Fermentation environment detection method, device and system based on digital twinning and medium

Abstract

The invention relates to a fermentation environment detection method, device and system based on digital twinning and a medium, and belongs to the field of fermentation. The method aims at solving the problem that traditional fermentation environment detection lacks accurate simulation and real-time feedback, and a virtual model accurately mapped with an actual fermentation environment is constructed by using a digital twinborn technology. The method comprises the following steps: deploying a sensor in a fermentation heap for collecting multi-dimensional parameters such as temperature and humidity in real time; transmitting the parameters to a digital twin model through a wireless network; in the digital twinborn model, simulating and updating a fermentation dynamic state according to a given formula; presetting a parameter threshold range, and triggering early warning when the data acquired by the sensor exceeds the threshold range; and judging whether fermentation is normal or not according to a preset rule and a model prediction result, and giving corresponding processing suggestions. The detection method realizes accurate monitoring and intelligent regulation and control of the fermentation environment, improves the fermentation efficiency and the product quality, and reduces the production cost.

Description

Technical Field

[0001] The present invention belongs to the field of monitoring the material fermentation process, and specifically relates to a method for detecting the fermentation environment based on digital twin, which can be applied to the technical fields of fermentation such as feed, feces, compost, etc. Background Art

[0002] In the field of traditional agricultural technologies, although fermentation technologies are widely used in the livestock and poultry breeding industry, such as the use of fermented feed and fermented feces organic fertilizers, there are still many problems. For example, when anaerobic fermentation of feed is carried out, the quality can only be judged after opening the bag after fermentation, and it is impossible to monitor in time during the fermentation process, resulting in difficulty in timely remedy when the fermented feed becomes moldy; there are problems in the aerobic fermentation of fecal compost, such as imperfect turning equipment, difficult control of turning time and frequency, resulting in low fermentation temperature, incomplete fecal maturation, and a large amount of toxic and harmful gases being produced to pollute the environment. In addition, traditional methods for detecting the fermentation environment are inconvenient to operate, slow in speed, non-real-time, and have a high labor cost, making it difficult to meet the requirements of modern fermentation production.

[0003] In addition, in the prior art, the detection of fermentation environment parameters mainly relies on the data of a single sensor, lacking a comprehensive and dynamic simulation process of the fermentation process, being unable to predict in advance the problems that may occur during the fermentation process, and the evaluation of fermentation quality is not accurate and timely enough. At the same time, traditional methods lack effective data analysis and processing means, making it difficult to achieve intelligent control and optimization of the fermentation environment.

[0004] Chinese Patent Application CN118956572A discloses a fermentation tank parameter detection device, method and system. In the technical solution provided by this disclosure, although it involves the process of sensor data acquisition of fermentation parameters of the fermentation tank, wireless transmission of data, and uploading data to the cloud, it does not disclose the specific method of data processing, including how to analyze these data, whether there are algorithms or models to predict the fermentation state, etc., and also lacks an alarm and adjustment mechanism for fermentation parameters. Summary of the Invention

[0005] The present invention aims to solve the technical problems of inconvenient operation, slow speed, non-real-time, and high labor cost in traditional fermentation environment detection, as well as the technical problems of lack of comprehensive, dynamic simulation and accurate evaluation of the fermentation process in the prior art. When solving the above technical problems, the present invention attempts to achieve real-time and accurate detection of the fermentation environment through the simulation of introducing digital twin algorithms, predict in advance the abnormal situations during the fermentation process and take timely measures, optimize the fermentation environment, and improve the fermentation effect.

[0006] In the first aspect, the present invention provides a method for detecting the fermentation environment based on digital twin, and the method includes:

[0007] Step S100: Real-time collect multi-dimensional parameters of the fermentation environment through sensors deployed in the fermentation pile body, where the multi-dimensional parameters include the temperature, humidity, ammonia concentration, hydrogen sulfide concentration, and oxygen concentration of the fermentation pile body;

[0008] Step S200: Transmit the multi-dimensional parameters to the digital twin model in the data processing system in real time through a wireless network;

[0009] Step S300: Preset the threshold range of the multi-dimensional parameters. When the data collected by the sensor exceeds the threshold range, a warning signal is triggered;

[0010] Step S400: Build the digital twin model corresponding to the actual fermentation environment in the data processing system. The digital twin model simulates and updates the dynamic changes of the fermentation process in real time according to the input multi-dimensional parameters. The construction and state update of the digital twin model are based on the formula S t+1 = f(S t , P t ), where S t+1 is the predicted state vector of the fermentation process at time t + 1, S t is the state vector of the fermentation process at time t, P t is the multi-dimensional parameter vector collected by the sensor at time t, and f is the mapping function of the digital twin model based on fermentation kinetics, which is determined according to the physical principles and chemical reaction mechanisms of the fermentation process;

[0011] Step S500: Judge whether the fermentation process is normal according to the preset rules and model prediction results, and give corresponding treatment suggestions.

[0012] Preferably, the calculation formula of S t is S t = a·y t +(1 - a)S t-1 , where S t is the smoothed value at time t, S t-1 is the smoothed value at time t - 1, y t is the actual value at time t, and a is the smoothing constant, and its value range is [0, 1].

[0013] Preferably, the f function adopts an LSTM dynamic prediction model and updates the model driven by real-time data.

[0014] Preferably, the sensor includes a probe sensor and a gas sensor. The probe sensor is in a probe structure and can be inserted into the fermentation pile body to a certain depth to collect the temperature and humidity in the fermentation pile body in real time; the gas sensor is used to collect the concentrations of ammonia, hydrogen sulfide, and oxygen in the fermentation pile body in real time.

[0015] Preferably, the wireless network is 4G, 5G, WiFi, LoRa, Sigfox or NB-IoT.

[0016] Preferably, when T < T min , T > T max , H < H min , H > H max , A > A th , S > S th or O < O th , a warning signal is triggered, where T is the temperature, H is the humidity, A is the ammonia concentration, S is the hydrogen sulfide concentration, O is the oxygen concentration, T min , T max , H min , H max , A th , S th , O th are the thresholds of the minimum temperature, maximum temperature, minimum humidity, maximum humidity, ammonia concentration, hydrogen sulfide concentration, and oxygen concentration preset according to different fermentation materials and process requirements.

[0017] Preferably, the treatment suggestions include: if it is determined that the temperature or humidity of the fermentation pile is too high, or the concentration of ammonia or hydrogen sulfide is too high, suggestions for increasing the ventilation frequency and time are given; if it is determined that the humidity of the fermentation pile is too low, suggestions for spraying water for humidification are given.

[0018] In a second aspect, the present invention provides a fermentation environment detection device based on digital twin, and the device includes:

[0019] Module M100, configured to: collect multi-dimensional parameters of the fermentation environment in real time through sensors deployed in the fermentation pile, and the multi-dimensional parameters include the temperature, humidity, ammonia concentration, hydrogen sulfide concentration, and oxygen concentration of the fermentation pile;

[0020] Module M200, configured to: transmit the multi-dimensional parameters to the digital twin model in the data processing system in real time through a wireless network;

[0021] Module M300, configured to: preset a threshold range for the multi-dimensional parameters, and trigger a warning signal when the data collected by the sensor exceeds the threshold range;

[0022] Module M400, configured to: construct the digital twin model corresponding to the actual fermentation environment in the data processing system, and the digital twin model simulates and updates the dynamic changes of the fermentation process in real time according to the input multi-dimensional parameters, and the construction and state update of the digital twin model are based on the formula S t+1 = f(S t , Pt ), where S t+1 is the predicted state vector of the fermentation process at time t + 1, and S t is the state vector of the fermentation process at time t, P t is the multi-dimensional parameter vector collected by the sensor at time t, and f is the mapping function of the digital twin model based on fermentation kinetics, which is determined according to the physical principles and chemical reaction mechanisms of the fermentation process;

[0023] Module M500 is used to: judge whether the fermentation process is normal according to preset rules and model prediction results, and give corresponding treatment suggestions.

[0024] In a third aspect, the present invention provides a computer system, including a processor, a memory, and a computer program stored on the memory and executable by the processor. When the processor runs the computer program, the method described in the first aspect of the present invention is implemented.

[0025] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, the method described in the first aspect of the present invention is implemented.

[0026] Due to the adoption of the above technical solutions, the present invention realizes real-time and accurate detection of the fermentation environment, can predict abnormal situations such as mildew in the fermentation process in advance, so as to take necessary measures in time, optimize and adjust the fermentation environment, improve the fermentation effect, shorten the fermentation cycle, and reduce environmental pollution, etc. In addition, due to the adoption of the technical means of remote monitoring, the present invention can also save the costs and time of researchers or production personnel for traveling to and from the fermentation site. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 : The step block diagram of the method of the present invention;

[0028] Figure 2 : The module block diagram of the device of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] To more clearly illustrate the features of the technical solution of the present invention, the present invention will be further described in detail below through specific embodiments and in conjunction with the drawings.

[0030] As a first embodiment, the present invention provides a method for detecting a fermentation environment based on digital twin, and the method includes:

[0031] Step S100: Real-time collect multi-dimensional parameters of the fermentation environment through sensors deployed in the fermentation pile body, and the multi-dimensional parameters include the temperature, humidity, ammonia concentration, hydrogen sulfide concentration, and oxygen concentration of the fermentation pile body;

[0032] Step S200: Transmit the multi-dimensional parameters to the digital twin model in the data processing system in real time through a wireless network;

[0033] Step S300: Preset the threshold range of the multi-dimensional parameters. When the data collected by the sensor exceeds the threshold range, a warning signal is triggered;

[0034] Step S400: Build the digital twin model corresponding to the actual fermentation environment in the data processing system. The digital twin model simulates and updates the dynamic changes of the fermentation process in real time according to the input multi-dimensional parameters. The construction and state update of the digital twin model are based on the formula S t+1 = f(S t , P t ), where S t+1 is the state vector of the predicted fermentation process at time t+1, S t is the state vector of the fermentation process at time t, P t is the multi-dimensional parameter vector collected by the sensor at time t, and f is the mapping function of the digital twin model based on fermentation kinetics. This function is determined according to the physical principles and chemical reaction mechanisms of the fermentation process;

[0035] Step S500: Judge whether the fermentation process is normal according to the preset rules and the model prediction results, and give corresponding treatment suggestions.

[0036] Figure 1 The block diagrams of the execution steps in the above method embodiments are given.

[0037] The written order between and within each step in the above method embodiments is only for the convenience of expression and understanding, but it does not limit the scope of protection required by the technical solution of the present invention. For example: The construction of the digital twin model in Step S400 can be prior to Steps S100, S200, and S300. For the execution of other specific operation steps, there are still other transformation forms different from the current specific embodiments on the premise that the technical problems of the present invention can be solved and the technical effects of the present invention can be achieved.

[0038] According to Figure 1As shown, in the fermentation environment detection method of this application, the following technical means are generally used: sensor data acquisition, wireless network transmission, data warning, digital twin model construction and analysis, software recognition and determination. Further, by deploying a variety of sensors in the fermentation pile to collect multi-dimensional parameters of the fermentation environment in real time, the collected data is transmitted to the data processing system through a wireless network. A digital twin model is constructed in the data processing system, the output results of the digital twin model are analyzed and judged, abnormal situations in the fermentation process are identified and warnings are issued, and the fermentation environment is optimized according to the warning results.

[0039] As the core data processing and analysis layer of this application, in the cloud server or local control system, a digital twin model corresponding to the actual fermentation environment is constructed, and the data collected by the sensor is input into the digital twin model in real time. The model is based on the input data and through the formula S t+1 =f(S t ,P t ), simulates the dynamic changes of the fermentation process. By analyzing and judging the output results of the digital twin model, abnormal situations in the fermentation process are identified and warnings are issued. This function is determined according to the physical principles and chemical reaction mechanisms of the fermentation process. Through the analysis and prediction of the digital twin model, possible problems in the fermentation process are predicted, and measures are taken in advance for adjustment.

[0040] The core algorithm of the digital twin model is based on machine learning or deep learning algorithms. For example, a neural network model is used to model and predict the fermentation process. The input of the model is the data collected by the sensor, and the output is the prediction results of the fermentation process, such as fermentation progress, fermentation quality, etc.

[0041] In the formula S t+1 =f(S t ,P t ), according to the current state S t and parameter P t , the state vector S t+1 of the next moment is predicted through the mapping function f. The formula shows that the predicted state S t+1 of the next moment is jointly determined by the current state S t and the latest sensor data P t . Specific explanations of f, S t and P t are as follows.

[0042] The mapping function represented by f is determined by the physical principles and chemical reaction mechanisms of the fermentation process. As the core branch discipline of fermentation engineering, fermentation kinetics includes the study of microbial growth kinetics (rates and influencing factors at different growth stages), substrate metabolism kinetics (relationship between substrate consumption and product formation), product formation kinetics (product synthesis rate and regulation mechanism), etc.

[0043] The state vector represented by S t is a set of temperature, humidity, ammonia concentration, hydrogen sulfide concentration, and oxygen concentration of the fermentation process at time t. These parameters together describe the current state of the fermentation process. In fermentation kinetics, they are all independent variables that affect the reaction rate and product quality. Suppose at a certain moment, the temperature of the fermentation pile is 55 °C, the humidity is 65%, the ammonia concentration is 10 ppm, the hydrogen sulfide concentration is 1 ppm, and the oxygen concentration is 8%, then S t = [55, 65, 10, 1, 8].

[0044] The parameter vector represented by P t is a set of original data actually collected by the sensors at time t. They are the input parameters that drive the update of the digital twin model. These parameters directly come from the sensors deployed in the fermentation device, including the temperature and humidity collected by the probe sensors and the ammonia, hydrogen sulfide, and oxygen concentrations collected by the gas sensors. Obviously, the sensor parameters need to meet the requirements in terms of real-time, multi-source heterogeneity, quantifiability, etc. Suppose in a certain sampling, the temperature collected by the sensor is 55.5 °C, the humidity is 70%, the ammonia concentration is 20 ppm, the hydrogen sulfide concentration is 2 ppm, and the oxygen concentration is 9%, then P t = [55.5, 70, 20, 2, 9].

[0045] It should be noted that the state vector represented by S here t is the "state that the model thinks the fermentation system should be in", that is, the theoretical value; while the parameter vector represented by P t is the "state actually detected by the sensor told to the model", that is, the measured value. The former is used to predict the future state S t+1 ; while the latter is used to calibrate the model and correct the deviation of S t . The reliability of the former depends on the model accuracy, while the reliability of the latter depends on the sensor accuracy. The two interact dynamically through the digital twin model to achieve the closed-loop calibration of theoretical prediction and actual data.

[0046] The formula S here t+1 = f(S t , P t ), describes the core logic of the digital twin model, that is: through the state and real-time parameters at the current moment, dynamically predict the fermentation state at the next moment. For example, if the current temperature is too high (S tin the case of abnormal temperature values), the model will calculate the possible future mildew risk (S t ) based on the sensor data (P t+1 ).

[0047] The present invention attempts to introduce digital twin technology into the field of fermentation environment detection. It is constructed based on the physical principles and chemical reaction mechanisms of the fermentation process. By constructing a digital twin model and analyzing and processing multi-dimensional data collected by sensors, real-time simulation and prediction of the fermentation process are achieved. Traditional fermentation environment detection methods mainly rely on the real-time monitoring of single-sensor data and threshold judgment, lacking comprehensive and dynamic simulation of the fermentation process and being unable to predict future changes in the fermentation process. However, this solution realizes real-time evaluation and early warning of fermentation quality by constructing a specific digital twin model, simulating the dynamic changes of the fermentation process through virtual mapping of the fermentation environment based on historical data and real-time data, discovering potential problems in advance, and providing a more scientific basis for the optimization of the fermentation process.

[0048] Moreover, the present invention combines various technical means, such as sensor data acquisition, wireless network transmission, data early warning, and software identification and determination, etc., to form a complete fermentation environment detection system, which can reflect various state changes in the fermentation process in real time and accurately, thereby predicting possible problems in advance and taking measures in a timely manner. Each technical means cooperates with each other and works synergistically, improving the accuracy and real-time performance of fermentation environment detection, reducing labor costs, and improving fermentation efficiency and quality.

[0049] As a preferred embodiment, the calculation formula of the S t is S t = a·y t +(1 - a)S t-1 , where S t is the smoothed value at time t, S t-1 is the smoothed value at time t - 1, y t is the actual value at time t, and a is the smoothing constant, and its value range is [0, 1].[[]END]]

[0050] In this preferred embodiment, the calculation method of S t in the function f is defined.

[0051] The calculation formula here is S t = a·y t +(1 - a)S t-1, the exponential smoothing method (ES, Exponential Smoothing) is used. It predicts the future of a phenomenon by calculating the exponential smoothing value and cooperating with a certain time series prediction model. The principle is that the exponential smoothing value of any period is the weighted average of the actual observation value of this period and the exponential smoothing value of the previous period. Through the exponential smoothing method, dynamic weights are assigned to historical data, enhancing the response ability to sudden fluctuations, and at the same time, the prediction accuracy of S t+1 can also be optimized.

[0052] As a preferred embodiment, the f function adopts an LSTM dynamic prediction model and updates the model with real-time data.

[0053] In this preferred embodiment, the specific prediction model is defined.

[0054] LSTM (Long Short-Term Memory) is a special type of RNN (Recurrent Neural Network) and is used to process data with time series characteristics.

[0055] In this application, LSTM is used to process the time series data collected by sensors. Its goal is to solve the problem of gradient disappearance existing in traditional RNN during long sequence training, enabling it to capture long-term dependencies in the sequence. In other words, the LSTM prediction model has certain advantages in the time series prediction of sensor data in this application. For example, it can handle the parameter change correlations spanning several hours or even days during the fermentation process (such as temperature fluctuations may affect the current mold growth rate); it supports multi-variable input, can simultaneously process multi-dimensional sensor data such as temperature, humidity, and gas concentration, and can perform multi-parameter fusion analysis; it predicts future states and triggers warnings by learning patterns in historical data (such as abnormal characteristics of feed fermentation failure), and has dynamic adaptability, etc.

[0056] LSTM controls the flow of information through a gating mechanism, mainly including three key gate structures:

[0057] First, the forget gate: determines whether to retain or discard historical information in the fermentation state, such as judging which temperature fluctuations are noise during the fermentation process.

[0058] Second, the input gate: screens the part of the current input information that needs to be stored in the fermentation state, such as screening the effective parameters collected by sensors.

[0059] Third, the output gate: generates the output at the current moment based on the fermentation state and the current input, such as predicting the fermentation state one hour later.

[0060] As a preferred embodiment, the sensor includes a probe sensor and a gas sensor. The probe sensor is in the form of a probe and can be inserted to a certain depth in the middle of the fermentation pile for real-time collection of the temperature and humidity in the fermentation pile. The gas sensor is used for real-time collection of the concentrations of ammonia, hydrogen sulfide, and oxygen in the fermentation pile.

[0061] In this preferred embodiment, as the sensor data acquisition layer of the present invention, a variety of sensors are deployed at different positions of the fermentation pile, including a probe sensor and a gas detection module sensor, for real-time collection of various parameters of the fermentation environment.

[0062] As a preferred embodiment, the wireless network is 4G, 5G, WiFi, LoRa, Sigfox, or NB-IoT.

[0063] In this preferred embodiment, as the data transmission layer of the present invention, the data collected by the sensors is transmitted to the digital twin model in the data processing system through wireless networks such as 4G, 5G, WiFi, LoRa, Sigfox, and NB-IoT, and the stability and real-time nature of the data transmission are ensured.

[0064] It is worth mentioning that LoRa (Long Range), as a low-power wide-area network (LPWAN) technology, can be used for long-distance radio transmission. It uses spread-spectrum technology to allow long-distance (several kilometers or even farther) communication at extremely low power consumption, and is particularly suitable for application scenarios that require battery power supply and long communication distances, especially the Internet of Things working scenario for fermentation environment detection in the present invention.

[0065] As a preferred embodiment, when T < T min , T > T max , H < H min , H > H max , A > A th , S > S th or O < O th , a warning signal is triggered, where T is the temperature, H is the humidity, A is the ammonia concentration, S is the hydrogen sulfide concentration, O is the oxygen concentration, and T min , T max , H min , H max , A th , S th , O th are the thresholds of the minimum temperature, maximum temperature, minimum humidity, maximum humidity, ammonia concentration, hydrogen sulfide concentration, and oxygen concentration preset according to different fermentation materials and process requirements.

[0066] In this preferred embodiment, the threshold ranges of various parameters are set, and when the data collected by the sensors exceeds the threshold ranges, the system immediately issues a warning signal.

[0067] As a preferred embodiment, the processing suggestions include: if it is determined that the temperature or humidity of the fermentation heap is too high, or the concentration of ammonia or hydrogen sulfide is too high, suggestions for increasing the ventilation frequency and time are given; if it is determined that the humidity of the fermentation heap is too low, suggestions for spraying water for humidification are given.

[0068] In this preferred embodiment, as the control and feedback layer, according to the preset rules and model prediction results, it is judged whether the fermentation process is normal, and corresponding processing suggestions are given. For example, increasing the ventilation volume, necessary humidification measures, etc., to optimize the fermentation environment and ensure the smooth progress of the fermentation process.

[0069] As a second embodiment, the present invention provides a fermentation environment detection device based on digital twin, and the device includes:

[0070] Module M100, configured to: collect multi-dimensional parameters of the fermentation environment in real time through sensors deployed in the fermentation heap, and the multi-dimensional parameters include the temperature, humidity, ammonia concentration, hydrogen sulfide concentration, and oxygen concentration of the fermentation heap;

[0071] Module M200, configured to: transmit the multi-dimensional parameters to the digital twin model in the data processing system in real time through a wireless network;

[0072] Module M300, configured to: preset a threshold range for the multi-dimensional parameters, and trigger a warning signal when the data collected by the sensor exceeds the threshold range;

[0073] Module M400, configured to: construct the digital twin model corresponding to the actual fermentation environment in the data processing system, and the digital twin model simulates and updates the dynamic changes of the fermentation process in real time according to the input multi-dimensional parameters. The construction and state update of the digital twin model are based on the formula S t+1 = f(S t , P t ), where S t+1 is the state vector of the predicted fermentation process at time t + 1, S t is the state vector of the fermentation process at time t, P t is the multi-dimensional parameter vector collected by the sensor at time t, and f is the mapping function of the digital twin model based on fermentation kinetics, and this function is determined according to the physical principles and chemical reaction mechanisms of the fermentation process;

[0074] Module M500, configured to: judge whether the fermentation process is normal according to the preset rules and model prediction results, and give corresponding processing suggestions.

[0075] Figure 2 The block diagrams of the constituent modules in the above device embodiments are given.

[0076] As a third embodiment, the present invention provides a computer system, including a processor, a memory, and a computer program stored on the memory and executable by the processor. When the processor runs the computer program, the method described in the first embodiment of the present invention is implemented.

[0077] As a fourth embodiment, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, the method described in the first embodiment of the present invention is implemented.

[0078] Finally, it should be noted that although the present invention has been described by way of specific embodiments, it does not constitute a limitation to the scope of patent protection of the present invention. Those skilled in the art should understand that various equivalent substitutions and optimizations can still be made to the specific embodiments of the present invention, and any substitution and improvement without departing from the spirit of the present invention should be covered within the scope of patent protection of the present invention.

Claims

1. A fermentation environment detection method based on digital twin, characterized in that: The method comprises: Step S100: collecting multi-dimensional parameters of the fermentation environment in real time through sensors deployed in the fermentation stack, wherein the multi-dimensional parameters include the temperature, humidity, ammonia concentration, hydrogen sulfide concentration and oxygen concentration of the fermentation stack; Step S200: transmitting the multi-dimensional parameters to the digital twin model in the data processing system in real time via a wireless network; Step S300: Preset a threshold range of the multi-dimensional parameter, and trigger an early warning signal when the data collected by the sensor exceeds the threshold range; Step S400: construct the digital twin model corresponding to the actual fermentation environment in the data processing system. The digital twin model simulates and updates the dynamic changes of the fermentation process in real time according to the input multi-dimensional parameters. The construction and status update of the digital twin model are based on formula S t+1 =f(S t ,P t ), where S t+1 is the predicted state vector of the fermentation process at time t+1, S t is the state vector of the fermentation process at time t, P t is the multi-dimensional parameter vector collected by the sensor at time t, and f is the mapping function of the digital twin model based on fermentation kinetics, which is determined according to the physical principles and chemical reaction mechanisms of the fermentation process; Step S500: According to the preset rules and model prediction results, determine whether the fermentation process is normal and give corresponding treatment suggestions.

2. The method according to claim 1, characterized in that The S t The calculation formula is S t =a·y t +(1-a)S t-1 , where S t is the smoothed value at time t, S t-1 is the smoothed value at time t-1, y t is the actual value at time t, a is a smoothing constant, and its value range is [0,1].

3. The method according to claim 1 or 2, characterized in that: The f function adopts the LSTM dynamic prediction model and drives the model update with real-time data.

4. The method according to claim 1, characterized in that: The sensor includes a probe sensor and a gas sensor. The probe sensor is a probe structure that can be inserted into the middle of the fermentation stack to a certain depth for real-time collection of the temperature and humidity in the fermentation stack; the gas sensor is used for real-time collection of the concentrations of ammonia, hydrogen sulfide and oxygen in the fermentation stack.

5. The method according to claim 1, characterized in that The wireless network is 4G, 5G, WiFi, LoRa, Sigfox or NB-IoT.

6. The method according to claim 1, characterized in that When T<T min 、T>T max 、H<H min 、H>H max 、A>A th 、S>S th Or O<O th When the temperature is 0.0400, the warning signal is triggered, where T is the temperature, H is the humidity, A is the ammonia concentration, S is the hydrogen sulfide concentration, O is the oxygen concentration, and T min , T max , H min , H max , A th , S th , O th The thresholds of minimum temperature, maximum temperature, minimum humidity, maximum humidity, ammonia concentration, hydrogen sulfide concentration, and oxygen concentration are pre-set according to different fermentation materials and process requirements.

7. The method according to claim 1, characterized in that The treatment suggestions include: if it is judged that the temperature or humidity of the fermentation stack is too high, or the concentration of ammonia or hydrogen sulfide is too high, a suggestion to increase the ventilation frequency and time is given; if it is judged that the humidity of the fermentation stack is too low, a suggestion to spray water for humidification is given.

8. A fermentation environment detection device based on digital twin, characterized in that: The device comprises: Module M100 is used to collect multi-dimensional parameters of the fermentation environment in real time through sensors deployed in the fermentation stack, wherein the multi-dimensional parameters include the temperature, humidity, ammonia concentration, hydrogen sulfide concentration and oxygen concentration of the fermentation stack; Module M200 is used to: transmit the multi-dimensional parameters to the digital twin model in the data processing system in real time via a wireless network; Module M300 is used to: preset a threshold range of the multi-dimensional parameter, and trigger an early warning signal when the data collected by the sensor exceeds the threshold range; Module M400 is used to: construct the digital twin model corresponding to the actual fermentation environment in the data processing system, and the digital twin model simulates and updates the dynamic changes of the fermentation process in real time according to the input multi-dimensional parameters. The construction and status update of the digital twin model are based on formula S t+1 =f(S t ,P t ), where S t+1 is the predicted state vector of the fermentation process at time t+1, S t is the state vector of the fermentation process at time t, P t is the multi-dimensional parameter vector collected by the sensor at time t, and f is the mapping function of the digital twin model based on fermentation kinetics, which is determined according to the physical principles and chemical reaction mechanisms of the fermentation process; Module M500 is used to: determine whether the fermentation process is normal based on preset rules and model prediction results, and give corresponding treatment suggestions.

9. A computer system comprising a processor, a memory, and a computer program stored in the memory and executable by the processor, characterized in that: When the processor runs the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images