Method, device, equipment, storage medium and program product for controlling soy sauce fermentation

Abstract

The application discloses a soy sauce fermentation control method, device, equipment, storage medium and program product, relates to the automatic control technical field, and comprises the following steps: collecting multi-source metabolic parameters of a soy sauce fermentation system according to a preset period to form a time sequence data sequence; determining a current fermentation stage of the soy sauce fermentation system based on change information of at least one metabolic parameter in the time sequence data sequence; inputting the time sequence data sequence and the identifier of the current fermentation stage into a pre-trained prediction model to obtain a starting probability of the soy sauce fermentation system entering a next fermentation stage and a starting time window of the next fermentation stage; determining a target temperature control interval of the soy sauce fermentation system based on the current fermentation stage, the starting probability and the starting time window, and controlling the temperature of the soy sauce fermentation system according to the target temperature control interval. The application improves the fermentation quality of soy sauce.

Description

Technical Field

[0001] This application relates to the field of automation control technology, and in particular to control methods, devices, equipment, storage media, and program products for soy sauce fermentation. Background Technology

[0002] Currently, the automation control of soy sauce fermentation mainly adopts technical solutions based on distributed control systems or programmable logic controllers. These automation systems transform manual inspection operations into automatic instrument control, reducing the temperature fluctuation range inside the tank and lowering the steam consumption per tank, thus becoming a standardized control mode widely adopted in the industry.

[0003] However, this control mode is essentially still a program control based on a fixed time schedule. The control system only drives the measured temperature to approach the preset static setpoint. When there are batch fluctuations in raw material composition, bacterial activity or environmental conditions, the system still mechanically executes the original time-temperature curve, which can easily lead to process mismatches such as premature cooling before the enzymatic reaction is fully completed or maintaining high temperature even after the microbial metabolism has declined, thus affecting product quality. Summary of the Invention

[0004] The main objective of this application is to provide a method, apparatus, equipment, storage medium, and program product for controlling soy sauce fermentation, aiming to solve the technical problem that controlling the fermentation temperature of soy sauce based on a fixed timetable leads to an impact on product quality.

[0005] To achieve the above objectives, this application proposes a method for controlling soy sauce fermentation, the method comprising: Multi-source metabolic parameters of the soy sauce fermentation system are collected according to a preset period to form a time-series data sequence, wherein the multi-source parameters are used to characterize the fermentation state of the soy sauce fermentation system. Based on the change information of at least one metabolic parameter in the time series data sequence, the current fermentation stage of the soy sauce fermentation system is determined; The time series data sequence and the identifier of the current fermentation stage are input into the pre-trained prediction model to obtain the start probability of the soy sauce fermentation system entering the next fermentation stage and the start time window of the next fermentation stage. The prediction model is trained with the time series data sequence of the historical fermentation system before the stage switch as input data and the start probability and start time window corresponding to the stage switch as labels. Based on the current fermentation stage, the start probability, and the start time window, the target temperature control range of the soy sauce fermentation system is determined, and the temperature of the soy sauce fermentation system is controlled according to the target temperature control range.

[0006] Furthermore, to achieve the above objectives, this application also proposes a control device for soy sauce fermentation, the control device comprising: The acquisition module is used to acquire multi-source metabolic parameters of the soy sauce fermentation system according to a preset period to form a time-series data sequence, wherein the multi-source parameters are used to characterize the fermentation state of the soy sauce fermentation system; The determination module is used to determine the current fermentation stage of the soy sauce fermentation system based on the change information of at least one metabolic parameter in the time-series data sequence. The prediction module is used to input the time series data sequence and the identifier of the current fermentation stage into a pre-trained prediction model to obtain the start probability of the soy sauce fermentation system entering the next fermentation stage and the start time window of the next fermentation stage. The prediction model is trained using the time series data sequence of the historical fermentation system before the stage switch as input data and the start probability and start time window corresponding to the stage switch as labels. The control module is used to determine the target temperature control range of the soy sauce fermentation system based on the current fermentation stage, the start probability, and the start time window, and to control the temperature of the soy sauce fermentation system according to the target temperature control range. In addition, to achieve the above objectives, this application also proposes a control device for soy sauce fermentation, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the soy sauce fermentation control method as described above.

[0007] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the soy sauce fermentation control method described above.

[0008] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the soy sauce fermentation control method described above.

[0009] This application constructs a continuous monitoring capability for the biochemical processes within the fermentation system by collecting multi-source metabolic parameters according to a preset cycle and forming a time-series data sequence. This replaces the control input that relies solely on comparing a single temperature variable with a preset time curve in the traditional mode. This enables the control system to acquire multi-dimensional information on key states such as enzyme activity and microbial metabolic intensity, providing a data foundation for solving the problem of the disconnect between control commands and the actual fermentation state.

[0010] Secondly, by determining the current fermentation stage based on the changes in metabolic parameters in the time-series data, objective and dynamic identification of the fermentation process is achieved. Traditional fixed-time control implicitly assumes that the same time corresponds to the same fermentation stage, while this application, by analyzing parameter change characteristics in real time, can accurately identify the actual physiological stage of fermentation, such as the main fermentation period and the post-ripening period, so that the control system directly responds to the intrinsic biochemical processes of the fermentation system.

[0011] By inputting time-series data sequences and current fermentation stage identifiers into a pre-trained prediction model, the probability of entering the next fermentation stage and the start-up time window are obtained. This introduces an advanced prediction function for the evolution of the fermentation process. The prediction model is trained based on historical fermentation data and can learn the correlation between the change patterns of multi-source parameters before different stage transitions and the timing of stage transitions. This allows the system to not only understand the current state but also predict the possible recent stage transition trends and time ranges, providing a forward-looking basis for control decisions.

[0012] Ultimately, the target temperature control range is dynamically determined based on the current fermentation stage, the predicted start-up probability, and the time window, and temperature control is implemented, forming a closed-loop adaptive regulation. The setting of the temperature control range is no longer static and pre-fixed, but is adjusted in real time according to the actual fermentation process and predicted transition points, thereby ensuring that the temperature conditions always match the immediate enzymatic reaction requirements or microbial metabolic activity of the system, avoiding mismatches such as cooling down before full fermentation or maintaining high temperatures after degradation.

[0013] Therefore, by constructing an adaptive control system that can respond to changes in the internal state of the fermentation system, this application cuts off the dependence of temperature control on absolute time and instead relies on metabolic parameters that reflect the biochemical process and their evolution. Thus, even when there are batch fluctuations in raw materials, strains, or the environment, the temperature control strategy can still be synchronized with the actual fermentation physiological state, thereby improving product quality. Attached Figure Description

[0014] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A schematic flowchart of the control method for soy sauce fermentation provided in Embodiment 1 of this application; Figure 2This is a flowchart illustrating Embodiment 2 of the soy sauce fermentation control method of this application. Figure 3 This is a schematic diagram of the model structure involved in an embodiment of the soy sauce fermentation control method of this application; Figure 4 This is a flowchart illustrating Embodiment 3 of the soy sauce fermentation control method of this application. Figure 5 This is a simplified flowchart illustrating an embodiment of the soy sauce fermentation control method of this application; Figure 6 This is a schematic diagram of the module structure involved in an embodiment of the soy sauce fermentation control method of this application; Figure 7 This is a schematic diagram of the module structure of the control device for soy sauce fermentation according to an embodiment of this application; Figure 8 This is a schematic diagram of the hardware operating environment involved in the soy sauce fermentation control method in the embodiments of this application.

[0017] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0019] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0020] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or soy sauce fermentation control device capable of the above functions. The following description uses a soy sauce fermentation control device as an example to illustrate this embodiment and the subsequent embodiments.

[0021] Based on this, the embodiments of this application provide a method for controlling soy sauce fermentation, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the soy sauce fermentation control method of this application. In this embodiment, the soy sauce fermentation control method includes steps S10 to S40: Step S10: Collect multi-source metabolic parameters of the soy sauce fermentation system according to a preset cycle to form a time-series data sequence, wherein the multi-source parameters are used to characterize the fermentation state of the soy sauce fermentation system; The soy sauce fermentation system refers to the microbial fermentation process used to produce soy sauce and its physicochemical environment. It typically includes the fermentation mash, microbial community, fermentation tank, and related environmental parameters. Multi-source metabolic parameters are various physical, chemical, or biological indicators that reflect microbial physiological activity and changes in mash composition during soy sauce fermentation. The dynamic changes of these parameters can comprehensively characterize the overall state of the fermentation system. Time-series data sequences are a series of data points formed by continuously collecting multi-source metabolic parameters at preset time intervals. These data points are arranged in chronological order and can show the trend of parameter changes over time.

[0022] In practice, parameters can be collected in various ways. For example, samples can be taken manually at regular intervals and tested using laboratory analytical equipment to obtain parameters such as total acidity and reducing sugar content. These data can then be manually entered into the system to form a time-series data sequence. Alternatively, basic online sensors can be deployed, such as those used to measure the conductivity or turbidity of the fermentation broth. These sensors output signals at a fixed frequency, which are received and stored by the data acquisition module, thereby constructing a time-series data sequence that reflects the dynamic changes in the fermentation process.

[0023] Step S20: Determine the current fermentation stage of the soy sauce fermentation system based on the change information of at least one metabolic parameter in the time-series data sequence; In one implementation, a simple threshold rule can be set to identify the stage. For example, when the value of a key parameter first reaches or exceeds a preset threshold, the system determines that the fermentation system has moved from the initial stage to the next stage. In another feasible implementation, the stage can also be identified based on the changing trend of at least one metabolic parameter in the time-series data sequence. For example, when the value of a key parameter continues to decrease, the system determines that the fermentation system has moved from the initial stage to the next stage.

[0024] Step S30: Input the time series data sequence and the identifier of the current fermentation stage into the pre-trained prediction model to obtain the start probability of the soy sauce fermentation system entering the next fermentation stage and the start time window of the next fermentation stage. The prediction model is trained using the time series data sequence of the historical fermentation system before the stage switch as input data and the start probability and start time window corresponding to the stage switch as labels.

[0025] The pre-trained predictive model is a mathematical model trained and optimized using machine learning algorithms based on a large amount of historical fermentation data. This model is capable of recognizing fermentation patterns and predicting future events. It is trained and possesses predictive capabilities before practical application. The initiation probability is the likelihood, expressed numerically, of the soy sauce fermentation system entering the next fermentation stage at a specific time point, typically between 0 and 1. The initiation time window, provided by the predictive model, represents the time range within which the soy sauce fermentation system is most likely to enter the next fermentation stage, providing temporal guidance for subsequent control decisions.

[0026] By inputting the time-series data sequence and the identifier of the current fermentation stage into a pre-trained prediction model, the probability of the soy sauce fermentation system entering the next fermentation stage and the start time window of the next fermentation stage are obtained. This prediction model has been trained with a large amount of historical fermentation data before being put into use. For example, data from a large number of successful fermentation batches can be collected, the parameter sequence before each stage switch can be recorded, and the actual switch time and corresponding confidence level can be marked. In practical applications, the currently collected time-series data sequence, along with the information of the current fermentation stage identified by the system, can be directly input into a rule-based expert system or a simple statistical model. This model compares the current data with historical patterns and outputs a rough estimate and probability of when the next stage will start.

[0027] Step S40: Based on the current fermentation stage, start probability, and start time window, determine the target temperature control range of the soy sauce fermentation system, and control the temperature of the soy sauce fermentation system according to the target temperature control range.

[0028] The target temperature control range is a suitable temperature range set for the soy sauce fermentation system based on the current fermentation stage, the predicted start-up probability, and the start-up time window, in order to optimize microbial activity and biochemical reaction processes.

[0029] In this embodiment, based on the current fermentation stage, start probability, and start time window, the target temperature control range of the soy sauce fermentation system is determined, and the temperature of the soy sauce fermentation system is maintained within the target temperature control range. Specifically, a fixed temperature range can be preset for each fermentation stage. When the system determines that it is currently in a certain stage, and the start probability given by the prediction model reaches a certain level and is within the start time window, the system will directly switch to the preset temperature range corresponding to the next stage. For example, if the current stage is protein hydrolysis, the preset temperature is 30-32℃; when the prediction model indicates that the flavor generation stage is about to begin, the system will adjust the target temperature control range to 35-37℃ and maintain the temperature in the fermentation tank within this range through heating or cooling devices.

[0030] Understandably, traditional soy sauce fermentation control methods primarily rely on preset time-temperature curves for mechanical control. Regardless of the actual metabolic state of the fermentation system, the system adjusts the temperature at fixed time points. In contrast, this embodiment achieves dynamic sensing of the actual state of the fermentation system by real-time acquisition of multi-source metabolic parameters. By monitoring parameters such as pH and carbon dioxide concentration, the system can accurately identify the current fermentation stage, rather than simply relying on the passage of time. This real-time stage determination capability allows the temperature control strategy to be closely coupled with the actual metabolic process of the microorganisms.

[0031] Furthermore, this embodiment introduces a pre-trained predictive model that can proactively predict the timing of fermentation stage transitions and provide start-up probabilities and start-up time windows. In the example, the system not only knows that it is currently in the flavor generation stage but can also predict when it will enter the maturation stage. This predictive capability allows the control system to plan and adjust temperature strategies in advance, avoiding process mismatches caused by delayed responses in traditional methods. It can provide early warnings and adjust temperatures to ensure that stage switching occurs at the optimal time. Thus, by adaptively coupling temperature control with dynamically changing microbial metabolic inflection points, this embodiment can solve problems such as poor batch-to-batch consistency of product quality, long fermentation cycles, and low energy efficiency.

[0032] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2 At least one metabolic parameter includes pH and carbon dioxide concentration; step S20, determining the current fermentation stage of the soy sauce fermentation system based on the change information of at least one metabolic parameter in the time-series data sequence, includes: Step S201: Determine the flavor maturity index based on pH value, wherein the flavor maturity index is used to characterize the flavor substance generation state of the soy sauce fermentation system. pH is an indicator of the acidity or alkalinity of a solution. During soy sauce fermentation, the metabolic activities of microorganisms produce organic acids, causing the pH to drop, making it an important indicator of the fermentation process. Carbon dioxide concentration reflects the respiration intensity and metabolic activity of microorganisms, directly reflecting the vigorousness of fermentation. pH and carbon dioxide concentrations can be monitored in real-time by online sensors, transmitting the data to the control system. Alternatively, samples can be taken periodically and analyzed using laboratory instruments to obtain pH and carbon dioxide concentration data.

[0033] The flavor maturity index is a comprehensive indicator designed to quantify the generation and accumulation of flavor compounds during soy sauce fermentation. Specifically, it combines multiple key parameters such as pH value, amino acid nitrogen concentration, and 4-ethylguaiacol concentration to more comprehensively and accurately reflect the degree of flavor formation in soy sauce. The index can be calculated using a preset mathematical model or empirical formula. For example, a weighted average method can be used, assigning different weights to pH value, amino acid nitrogen concentration, and 4-ethylguaiacol concentration according to their importance to flavor contribution. Alternatively, a machine learning model can be used, taking pH value, amino acid nitrogen concentration, and 4-ethylguaiacol concentration from historical fermentation data as input and using the flavor maturity level assessed by experts as a label for training, thereby predicting the current flavor maturity index.

[0034] Step S202: Calculate the rate of change of flavor maturity index and the rate of decrease of pH value, and calculate the carbon dioxide release rate based on carbon dioxide concentration; The rate of change of the flavor maturity index, the rate of decrease in pH, and the rate of carbon dioxide release are dynamic indicators used to capture the trends and intensity of parameters during fermentation. The rate of change of the flavor maturity index reflects the speed of flavor compound formation; the rate of decrease in pH indicates the rate of acidic substance accumulation; and the rate of carbon dioxide release quantifies the intensity of microbial respiration. Dynamic indicators are more sensitive to the transitions between fermentation stages than static values. These rates can be obtained by differential calculation or sliding window averaging of time-series data. For example, the rate of change of the flavor maturity index can be calculated as the difference between the index at the current moment and the previous preset time interval divided by the time interval. Alternatively, curve fitting can be used to fit time-series data over a period of time, and then the instantaneous or average rate of change can be obtained by differentiation.

[0035] Step S203: If the rate of change of the flavor maturity index is greater than the first threshold, then the current fermentation stage is determined to be the protein hydrolysis period. When the rate of change of the flavor maturity index rises rapidly and exceeds the first threshold, it indicates that the protein hydrolysis and the generation of flavor precursors are vigorous, and the fermentation system is identified as being in the protein hydrolysis phase.

[0036] Step S204: If the rate of change of the flavor maturity index is less than or equal to the first threshold, the rate of decrease of pH value is less than the second threshold, and the carbon dioxide release rate is greater than the third threshold, then the current fermentation stage is determined to be the flavor generation period. When the rate of change of the flavor maturity index slows down, while the rate of pH decreases and the rate of carbon dioxide release remains at a high level, it indicates that the synthesis of flavor substances has entered an active stage, and the fermentation system is identified as the flavor generation period at this time.

[0037] Step S205: If the rate of change of the flavor maturity index is less than the fourth threshold, then the current fermentation stage is determined to be the maturation period, wherein the first threshold is greater than the fourth threshold.

[0038] When the rate of change of the flavor maturity index decreases below the fourth threshold, it indicates that fermentation has entered the stable and aging stage, and is identified as the maturation period.

[0039] The first, second, third, and fourth thresholds are preset critical values ​​used to distinguish different fermentation stages. These thresholds can be determined based on extensive historical fermentation data and expert experience, representing typical boundaries of parameter changes between different fermentation stages. By comparing these thresholds, the system can automatically determine the specific stage of fermentation currently in operation. Thresholds can be established through statistical analysis of historical fermentation batch data, such as cluster analysis or decision tree algorithms, to identify the boundary points of parameter changes between different fermentation stages. Alternatively, these thresholds can be manually set and adjusted using expert systems or rule-based inference engines, combined with the experience and knowledge of fermentation engineers, and iteratively optimized in practical applications.

[0040] The protein hydrolysis stage, flavor formation stage, and maturation stage are typical phases in the soy sauce fermentation process, each with its unique biochemical reactions and microbial activity characteristics. The protein hydrolysis stage primarily involves protein decomposition; the flavor formation stage focuses on the synthesis of flavor compounds; and the maturation stage emphasizes the stabilization and maturation of flavor compounds. By using the aforementioned parameters and thresholds, the fermentation process is dynamically categorized into these predefined stages.

[0041] In this embodiment, the current fermentation stage is determined by multiple source parameters, which can overcome the limitations of single parameter or static parameter determination and achieve accurate and dynamic identification of the fermentation stage of soy sauce. The accurate stage identification provides a solid foundation for subsequent temperature control based on prediction models, enabling temperature regulation to be closely coupled with the actual metabolic activities of microorganisms, thereby optimizing the fermentation process.

[0042] Step S30, which involves inputting the time-series data sequence and the identifier of the current fermentation stage into a pre-trained prediction model to obtain the initiation probability of the soy sauce fermentation system entering the next fermentation stage and the initiation time window of the next fermentation stage, includes: Step S301: Determine the input data sequence from the time-series data sequence, wherein the input data sequence is the data at the current time and the data at a preset time length before the current time; From the time-series data sequence formed by continuously collected multi-source metabolic parameters of the soy sauce fermentation system, a segment of data containing the current moment and a preset time length is extracted to form an input data sequence focusing on the recent fermentation state. This allows for the selection of a representative, fixed-length data segment as model input, focusing on the recent trends in fermentation state and filtering out premature or irrelevant historical data, thereby improving the model's real-time performance and prediction accuracy. Specifically, this can be achieved through a sliding window mechanism, extracting data points containing the current moment and N time steps prior to it from the complete time-series data sequence at each prediction moment, forming an input sequence of length N+1; alternatively, the input data sequence can be dynamically queried from a database based on a preset time length, without any restrictions.

[0043] Step S302: Input the input data sequence into the feature extraction layer to obtain a high-dimensional feature vector; The input data sequence is then fed into the feature extraction layer of the prediction model. The filtered and truncated input data sequence is fed into the feature extraction layer of the prediction model for processing. Its function is to use the powerful learning ability of the feature extraction layer to automatically learn from the original time series data and generate a high-dimensional feature vector containing rich semantic information. This vector can effectively represent the dynamic changes of the current and recent fermentation state.

[0044] Specifically, the process can be as follows: the input data sequence can be directly used as the input of a recurrent neural network. The LSTM (Long Short-Term Memory) unit will gradually update its internal state when processing the sequence data, and the final output hidden state or cell state can be used as a high-dimensional feature vector; or, the input data sequence can be processed through a one-dimensional convolutional layer to perform multiple convolution and pooling operations, and the resulting feature map can be flattened and used as a high-dimensional feature vector.

[0045] Step S303: Input the identifier of the current fermentation stage into the stage coding layer to obtain the stage coding vector; The identifier of the current fermentation stage is input into the stage encoding layer of the prediction model. Its role is to transform this discrete, non-numerical stage information into a continuous numerical vector that the model can understand and process, namely the stage encoding vector, thereby providing the model with important contextual information. Specifically, the stage identifier can be directly obtained by looking up the corresponding stage encoding vector through a pre-trained embedding layer; or, the stage identifier can be first converted into a one-hot encoded form, and then dimensionality reduction and feature learning can be performed through a fully connected layer to obtain the stage encoding vector.

[0046] Step S304: Concatenate the stage encoding vector with the high-dimensional feature vector to obtain the concatenated vector; The stage encoding vector is concatenated with the high-dimensional feature vector extracted from the time-series data to form a comprehensive concatenated vector. Combining the high-dimensional feature vector extracted from the time-series data with the stage encoding vector representing the current fermentation stage effectively integrates information from two different sources, both crucial for prediction, forming a more comprehensive and expressive concatenated vector. This allows subsequent output layers to make predictions based on richer information. Specifically, this can be achieved by directly concatenating the two vectors dimensionally.

[0047] Step S305: Input the spliced ​​vector into the output layer to obtain the initiation probability, target time, and initiation time window of the next fermentation stage; The concatenated vector, integrating temporal features and stage information, is fed into the output layer of the prediction model. The output layer then performs a final nonlinear transformation and mapping on the concatenated vector to generate key prediction results regarding the switching of the next fermentation stage. Specifically, the output layer may contain three independent fully connected sublayers, used to predict the initiation probability, target time, and initiation time window, respectively.

[0048] In one feasible embodiment, please refer to Figure 3The schematic diagram of the model structure shown illustrates that the prediction model comprises an input layer, a feature extraction layer, a stage encoding layer, and an output layer. The input layer, as the starting point of the prediction model, primarily receives external input data. This can be achieved by using preprocessed time-series data sequences as the initial input, or by directly receiving raw time-series data and converting it into a tensor format that the model can process internally. The feature extraction layer aims to automatically learn and extract deep features meaningful for the prediction task from the input data. It can employ recurrent neural networks and their variants, such as Long Short-Term Memory (LSTM) networks or Gated Recurrent Units (GRUs), to capture temporal dependencies in the time-series data; alternatively, it can use a one-dimensional convolutional neural network (1DCNN) to extract local features at different scales using convolutional kernels of varying sizes. The stage encoding layer transforms the discrete identifiers of the current fermentation stage into continuous, low-dimensional vector representations, i.e., stage-encoded vectors. This can be achieved using an embedding layer, mapping each fermentation stage to a unique vector space; or by using one-hot encoding followed by a fully connected layer, converting the high-dimensional sparse representation into a dense vector representation. The output layer, as the final part of the prediction model, generates the final prediction result based on the features and stage information learned by the model. It can consist of one or more fully connected layers combined with an appropriate activation function. Specifically, the predictive model works as follows: Input Layer: Receives multi-dimensional time-series data (temperature, pH, CO2, ORP (Oxidation-Reduction Potential)), with a 72-hour sliding window, and performs normalization processing; Feature Extraction Layer: Uses a 1D convolutional neural network to extract local features, and then uses LSTM to capture long-term dependencies; Stage Encoding Layer: Embeds the current stage label (e.g., proteolysis period) into a vector, and concatenates it with the feature vector; Output Layer: The fully connected layer outputs two results: the probability of starting the next stage (0~1) and the start time window (±2 hours).

[0049] This embodiment introduces a hierarchical prediction model structure, particularly a feature extraction layer and a stage encoding layer. The model can focus on capturing subtle dynamic trends from time-series data of multi-source metabolic parameters and transforming discrete fermentation stage information into a continuous vector representation that the model can understand. These two types of information are deeply fused through a concatenation operation, forming a comprehensive and expressive concatenated vector. This greatly enriches the input information for model prediction. The information fusion mechanism enables the prediction model to more accurately understand the complex state of the soy sauce fermentation system and make more precise predictions about the initiation probability, target time, and initiation time window of fermentation stage switching. Compared to models that rely solely on time-series data or stage information, this embodiment can more fully capture the dynamic changes and stage characteristics of the fermentation process, thereby improving the accuracy and real-time performance of predictions. This provides a reliable basis for subsequent temperature control decisions, allowing temperature control strategies to be more adaptively coupled with microbial metabolic inflection points, avoiding process mismatches caused by inaccurate predictions. This, in turn, helps improve the consistency of soy sauce product quality, shorten the fermentation cycle, and optimize energy utilization efficiency.

[0050] Based on the first and / or second embodiments of this application, in the third embodiment of this application, the content that is the same as or similar to the first and / or second embodiments described above can be referred to the above description and will not be repeated hereafter. Based on this, please refer to... Figure 4 Step S40, the step of determining the target temperature control range of the soy sauce fermentation system based on the current fermentation stage, start probability, and start time window, includes: Step S401: Obtain the preset first temperature range corresponding to the current fermentation stage, and divide the first temperature range into multiple temperature sub-intervals; The system obtains a preset first temperature range corresponding to the current fermentation stage and divides this range into multiple temperature sub-intervals. This provides an initial, explorable temperature operating range for the current fermentation stage, and further refinement of the sub-intervals provides finer granularity for subsequent temperature optimization. The preset first temperature range can be determined based on historical experience or process requirements, for example, by consulting a fermentation stage-temperature range mapping table pre-stored in a database and dividing it into multiple temperature sub-intervals at equal intervals. Alternatively, it can be dynamically generated based on the characteristics of the current fermentation stage using an expert system or rule-based inference engine, and then recursively divided into smaller temperature sub-intervals using an adaptive algorithm.

[0051] Step S402: The temperature of the soy sauce fermentation system is adjusted to each temperature sub-range in turn, and the microbial activity intensity score corresponding to each temperature sub-range is calculated. The control system can progressively adjust the fermenter temperature to the center or boundary value of each temperature sub-range in a predetermined sequence, and maintain this temperature within each sub-range for a period of time to ensure the system reaches a stable state. During this stabilization period, relevant metabolic parameters are collected, and a microbial activity intensity score is calculated based on a preset scoring model. Alternatively, more intelligent exploration strategies can be employed, such as Bayesian optimization or reinforcement learning methods, to dynamically adjust the next temperature sub-range to be explored based on the score results of the previous sub-range, thereby finding the optimal temperature more quickly.

[0052] Step S403: If the start probability is greater than or equal to the preset confidence threshold and the current time is within the start time window, then the preset second temperature range corresponding to the next fermentation stage is determined as the target temperature control range. Utilizing the reliable output of the predictive model, when the prediction results are highly reliable and the timing is appropriate, the fermentation process is directly driven into the next stage of temperature control. When the start-up probability output by the predictive model reaches or exceeds a preset confidence threshold (e.g., 0.8), and the current system time falls within the predicted start-up time window, the system retrieves the second temperature range corresponding to the next fermentation stage from the preset process parameter database and sets it as the new target temperature control interval. Furthermore, a dynamically adjusted confidence threshold can be used, for example, setting it lower in the early stages of fermentation and appropriately increasing it in the later stages.

[0053] Step S404: If the activation probability is less than the confidence threshold and / or the current time has not reached the activation time window, then the temperature sub-interval with the highest microbial activity intensity score among the various temperature sub-intervals is determined as the target temperature control interval.

[0054] If the activation probability output by the predictive model is lower than the preset confidence threshold, or if the current time has not yet entered the predicted activation time window, the system will not adopt the temperature recommendation for the next stage. Based on the microbial activity intensity scores obtained from previous exploration and evaluation of each temperature sub-interval, the system selects the temperature sub-interval with the highest score and sets it as the new target temperature control interval for the current stage. Furthermore, a certain hysteresis or smoothing mechanism can be introduced to avoid frequent and large-scale temperature adjustments.

[0055] This embodiment introduces a refined exploration of the temperature range of the current fermentation stage and a microbial activity intensity scoring mechanism. This allows the system to adaptively select the temperature sub-range most favorable to microbial activity within the current stage as the target temperature control range based on real-time feedback data when predictions are uncertain. This avoids the risk of blindly switching to the next stage temperature range under uncertain conditions, ensuring that the fermentation process operates under optimal or near-optimal temperature conditions at all times. Therefore, this embodiment improves the robustness and adaptability of the soy sauce fermentation process, helps stabilize fermentation efficiency, and ensures the consistency of the final product quality. By combining accurate identification of fermentation stages and prediction of stage transitions, this embodiment further refines the temperature control strategy, enabling it to be more closely coupled with the actual metabolic state of the microorganisms, thereby achieving a more intelligent and efficient soy sauce fermentation process.

[0056] In one feasible embodiment, step S401, the step of calculating the microbial activity intensity score corresponding to each of the temperature sub-intervals, includes: Step S4011: After determining that the temperature of the soy sauce fermentation system is within the temperature sub-range, the average temperature, average pH value, average carbon dioxide concentration and average redox potential of the soy sauce fermentation system are collected. After determining that the temperature of the soy sauce fermentation system is within a certain temperature sub-range, key parameters are collected. For example, the temperature of the fermentation system can be continuously monitored. When it remains stable within the target temperature sub-range for a period of time (e.g., more than 30 minutes), the system automatically triggers data acquisition. Alternatively, the data acquisition module can be activated after the temperature control system adjusts the temperature of the fermentation system to the target temperature sub-range and reaches a stable state.

[0057] The average temperature, average pH, average carbon dioxide concentration, and average redox potential of the soy sauce fermentation system are collected. These parameters are key indicators of microbial life activities and can comprehensively reflect the microenvironment in which the microorganisms live. By collecting average values, transient noise and short-term fluctuations can be effectively filtered out, obtaining more representative environmental data. Specifically, multiple sensors can be arranged in the fermenter, including temperature sensors, pH sensors, carbon dioxide concentration sensors, and redox potential sensors. These sensors sample data at a preset frequency (e.g., every 5 minutes) and average the sampled data within a set time window (e.g., 1 hour) to obtain the average value. Alternatively, an integrated multi-parameter probe can be used to continuously collect parameter values ​​for a period of time after the temperature stabilizes, and the control unit can calculate the average value.

[0058] Step S4012: Based on the average temperature, the average pH value, the average carbon dioxide concentration, and the average redox potential, calculate the temperature adaptation score, acid-base adaptation score, respiratory activity score, and redox adaptation score of the soy sauce fermentation system. Based on the collected average temperature, average pH, average carbon dioxide concentration, and average redox potential, the temperature adaptation score, acid-base adaptation score, respiratory activity score, and redox adaptation score of the soy sauce fermentation system are calculated. This transforms the original environmental parameters into quantitative indicators of the microorganisms' adaptation to these parameters, enabling unified comparison and evaluation of data from different dimensions. The specific calculation method is not limited here. For example, an adaptation function can be preset for each parameter; the average temperature can be substituted into the temperature adaptation function to obtain the temperature adaptation score, the average pH value into the pH adaptation function to obtain the acid-base adaptation score, and so on. Alternatively, the calculated temperature adaptation score, acid-base adaptation score, respiratory activity score, and redox adaptation score of the soy sauce fermentation system can be normalized to obtain the corresponding scores. The specific settings can be configured according to actual needs and are not limited here.

[0059] Step S4013: According to preset weighting coefficients, the temperature adaptation score, the acid-base adaptation score, the respiratory activity score, and the redox adaptation score are weighted and summed to obtain the microbial activity intensity score of the temperature sub-interval.

[0060] By introducing weighting coefficients, this embodiment can flexibly adjust the contribution of each adaptation score to the total score according to the sensitivity of different fermentation stages or specific microorganisms to various environmental factors, thereby more accurately reflecting the overall activity intensity of microorganisms.

[0061] The preset weighting coefficients are empirical values. For example, based on historical fermentation data or expert experience, different weighting values ​​can be set for temperature adaptation, acid-base adaptation, respiratory activity, and redox adaptation. The sum of these weighting values ​​is usually 1. Alternatively, different weighting coefficients can be set according to the fermentation stage. In some fermentation stages, temperature may be more critical, so the weight of temperature adaptation will be higher; while in other stages, pH or redox potential may be more decisive, so their corresponding weights will be increased accordingly.

[0062] Specifically, the activity intensity scoring formula can be:

[0063] Wherein: f1(T): temperature adaptation function, peak value around 30℃; f2(pH): acid-base adaptation function, peak value between pH 4.5 and 5.0; f3(CO2): respiratory activity function, increases with CO2 release; f4(ORP): redox adaptation function, better in the negative region; wi: weighting coefficient, weight w1+w2+w3+w4=1, can be adjusted by the user or learned automatically.

[0064] In one feasible embodiment, this embodiment further proposes a step of calculating the temperature adaptation score, acid-base adaptation score, respiratory activity score, and redox adaptation score of a soy sauce fermentation system based on average temperature, average pH value, average carbon dioxide concentration, and average redox potential. This includes: substituting the average temperature, average pH value, average carbon dioxide concentration, and average redox potential into a preset adaptation function to calculate the temperature adaptation score, acid-base adaptation score, respiratory activity score, and redox adaptation score; wherein the adaptation function includes: a peak function for calculating the temperature adaptation score, the peak value of which corresponds to a preset optimal temperature; a bell-shaped function for calculating the acid-base adaptation score, the peak range of which corresponds to a preset optimal pH range; an increasing function for calculating the respiratory activity score, the value of which increases with increasing carbon dioxide concentration; and a decreasing function for calculating the redox adaptation score, the value of which decreases with increasing redox potential.

[0065] Specifically, the peak function used to calculate the temperature adaptation score is a mathematical function that reaches a maximum value (peak) at a specific input value and gradually decreases on both sides of this value. It is used to simulate the temperature adaptation of microorganisms, that is, the activity of microorganisms is highest at a certain optimal temperature, and their activity decreases when deviating from this temperature. For example, a Gaussian function can be used as the peak function, with its mean set to the preset optimal temperature and its standard deviation reflecting the sensitivity of microorganisms to temperature changes; or a quadratic function or a piecewise linear function can be used to approximate it.

[0066] The bell-shaped function used to calculate pH adaptation scores is a function that maintains a high value within a specific range and decreases symmetrically on both sides of that range. It describes the pH adaptation of microorganisms, that is, the activity of microorganisms is high and stable within a certain pH range (optimal pH range), and the activity decreases when deviating from this range. For example, a bell-shaped function can be constructed by using a combination of sigmoid functions or double sigmoid functions to maintain a high adaptation score within a preset optimal pH range; or a piecewise function can be used to set a constant high adaptation score within the optimal pH range, while outside the optimal pH range, the adaptation score decreases linearly or non-linearly with the degree of pH deviation.

[0067] The increasing function used to calculate the respiratory activity fraction is a function whose output value increases with the input value. It reflects the positive correlation between microbial respiratory activity and carbon dioxide concentration. That is, the higher the carbon dioxide concentration, the more vigorous the microbial respiration and the stronger the activity. For example, a linear function or an exponential function can be used as the increasing function; or a logarithmic function or a power function can be used. The activity fraction increases faster when the carbon dioxide concentration is low and slows down when the concentration is high.

[0068] The decay function used to calculate the redox fitness score is a function whose output value decreases as the input value increases. It describes the adaptation of microorganisms to redox potential. That is, the higher the redox potential (the stronger the environmental oxidizing power), the more likely the microbial activity will be inhibited and the worse the adaptation. For example, an exponential decay function or a sigmoid decay function can be used; or an inverse proportional function or a piecewise linear function can be used. The fitness score is higher when the redox potential is low and gradually decreases as the potential increases.

[0069] This embodiment collects multiple parameters, including average temperature, average pH, average carbon dioxide concentration, and average redox potential, and converts them into their respective fitness scores. This embodiment can reflect the physiological activity of microorganisms in detail from multiple dimensions. Furthermore, by introducing preset weighting coefficients to weight and sum these fitness scores, this embodiment can flexibly adjust the importance of various environmental factors according to the actual needs of the fermentation stage, thereby obtaining a more accurate and comprehensive microbial activity intensity score. This refined evaluation mechanism enables more accurate identification of the temperature environment most conducive to microbial growth and metabolism when determining the target temperature control range, thereby optimizing the temperature control strategy, helping to improve batch consistency in the soy sauce fermentation process, shorten the fermentation cycle, improve energy utilization efficiency, and ultimately improve the quality of the final product.

[0070] Based on the first, second, and / or third embodiments of this application, in the fourth embodiment of this application, the content that is the same as or similar to the first, second, and / or third embodiments described above can be referred to the above description and will not be repeated hereafter. In addition, the method further includes: Step S50: Monitor the temperature, pH value, and redox potential of the soy sauce fermentation system; Continuous monitoring of temperature, pH, and redox potential in the soy sauce fermentation system allows for real-time acquisition of key physicochemical parameters, providing a data foundation for determining whether the fermentation state is abnormal. Temperature is a core factor affecting microbial enzyme activity and metabolic rate; pH reflects the acid-base environment of the fermentation broth, directly related to microbial growth and flavor compound formation; and redox potential characterizes the redox state of the system, crucial for the activity of anaerobic or facultative anaerobic microorganisms.

[0071] Step S60: If the temperature, pH value, or redox potential is abnormal, an abnormal response action is triggered, wherein the abnormal response action includes pausing the heating, injecting buffer solution, and / or activating inert gas protection. Based on real-time monitoring data, the system determines whether the temperature, pH value, or redox potential is abnormal. This allows the system to assess whether the fermentation system deviates from its normal operating range based on the monitored parameter values ​​and decide whether corrective measures are necessary. The logic for anomaly detection is typically based on preset thresholds or dynamic models. For example, fixed thresholds can be set; when the temperature exceeds a preset safe upper or lower limit, the pH value is below or above the safe range, or the redox potential exceeds a specific threshold, the system determines it to be abnormal. These thresholds can be set based on historical fermentation data and process requirements. Alternatively, a machine learning-based anomaly detection model can be used. This model learns parameter fluctuation patterns during normal fermentation and can identify parameter changes that deviate from the normal pattern, thus more intelligently identifying abnormal situations.

[0072] If an anomaly is detected, the system will trigger an anomaly response. These responses are pre-set and take different corrective measures for different types of anomalies, aiming to quickly stabilize the fermentation environment, prevent further deterioration, and thus protect the stability of the fermentation system and product quality. For example, pausing heating can be achieved by closing the steam valve or stopping the heating device, suitable for situations where the temperature is too high. Injecting buffer solution can be done by automatically metering pumping a preset concentration of acidic or alkaline solution into the fermenter to adjust the pH value. Activating inert gas protection can be done by controlling the valve to introduce inert gases such as nitrogen or carbon dioxide into the fermenter to lower the redox potential and inhibit excessive oxidation. Anomaly response actions can also be implemented through linkage control; for example, when an abnormal temperature rise is detected, in addition to pausing heating, cooling water circulation can also be started simultaneously.

[0073] Step S70: After the abnormal response action is executed, if the temperature, pH value and redox potential return to normal, the temperature of the soy sauce fermentation system is controlled according to the target temperature control range before the abnormal response action was triggered.

[0074] After executing the abnormal response action, the system can continuously monitor the parameters. Once all abnormal parameters return to the preset normal range, the abnormal response state will be automatically lifted, and the target temperature control range determined by the prediction model before the abnormality was triggered will be reactivated to continue temperature adjustment. This ensures that after the abnormal situation is effectively handled and normal is restored, the fermentation process can smoothly return to the original intelligent control strategy, avoiding interruption or deviation from the predetermined track due to abnormal handling.

[0075] This embodiment, through real-time monitoring of key parameters, can promptly detect and assess abnormal states in the fermentation system and automatically trigger corresponding corrective measures based on the type of abnormality, such as pausing heating, injecting buffer solution, or activating inert gas protection. This quickly stabilizes the fermentation environment and prevents damage to the fermentation process caused by abnormalities. After the abnormality is handled, the system can intelligently determine whether the parameters have returned to normal and seamlessly revert to the intelligent control strategy based on dynamically adjusting the temperature during the fermentation stage, ensuring the continuity and stability of the fermentation process. This not only avoids the lag and uncertainty of manual intervention and reduces the risk of fermentation failure due to parameter malfunction, but also ensures the consistency of batch quality of soy sauce products, improving production efficiency and process robustness.

[0076] In one feasible embodiment, the method further includes: if the same anomaly is triggered consecutively a preset number of times within a preset monitoring period, then switching to manual control mode. The monitoring period refers to a time window used to assess the frequency and duration of abnormal events, defining how long the system counts and judges abnormal events. It can be a fixed time length or a dynamically adjusted time length. The same anomaly refers to abnormal events of the same type or cause identified by the system during the fermentation process, ensuring that the system counts and responds to specific, recurring problems, rather than confusing different types of anomalies. For example, it could refer to an anomaly caused by a parameter such as temperature, pH value, or redox potential continuously exceeding its safe range, or an anomaly caused by a specific sensor malfunction or control actuator failure. Continuous triggering refers to the repeated occurrence or persistence of the same abnormal event within a set period without resolution or restoration to normal. It is used to assess the persistence of the abnormality and the effectiveness of the automatic control system in handling it. Specifically, it can mean that within the monitoring period, the system detects the same abnormal event in an abnormal state on every sampling or inspection, or that even if there is a brief recovery, the recovery time does not reach the preset stable duration, and then it triggers again. The preset number of occurrences serves as a critical condition for determining whether manual intervention is required by the automatic control system. This can be a fixed integer value or a configurable parameter, allowing operators or administrators to adjust it based on experience or the characteristics of the fermentation batch.

[0077] This embodiment intelligently monitors the number of consecutive triggers of the same anomaly within a set period. This allows for the accurate identification of persistent anomalies that are difficult for automatic control systems to handle, avoiding excessive intervention due to single, occasional anomalies, while ensuring timely transfer of control to operators when necessary. This mechanism enables operators to manually intervene in complex or persistent problems, thereby preventing the fermentation process from spiraling out of control or product quality deterioration. Compared to solutions relying solely on automatic anomaly responses, this embodiment improves the stability and reliability of the fermentation process, ensures batch-to-batch consistency of soy sauce products, and reduces resource waste caused by prolonged, ineffective automatic intervention. In one feasible embodiment, the method further includes: responding to a parameter adjustment instruction sent by a user terminal, wherein the parameter adjustment instruction is used to modify at least one of the preset temperature range, stage identification threshold, scoring weight, or model confidence threshold; when performing the decision and execution steps, using user-adjusted parameters that have undergone validity verification to overwrite the default parameters in the system database, wherein validity verification is to check the parameters input by the user to ensure that they conform to preset logical rules and physical constraints, such as checking whether the temperature range is within a reasonable range, whether the threshold is a valid value, etc., and the verification may include range verification, data type verification, logical consistency verification, etc.

[0078] In this embodiment, the user can flexibly adjust key control parameters according to the actual production situation, so that the control strategy can better match the dynamic changes of the fermentation system, thereby improving the adaptability and control precision of the soy sauce fermentation process. This helps to avoid process mismatches such as premature cooling before the enzymatic reaction is fully completed or maintaining high temperature after the microbial metabolism has declined, thereby improving the quality consistency of the final product, shortening the fermentation cycle, and optimizing energy utilization efficiency.

[0079] For example, to help understand the implementation flow of the soy sauce fermentation control method obtained in this embodiment combined with the above embodiment one, please refer to... Figure 5 , Figure 5 A simplified flowchart of a method for controlling soy sauce fermentation is provided, specifically: 1. Process startup and initialization: First, load the preset temperature range, stage division threshold, AI model parameters, scoring weights and other configuration files to complete the initialization preparation.

[0080] 2. Multi-source data acquisition: Timely reading of parameters such as temperature, pH, CO2, and ORP of the fermentation system, and integrating them into the current state vector (that is, collecting multi-source metabolic parameters of the soy sauce fermentation system according to a preset cycle to form a time-series data sequence).

[0081] 3. Fermentation stage identification (i.e., determining the current fermentation stage of the soy sauce fermentation system based on the change information of at least one metabolic parameter in the time series data sequence): the stage is determined according to the indicators in the state vector: if the growth rate of FMI (Flavor Maturity Index) is greater than the threshold α, it is determined to be the protein hydrolysis stage; if the pH decrease rate is greater than the threshold β and the CO2 release is greater than the threshold γ, it is determined to be the flavor generation stage; if the FMI tends to be stable, it is determined to be the maturation stage.

[0082] 4. AI Stage Transition Prediction (i.e., inputting the time series data sequence and the identifier of the current fermentation stage into the pre-trained prediction model to obtain the start probability of the soy sauce fermentation system entering the next fermentation stage and the start time window of the next fermentation stage): Input the current state vector + historical 72-hour data into the AI ​​model and output the start probability P of the next stage; if P≥0.85 and is within the allowable time window, it is marked as a suggested transition.

[0083] 5. Temperature range assessment: Based on the current stage, a preset temperature range is matched (e.g., 28-32℃ for the maturation stage). This range is divided into multiple sub-ranges (e.g., 28-29℃, 29-30℃, 30-32℃). The temperature is adjusted to each sub-range in turn and maintained for 1 hour. The microbial activity intensity score S is recorded. The sub-range with the highest score is selected as the current optimal temperature range (that is, the preset first temperature range corresponding to the current fermentation stage is obtained, and the first temperature range is divided into multiple temperature sub-ranges; the temperature of the soy sauce fermentation system is adjusted to each of the temperature sub-ranges in turn, and the microbial activity intensity score corresponding to each of the temperature sub-ranges is calculated).

[0084] 6. Decision fusion and instruction issuance: If the suggested transition is successful, the target temperature range for the next stage is adopted as the target temperature interval; if it is not successful, the current optimal temperature interval is adopted as the target temperature interval; then, an instruction is issued to the temperature control system to perform temperature switching or maintenance (that is, if the start probability is greater than or equal to the preset confidence threshold and the current time is within the start time window, then the preset second temperature range corresponding to the next fermentation stage is determined as the target temperature control interval; if the start probability is less than the confidence threshold and / or the current time has not reached the start time window, then the temperature sub-interval with the highest microbial activity intensity score among the various temperature sub-intervals is determined as the target temperature control interval).

[0085] 7. Anomaly detection and user configuration: If parameters exceed safety thresholds (e.g., temperature > 38℃) during the process, an alarm is triggered and a suggestion to pause heating is provided; simultaneously, user terminals can remotely adjust preset temperature ranges, stage thresholds, and scoring weights, with user parameters having higher priority than database defaults (i.e., monitoring the temperature, pH, and redox potential of the soy sauce fermentation system; if the temperature, pH, or redox potential is abnormal, an anomaly response action is triggered, including pausing heating, injecting buffer solution, and / or activating inert gas protection; after executing the anomaly response action, if the temperature, pH, and redox potential return to normal, the temperature of the soy sauce fermentation system is controlled according to the target temperature control range before the anomaly response action was triggered).

[0086] Reference Figure 6 The control system architecture for soy sauce fermentation can be as follows: 1. Data acquisition unit (A): Acquires real-time data from sensors (temperature, pH, CO2, ORP) deployed inside the fermentation tank; 2. Stage identification unit (B): Receives the output from the data acquisition unit and determines the current fermentation stage (such as protein hydrolysis, flavor generation, and maturation) based on preset rules or machine learning models; 3. AI prediction unit (C): Receives the stage identification results and historical data, runs the trained AI model, and outputs the probability of starting the next stage and the recommended time window; 4. Temperature range evaluation unit (D): Matches the preset temperature range to the current stage, divides it into N temperature ranges, adjusts the temperature sequentially, and records the microbial activity intensity score; 5. Target execution unit (E): Combines the AI ​​prediction results and the score results to determine the final target temperature range and sends control commands to the temperature control system; 6. User terminal (F): Supports remote adjustment of preset temperature range, stage division threshold, score weight, and other parameters, with higher priority than the database default value. The user terminal (F) communicates bidirectionally with all units for parameter coverage.

[0087] The following describes the control process of soy sauce fermentation using a specific implementation method: I. Data Acquisition and Preprocessing.

[0088] (I) Sensor Deployment: Four types of monitoring sensors were deployed at different heights in the fermenter, namely 0.5 meters, 1.5 meters, and 2.5 meters. The specifications of each sensor are as follows: The temperature sensor is a PT100 (Platinum Resistance Thermometer with R0=100Ω), a platinum resistance type sensor with a measurement accuracy of ±0.1℃. The sampling interval is set to 5 minutes to ensure real-time capture of temperature changes. The pH sensor uses a glass electrode type, with a measurement range of 2 to 12 and a measurement accuracy of ±0.05, which can accurately reflect the acid-base changes of the fermentation system. The CO2 concentration sensor is an infrared absorption type, with a measurement range of 0 to 5000ppm and an accuracy of ±2% of full scale (FS), which can effectively monitor the release of CO2 during fermentation. The ORP sensor uses a platinum electrode material, with a measurement range of -1000mV to +1000mV and an accuracy of ±5mV, which is used to reflect the redox state of the fermentation system.

[0089] (II) Data Preprocessing: The collected raw data undergoes a three-step preprocessing process to ensure data quality and the effectiveness of subsequent model input: Sliding window averaging: The mean of the collected parameter data is calculated within a 5-minute period to eliminate instantaneous fluctuations and improve data stability. Outlier removal: The 3σ principle is used to filter out outliers, removing data that deviates from the mean of the data set by three times the standard deviation, preventing invalid data from interfering with subsequent analysis. Normalization: All parameters processed in the first two steps are uniformly scaled to the range [0,1] to facilitate data reading and analysis by the AI ​​model. After the above preprocessing, the final output is a current state vector containing the current temperature, current pH value, current CO2 concentration, and current ORP value.

[0090] III. Fermentation Stage Identification. Taking high-salt, thin-state soy sauce fermentation as an example, the fermentation process is mainly divided into three core stages. The core characteristics and criteria for each stage are as follows: **Protein Hydrolysis Stage:** The core characteristic of this stage is the decomposition of large protein molecules, a rapid decrease in pH, and a high FMI (Flavor Intensity Moisture) growth rate. The criterion is an FMI growth rate exceeding 0.01 per hour. **Flavor Development Stage:** This stage is characterized by the accumulation of amino acids and organic acids. The pH tends to stabilize, while CO2 release increases significantly. The criterion is a pH decrease rate below 0.005 per hour and a CO2 release exceeding 1.2 liters per liter per hour (L / L). h). Maturation stage: The core characteristics of this stage are that flavor compounds tend to stabilize, FMI changes tend to be stable, and the metabolic activity of microorganisms decreases. The criterion is that the FMI change rate is less than 0.001 per hour.

[0091] The system has a clearly defined phase identification logic: First, it acquires three indicators: FMI growth rate, pH decrease rate, and CO2 release. If the FMI growth rate is greater than 0.01 / h, it is determined that the current stage is the proteolysis phase; if the proteolysis phase conditions are not met, but the pH decrease rate is less than 0.005 / h and the CO2 release is greater than 1.2 L / L, the system is considered to be in the proteolysis phase. If both conditions are met, the system determines that the current stage is in the flavor development phase; if neither condition is met, the system determines that the current stage is in the maturation phase. This logic ultimately outputs the corresponding label for the current fermentation stage.

[0092] IV. Prediction Module. The prediction model architecture adopts a Transformer-LSTM hybrid structure. This architecture can simultaneously extract local features from fermentation data and model the dependencies of time-series data, improving prediction accuracy. The dataset used for model training comes from the complete fermentation records of 120 batches of high-salt, thin-state soy sauce over the past 3 years; the data sampling frequency is once every 5 minutes, accumulating approximately 1 million valid data points; before being used for model training, these data also need to undergo a series of preprocessing operations, including 72-hour sliding window processing, normalization, noise reduction, and labeling of FMI inflection points.

[0093] The training label definition (FMI inflection point) clarifies the calculation method of FMI: FMI is obtained by adding 0.4 times the amino acid nitrogen, 0.35 times the 4-ethylguaiacol, and 0.25 times the pH value. The weight of 0.4 reflects the contribution of umami, 0.35 reflects the contribution of aroma, and 0.25 reflects the acid-base environment adaptability. Simultaneously, the marking rules for FMI inflection points are set: when the rate of change of FMI is greater than or equal to 0.01 / h, the time point is marked as the starting point of flavor transition; when the rate of change of FMI is less than or equal to 0.001 / h, the time point is marked as the end point of flavor maturation. These marked points serve as the core labels for model training.

[0094] The model employs a joint loss function that combines weighted binary cross-entropy and mean absolute error. Specifically, the total loss Loss is equal to 0.7 times the binary cross-entropy (BCE, used to optimize the prediction accuracy of the next stage initiation probability P), plus 0.3 times the mean absolute error (MAE, used to optimize the prediction accuracy of the recommendation start time window). Through this weight allocation, a balanced optimization of the classification task (probability prediction) and the regression task (time prediction) is achieved.

[0095] To meet real-time control requirements, the prediction model is deployed on edge computing nodes (such as Jetson AGX Orin). The time for a single inference is less than 50 milliseconds, while the system control cycle is 5 minutes. The model inference speed can fully meet the system's real-time requirements. Finally, it outputs the probability of starting the next stage (within the range of 0 to 1) and the recommended time window (±2 hours before and after the current time).

[0096] V. Temperature Range Assessment Module.

[0097] (I) Temperature Range Division Rules: The system will match the corresponding preset temperature range based on the currently identified fermentation stage, and uniformly divide the temperature range of each stage into 3 sub-ranges. The specific division details are as follows: The preset temperature range for the protein hydrolysis stage is 28℃ to 32℃, and the width of each sub-range is 1.33℃. The preset temperature range for the flavor development stage is 29℃ to 30℃, and the width of each sub-range is 0.33℃. The preset temperature range for the maturation stage is 30℃ to 32℃, and the width of each sub-range is 0.67℃.

[0098] (II) The microbial activity intensity score S is calculated using a weighted summation of four parameters. The specific formula is the sum of the fitness functions of each of the four parameters multiplied by their corresponding weights. The fitness functions of each parameter and their explanations are as follows: The temperature fitness function (corresponding to the temperature parameter) uses a Gaussian distribution function, with a peak value around 30℃, reflecting the temperature adaptation characteristics of fermenting microorganisms. The acid-base fitness function (corresponding to the pH parameter) uses a bell-shaped curve, with a peak value between pH 4.5 and 5.0, matching the optimal acid-base environment for microbial growth. The respiratory activity function (corresponding to the CO2 parameter) uses a linearly increasing function, with the score increasing with CO2 release, reflecting the respiratory metabolic activity of the microorganisms. The redox fitness function (corresponding to the ORP parameter) uses a negative exponential function, favoring the negative value region, where the score is higher, consistent with the redox optimum of the fermentation system. Regarding weight allocation, the default weights are set as follows: temperature weight 0.3, pH weight 0.3, CO2 weight 0.2, and ORP weight 0.2. Users can also adjust these weights through their terminal devices, and the system also has the function of automatically optimizing the weights using gradient descent.

[0099] The temperature range scoring follows a fixed procedure: First, within each temperature sub-range, the fermentation temperature is stably controlled and run for 1 hour; then, the average temperature, average pH value, average CO2 release, and average ORP value are recorded during that hour; these average parameters are substituted into the scoring formula to calculate the microbial activity intensity score S corresponding to that sub-range; finally, from all the sub-range scores, the sub-range with the highest score is selected as the current optimal temperature range and output.

[0100] VI. Decision Integration and Target Execution. The system adopts a comprehensive decision-making logic, integrating AI prediction results and temperature range scoring results to determine the final target temperature range: when the probability P of starting the next stage output by the AI ​​model is greater than or equal to 0.85, and the current time is within the recommended time window, the target temperature range of the next fermentation stage is taken as the final target temperature range; if the above two conditions are not met, the current optimal temperature range (i.e., the sub-range with the highest score) is taken as the final target temperature range.

[0101] After the decision is made, the system sends a control command containing the target temperature range to the temperature control system (PLC or DCS system). The actuators of the temperature control system, including the heater and the cooling fan, will start the PID (Proportional-Integral-Derivative) control algorithm to stabilize the actual temperature of the fermentation system within the target temperature range. The temperature control accuracy can reach ±0.5℃, and finally outputs a temperature control command containing the minimum and maximum temperature values.

[0102] VII. User Terminal and Parameter Adjustment Mechanism. The user terminal has three core functions: remote monitoring, which can display the fermentation system's core parameters such as temperature, pH, CO2, and ORP in real time, and also view the FMI change trend, current fermentation stage, and target temperature range, achieving comprehensive remote control of the fermentation process. Parameter adjustment, which supports users to remotely adjust three types of key parameters: preset temperature range (e.g., adjusting the preset temperature range for the maturation period from 30℃–32℃ to 29℃–31℃), stage division threshold (e.g., adjusting the FMI growth rate threshold from 0.01 / h to 0.008 / h), and scoring weight (e.g., adjusting the weight corresponding to temperature from 0.3 to 0.4). Safety mechanism, with multiple safety safeguards, requires secondary confirmation when adjusting key parameters to prevent accidental operation; it also sets upper and lower limits for various parameters (e.g., temperature cannot be set below 25℃ or above 38℃); when abnormal parameters are detected (e.g., temperature exceeds 38℃), the system will trigger the corresponding alarm and provide a prompt to pause the heating process.

[0103] User-defined parameters have higher priority than database default parameters. The specific execution logic is as follows: After obtaining the user input parameters and database default parameters, the system will iterate through each user input parameter one by one. For each user input parameter, the system first performs a validity check to determine whether it is within a reasonable range. If the parameter is valid, the user input parameter is directly used as the final effective parameter. If the user input parameter is invalid, the corresponding database default parameter is retained. All parameter adjustment records are stored in the system audit log for easy traceability and querying later. Finally, the integrated final effective parameter set is output.

[0104] VIII. Safety and Anomaly Detection Mechanism. The system sets clear safety thresholds and corresponding alarm actions for key monitoring parameters: Temperature parameter: When the fermentation system temperature exceeds 38℃, the system immediately triggers an alarm suggesting a pause in heating and automatically stops the operation of the heating equipment to prevent excessive temperature from affecting fermentation quality. pH parameter: When the pH value is below 3.5 or above 6.5, the system triggers an acid-base anomaly alarm and automatically starts the buffer injection program to adjust the acid-base balance of the fermentation system. ORP parameter: When the ORP value exceeds 500mV, the system triggers an excessive oxidation alarm and automatically starts the nitrogen protection program to alleviate the excessive oxidation state of the fermentation system.

[0105] Once the detected abnormal parameters return to the safe threshold range, i.e. the abnormality is resolved, the system will automatically restore to the set parameters from the most recent normal operation and continue fermentation temperature control. If the system detects three abnormal situations consecutively, it will be forced to switch to manual mode, stop automatic control, and wait for staff to intervene and investigate manually.

[0106] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the soy sauce fermentation control method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0107] This application also provides a control device for soy sauce fermentation; please refer to [reference needed]. Figure 7 The soy sauce fermentation control device includes: The acquisition module 10 is used to acquire multi-source metabolic parameters of the soy sauce fermentation system according to a preset period to form a time-series data sequence, wherein the multi-source parameters are used to characterize the fermentation state of the soy sauce fermentation system. The determination module 20 is used to determine the current fermentation stage of the soy sauce fermentation system based on the change information of at least one metabolic parameter in the time series data sequence. The prediction module 30 is used to input the time series data sequence and the identifier of the current fermentation stage into a pre-trained prediction model to obtain the start probability of the soy sauce fermentation system entering the next fermentation stage and the start time window of the next fermentation stage. The prediction model is trained using the time series data sequence of the historical fermentation system before the stage switch as input data and the start probability and start time window corresponding to the stage switch as labels. The control module 40 is used to determine the target temperature control range of the soy sauce fermentation system based on the current fermentation stage, the start probability and the start time window, and to control the temperature of the soy sauce fermentation system according to the target temperature control range.

[0108] The soy sauce fermentation control device provided in this application, employing the soy sauce fermentation control method described in the above embodiments, can solve the technical problem that controlling the soy sauce fermentation temperature based on a fixed timetable leads to an impact on product quality. Compared with the prior art, the beneficial effects of the soy sauce fermentation control device provided in this application are the same as those of the soy sauce fermentation control method provided in the above embodiments, and other technical features in the soy sauce fermentation control device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0109] This application provides a control device for soy sauce fermentation, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the soy sauce fermentation control method in the above embodiment 1.

[0110] The following is for reference. Figure 8 The diagram illustrates a structural schematic of a control device suitable for implementing the soy sauce fermentation embodiments of this application. The control device for soy sauce fermentation in these embodiments may include, but is not limited to, mobile terminals such as laptops and fixed terminals such as central control devices. Figure 8 The control device for soy sauce fermentation shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0111] like Figure 8As shown, the control device for soy sauce fermentation may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 or a program loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the soy sauce fermentation control device. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: input devices 1007 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. The communication device 1009 allows the control equipment for soy sauce fermentation to communicate wirelessly or wiredly with other equipment to exchange data. Although the figure shows control equipment for soy sauce fermentation with various systems, it should be understood that it is not required to implement or possess all of the systems shown. More or fewer systems may be implemented alternatively.

[0112] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0113] The soy sauce fermentation control device provided in this application, employing the soy sauce fermentation control method described in the above embodiments, can solve the technical problem that controlling the soy sauce fermentation temperature based on a fixed timetable leads to an impact on product quality. Compared with the prior art, the beneficial effects of the soy sauce fermentation control device provided in this application are the same as those of the soy sauce fermentation control method provided in the above embodiments, and other technical features of this soy sauce fermentation control device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0114] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0115] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0116] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the soy sauce fermentation control method in the above embodiments.

[0117] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0118] The aforementioned computer-readable storage medium may be included in the control equipment for soy sauce fermentation; or it may exist independently and not be assembled into the control equipment for soy sauce fermentation.

[0119] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the soy sauce fermentation control device, cause the soy sauce fermentation control device to: (independent scheme).

[0120] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0121] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0122] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0123] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described soy sauce fermentation control method. This solves the technical problem that controlling the soy sauce fermentation temperature based on a fixed timetable affects product quality. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the soy sauce fermentation control method provided in the above embodiments, and will not be repeated here.

[0124] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the soy sauce fermentation control method described above.

[0125] The computer program product provided in this application can solve the technical problem that controlling the fermentation temperature of soy sauce based on a fixed timetable leads to an impact on product quality. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the soy sauce fermentation control method provided in the above embodiments, and will not be repeated here.

[0126] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for controlling soy sauce fermentation, characterized in that, The method for controlling the fermentation of soy sauce includes: Multi-source metabolic parameters of the soy sauce fermentation system are collected according to a preset period to form a time-series data sequence, wherein the multi-source parameters are used to characterize the fermentation state of the soy sauce fermentation system; Based on the change information of at least one metabolic parameter in the time series data sequence, the current fermentation stage of the soy sauce fermentation system is determined; The time series data sequence and the identifier of the current fermentation stage are input into the pre-trained prediction model to obtain the start probability of the soy sauce fermentation system entering the next fermentation stage and the start time window of the next fermentation stage. The prediction model is trained with the time series data sequence of the historical fermentation system before the stage switch as input data and the start probability and start time window corresponding to the stage switch as labels. Based on the current fermentation stage, the start probability, and the start time window, the target temperature control range of the soy sauce fermentation system is determined, and the temperature of the soy sauce fermentation system is controlled according to the target temperature control range.

2. The method for controlling soy sauce fermentation as described in claim 1, characterized in that, The at least one metabolic parameter includes pH value and carbon dioxide concentration; The step of determining the current fermentation stage of the soy sauce fermentation system based on the change information of at least one metabolic parameter in the time-series data sequence includes: The flavor maturity index is determined based on the pH value, wherein the flavor maturity index is used to characterize the flavor substance formation state of the soy sauce fermentation system. Calculate the rate of change of the flavor maturity index and the rate of decrease of the pH value, and calculate the carbon dioxide release rate based on the carbon dioxide concentration; If the rate of change of the flavor maturity index is greater than the first threshold, then the current fermentation stage is determined to be the protein hydrolysis period; If the rate of change of the flavor maturity index is less than or equal to the first threshold, the rate of decrease of the pH value is less than the second threshold, and the carbon dioxide release rate is greater than the third threshold, then the current fermentation stage is determined to be the flavor generation period. If the rate of change of the flavor maturity index is less than the fourth threshold, then the current fermentation stage is determined to be the maturation period, wherein the first threshold is greater than the fourth threshold.

3. The method for controlling soy sauce fermentation as described in claim 1, characterized in that, The prediction model includes: an input layer, a feature extraction layer, a stage coding layer, and an output layer; The step of inputting the time-series data sequence and the identifier of the current fermentation stage into a pre-trained prediction model to obtain the initiation probability of the soy sauce fermentation system entering the next fermentation stage and the initiation time window of the next fermentation stage includes: An input data sequence is determined from the time-series data sequence, wherein the input data sequence is data at the current time and the time preceding the current time by a preset time length; The input data sequence is input into the feature extraction layer to obtain a fused feature vector; The identifier of the current fermentation stage is input into the stage coding layer to obtain the stage coding vector; The stage encoding vector is concatenated with the fused feature vector to obtain the concatenated vector; The spliced ​​vector is input into the output layer to obtain the initiation probability, target time, and initiation time window of the next fermentation stage.

4. The method for controlling soy sauce fermentation as described in claim 1, characterized in that, The step of determining the target temperature control range of the soy sauce fermentation system based on the current fermentation stage, the start probability, and the start time window includes: Obtain the preset first temperature range corresponding to the current fermentation stage, and divide the first temperature range into multiple temperature sub-intervals; The temperature of the soy sauce fermentation system is adjusted to each of the temperature sub-ranges in sequence, and the microbial activity intensity score corresponding to each of the temperature sub-ranges is calculated. If the start probability is greater than or equal to a preset confidence threshold and the current time is within the start time window, then the preset second temperature range corresponding to the next fermentation stage is determined as the target temperature control range. If the activation probability is less than the confidence threshold and / or the current time has not reached the activation time window, then the temperature sub-interval with the highest microbial activity intensity score among the various temperature sub-intervals is determined as the target temperature control interval.

5. The method for controlling soy sauce fermentation as described in claim 4, characterized in that, The step of calculating the microbial activity intensity score corresponding to each of the temperature sub-intervals includes: After determining that the temperature of the soy sauce fermentation system is within the specified temperature sub-range, the average temperature, average pH value, average carbon dioxide concentration, and average redox potential of the soy sauce fermentation system are collected. Based on the average temperature, average pH value, average carbon dioxide concentration and average redox potential, the temperature adaptation score, acid-base adaptation score, respiratory activity score and redox adaptation score of the soy sauce fermentation system are calculated. According to preset weighting coefficients, the temperature adaptation score, acid-base adaptation score, respiratory activity score, and redox adaptation score are weighted and summed to obtain the microbial activity intensity score of the temperature sub-interval.

6. The method for controlling soy sauce fermentation as described in any one of claims 1 to 5, characterized in that, The method further includes: Monitor the temperature, pH value, and redox potential of the soy sauce fermentation system; If the temperature, pH value, or redox potential is abnormal, an abnormal response action is triggered, wherein the abnormal response action includes pausing the heating, injecting buffer solution, and / or activating inert gas protection. If the temperature, pH value, and redox potential return to normal after the abnormal response action is executed, the temperature of the soy sauce fermentation system is controlled according to the target temperature control range before the abnormal response action was triggered.

7. A control device for soy sauce fermentation, characterized in that, The soy sauce fermentation control device includes: The acquisition module is used to acquire multi-source metabolic parameters of the soy sauce fermentation system according to a preset period to form a time-series data sequence, wherein the multi-source parameters are used to characterize the fermentation state of the soy sauce fermentation system; The determination module is used to determine the current fermentation stage of the soy sauce fermentation system based on the change information of at least one metabolic parameter in the time-series data sequence. The prediction module is used to input the time series data sequence and the identifier of the current fermentation stage into a pre-trained prediction model to obtain the start probability of the soy sauce fermentation system entering the next fermentation stage and the start time window of the next fermentation stage. The prediction model is trained using the time series data sequence of the historical fermentation system before the stage switch as input data and the start probability and start time window corresponding to the stage switch as labels. The control module is used to determine the target temperature control range of the soy sauce fermentation system based on the current fermentation stage, the start probability, and the start time window, and to control the temperature of the soy sauce fermentation system according to the target temperature control range.

8. A control device for soy sauce fermentation, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the control method for soy sauce fermentation as described in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the soy sauce fermentation control method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the soy sauce fermentation control method as described in any one of claims 1 to 6.

Images