight between the kth neuron node in the fourth layer of the antecedent network and the ith neuron node in the fifth layer; a i represents the membership degree of the i-th input variable; Finally, we get: ; ; Where E is the loss function; Represents the error term of the i-th neuron node in the k-th layer; x i represents the value of the i-th input variable; c ik represents the central value of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network; c jk represents the central value of the k-th Gaussian membership function corresponding to the j-th input used in the second layer of the antecedent network; qik represents the width of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network; The central value c of the k-th Gaussian membership function corresponding to the i-th input used in the second layer of the antecedent network ik and width value q ik The learning rule is: ; ; In the formula, represents the central value of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network during the l+1th Daqu fermentation controller iteration; represents the central value of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network during the l-th iteration of the Daqu fermentation controller; represents the width of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network during the l+1th Daqu fermentation controller iteration; represents the width of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network during the l-th iteration of the Daqu fermentation controller; r is the learning rate; E is the loss function.

[0017] The beneficial effects of the present invention are: The present invention provides a kind of Daqu fermentation control method and Daqu fermentation system, which adopts an adaptive weighted fusion algorithm based on deep learning to fuse temperature detection data and detection data, and then proposes a Daqu fermentation controller based on swarm intelligence optimization strategy by fusing fuzzy logic control algorithm and neural network, and defines the associated parameters of the input conditions and output results of the fermentation process as individuals in the search space, and performs global optimization on the antecedent and post-part networks of the Daqu fermentation network model, and combines the actual characteristics of the qu room itself with the temperature and humidity characteristics of the Daqu fermentation environment, and applies them to the Daqu fermentation process. The optimized Daqu fermentation controller can meet the control requirements of the Daqu fermentation system for temperature and humidity, realize more accurate start and stop time of actuators such as internal circulation fans, external circulation fans, electric windows, heaters, humidifiers, etc., ensure the accuracy of the temperature and humidity control of the qu room environment and the good followability of the environmental temperature and humidity, improve the fermentation quality, reduce energy consumption, and ensure the fermentation quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application 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 application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 This is a front view of the curved room in Example 1.

[0020] Figure 2 It is a top view of the curved room in Example 1.

[0021] Figure 3 Schematic diagram of the process of controlling the fermentation of Daqu in Example 2.

[0022] Figure 4 This is a schematic diagram of the process of optimizing the Daqu fermentation controller using the swarm intelligence optimization strategy in Example 2.

[0023] Figure 5 This is a schematic diagram of the processing flow of temperature detection data or humidity detection data in Example 2.

[0024] The accompanying drawings are marked as follows: 100- orifice plate, 200- electric window, 300- humidifier, 400- external circulation fan, 500- internal circulation fan, 600- heater, 700- curved rack, 800- temperature sensor, 900- humidity sensor. DETAILED DESCRIPTION

[0025] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0026] The disclosure below provides many different embodiments or examples to realize different structures of the present invention. In order to simplify the disclosure of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present invention.

[0027] The embodiments of the invention are described in detail below with reference to the accompanying drawings.

[0028] Example 1

[0029] A Daqu fermentation system comprises: a qu room, an actuator, a temperature sensor, a humidity sensor and a host computer.

[0030] Wherein, the actuators are dispersedly arranged in the curved room; wherein, the actuators include a heater, a humidifier, an electric window, an internal circulation fan and an external circulation fan.

[0031] The temperature sensors are dispersedly arranged in the qu room to collect temperature detection data at different monitoring points in the qu room. The temperature accuracy and range are: temperature ±0.3℃, 0℃~80℃; The humidity sensors are dispersedly arranged in the qu room to collect humidity detection data at different monitoring points in the qu room, wherein the humidity accuracy and range are: humidity ±3%, 10%RH~100%RH.

[0032] The Daqu fermentation controller is constructed based on the fusion of fuzzy logic control algorithm and neural network, and is optimized by swarm intelligence optimization strategy; the optimized Daqu fermentation controller uses an adaptive weighted fusion algorithm based on deep learning to perform data fusion processing on the temperature detection data and humidity detection data, and then outputs the temperature deviation and temperature deviation change rate between the actual temperature value of the qu room and the target temperature value, as well as the humidity deviation and humidity deviation change rate between the actual humidity value of the qu room and the target humidity value as input signals, and then outputs the opening time of each actuator as an output signal. Among them, the temperature deviation change rate is the ratio of the temperature deviation to the target temperature value, and the humidity deviation change rate is the ratio of the humidity deviation to the target humidity value.

[0033] The host computer receives all the output signals and converts the output signals into corresponding Boolean signals one by one, and then the host computer controls the execution start or stop operation of the corresponding actuator according to the Boolean signal. At the same time, the host computer also has a visualization module; the visualization module is used for visual display of chart data, such as the temperature curve and humidity curve of each monitoring point.

[0034] Figure 1 It is a front view of the curved room in Example 1. Figure 2 is a top view of the curved room in Example 1. Figure 1 and Figure 2In the arrangement structure of the qu room shown in FIG, four orifice plates 100 are arranged on the left and right side walls of the qu room, an electric window 200 is arranged on both sides of the rear side wall and in the middle of the front side wall of the qu room, a door is arranged in the middle of the rear side wall of the qu room, a humidifier 300 is arranged between the electric window 200 and the door on the rear side wall of the qu room, an external circulation air 400 and an internal circulation fan 500 are arranged on both sides of the front side wall of the qu room, and a heater 600 is arranged between the external circulation air 400 and the electric window 200 on the front side wall of the qu room. Four qu racks 700 are arranged in the qu room; each qu rack 700 has three layers, and 16 pieces of Daqu are placed on each layer, for a total of 192 pieces of Daqu. Temperature sensors 800 and humidity sensors 900 are evenly distributed on the upper, middle and lower layers of the qu room, and 5 monitoring points are arranged on each layer; the acquisition frequency of each temperature sensor 800 and humidity sensor 900 is to record the collected temperature and humidity data once every 6 minutes.

[0035] When controlling Daqu fermentation, the theoretical curve of temperature and humidity of Daqu fermentation is used as the control target. The temperature and humidity data of the Daqu fermentation room after the adaptive weighted fusion algorithm are used as the input data of the Daqu fermentation controller. The output realizes the start and stop time of the internal circulation fan, external circulation fan, electric window, heater, humidifier and other actuators, and the upper computer controls the execution of the start or stop operation of the corresponding actuator.

[0036] Wherein, the Daqu fermentation controller includes a front-end network and a back-end network.

[0037] The first layer of the antecedent network is composed of an input layer, and each node in the input layer is connected to each component of the input vector and transmitted to the next layer; the input vector is the input signal, namely the temperature deviation and the temperature deviation change rate, as well as the humidity deviation and the humidity deviation change rate; the second layer of the antecedent network is composed of a fuzzification layer, and the fuzzification layer fuzzifies the input layer through a Gaussian membership function; the third layer of the antecedent network is composed of a rule layer, and the rule layer is composed of 49*49 neuron nodes, and each neuron node of the rule layer represents a fuzzy rule; the fourth layer of the antecedent network is composed of a normalization layer, and the normalization layer is used to perform normalization calculations and serves as the connection weight of the third layer of the consequent network; the number of neuron nodes in the normalization layer is 49*49.

[0038] The first layer of the consequent network is composed of the input layer, and transmits the input vector to the next layer; the second layer of the consequent network is composed of a rule calculation layer, and the rule calculation layer is composed of 49*49 neuron nodes, and each neuron node of the rule calculation layer represents a calculation rule; the rule calculation layer is used to calculate the consequent of the fuzzy rule; the third layer of the consequent network is composed of an output layer, and the output layer performs weighted summation on the output of the normalization layer and the output of the rule calculation layer.

[0039] Example 2

[0040] Figure 3 Schematic diagram of the process of controlling the fermentation of Daqu in Example 2. Figure 3 As shown in, the Daqu fermentation control method comprises the following steps: Step S1, constructing a Daqu fermentation controller based on fuzzy logic control algorithm and neural network fusion, and optimizing the Daqu fermentation controller using swarm intelligence optimization strategy; Step S2, collecting temperature detection data and humidity detection data from different monitoring points in the Qu room, and using an adaptive weighted fusion algorithm based on deep learning to perform data fusion processing on the temperature detection data and the humidity detection data, to obtain the fused actual temperature value of the Qu room and the actual humidity value of the Qu room, and then outputting the temperature deviation and the temperature deviation change rate between the actual temperature value of the Qu room and the target temperature value, and outputting the humidity deviation and the humidity deviation change rate between the actual humidity value of the Qu room and the target humidity value; wherein the temperature deviation change rate is the ratio of the temperature deviation to the target temperature value, and the humidity deviation change rate is the ratio of the humidity deviation to the target humidity value; Step S3, using the temperature deviation and the temperature deviation change rate, as well as the humidity deviation and the humidity deviation change rate as input signals to be introduced into the optimized Daqu fermentation controller for learning and training, and outputting the opening time of each actuator as an output signal; wherein the actuator includes a heater, a humidifier, an electric window, an internal circulation fan and an external circulation fan; Step S4, the host computer receives all the output signals and converts the output signals into corresponding Boolean signals one by one, and then the host computer controls the corresponding actuator to start or stop according to the Boolean signals.

[0041] The Daqu fermentation control method in this embodiment adopts an adaptive weighted fusion algorithm based on deep learning to fuse the temperature detection data and the detection data, and then proposes a Daqu fermentation controller based on a swarm intelligence optimization strategy by fusing the fuzzy logic control algorithm and the neural network. By defining the associated parameters of the input conditions and output results of the fermentation process as individuals in the search space, the antecedent and the post-part networks of the Daqu fermentation network model are globally optimized, and the actual characteristics of the qu room itself and the temperature and humidity characteristics of the Daqu fermentation environment are combined and applied to the Daqu fermentation process. The optimized Daqu fermentation controller can meet the temperature and humidity control requirements of the Daqu fermentation system, realize the start and stop time of the actuators such as the internal circulation fan, the external circulation fan, the electric window, the heater, and the humidifier more accurately, ensure the accuracy of the temperature and humidity control of the qu room environment and the good follow-up of the environmental temperature and humidity, improve the fermentation quality, reduce energy consumption, and ensure the fermentation quality.

[0042] It should be noted that the order of step S1 and step S2 in this embodiment can be reversed or the two steps can be executed simultaneously.

[0043] In this embodiment, the Daqu fermentation controller based on the fusion of fuzzy logic control algorithm and neural network includes a front-end network and a back-end network. Among them, the first layer of the front-end network is composed of an input layer, and each node in the input layer is connected to each component of the input vector and transmitted to the next layer; the input vector is the input signal, that is, the temperature deviation and the temperature deviation change rate, and the humidity deviation and the humidity deviation change rate; the second layer of the front-end network is composed of a fuzzification layer, and the fuzzification layer fuzzifies the input layer through a Gaussian membership function; the third layer of the front-end network is composed of a rule layer, and the rule layer is composed of 49*49 neuron nodes, and each neuron node of the rule layer represents a fuzzy rule; the fourth layer of the front-end network is composed of a normalization layer, and the normalization layer is used to perform normalization calculations and serve as the connection weight of the third layer of the back-end network; the number of neuron nodes in the normalization layer is 49*49.

[0044] The first layer of the consequent network is composed of the input layer, and transmits the input vector to the next layer; the second layer of the consequent network is composed of a rule calculation layer, and the rule calculation layer is composed of 49*49 neuron nodes, and each neuron node of the rule calculation layer represents a calculation rule; the rule calculation layer is used to calculate the consequent of the fuzzy rule; the third layer of the consequent network is composed of an output layer, and the output layer performs weighted summation on the output of the normalization layer and the output of the rule calculation layer.

[0045] Figure 4 The figure is a schematic diagram of the process of optimizing the Daqu fermentation controller using the swarm intelligence optimization strategy in Example 2. Figure 4 As shown in , the specific process of optimizing the Daqu fermentation controller by using the swarm intelligence optimization strategy in step S1 is: Step S11, initializing the Daqu fermentation controller; wherein the Daqu fermentation controller includes a front-end network and a back-end network; Assume that there are m individuals in the D-dimensional search space, and the specific meaning of each individual is the parameter combination of the antecedent network and the consequent network; the fitness of each individual in the swarm intelligence optimization strategy is the difference between the start and stop time of the actuator actually output and the expected start and stop time, and the position and speed of the individual are updated according to the fitness of each individual; wherein, the speed and position of the individual are updated by the individual extreme value and the group extreme value using the following formula: ; ; In the formula, It represents the speed of the i-th individual in the k+1th iteration of the swarm intelligence optimization strategy in the j-th dimension in the D-dimensional search space; It represents the speed of the kth iteration of the swarm intelligence optimization strategy of the i-th individual in the j-th dimension in the D-dimensional search space; represents the inertia weight; c1 and c2 represent acceleration constants; r1 and r2 represent random numbers in the interval [0, 1]; It represents the optimal position of the i-th individual in the k-th iteration of the swarm intelligence optimization strategy on the j-th dimension in the D-dimensional search space; Represents the position of the i-th individual at the k-th iteration of the swarm intelligence optimization strategy on the j-th dimension in the D-dimensional search space; It represents the position of the i-th individual at the k+1th iteration of the swarm intelligence optimization strategy on the j-th dimension in the D-dimensional search space; It represents the global optimal position of the i-th individual in the k-th iteration of the swarm intelligence optimization strategy in the D-dimensional search space; Step S12, presetting the maximum number of iterations of the swarm intelligence optimization strategy, and evaluating the fitness of each individual after each iteration; after reaching the preset maximum number of iterations, terminating the iteration and executing the next step; wherein the smaller the fitness, the closer to the optimal solution; Step S13, presetting the maximum number of iterations of the Daqu fermentation controller, selecting the optimal value of the individual as the combined parameter weight of the antecedent network and the consequent network and importing it into the Daqu fermentation controller; after reaching the preset maximum number of iterations, terminating the iteration to obtain the optimized Daqu fermentation controller.

[0046] Figure 5 FIG. 2 is a schematic diagram of the processing flow of temperature detection data or humidity detection data in Example 2. Figure 5 As shown in , in step S2 of this embodiment, before performing the fusion process, the collected temperature detection data or the humidity detection data is firstly subjected to consistency detection to improve the accuracy of data fusion.

[0047] More specifically, the specific process of performing consistency detection on the temperature detection data in step S2 is: determine one by one whether the difference between the temperature detection data currently collected by the same temperature sensor and the temperature detection data collected at the previous moment is less than or equal to the corresponding temperature monitoring threshold; if it is less than or equal to the corresponding temperature monitoring threshold, perform subsequent fusion processing on the temperature detection data currently collected by the same temperature sensor; if it is greater than the corresponding temperature monitoring threshold, take the temperature detection average value of all the temperature detection data collected by the same temperature sensor to perform subsequent fusion processing; The specific process of performing consistency detection on the humidity detection data in step S2 is: determine one by one whether the difference between the humidity detection data currently collected by the same humidity sensor and the humidity detection data collected at the previous moment is less than or equal to the corresponding humidity monitoring threshold; if it is less than or equal to the corresponding humidity monitoring threshold, perform subsequent fusion processing on the humidity detection data currently collected by the same humidity sensor; if it is greater than the corresponding humidity monitoring threshold, take the humidity detection average value of all the humidity detection data collected by the same humidity sensor to perform subsequent fusion processing.

[0048] For example, in the koji room in Example 1, a total of 15 temperature sensors are set, and each temperature sensor detects the environment every 10 seconds during the fermentation process of Daqu, forming a continuous temperature detection data. The method for determining its consistency is to determine whether the difference between the latest detection value of the same temperature sensor and the detection value at the previous moment is less than or equal to the temperature monitoring threshold. Similarly, the detection data of the 15 humidity sensors set in the koji room are also executed in a similar manner.

[0049] In this embodiment, in step S2, the temperature detection data is subjected to data fusion processing, and the actual temperature value of the fused qufang is obtained by the following formula: ; ; ; In the formula, Indicates the actual temperature value of the fused qufang; Indicates the latest temperature detection data of the p-th temperature sensor among the n temperature sensors or the temperature detection average value of all temperature detection data of the p-th temperature sensor; represents the variance of the temperature detection data of the pth temperature sensor among the n temperature sensors; Represents the average value of the temperature detection data of n temperature sensors; Similarly, when the fusion processing of the humidity detection data is performed in step S2, reference is made to the fusion processing process of the temperature detection data.

[0050] In this embodiment, when learning and training are performed in the Daqu fermentation controller in step S3, the back propagation algorithm is used to change the weights of the second layer of the successor network in the Daqu fermentation controller, and the center value and width of the Gaussian membership function used in the second layer of the predecessor network in the Daqu fermentation controller are adjusted, so that the Daqu fermentation controller can stably track the optimization target value.

[0051] Furthermore, the specific process of learning and training in the Daqu fermentation controller in step S3 is as follows: The learning algorithm for the weights of the second layer of the subsequent network is: ; ; Where E is the loss function; is the weight change between the kth output and the ith input in the pth calculation rule in the consequent network; is the output of the pth calculation rule in the consequent network; is the output of the kth neuron node in the pth calculation rule in the consequent network; is the expected output of the pth calculation rule in the consequent network; is the membership degree of the jth fuzzy set; is the component of the i-th input vector; r is the learning rate; l is the iteration number of the Daqu fermentation controller; In parameter value After determination, the influence of the error on each layer in the previous network is checked, and the derived formula is as follows: ; ; ; In the formula, represents the error term of the i-th neuron node in the fifth layer of the antecedent network; the fifth layer of the antecedent network consists of the output layer; Represents the error term of the kth neuron node in the fourth layer of the antecedent network; Represents the error term of the kth neuron node in the third layer of the antecedent network; t i represents the expected output of the i-th output node; y i Represents the actual output of the i-th output node; y ik Represents the connection weight between the kth neuron node in the fourth layer of the antecedent network and the ith neuron node in the fifth layer; a i represents the membership degree of the i-th input variable; Finally, we get: ; ; Where E is the loss function; Represents the error term of the i-th neuron node in the k-th layer; x i represents the value of the i-th input variable; c ik represents the central value of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network; c jk represents the central value of the k-th Gaussian membership function corresponding to the j-th input used in the second layer of the antecedent network; q ik represents the width of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network; The central value c of the k-th Gaussian membership function corresponding to the i-th input used in the second layer of the antecedent network ik and width value q ik The learning rule is: ; ; In the formula, represents the central value of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network during the l+1th Daqu fermentation controller iteration; represents the central value of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network during the l-th iteration of the Daqu fermentation controller; represents the width of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network during the l+1th Daqu fermentation controller iteration; represents the width of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network during the l-th iteration of the Daqu fermentation controller; r is the learning rate; E is the loss function.

Claims

1. A Daqu fermentation system, characterized in that: include: Qufang; Actuators, the actuators are dispersedly arranged in the curved room; wherein the actuators include a heater, a humidifier, an electric window, an inner circulation fan and an outer circulation fan; Temperature sensors are dispersedly arranged in the qu room to collect temperature detection data at different monitoring points in the qu room; Humidity sensors are dispersedly arranged in the qu room to collect humidity detection data at different monitoring points in the qu room; A Daqu fermentation controller, the Daqu fermentation controller is constructed based on fuzzy logic control algorithm and neural network fusion, and is optimized by swarm intelligence optimization strategy; the optimized Daqu fermentation controller uses an adaptive weighted fusion algorithm based on deep learning to perform data fusion processing on the temperature detection data and humidity detection data, and then outputs the temperature deviation and temperature deviation change rate between the actual temperature value of the qu room and the target temperature value, as well as the humidity deviation and humidity deviation change rate between the actual humidity value of the qu room and the target humidity value as input signals, and outputs the opening time of each of the actuators as output signals; wherein the temperature deviation change rate is the ratio of the temperature deviation to the target temperature value, and the humidity deviation change rate is the ratio of the humidity deviation to the target humidity value; The host computer receives all the output signals and converts the output signals into corresponding Boolean signals one by one, and then the host computer controls the execution start or stop operation of the corresponding actuator according to the Boolean signals.

2. The Daqu fermentation system according to claim 1, characterized in that: The Daqu fermentation controller includes a front-end network and a back-end network; The first layer of the antecedent network is composed of an input layer, and each node in the input layer is connected to each component of an input vector and transmitted to the next layer; the input vector is the input signal; The second layer of the antecedent network is composed of a fuzzification layer, and the fuzzification layer fuzzifies the data of the input layer through a Gaussian membership function; The third layer of the antecedent network is composed of a rule layer, and the rule layer is composed of 49*49 neuron nodes, and each neuron node of the rule layer represents a fuzzy rule; The fourth layer of the antecedent network is composed of a normalization layer, which is used to perform normalization calculations and serves as the connection weight of the third layer of the consequent network; the number of neuron nodes in the normalization layer is 49*49; The first layer of the subsequent network consists of the input layer and transmits the input vector to the next layer; The second layer of the consequent network is composed of a rule calculation layer, which is composed of 49*49 neuron nodes, each of which represents a calculation rule; the rule calculation layer is used to calculate the consequent of the fuzzy rule; The third layer of the subsequent network consists of an output layer, which performs weighted summation of the output of the normalization layer and the output of the rule calculation layer.

3. The Daqu fermentation system according to claim 1, characterized in that: The host computer also has a visualization module; the visualization module is used for visual display of chart data.

4. A method for controlling Daqu fermentation, characterized in that: The following steps are involved: Step S1, constructing a Daqu fermentation controller based on the fusion of fuzzy logic control algorithm and neural network, and optimizing the Daqu fermentation controller by adopting swarm intelligence optimization strategy; Step S2, collecting temperature detection data and humidity detection data from different monitoring points in the Qu room, and using an adaptive weighted fusion algorithm based on deep learning to perform data fusion processing on the temperature detection data and the humidity detection data, to obtain the fused actual temperature value of the Qu room and the actual humidity value of the Qu room, and then outputting the temperature deviation and the temperature deviation change rate between the actual temperature value of the Qu room and the target temperature value, and outputting the humidity deviation and the humidity deviation change rate between the actual humidity value of the Qu room and the target humidity value; wherein the temperature deviation change rate is the ratio of the temperature deviation to the target temperature value, and the humidity deviation change rate is the ratio of the humidity deviation to the target humidity value; Step S3, using the temperature deviation and the temperature deviation change rate, as well as the humidity deviation and the humidity deviation change rate as input signals to be introduced into the optimized Daqu fermentation controller for learning and training, and outputting the opening time of each actuator as an output signal; wherein the actuator includes a heater, a humidifier, an electric window, an internal circulation fan and an external circulation fan; Step S4, the host computer receives all the output signals and converts the output signals into corresponding Boolean signals one by one, and then the host computer controls the corresponding actuator to start or stop according to the Boolean signals.

5. The Daqu fermentation control method according to claim 4, characterized in that: The specific process of optimizing the Daqu fermentation controller by using the swarm intelligence optimization strategy in step S1 is as follows: Step S11, initializing the Daqu fermentation controller; wherein the Daqu fermentation controller includes a front-end network and a back-end network; Assume that there are m individuals in the D-dimensional search space, and the specific meaning of each individual is the parameter combination of the antecedent network and the consequent network; the fitness of each individual in the swarm intelligence optimization strategy is the difference between the start and stop time of the actuator actually output and the expected start and stop time, and the position and speed of the individual are updated according to the fitness of each individual; wherein, the speed and position of the individual are updated by the individual extreme value and the group extreme value using the following formula: ; ; In the formula, It represents the speed of the i-th individual in the k+1th iteration of the swarm intelligence optimization strategy in the j-th dimension in the D-dimensional search space; It represents the speed of the kth iteration of the swarm intelligence optimization strategy of the i-th individual in the j-th dimension in the D-dimensional search space; represents the inertia weight; c1 and c2 represent acceleration constants; r1 and r2 represent random numbers in the interval [0, 1]; It represents the optimal position of the i-th individual in the k-th iteration of the swarm intelligence optimization strategy on the j-th dimension in the D-dimensional search space; Represents the position of the i-th individual at the k-th iteration of the swarm intelligence optimization strategy on the j-th dimension in the D-dimensional search space; It represents the position of the i-th individual at the k+1th iteration of the swarm intelligence optimization strategy on the j-th dimension in the D-dimensional search space; It represents the global optimal position of the i-th individual in the k-th iteration of the swarm intelligence optimization strategy in the D-dimensional search space; Step S12, presetting the maximum number of iterations of the swarm intelligence optimization strategy, and evaluating the fitness of each individual after each iteration; after reaching the preset maximum number of iterations, terminating the iteration and executing the next step; Step S13, presetting the maximum number of iterations of the Daqu fermentation controller, selecting the optimal value of the individual as the combined parameter weight of the antecedent network and the consequent network and importing it into the Daqu fermentation controller; after reaching the preset maximum number of iterations, terminating the iteration to obtain the optimized Daqu fermentation controller.

6. The Daqu fermentation control method according to claim 5, characterized in that: In the step S2, before performing the fusion process, the collected temperature detection data or the collected humidity detection data is firstly subjected to a consistency check.

7. The Daqu fermentation control method according to claim 6, characterized in that: The specific process of performing consistency detection on the temperature detection data in step S2 is: Determine one by one whether the difference between the temperature detection data currently collected by the same temperature sensor and the temperature detection data collected at the last moment is less than or equal to the corresponding temperature monitoring threshold; If it is less than or equal to the corresponding temperature monitoring threshold, performing subsequent fusion processing on the temperature detection data currently collected by the same temperature sensor; If it is greater than the corresponding temperature monitoring threshold, taking the temperature detection average value of all the temperature detection data collected by the same temperature sensor to perform subsequent fusion processing; The specific process of performing consistency detection on the humidity detection data in step S2 is: Determine one by one whether the difference between the humidity detection data currently collected by the same humidity sensor and the humidity detection data collected at the last moment is less than or equal to the corresponding humidity monitoring threshold; If it is less than or equal to the corresponding humidity monitoring threshold, performing subsequent fusion processing on the humidity detection data currently collected by the same humidity sensor; If it is greater than the corresponding humidity monitoring threshold, the humidity detection average value of all the humidity detection data collected by the same humidity sensor is taken to perform subsequent fusion processing.

8. The Daqu fermentation control method according to claim 7, characterized in that: In step S2, the temperature detection data is subjected to data fusion processing, and the actual temperature value of the fused room is Obtained by the following formula: ; ; ; In the formula, Indicates the actual temperature value of the fused qufang; Indicates the latest temperature detection data of the p-th temperature sensor among the n temperature sensors or the temperature detection average value of all temperature detection data of the p-th temperature sensor; represents the variance of the temperature detection data of the pth temperature sensor among the n temperature sensors; Represents the average value of the temperature detection data of n temperature sensors; Similarly, when the fusion processing of the humidity detection data is performed in step S2, reference is made to the fusion processing process of the temperature detection data.

9. The Daqu fermentation control method according to claim 8, characterized in that: When learning and training are performed in the Daqu fermentation controller in step S3, a back propagation algorithm is used to change the weights of the second layer of the successor network in the Daqu fermentation controller, and the center value and width of the Gaussian membership function used in the second layer of the predecessor network in the Daqu fermentation controller are adjusted, so that the Daqu fermentation controller can stably track the optimization target value.

10. The Daqu fermentation control method according to claim 9, characterized in that: The specific process of learning and training in the Daqu fermentation controller in step S3 is as follows: The learning algorithm for the weights of the second layer of the subsequent network is: ; ; Where E is the loss function; is the weight change between the kth output and the ith input in the pth calculation rule in the consequent network; is the output of the pth calculation rule in the consequent network; is the output of the kth neuron node in the pth calculation rule in the consequent network; is the expected output of the pth calculation rule in the consequent network; is the membership degree of the jth fuzzy set; is the component of the i-th input vector; r is the learning rate; l is the iteration number of the Daqu fermentation controller; In parameter value After determination, the influence of the error on each layer in the previous network is checked, and the derived formula is as follows: ; ; ; In the formula, Represents the error term of the i-th neuron node in the fifth layer of the antecedent network; The fifth layer of the antecedent network consists of the output layer; Represents the error term of the kth neuron node in the fourth layer of the antecedent network; Represents the error term of the kth neuron node in the third layer of the antecedent network; t i represents the expected output of the i-th output node; y i Represents the actual output of the i-th output node; y ik Represents the connection weight between the kth neuron node in the fourth layer of the antecedent network and the ith neuron node in the fifth layer; a i represents the membership degree of the i-th input variable; Finally, we get: ; ; Where E is the loss function; Represents the error term of the i-th neuron node in the k-th layer; x i represents the value of the i-th input variable; c ik represents the central value of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network; c jk represents the central value of the k-th Gaussian membership function corresponding to the j-th input used in the second layer of the antecedent network; q ik represents the width of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network; The central value c of the k-th Gaussian membership function corresponding to the i-th input used in the second layer of the antecedent network ik and width value q ik The learning rule is: ; ; In the formula, represents the central value of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network during the l+1th Daqu fermentation controller iteration; represents the central value of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network during the l-th iteration of the Daqu fermentation controller; represents the width of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network during the l+1th Daqu fermentation controller iteration; represents the width of the k-th Gaussian membership function corresponding to the ith input used in the second layer of the antecedent network during the l-th iteration of the Daqu fermentation controller; r is the learning rate; E is the loss function.

Images