A mushroom residue fermentation ingredient control method and ingredient device based on AI dynamic proportioning

Abstract

The application discloses a kind of based on AI dynamic proportioning's fungus mushroom residue fermentation ingredient control method and ingredient device, it is related to biological fermentation technical field.The application includes: S1: multi-modal data acquisition: by near infrared spectrometer, metal oxide semiconductor gas sensor array and microwave resonance moisture sensor, obtain cellulose / lignin spatial distribution, spoilage characteristic and humidity gradient;S2: parameter determination: according to target threshold value, the three-dimensional feature of original fungus mushroom residue pile and time sequence result, determine the optimal parameter;S3: adjustment control: according to the optimal solution, real-time adjustment is carried out to fermentation temperature, auxiliary material addition and aeration amount.The application combines spatial feature and time sequence data, realizes the digital twin modeling of fermentation process, simultaneously by the three-dimensional target space of composting degree-conversion rate-energy consumption constructed, in combination with K-means clustering algorithm and B spline curvature optimization, determine the optimal parameter, shorten the fermentation parameter adjustment response time.

Description

Technical Field

[0001] This invention relates to the field of bio-fermentation technology, specifically to a method and apparatus for controlling the fermentation of mushroom residue based on AI dynamic proportioning. Background Technology

[0002] With the acceleration of agricultural modernization in my country, the edible fungi industry, as an important component of efficient ecological agriculture, has experienced rapid development in recent years. However, its large-scale production process also generates a large amount of waste—mushroom residue (also known as mushroom bag waste). According to incomplete statistics, the amount of mushroom residue generated in my country each year exceeds ten million tons. This type of waste is rich in organic matter, cellulose, lignin, and small amounts of nutrients such as nitrogen, phosphorus, and potassium, and has high resource utilization value.

[0003] Currently, the main methods for treating mushroom residue include direct return to the field, incineration for power generation, feed production, and composting. Composting is an environmentally friendly and resource-efficient method that can convert mushroom residue into stable, harmless organic fertilizer applicable to farmland. However, in practice, due to the high carbon-to-nitrogen ratio (C / N > 30:1), simple structure, and unstable moisture content of mushroom residue, composting alone often suffers from slow fermentation initiation, high ammonia release, and long maturation periods, severely impacting its resource utilization efficiency and product quality.

[0004] Chinese invention patent CN120108534A discloses a method, system, and storage medium for dynamic feed formulation adjustment based on online near-infrared detection. The method includes the following steps: obtaining the required nutrient content of the formulated feed based on the feed formula and the standard nutrient content of each feed ingredient; detecting the actual nutrient content of each feed ingredient in real time using a near-infrared sensor; optimizing and adjusting the actual ratio of each ingredient in the feed formula based on the actual nutrient content and the required nutrient content of the formulated feed; and obtaining the corresponding amount of ingredients for feed formulation based on the actual ratio of each ingredient. This invention, based on near-infrared sensing technology, obtains the actual nutrient content of each ingredient in the feed ingredients in real time and adjusts the amount of each ingredient in the feed formula accordingly. This ensures that the nutrient content of the formulated feed remains consistent with the feed formula, avoiding waste of raw materials while guaranteeing that animals receive sufficient nutrients for growth.

[0005] In the process of utilizing mushroom residue as a resource, it can be used to produce various products such as organic fertilizer, animal feed, and biofuel. However, in organic fertilizer production, the focus is on the maturity of the material and its nutrient release characteristics. In feed substrate preparation, the focus is on increasing protein content and the effect of toxin degradation. When used for biofuel, the focus is on cellulose conversion rate and calorific value. Therefore, there are significant differences in the physicochemical properties of fermentation products in different application scenarios. Most existing batching systems use fixed ratios or manual adjustments based on experience, which cannot be dynamically optimized according to the target application, making it difficult for the same equipment to flexibly adapt to various application needs. Summary of the Invention

[0006] The purpose of this invention is to provide a method and apparatus for controlling the fermentation of mushroom residue based on AI dynamic proportioning, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for controlling the fermentation of mushroom residue based on AI dynamic proportioning, comprising:

[0008] S1: Multimodal data acquisition: Spatial distribution of cellulose / lignin, decay characteristics and humidity gradient are obtained through near-infrared spectrometer, metal oxide semiconductor gas sensor array and microwave resonant moisture sensor;

[0009] S2: Parameter Determination: Based on the target threshold, the three-dimensional characteristics of the original mushroom residue compost, and the time series results, the optimal parameters are determined, including:

[0010] S2.1: Data preprocessing: Based on the spatial distribution of cellulose / lignin, decay characteristics and humidity gradient, the three-dimensional features and time-series results of the original mushroom residue compost are determined by using a 3D-CNN model and an LSTM network model.

[0011] S2.2: Determine the target weight: Based on the spatial distribution of the fermenters, set the categories of the K-means clustering algorithm and obtain the candidate solutions for each category. At the same time, set the target performance through the purpose-performance index mapping library. Based on the candidate solutions and the target performance, determine the comprehensive score of the candidate solutions.

[0012] S2.3: Constructing a three-dimensional target space: Using the normalized maturity prediction value as the X-axis, the normalized conversion rate value as the Y-axis, and the normalized energy consumption as the Z-axis, a three-dimensional target space is constructed.

[0013] S2.4: Determine the optimal solution: Based on the three-dimensional target space and the comprehensive score, obtain the comprehensive score of the frontier solution, and determine the three frontier solution comprehensive scores with the largest values. At the same time, obtain the curvature value of the comprehensive score of the frontier solution through B-spline fitting. The frontier solution corresponding to the maximum curvature value is the optimal solution.

[0014] S3: Regulation and control: Based on the optimal solution, the fermentation temperature, auxiliary material addition, and aeration rate are adjusted in real time.

[0015] Furthermore, the spatial distribution of cellulose / lignin, decay characteristics, and humidity gradient were obtained, including:

[0016] S1.1: Near-infrared spectral scanning: The mushroom residue is scanned by a near-infrared spectral probe equipped with an automatic rotating gimbal, and the azimuth and elevation angles of each spectral point are obtained. Based on the laser ranging data, the three-dimensional rectangular coordinates of the spectral points are determined, a spatial distribution model is constructed, and the grid data in the spatial distribution model is normalized. A heat map is generated based on the RGB values ​​of different colors.

[0017] S1.2: VOCs fingerprint detection: The gas in the fermenter is monitored by a metal oxide semiconductor gas sensor array, and a fingerprint database is constructed based on the monitored gas content;

[0018] S1.3: Moisture Gradient Monitoring: Microwave resonant moisture sensors are installed in different areas on the side wall of the fermenter to acquire the vertical moisture gradient difference and the abnormal mid-layer moisture value within the fermenter. These values ​​are then compared with a preset gradient threshold range and a preset anomaly threshold, respectively. Based on the comparison results, the fermentation state is determined. Specifically:

[0019] When the vertical moisture gradient difference is within a preset gradient threshold range and the abnormal value of the middle layer moisture is not less than a preset abnormal threshold, the fermentation state is normal fermentation; otherwise, the fermentation state is abnormal fermentation.

[0020] Furthermore, the lignin content of the grid data in the heatmap is obtained, and the lignin content is compared with a preset content threshold. Based on the comparison result, the marked voxel grid is determined, specifically as follows:

[0021] When the lignin content is greater than a preset content threshold, the corresponding voxel grid is marked; otherwise, the corresponding voxel grid is not marked.

[0022] Simultaneously, the interconnected marked voxel meshes are connected to determine the volume of the connected regions. This volume is then compared with a preset volume threshold. Based on the comparison result, priority processing regions are determined, specifically:

[0023] When the volume of the connected region is greater than a preset volume threshold, the corresponding connected region is the priority processing region; otherwise, the corresponding connected region is not the priority processing region.

[0024] Furthermore, the three-dimensional characteristics and temporal results of the original mushroom residue compost were determined, including:

[0025] S2.1.1: Determine the three-dimensional structural features: The priority processing area is resampled using a trilinear interpolation algorithm, and the normalized cellulose content and normalized lignin content are obtained based on the resampled voxel grid. At the same time, the three-dimensional structural features are obtained by outputting the data through a 3D-CNN model.

[0026] S2.1.2: Time-series prediction: Based on the fingerprint database and fermentation status, the normalized gas content and normalized moisture content are obtained, and the predicted value of decomposition degree and the amount of cellulose are obtained by inputting the LSTM network model.

[0027] Furthermore, based on the Z-value of the three-dimensional target space, the data in the three-dimensional target space is divided into multiple layers of data, and the comprehensive score of the leading edge solution for each layer is as follows:

[0028] ;

[0029] in: For the comprehensive scoring of cutting-edge solutions, This is the scoring weighting coefficient. For comprehensive scoring, As the frontier level, This is the hierarchical attenuation factor.

[0030] Furthermore, based on the optimal auxiliary material ratio and fermentation tank volume in the optimal solution, the optimal total amount of auxiliary materials is determined. At the same time, the optimal total amount of auxiliary materials is divided into multiple auxiliary material portions, and the auxiliary material portions are added through a screw conveyor and a spray gun.

[0031] Furthermore, based on the optimal aeration rate and air path composition in the optimal solution, the aeration rate of each air path is divided. Simultaneously, the average flow rate of the aeration valve is adjusted according to its flow rate, opening duration, and total duration. Specifically:

[0032] ;

[0033] in: The periodic average flow rate The pulse frequency, This represents the valve's maximum flow rate. Duration of a single activation. This represents the total cycle duration.

[0034] Furthermore, when adjusting the fermentation temperature, auxiliary material addition, and aeration rate in real time, the probability of toxin accumulation inside the fermenter is obtained, and the cellulase and fermentation time are adjusted in real time based on the toxin accumulation probability, including:

[0035] S3.1: Determining the probability of toxin accumulation: Based on the temperature, gas concentration, and moisture content of the fermenter, the data is discretized into multiple discrete temperature intervals, discrete gas intervals, and discrete moisture intervals. The probability of toxin accumulation is determined based on the toxin probabilities corresponding to each of these intervals. Specifically:

[0036] ;

[0037] in: This represents the final probability of toxin accumulation. As the baseline conditional probability, Let i be the weight of the i-th influencing factor. The total number of influencing factors. Index of influencing factors;

[0038] S3.2: Determine the risk level: Compare the cumulative probability of the toxin with a preset probability threshold range, and determine the risk level based on the comparison result; specifically:

[0039] When the cumulative probability of the toxin is less than the lower limit of the preset probability threshold range, the corresponding risk level is low risk; when the cumulative probability of the toxin is within the preset probability threshold range, the corresponding risk level is medium risk; when the cumulative probability of the toxin is greater than the upper limit of the preset probability threshold range, the corresponding risk level is high risk.

[0040] S3.3: Determine emergency measures: When the risk level is medium risk, increase the cellulase content and increase the stirring speed; when the risk level is high risk, extend the fermentation time and inject lime water into the fermentation tank. At the same time, obtain the corresponding toxin accumulation probability in real time, and repeat steps S3.1-S3.3 according to the toxin accumulation probability until the risk level is low risk.

[0041] Furthermore, based on the difference between the toxin accumulation probability and the lower limit of a preset probability threshold range, the cellulase dosage and fermentation extension time are determined, specifically as follows:

[0042] ;

[0043] in: This refers to the dosage of cellulase. Basic dosage This is the dose-probability coefficient. The intervention threshold, This represents the final probability of toxin accumulation. To extend the fermentation time, This is the risk-time conversion factor. Extend the duration to the maximum baseline.

[0044] A mushroom residue fermentation feed mixing device based on AI dynamic proportioning uses any one of the above-mentioned mushroom residue fermentation feed mixing control methods based on AI dynamic proportioning.

[0045] Compared with the prior art, the beneficial effects of the present invention are:

[0046] Firstly, this invention extracts spatial features through a 3D-CNN model, acquires temporal data through an LSTM network model, and combines spatial features and temporal data to achieve digital twin modeling of the fermentation process. At the same time, by constructing a three-dimensional target space of maturity-conversion rate-energy consumption, and combining K-means clustering algorithm and B-spline curvature optimization, the optimal parameters are determined, thereby shortening the response time for adjusting fermentation parameters.

[0047] Secondly, this invention, through the coordinated operation of a near-infrared spectrometer, a metal oxide semiconductor gas sensor array, and a microwave resonant moisture sensor, can acquire in real time the spatial distribution of cellulose / lignin, VOC fingerprint characteristics, and three-dimensional moisture gradient data of mushroom residue. This not only allows for the construction of a complete material characteristic database but also solves the problem of the one-sidedness of traditional single-sensor monitoring data.

[0048] Thirdly, this invention, through normalized energy consumption analysis and frontier solution optimization algorithm, can improve cellulose conversion rate and reduce unit processing energy consumption while ensuring that the degree of decomposition meets the standard.

[0049] Fourthly, this invention uses a discretized toxin accumulation probability model to determine the risk level and automatically increases the amount of cellulase added, increases the stirring speed, extends the fermentation time, and injects lime water, thereby reducing the fermentation failure rate.

[0050] Fifthly, this invention, through its application-performance index mapping library, allows for free switching between production modes such as organic fertilizer, feed, and biofuel, thereby not only improving equipment utilization and product qualification rate but also increasing system capacity. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating the method for controlling the fermentation and batching of mushroom residue in this invention.

[0052] Figure 2 This is a temperature control curve diagram in the present invention;

[0053] Figure 3 This is a comparison chart of aeration flow rate stability in this invention;

[0054] Figure 4 This is a comparison chart of indicator data in this invention;

[0055] Figure 5 This is a graph showing the dynamic changes in the probability of toxins and the changes in automatic intervention in this invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] In the process of resource utilization of mushroom residue, it can be used to produce various products such as organic fertilizer, animal feed, and biofuel. However, in organic fertilizer production, the focus is on the maturity of the material and its nutrient release characteristics. In feed substrate preparation, the focus is on increasing protein content and the effect of toxin degradation. When used for biofuel, the focus is on cellulose conversion rate and calorific value. Therefore, there are significant differences in the physicochemical properties of fermentation products in different application scenarios. Most existing batching systems adopt fixed ratios or manual adjustment based on experience, which cannot be dynamically optimized according to the target use, making it difficult for the same equipment to flexibly adapt to various application needs. This application uses a near-infrared spectrometer, a gas sensor array, and a moisture sensor to collect real-time data on the spatial distribution of cellulose / lignin, decay characteristics, and humidity gradient of mushroom residue. At the same time, it uses 3D-CNN and LSTM models to analyze the three-dimensional characteristics of the material and the changes in fermentation time, and constructs a three-dimensional target space with maturity, conversion rate, and energy consumption as dimensions. Simultaneously, the optimal fermentation parameters were determined using K-means clustering and B-spline fitting algorithms. By obtaining the toxin accumulation probability in real time, the addition of cellulase and fermentation time were dynamically adjusted, thereby achieving intelligent optimization control of the fermentation process.

[0058] Example 1

[0059] refer to Figures 1-4 This embodiment provides a method for controlling the fermentation of mushroom residue based on AI dynamic proportioning. The method includes the following steps:

[0060] Step S1: Multimodal data acquisition. This involves scanning the mushroom residue using a near-infrared spectrometer to obtain the spatial distribution of cellulose / lignin. Simultaneously, a metal oxide semiconductor gas sensor array and a microwave resonant moisture sensor are used to acquire the putrefaction characteristics and humidity gradient inside the fermenter. Details are as follows:

[0061] Step S1.1: Near-infrared spectroscopy scanning. A high-resolution near-infrared spectral probe (e.g., an InGaAs detector) with a wavelength range of 900-2500 nm is installed above the feed inlet of the fermenter. An automatic rotating gimbal is installed below the high-resolution near-infrared spectral probe. The automatic rotating gimbal is a two-dimensional rotating gimbal driven by a stepper motor, capable of horizontal rotation and pitch adjustment. In other words, the rotation of the automatic rotating gimbal drives the high-resolution near-infrared spectral probe to rotate, thereby scanning the mushroom residue on the conveyor belt from multiple angles.

[0062] Furthermore, a high-resolution near-infrared spectral probe is used to perform a spiral scan of the raw material surface at a linear velocity of 10 cm / s, acquiring 5 spectral points per square centimeter. Simultaneously, a gimbal encoder acquires the azimuth and elevation angles of each spectral point, and based on laser ranging data, the corresponding three-dimensional Cartesian coordinates of each spectral point are determined. Specifically, a spatial distribution model is generated using the Kriging interpolation algorithm and the corresponding three-dimensional Cartesian coordinates of each spectral point.

[0063] Furthermore, the grid data in the spatial distribution model is normalized to obtain the corresponding normalized values. Simultaneously, based on the normalized value of each grid data point and the RGB value of each color, the color corresponding to each grid data point is determined, thereby generating the corresponding heatmap.

[0064] In this embodiment, the three-dimensional grid data in the heat map is scanned, and the lignin content corresponding to each grid data is determined. Simultaneously, the obtained lignin content is compared with a preset content threshold (set specifically according to the specific data; this embodiment will not elaborate on it in detail, for example, 25%), and the corresponding voxel grid is marked based on the comparison result. Specifically:

[0065] When the obtained lignin content is greater than the preset content threshold (25%), the voxel grid corresponding to that lignin content is marked. Conversely, when the obtained lignin content is not greater than the preset content threshold (25%), the voxel grid corresponding to that lignin content is not marked.

[0066] Furthermore, based on the marked voxel meshes, the interconnected voxel meshes are connected to obtain multiple connected regions, and the volume size of each connected region is determined, specifically:

[0067] ;

[0068] in: Let V be the volume of the connected region. The number of voxels marked within a connected region. Let be the volume of the voxel mesh.

[0069] In this embodiment, the obtained connected region volume is compared with a preset volume threshold (the specific setting depends on the specific data; this implementation example will not elaborate on it, for example, 10cm). 3 The comparison is performed, and based on the comparison results, the priority processing area is determined. Specifically:

[0070] When the volume of the obtained connected region is greater than a preset volume threshold (e.g., 10cm) 3 When the volume of the obtained connected region is less than or equal to a preset volume threshold (e.g., 10cm), the corresponding connected region is the priority processing region. Conversely, when the volume of the obtained connected region is less than or equal to a preset volume threshold (e.g., 10cm), the priority processing region is determined by the volume of the connected region. 3 When ), the corresponding connected region is not a priority region for processing.

[0071] Step S1.2: VOCs fingerprint detection. Six mounting holes are evenly distributed on the top cover of the fermenter, and each hole is equipped with a metal oxide semiconductor gas sensor array to monitor ammonia, sulfide, and alcohol / aldehyde gases within the fermenter.

[0072] In this embodiment, a pulse-type gas pump is used to extract gas from the fermenter, and the extracted gas is transmitted to a metal-oxide-semiconductor gas sensor array to obtain the corresponding gas content. It is worth noting that in this embodiment, gas is extracted from the fermenter every 30 minutes.

[0073] Furthermore, based on the obtained gas content in the fermenter, the corresponding contents of ethanol, propionaldehyde, ammonia, hydrogen sulfide, butyric acid, and dimethyl disulfide were determined. Simultaneously, a fingerprint database was constructed based on the different gas contents, as shown in Table 1 below.

[0074] Table 1: Fingerprint Database Tables

[0075] Corruption levels Gas combination Early Ethanol and propionaldehyde Mid-term Ammonia and hydrogen sulfide Late Butyric acid and dimethyl disulfide

[0076] Step S1.3: Moisture gradient monitoring. Three sets of microwave resonant moisture sensors are installed on the side wall of the fermenter, with a vertical spacing of 30cm between each set. Specifically, the upper sensor monitors the surface area of ​​the fermenter, the middle sensor monitors the core of the fermentation pile, and the lower sensor monitors the bottom material.

[0077] Furthermore, during the monitoring of moisture content within the fermentation tank, the emitter of the microwave resonant moisture sensor emits a microwave signal that penetrates the interior of the fermentation tank and is received by the corresponding receiver of the microwave resonant moisture sensor. The moisture content within the fermentation tank is then determined based on the microwave signal received by the receiver.

[0078] In this embodiment, the moisture content at different locations in the upper, middle, and lower layers of the fermenter is determined based on the moisture content obtained from each set of microwave resonant moisture sensors. More specifically, based on the moisture content at different locations in the upper, middle, and lower layers of the fermenter, the corresponding vertical moisture gradient difference and the middle layer moisture anomaly value are obtained, specifically as follows:

[0079] ;

[0080] in: For vertical moisture gradient difference, This is an anomaly in mid-layer moisture. The moisture content of the lower layer, The moisture content of the upper layer, This represents the water content of the middle layer.

[0081] Furthermore, the obtained vertical moisture gradient difference and mid-layer moisture anomaly values ​​are compared with a preset gradient threshold range (specifically set based on the specific data, not elaborated in this implementation example, e.g., 5%-8%) and a preset anomaly threshold (specifically set based on the specific data, not elaborated in this implementation example, e.g., 3%), and the fermentation state within the fermenter is determined based on the comparison results. Specifically:

[0082] When the obtained vertical moisture gradient difference is within the preset gradient threshold range (e.g., 5%-8%), and the obtained mid-layer moisture anomaly value is not less than the preset anomaly threshold (e.g., 3%), the fermentation state in the fermenter is normal fermentation. Otherwise, the fermentation state in the fermenter is abnormal fermentation.

[0083] Step S2: Parameter Determination. Based on the priority processing area determined in Step S1.1, the fingerprint database constructed in Step S1.2, and the fermentation state determined in Step S1.3, the three-dimensional characteristics and temporal results of the original mushroom residue compost are determined, and the corresponding optimal parameters are determined according to the set target threshold. Details are as follows:

[0084] Step S2.1: Data Preprocessing. Based on the priority processing region determined in Step S1.1, the fingerprint database constructed in Step S1.2, and the fermentation state determined in Step S1.3, the corresponding three-dimensional features and time-series results are obtained using a 3D-CNN model and an LSTM network model. Details are as follows:

[0085] Step S2.1.1: Determine the three-dimensional structural features. Based on the priority processing area determined in the heat map in step S1.1, a trilinear interpolation algorithm is used to resample the data into a 1 cm³ standard voxel grid, where each voxel includes cellulose content, lignin content, and temperature value. Further, based on the resampled voxel grid, the maximum and minimum measured values ​​of cellulose and lignin content are determined. Then, based on these maximum and minimum measured values, the cellulose and lignin content corresponding to each voxel are normalized to obtain the corresponding normalized cellulose and lignin content.

[0086] Furthermore, based on the normalized cellulose content, normalized lignin content, and the temperature corresponding to the infrared thermography, three data channels are set in the 3D-CNN model, with each data channel corresponding to a specific data type. Specifically, the 3D-CNN model outputs 128-dimensional structural features. The first 32 dimensions represent the spatial distribution of cellulose, the middle 64 dimensions represent the lignin-temperature coupling characteristics, and the last 32 dimensions represent the structural heterogeneity characteristics.

[0087] Step S2.1.2: Time series prediction. Based on the fingerprint database constructed in step S1.2 and the fermentation state determined in step S1.3, the gas content obtained by the metal oxide semiconductor gas sensor array and the moisture content obtained by the microwave resonant moisture sensor are both normalized to obtain the corresponding normalized gas content and normalized moisture content.

[0088] Furthermore, normalized gas content and normalized moisture content are both used as inputs to the LSTM network model, and the corresponding predicted values ​​of decomposition degree and the amount of cellulose remaining are obtained as outputs.

[0089] Step S2.2: Determine the target weights. Based on the spatial distribution of the fermentation tank, the original mushroom residue is divided into five categories: the core area, the tank wall area, the lower and middle layers, the surface area, and the remaining area. Further, based on the 128-dimensional structural features obtained in step S2.1.1, candidate solutions corresponding to each category are obtained using the K-means clustering algorithm.

[0090] Furthermore, the loading thresholds for organic fertilizer, feed, and biofuel are set through a defined use-performance index mapping library. Specifically, when setting the loading threshold for organic fertilizer, the corresponding degree of decomposition is not less than 80%, and the proportion of available nitrogen is not less than 30%. When setting the loading threshold for feed, the corresponding crude protein content is not less than 18%, and the aflatoxin degradation rate is not less than 95%. When setting the loading threshold for biofuel, the corresponding cellulose conversion rate is not less than 75%, and the calorific value is not less than 16 MJ / kg. In this embodiment, the target use is set through the HMI interface, that is, the performance level corresponding to the target use is determined according to the defined use-performance index mapping library.

[0091] Specifically, based on the performance requirements corresponding to the set target purpose and the candidate solutions for each partition category, a comprehensive score is determined for each candidate solution. It's worth noting that there are multiple candidate solutions for each partition category. This means that the comprehensive scores of each candidate solution are compared, and the candidate solution with the highest comprehensive score is determined. The candidate solution corresponding to the highest comprehensive score is the optimal solution for that partition category.

[0092] In this embodiment, the formula for obtaining the comprehensive score is as follows:

[0093] ;

[0094] in: For comprehensive scoring, As a weight for maturity, The degree of decomposition deviation, As a conversion rate weight, For cellulose conversion rate, As energy consumption weight, This is the energy consumption coefficient.

[0095] Step S2.3: Construct a three-dimensional target space. This involves obtaining the maximum and minimum predicted decomposition values ​​based on the decomposition maturity predictions obtained through the LSTM network model in step S2.1.2. Then, each predicted decomposition maturity value is normalized based on these two values ​​to obtain a normalized predicted decomposition maturity value.

[0096] Furthermore, based on the spatial distribution features of cellulose obtained through the 3D-CNN model in step S2.1.1, the hourly variation of the corresponding spatial distribution features of cellulose is obtained, and the corresponding conversion rate normalization value is obtained through the inverse Sigmoid transform algorithm, specifically:

[0097] ;

[0098] in: This is the normalized value of the conversion rate. This is the original value for the cellulose conversion rate. This is a smoothing factor.

[0099] Furthermore, the unit energy consumption of the fermenter during the fermentation process is obtained through a power sensor (such as a smart meter). Specifically, within a preset time period, the unit energy consumption of the fermenter during the fermentation of the same raw material is obtained, and the minimum unit energy consumption within the preset time period is determined. Simultaneously, based on the determined minimum unit energy consumption, the obtained total energy consumption is normalized to obtain the normalized energy consumption.

[0100] In this embodiment, the obtained normalized maturity prediction value is used as the X-axis, the obtained normalized conversion rate value is used as the Y-axis, and the obtained normalized energy consumption is used as the Z-axis to construct the corresponding three-dimensional target space.

[0101] Step S2.4: Determine the optimal solution. Based on the 3D target space and Z-value obtained in Step S2.3, the data in the 3D target space is divided into multiple layers, and the comprehensive score of the frontier solution corresponding to each layer is obtained. Specifically:

[0102] ;

[0103] in: For the comprehensive scoring of cutting-edge solutions, This is the scoring weighting coefficient. For comprehensive scoring, As the frontier level, This is the hierarchical attenuation factor.

[0104] Furthermore, all the comprehensive scores of the frontier solutions are compared, and the three with the highest comprehensive scores are determined. Simultaneously, based on the data size corresponding to the comprehensive scores of the frontier solutions, the corresponding curvature values ​​are obtained through B-spline fitting, specifically:

[0105] ;

[0106] in: The curvature value, The second derivative of the solution at the frontier. It is the first derivative of the solution at the frontier.

[0107] In this embodiment, the curvature value is obtained based on the front solution corresponding to the comprehensive score of the three largest front solutions. At the same time, the three curvature values ​​are compared to determine the largest curvature value. The front solution corresponding to the largest curvature value is the optimal solution.

[0108] Step S3: Adjustment and Control. Based on the optimal solutions determined in Step S2.4—namely, the optimal temperature, optimal auxiliary material ratio, and optimal aeration rate—the temperature, auxiliary material addition, and aeration rate are adjusted in real time during the fermentation process. Specifically, based on the optimal temperature and real-time temperature, the total fermentation time is divided into multiple fermentation time periods. Simultaneously, based on the fermentation time interval and temperature rise interval corresponding to each fermentation time period, the corresponding temperature rise rate for each fermentation time period is determined. In other words, according to the determined temperature rise rate, corresponding temperature operations are performed within each fermentation time period.

[0109] Furthermore, based on the optimal proportions of auxiliary materials and the volume of the fermentation tank, the optimal total amount of auxiliary materials is determined and divided into multiple smaller portions. These smaller portions are then added to the fermentation tank via a screw conveyor and spray guns. Specifically, during the addition of the smaller portions, the materials are dissolved before being atomized and sprayed through the spray guns for both the initial and final additions. The remaining auxiliary materials are added directly to the fermentation tank via the screw conveyor.

[0110] Furthermore, based on the optimal aeration rate, the aeration rate of each air path is divided according to the air path composition: the aeration rate of the bottom air path is set to 70% of the total aeration rate, and the aeration rate of the side wall air paths is set to 30% of the total aeration rate. Simultaneously, the average flow rate of the aeration valve is adjusted according to the flow rate, opening duration, and total duration of the aeration valve, specifically as follows:

[0111] ;

[0112] in: The periodic average flow rate The pulse frequency, This represents the valve's maximum flow rate. Duration of a single activation. This represents the total cycle duration.

[0113] refer to Figure 2 , Figure 2 This is the temperature control curve in this embodiment, generated by... Figure 2It can be seen that during the high-temperature stage of 55℃-60℃, the temperature fluctuation under AI dynamic control is within ±1℃, while the fluctuation under traditional control reaches ±5℃. Furthermore, in the later stage of fermentation, at 72 hours, AI control maintains a temperature of 55℃, while traditional control has reduced it to 45℃. In other words, AI control can accurately respond to the thermodynamic requirements of the fermentation process, reducing the impact of temperature fluctuations on microbial activity.

[0114] refer to Figure 3 , Figure 3 This is a comparison chart of aeration flow rate stability in this embodiment, provided by... Figure 3 It can be seen that the flow rate fluctuation range decreased from ±0.3 m³ / h in traditional aeration to ±0.1 m³ / h, and the average flow rate increased from 0.5 m³ / h to 0.6 m³ / h, an increase of 20%. In other words, AI control can stabilize dissolved oxygen levels, promote the growth of aerobic microorganisms, and avoid local hypoxia or over-aeration.

[0115] refer to Figure 4 , Figure 4 This is a dynamic change graph of toxin probability and automatic intervention in this embodiment, generated by... Figure 4 It can be seen that: the maturity rate of composting has increased from 75% in the traditional process to 98.5%, an improvement of 31%, thus ensuring the quality of compost products. The cellulose conversion rate has increased from 60% in the traditional process to 78.5%, an improvement of 30%, thus improving the degradation efficiency of organic waste. Unit energy consumption has decreased from 0.5 kWh / kg in the traditional process to 0.35 kWh / kg, a reduction of 30%, thus reducing operating costs. The fermentation cycle has decreased from 20 days in the traditional process to 17 days, a reduction of 15%, thus improving production efficiency. Labor costs have decreased from 100% in the traditional process to 15%, a reduction of 85%, thus reducing reliance on labor.

[0116] This embodiment also provides a mushroom residue fermentation batching device based on AI dynamic proportioning, which uses the above-mentioned mushroom residue fermentation batching control method based on AI dynamic proportioning.

[0117] Example 2

[0118] This embodiment provides a method for controlling the fermentation of mushroom residue based on AI dynamic proportioning. The specific implementation method is the same as in Embodiment 1, except that, during real-time monitoring of the fermentation products inside the fermenter, the probability of toxin accumulation inside the fermenter is obtained in real time, and the cellulase and fermentation time are adjusted in real time based on the toxin accumulation probability. The invention will be illustrated below with specific examples of this embodiment.

[0119] In this embodiment, the cellulase and fermentation time are adjusted in real time according to the probability of toxin accumulation, as follows:

[0120] Step S3.1: Determine the probability of toxin accumulation. This involves discretizing the temperature, gas concentration, and moisture content within the fermenter based on the different fermenting materials, dividing them into multiple discrete temperature, gas, and moisture intervals. Simultaneously, based on these discrete intervals, the probability of toxin accumulation corresponding to each interval is determined.

[0121] In this embodiment, the temperature inside the fermenter is acquired using a high-resolution near-infrared spectral probe, the gas content inside the fermenter is acquired using a metal oxide semiconductor gas sensor array, and the moisture content inside the fermenter is acquired using a microwave resonant moisture sensor. In other words, based on the acquired temperature, a corresponding discrete temperature range is determined, and the probability of toxins corresponding to that discrete temperature range is obtained. Based on the acquired gas content, a corresponding discrete gas range is determined, and the probability of toxins corresponding to that discrete gas range is obtained. Based on the acquired moisture content, a corresponding discrete moisture range is determined, and the probability of toxins corresponding to that discrete moisture range is obtained.

[0122] Furthermore, based on the toxin probability corresponding to the discrete temperature range, the toxin probability corresponding to the discrete gas range, and the toxin probability corresponding to the discrete moisture range, the corresponding toxin accumulation probability is determined, specifically as follows:

[0123] ;

[0124] in: This represents the final probability of toxin accumulation. As the baseline conditional probability, Let i be the weight of the i-th influencing factor. The total number of influencing factors. Index of influencing factors.

[0125] Step S3.2: Determine the risk level. This involves comparing the final cumulative toxin probability obtained in step S3.1 with a preset probability threshold range (the specific range is set based on the data; this example does not elaborate on it, but is for example, 15%-30%), and determining the corresponding risk level based on the comparison result. Specifically:

[0126] When the final cumulative toxin probability is less than the lower limit of the preset probability threshold range (e.g., 15%), the corresponding risk level is low, and monitoring continues. When the final cumulative toxin probability is within the preset probability threshold range (e.g., 15%-30%), the corresponding risk level is medium, and the cellulase content is increased, along with the stirring speed. When the final cumulative toxin probability is greater than the upper limit of the preset probability threshold range (e.g., 30%), the corresponding risk level is high, and the fermentation time is extended, with 5% lime water added to the fermenter for adjustment.

[0127] Step S3.3: Determine emergency measures. Specifically, when the risk level is medium risk, the cellulase dosage is determined based on the difference between the final toxin accumulation probability and the lower limit of the preset probability threshold range.

[0128] ;

[0129] in: This refers to the dosage of cellulase. Basic dosage This is the dose-probability coefficient. The intervention threshold, This represents the final probability of toxin accumulation.

[0130] Specifically, during the process of adding cellulase, the cellulase is added to the priority treatment area according to the priority treatment area determined in step S1.1.

[0131] Furthermore, when the risk level is high, the required extended fermentation time is determined based on the difference between the final toxin accumulation probability and the lower limit of the preset probability threshold range. Specifically:

[0132] ;

[0133] in: To extend the fermentation time, This is the risk-time conversion factor. Extend the duration to the maximum baseline. The intervention threshold, This represents the final probability of toxin accumulation.

[0134] Specifically, during the process of determining the dosage of cellulase and the fermentation time, the corresponding toxin accumulation probability is obtained in real time, and steps S3.1-S3.3 are repeated based on the obtained toxin accumulation probability until the corresponding risk level is low risk.

[0135] refer to Figure 5 , Figure 5 This is a dynamic change graph of toxin probability and automatic intervention in this embodiment, generated by... Figure 5 It can be seen that the probability of toxin accumulation fluctuates and increases over time. When the probability of toxin accumulation is between 15% and 30%, the corresponding risk level is medium risk, and when it exceeds 30%, the corresponding risk level is high risk. Furthermore, when the probability of toxin accumulation exceeds 15%, intervention will be triggered to prevent the risk from escalating.

[0136] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended embodiments and their equivalents.

Claims

1. A method for controlling the fermentation of mushroom residue based on AI dynamic proportioning, characterized in that, Including: S1: Multimodal data acquisition: Using a sensor network, acquire spatial distribution of cellulose / lignin, decay characteristics, and humidity gradients, including: S1.1: Near-infrared spectral scanning: The azimuth and elevation angles of the spectral points are obtained by scanning, and a spatial distribution model is constructed based on the three-dimensional rectangular coordinates of the spectral points. At the same time, a heat map is generated based on the RGB values ​​of different colors, and the lignin and cellulose contents of the grid data in the heat map are obtained to determine the marked voxel grid. Based on the interconnected marked voxel grids, the volume of the connected region and the priority processing region are determined. S1.2: VOCs fingerprint detection: Monitor the gas inside the fermenter and construct a fingerprint database based on the gas content; S1.3: Moisture gradient monitoring: Obtain the vertical moisture gradient difference and the abnormal value of the middle layer moisture in the fermenter, and compare the results with the preset gradient threshold range and the preset abnormal threshold to determine the fermentation status; S2: Parameter Determination: Based on the target threshold, the three-dimensional characteristics of the original mushroom residue compost, and the time series results, the optimal parameters are determined, including: S2.1: Data Preprocessing: Based on the spatial distribution of cellulose / lignin, decay characteristics, and humidity gradient, the three-dimensional characteristics and time-series results of the original mushroom residue compost are determined, including: S2.1.1: Determine the three-dimensional structural features: Resample the priority processing area using a trilinear interpolation algorithm to obtain the normalized cellulose content / lignin content, and output the obtained three-dimensional structural features through a 3D-CNN model; S2.1.2: Time series prediction: Based on the fingerprint database and fermentation status, the normalized gas content / moisture content is obtained, and the predicted value of decomposition degree and the amount of cellulose remaining are output through the LSTM network model; S2.2: Determine the target weight: Based on the spatial distribution of the fermenters, set the categories for the K-means clustering algorithm, obtain candidate solutions for each category, and determine the comprehensive score of the candidate solutions based on the candidate solutions and the set target performance, specifically as follows: S=ω1·(1-f1)+ω2·f2+ω3·(1-f3) Where: S is the comprehensive score, ω1 is the maturity weight, f1 is the maturity deviation, ω2 is the conversion rate weight, f2 is the cellulose conversion rate, ω3 is the energy consumption weight, and f3 is the energy consumption coefficient. S2.3: Constructing a three-dimensional target space: Obtaining the normalized maturity prediction value, and based on the three-dimensional structural features and the inverse Sigmoid transform algorithm, obtaining the normalized conversion rate value, specifically: Where: Y is the normalized value of conversion rate, r is the original value of cellulose conversion rate, and a is the smoothing factor; A three-dimensional target space is constructed by using the normalized maturity prediction value as the X-axis, the normalized conversion rate value as the Y-axis, and the normalized energy consumption as the Z-axis. S2.4: Determining the Optimal Solution: Based on the three-dimensional target space and the comprehensive score, obtain the comprehensive score of the frontier solution. Then, by fitting B-spline and summing the three largest comprehensive scores of the frontier solutions, obtain the curvature value of the comprehensive score of the frontier solution. The frontier solution corresponding to the largest curvature value is the optimal solution. Specifically, the comprehensive score of the frontier solution is as follows: S fro =b×S+(1-b)×(1-L×δ) Wherein: S fro Let b be the comprehensive score of the frontier solution, S be the comprehensive score, L be the frontier level, and δ be the level decay factor. S3: Regulation and control: Based on the optimal solution, the fermentation temperature, auxiliary material addition, and aeration rate are adjusted in real time.

2. The method for controlling the fermentation of mushroom residue based on AI dynamic proportioning according to claim 1, characterized in that, Based on the optimal auxiliary material ratio and fermentation tank volume in the optimal solution, the optimal total amount of auxiliary materials is determined. At the same time, the optimal total amount of auxiliary materials is divided into multiple auxiliary material portions, and the auxiliary material portions are added through a screw conveyor and a spray gun.

3. The method for controlling the fermentation of mushroom residue based on AI dynamic proportioning according to claim 1, characterized in that, Based on the optimal aeration rate and air path composition in the optimal solution, the aeration rate of each air path is divided. Simultaneously, the average flow rate of the aeration valve is adjusted according to its flow rate, opening duration, and total duration. Specifically: Among them: Q cyc f is the periodic average flow rate. cyc Q is the pulse frequency. max For the valve's maximum flow rate, t ON t represents the duration of a single activation. tot This represents the total cycle duration.

4. The method for controlling the fermentation of mushroom residue based on AI dynamic proportioning according to claim 1, characterized in that, When adjusting the fermentation temperature, additives, and aeration rate in real time, the probability of toxin accumulation inside the fermenter is obtained, and the cellulase and fermentation time are adjusted in real time based on the toxin accumulation probability, including: S3.1: Determining the probability of toxin accumulation: Based on the temperature, gas concentration, and moisture content of the fermenter, the data is discretized into multiple discrete temperature intervals, discrete gas intervals, and discrete moisture intervals. The probability of toxin accumulation is determined based on the toxin probabilities corresponding to each of these intervals. Specifically: Where: P final P represents the final probability of toxin accumulation. CPT As the baseline conditional probability, α i Let m be the weight of the i-th influencing factor, m be the total number of influencing factors, and i be the index of the influencing factor. S3.2: Determine the risk level: Compare the cumulative probability of the toxin with a preset probability threshold range, and determine the risk level based on the comparison result; specifically: When the cumulative probability of the toxin is less than the lower limit of the preset probability threshold range, the corresponding risk level is low risk; when the cumulative probability of the toxin is within the preset probability threshold range, the corresponding risk level is medium risk; when the cumulative probability of the toxin is greater than the upper limit of the preset probability threshold range, the corresponding risk level is high risk. S3.3: Determine emergency measures: When the risk level is medium risk, increase the cellulase content and increase the stirring speed; when the risk level is high risk, extend the fermentation time and inject lime water into the fermentation tank. At the same time, obtain the corresponding toxin accumulation probability in real time, and repeat steps S3.1-S3.3 according to the toxin accumulation probability until the risk level is low risk.

5. The method for controlling the fermentation of mushroom residue based on AI dynamic proportioning according to claim 4, characterized in that, The cellulase dosage and fermentation extension time are determined based on the difference between the toxin accumulation probability and the lower limit of the preset probability threshold range, specifically as follows: Where: D is the cellulase dosage, B is the basal dosage, k is the dose-probability coefficient, and T is the dose-probability coefficient. the P is the intervention threshold. final ΔT represents the final probability of toxin accumulation, ΔT represents the extended fermentation time, C represents the risk-time conversion coefficient, and H represents the maximum baseline extension time.

Images