Intelligent bionic cultivation control system and method for cordyceps sinensis based on Internet of Things

Abstract

The invention discloses a cordyceps sinensis intelligent bionic cultivation control system and method based on the Internet of Things, and the method comprises the steps: collecting the temperature and humidity, gas concentration, illumination intensity and larva behavior data of a cordyceps sinensis matrix through a sensor network, and carrying out the fusion through an edge calculation gateway, and generating a dynamic environment map; constructing a plateau ecological digital twinborn model based on the dynamic environment map in combination with historical meteorological data and a host-strain interaction rule, and generating a staged control time sequence; based on a staged control time sequence, temperature and humidity coupling compensation is carried out through a feedforward-feedback composite algorithm, and actuator linkage is adjusted through fuzzy PID; aI vision is used for identifying a larva infection window period and accurate spraying is carried out, and hypha infection is monitored in combination with a near infrared spectrum to dynamically adjust the CO2 concentration; and constructing an LSTM (Long Short Term Memory) model to predict the growth risk in combination with a dynamic environment map and insect body monitoring data, and actively intervening in an abnormal environment. According to the invention, high-precision, dynamic and intelligent regulation and control of the growth environment of cordyceps sinensis are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of cordyceps cultivation, and in particular to an intelligent bionic cultivation control system and method for cordyceps based on the Internet of Things. Background Art

[0002] At present, Cordyceps sinensis is a rare Chinese medicinal material. Due to over-exploitation, wild resources are endangered and the price is high. Artificial cultivation has become the core path for sustainable development. However, traditional cultivation technology has significant bottlenecks: the growth of Cordyceps sinensis depends on the low temperature, variable temperature, high humidity and host insects unique to the plateau, as well as the precise symbiotic relationship between the bat moth larvae and the strain. The existing semi-artificial model is difficult to dynamically reproduce the complex high-altitude ecological parameters. The artificial environment is extensively regulated, with large temperature fluctuations, high humidity deviations, and poor synergy between factors such as light and CO2, resulting in insufficient larval infection rate, high fruiting body deformity rate, and unstable content of active ingredients (cordycepin, polysaccharides); the core difficulty lies in the difficulty in accurately, stably and dynamically reproducing and maintaining the complex, variable and mutually coupled high-altitude bionic environment required for the growth of Cordyceps sinensis, especially the precise environmental regulation of key biological processes such as host insect infection of Cordyceps fungi, sclerotium development, and fruiting body formation.

[0003] The defects of existing Cordyceps cultivation technology mainly stem from the following aspects: 1. Insufficient environmental control precision. It relies on threshold switch control equipment, and the response lag leads to drastic fluctuations in temperature and humidity, which cannot meet the sensitive needs of Cordyceps fungus during the infection period; multiple factors are independently regulated, and synergistic effects are ignored; 2. Lack of biomimetic dynamics. The static environment cannot simulate key biological rhythms such as the day and night temperature difference and seasonal changes in the plateau, resulting in the obstruction of fruiting body formation; 3. Ignoring the physiological rhythm of the bat moth larvae, and failing to establish a fungus-insect interaction environment model. The above defects are essentially the combined result of the low precision and non-dynamic nature of environmental control and the lack of intelligent decision-making, which directly result in low larval survival rate, high fruiting body deformity rate, and reduced effective ingredients of Cordyceps sinensis; high energy consumption, high losses and labor-intensive supervision drive up production costs.

[0004] Therefore, it is necessary to improve the existing Internet of Things-based Cordyceps sinensis intelligent bionic cultivation control system and method to solve the above problems. Summary of the Invention

[0005] The present invention overcomes the shortcomings of the existing technology and provides an intelligent bionic cultivation control system and method for Cordyceps sinensis based on the Internet of Things, aiming to solve the problem in the existing technology that key biological processes are difficult to accurately adapt due to low precision, non-dynamic environmental control and non-intelligent decision-making.

[0006] To achieve the above-mentioned purpose, the technical solution adopted by the present invention is: an intelligent bionic cultivation control method of Cordyceps sinensis based on the Internet of Things, comprising:

[0007] S1. Collect data on Cordyceps sinensis substrate temperature and humidity, gas concentration, light intensity, and larval behavior through a sensor network, and generate a dynamic environmental map through edge computing gateway integration.

[0008] S2. Based on the dynamic environmental map, combined with historical meteorological data and host-strain interaction rules, a digital twin model of plateau ecology is constructed to generate a staged control time series;

[0009] S3, based on the staged control sequence, temperature and humidity coupling compensation is performed through the feedforward-feedback composite algorithm, and the actuator linkage is adjusted through fuzzy PID;

[0010] S4. Use AI vision to identify the larval infection window and accurately spray, and combine near-infrared spectroscopy to monitor mycelial infection and dynamically adjust CO2 concentration;

[0011] S5. Combining dynamic environmental maps and insect monitoring data, we build an LSTM model to predict growth risks and proactively intervene in abnormal environments.

[0012] In a preferred embodiment of the present invention, the host-strain interaction rules include four growth stages, and their environmental requirements include:

[0013] Low-temperature diapause: The temperature is maintained at 8–10°C and the CO2 concentration is maintained at ≤1000ppm to induce the bat moth larvae to enter diapause;

[0014] Mycelial infection period: humidity 92% ± 2%, CO2 1800–2000 ppm, O2 maintained at 18-21%, triggering the mycelium to secrete chitinase to infect the larval body wall;

[0015] Sclerotium formation period: O2 is increased to ≥20%, humidity is 90%±1%, and CO2 is reduced to 1200-1500ppm to promote sclerotium development;

[0016] Fruiting body induction period: day and night temperature difference ≥ 8℃, PAR>200μmol, daytime CO2 controlled at 1500-1700ppm, nighttime CO2 ≤ 2000ppm, simulating nocturnal respiratory accumulation in the plateau.

[0017] In a preferred embodiment of the present invention, the construction of the dynamic environment map includes:

[0018] The cultivation chamber space is divided into a 20cm×20cm×10cm three-dimensional grid. Sensor data is synchronized using the NTP protocol with an error of <10ms.

[0019] The temperature and humidity fields are reconstructed using the Kriging interpolation algorithm: Among them, Z(s0) is the predicted value of the unsampled point, λ i is the weight coefficient, Z(s i) is the actual value measured by the sensor;

[0020] The larval behavior coordinates were superimposed on the environmental heat map to form a location-environmental parameter-behavioral state association matrix.

[0021] In a preferred embodiment of the present invention, the actuator linkage avoidance strategy includes:

[0022] Heating operation: Start the heater immediately and delay humidification for 120 seconds;

[0023] Cooling operation: start the refrigerator immediately and humidify simultaneously;

[0024] Spray operation: Turn off the ventilation fan to prevent a sudden drop in humidity. After spraying, resume ventilation according to the CO2 compensation curve.

[0025] In a preferred embodiment of the present invention, the plateau ecological digital twin model is constructed by a radial basis function neural network:

[0026] Input the grid data of the dynamic environmental map, historical meteorological data and host-strain interaction rules, and output the four-dimensional environmental field:

[0027] Spatial dimension: Through a three-dimensional spatial grid display, the cultivation chamber space is divided into a 20cm×20cm×20cm cube grid. The spatial position corresponding to each grid point is the estimated value of environmental parameters such as temperature, humidity, CO2 / O2 gas concentration, etc. output at each grid point;

[0028] Time dimension: Through time series forecasting, the environment for the next 72 hours is predicted with a time step of 5 minutes. Combined with the time series patterns in historical meteorological data, the ARIMA time series forecasting algorithm is used to provide forward-looking information in the time dimension for the dynamic regulation of the Cordyceps sinensis growth environment.

[0029] Biological response dimension: obtained through the biological state probability, outputs the probability values ​​of two biological states: infection success rate and fruiting body differentiation probability. The probability values ​​are calculated based on the environmental-biological response surface fitting model.

[0030] The method for controlling the intelligent bionic cultivation of Cordyceps sinensis based on the Internet of Things according to claim 1 is characterized in that: the feedforward-feedback composite algorithm in step S3 dynamically corrects the temperature and humidity coupling effect;

[0031] Feedforward compensation: Real-time calculation of compensation value ΔH=m×(T 目标 -T 当前 ); where ΔH is the humidity compensation, T is the temperature, and m is the temperature coefficient;

[0032] Feedback control: Fuzzy PID is used to adjust the actuator. The fuzzy rule base is established based on 49 groups of working condition experiments. The input is temperature / humidity error and error change rate, and the output is PID control quantity. Among them, e(t) is the current error, K p , K i , K d are the proportional gain coefficient, integral gain coefficient and differential gain coefficient; is the sum of historical errors, compensating for long-term environmental drift; is the error change rate.

[0033] In a preferred embodiment of the present invention, in step S4, the larval surface characteristics are captured in real time: the transparency of the body wall after molting, the activity frequency; the YOLOv7 model is used to identify key biomarkers; and the AI ​​larval behavior analysis is performed:

[0034] Molting is marked by the peeling of the stratum corneum, resulting in surface transparency. Among them, I tra I is the reflected light intensity of the transparent area of ​​the insect body in the near-infrared band, reflecting the degree of chitin degradation; op P is the reflected light intensity of the opaque area of ​​the insect body in the visible light band, reflecting the pigment deposition state of the body wall; 阈值 is the set molting threshold. When the ratio exceeds this value, the mycelial infection process control is triggered. The precursor to rigidification is when the movement speed is continuously less than 0.1 mm / s and the body color RGB mean is greater than 150 and the variance is less than 10, which is considered to be the molting state.

[0035] The bacterial solution spray command was triggered within 48 hours after molting, with a spray volume of 0.5 ml / larva and a droplet size of ≤20 μm.

[0036] In a preferred embodiment of the present invention, the dynamic adjustment of CO2 includes: infrared spectroscopy monitoring the degradation of chitin in the insect body, when the degradation is greater than 70%, the mycelium is marked as successfully colonized, and the respiratory compensation formula ΔCO2 = k(T 目标 -T 当前 ) adjusts the CO2 concentration, and k is the respiration coefficient.

[0037] In a preferred embodiment of the present invention, the LSTM model of step S5 includes:

[0038] Input layer: dynamic environment map data, insect infestation progress, and actuator historical status;

[0039] Network architecture: Two-layer LSTM hidden layer, fully connected layer outputs risk probability P r ;

[0040] The loss function uses weighted cross entropy loss to strengthen the weight of high-risk samples. The total loss value is Among them, w iis the sample weight, N is the total sample size, y i is the true risk label, indicating the actual risk status of the i-th sample, To predict the probability of risk;

[0041] When the predicted risk is less than 0.3, the current control sequence is maintained; if it exceeds 0.3, an alarm is issued, that is, local adjustments are made through PID control and timely adjustments are made. If it exceeds 0.7, it is a high-risk stage and the current sequence needs to be interrupted, the emergency protocol is executed, and an alarm is sent to the administrator.

[0042] The present invention provides an intelligent bionic cultivation control system for Cordyceps sinensis based on the Internet of Things, comprising:

[0043] A sensor network module is used to collect data on Cordyceps sinensis substrate temperature and humidity, gas concentration, light intensity, and larval behavior;

[0044] Edge computing gateway module, used to fuse sensor data to generate dynamic environment maps;

[0045] A plateau ecological digital twin model construction module is used to combine dynamic environmental maps, historical meteorological data, and host-strain interaction rules to generate a staged control time series;

[0046] Feedforward-feedback composite control algorithm module is used to dynamically correct the temperature and humidity coupling effect to achieve precise environmental control;

[0047] AI visual recognition and near-infrared spectroscopy monitoring modules are used to identify the larval infection window and monitor hyphae infection dynamics;

[0048] LSTM risk prediction and intervention module, used to build an LSTM model to predict growth risks and proactively intervene in abnormal environments;

[0049] Actuator linkage control module, used to adjust actuators such as heaters, humidifiers, coolers, ventilation fans, etc. according to control instructions to achieve precise environmental control

[0050] The present invention solves the defects existing in the background technology and has the following beneficial effects:

[0051] (1) The present invention proposes an intelligent bionic cultivation control system and method for Cordyceps sinensis based on the Internet of Things. The system collects environmental and biological data through a sensor network and generates a dynamic environmental map through edge computing. A digital twin model of plateau ecology is constructed, and a phased control sequence is generated by combining historical meteorological data with host-strain interaction rules. A feedforward-feedback composite algorithm and fuzzy PID regulation are used to achieve precise environmental control. AI vision is used to identify the larval infection window period and the gas concentration is dynamically adjusted in combination with near-infrared spectroscopy monitoring. The LSTM model is used to predict growth risks and actively intervene in abnormal environments, thereby achieving high-precision, dynamic, and intelligent control of the Cordyceps sinensis growth environment, improving the larval infection rate and active ingredient content, and reducing the deformity rate and production costs.

[0052] (2) The present invention constructs a digital twin model of plateau ecology, combines dynamic environmental maps, historical meteorological data and host-strain interaction rules to generate a phased control sequence, and overcomes the difficulties of existing technologies in accurately, stably and dynamically reproducing and maintaining the complex, changeable and mutually coupled high-altitude bionic environment required for the growth of Cordyceps sinensis, especially the difficulty in accurately regulating the environment of key biological processes such as host insect infection with Cordyceps fungi, sclerotium development, and stroma formation. Compared with existing technologies, the present invention accurately reproduces the key biological processes of the interaction between host insects and strains, lays a dynamic environmental benchmark for the precise regulation of Cordyceps sinensis growth, improves the survival rate of larvae, and reduces production costs.

[0053] (3) The present invention dynamically corrects the temperature and humidity coupling effect through a feedforward-feedback composite algorithm, adopts fuzzy PID to adjust the actuator linkage, and formulates an actuator linkage avoidance strategy, thereby solving the problems of drastic fluctuations in temperature and humidity and the independent regulation of multiple factors ignoring the synergistic effect in the existing technology. Compared with the existing technology, the present invention effectively reduces the phenomenon of excessive environmental fluctuations caused by equipment oscillation, reduces energy consumption, provides a precise and stable environmental basis for host-strain interaction, and ensures the stability of the Cordyceps sinensis growth environment.

[0054] (4) The present invention uses AI vision to identify the larval infection window period and spray accurately, combines near-infrared spectroscopy to monitor the mycelial infection dynamics to adjust the CO2 concentration, and uses the LSTM model to predict growth risks and actively intervene in abnormal environments. It solves the problems of the existing technology that it is difficult to accurately control the infection timing and cannot timely monitor the mycelial infection dynamics and predict growth risks. Compared with the existing technology, it realizes intelligent management of the infection process, significantly improves the success rate of mycelial infection, comprehensively optimizes the host-strain interaction, ensures the stability of Cordyceps sinensis growth and the accumulation of active ingredients, reduces the incidence of diseases and pests, and improves the quality and yield of Cordyceps sinensis.

[0055] (5) The present invention collects environmental and biological data in a multi-dimensional and high-precision manner through a sensor network, and generates a dynamic environmental map through edge computing fusion, thereby solving the problems of insufficient environmental control accuracy, delayed response, and poor synergy caused by independent control of multiple factors in the existing technology. Compared with the existing technology, it provides a real-time and accurate monitoring basis for the growth environment of Cordyceps sinensis, effectively meets the sensitive needs of key stages such as the Cordyceps fungus infection period, increases the larval infection rate, reduces the fruiting body deformity rate, and stabilizes the content of active ingredients. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments described in the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive efforts.

[0057] Figure 1 is a flowchart of the steps of a preferred embodiment of the present invention;

[0058] Figure 2 is a flow chart of the AI ​​recognition algorithm of a preferred embodiment of the present invention;

[0059] Figure 3 is a digital twin output flow chart of a preferred embodiment of the present invention;

[0060] Figure 4 is a real-time intervention flow chart of a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0061] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0062] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0063] Application Overview:

[0064] The growth of Cordyceps sinensis involves the biological process of cross-species symbiosis and the physiological rhythm of the host insect: the bat moth larvae enter the diapause period at low temperatures of 8-10℃, at which time the Cordyceps mycelium can penetrate the body wall; if the temperature fluctuates by >1℃, the larval immune system is activated, resulting in infection failure; the strain development stage specificity: the formation of fruiting bodies requires a step-by-step environmental switch of "low temperature induction → variable temperature stimulation", and static constant temperature causes the development of sclerotia to stagnate; in addition, temperature and humidity are strongly coupled, but the control system does not introduce a feedforward compensation algorithm. When the heater is started, the humidity drops sharply, triggering the humidification command. The alternating actions of the two cause oscillatory fluctuations, far exceeding the tolerance threshold of Cordyceps; the day and night temperature difference in the plateau environment activates the cold-resistant genes of Cordyceps fungi through epigenetic regulation, while the artificial constant temperature environment inhibits this pathway, resulting in undifferentiated fruiting bodies.

[0065] To address the above-mentioned issues, this application proposes an IoT-based intelligent bionic cultivation control system and method for Cordyceps sinensis. Through multi-dimensional, high-precision environmental sensing technology, it captures subtle changes in the Cordyceps sinensis growth environment in real time. Edge computing gateways are used to fuse these data to form a dynamic environmental map, providing data support for precise control. At the same time, combined with the plateau ecological digital twin model, the key biological processes of the interaction between host insects and strains are accurately reproduced. Furthermore, a feedforward-feedback composite control algorithm effectively addresses the problem of strong coupling between temperature and humidity. AI visual recognition technology accurately grasps the larval infection window period. Near-infrared spectroscopy monitoring dynamically adjusts CO2 concentrations, optimizing host-strain interactions in all directions. An LSTM model predicts growth risks, proactively intervenes in abnormal environments, and reduces the incidence of pests and diseases.

[0066] Exemplary methods:

[0067] like Figure 1 As shown, the intelligent bionic cultivation control method of Cordyceps sinensis based on the Internet of Things includes the following steps:

[0068] S1. Collect data on Cordyceps sinensis substrate temperature and humidity, gas concentration, light intensity, and larval behavior through a sensor network, and generate a dynamic environmental map through edge computing gateway integration.

[0069] S2. Based on the dynamic environmental map, combined with historical meteorological data and host-strain interaction rules, a digital twin model of plateau ecology is constructed to generate a staged control time series;

[0070] S3, based on the staged control sequence, temperature and humidity coupling compensation is performed through the feedforward-feedback composite algorithm, and the actuator linkage is adjusted through fuzzy PID;

[0071] S4. Use AI vision to identify the larval infection window and accurately spray, and combine near-infrared spectroscopy to monitor mycelial infection and dynamically adjust CO2 concentration;

[0072] S5. Combining dynamic environmental maps and insect monitoring data, we build an LSTM model to predict growth risks and proactively intervene in abnormal environments.

[0073] In step S1, the sensor network includes: substrate temperature and humidity sensors, gas concentration sensors, multispectral light sensors, insect behavior cameras and near-infrared spectrum probes; the data acquisition frequency is not less than 1 Hz, and an improved Kalman filter algorithm is used through the edge computing gateway to achieve spatiotemporal alignment and noise suppression of multimodal data, generating a three-dimensional dynamic environmental map containing spatial gradient field information.

[0074] Since Cordyceps sinensis larvae must strictly maintain a low temperature of 8–10°C during the diapause period, humidity fluctuations exceeding ±2% will lead to failure of mycelial infection;

[0075] The matrix temperature and humidity sensor includes:

[0076] Temperature monitoring uses the DS18B20 high-precision digital sensor with a detection accuracy of ±0.1°C; the DS18B20 converts temperature into a digital signal through a single bus protocol;

[0077] Humidity monitoring uses the SHT45 industrial-grade humidity sensor with a detection accuracy of ±1.8% RH; the SHT45 measures relative humidity based on a capacitive polymer film.

[0078] The substrate temperature and humidity sensors are embedded in the substrate at depths of 5cm, 10cm, and 15cm in a layered manner to monitor soil temperature gradient and moisture distribution in real time.

[0079] During the mycelial infection period, CO2 needs to be maintained at 1800–2000 ppm and O2 at 18-21%. When CO2 is greater than 2200 ppm, mycelial activity is inhibited, and when O2 is less than 18%, the larvae will suffocate.

[0080] Gas concentration sensors include: CO2 and O2 sensors, using Senseair K30 infrared sensors, with detection accuracies of ±50ppm and ±0.1% for CO2 and O2, respectively. CO2 molecules absorb infrared light of a specific wavelength, and the concentration is calculated by light intensity attenuation; O2 is measured using an electrochemical method.

[0081] The gas concentration sensor is hung on the top of the Cordyceps sinensis bionic cabin, and every 10-12 2 Deploy one to avoid airflow dead spots.

[0082] Ultraviolet UV-B band stimulates the differentiation of Cordyceps sinensis fruiting bodies, with PAR intensity > 200 μmol / m 2 / s can accelerate the metabolism of the worm; but PAR<100μmol / m 2 / s causes fruiting body deformity;

[0083] Multispectral light sensors include: PAR sensor and UV sensor, with detection ranges of: photosynthetically active radiation 400–700nm, ultraviolet band 280–400nm;

[0084] Multispectral light sensors are evenly distributed 30 cm above the culture dish in the cabin. The PAR sensor quantifies the light quantum flux density through a silicon photodiode; the UV sensor uses a GaP photodiode to detect the ultraviolet intensity.

[0085] Larval behavioral data was collected using an insect behavior camera and near-infrared spectroscopy probe. The FLIR Blackfly S BFS-U3-16S2M insect behavior camera used 940nm infrared light to collect larval activity frequency, body color changes, and molting behavior. A whitish body coloration indicates the molting period, when the body wall is fragile and the insect is optimally infected by mycelium. The Ocean Insight HDX near-infrared spectroscopy probe used 900-1700nm near-infrared light to collect chitin degradation in the insect body. A chitin degradation rate greater than 70% indicates successful mycelial colonization and requires adjustment of the CO2 concentration to promote the differentiation of fruiting body primordia.

[0086] During the diapause period (8-10°C), the mucus secretion of the bat moth larvae decreases, and the body segments shrink, causing the activity frequency to drop to less than 0.1 times / minute. During the molting period, the body color changes from dark brown to grayish white, which is the key window for mycelial infection.

[0087] A 940nm infrared camera prevents visible light from interfering with larval rhythms. Combined with a circularly polarized light source, it eliminates mucus reflections and uses an AI algorithm to output activity frequency, body color RGB mean, and variance.

[0088] like Figure 2 As shown in the figure, specifically, the background difference method is used to extract the larval outline, the YOLOv7 model is used to detect the changes in the transparency of the larval surface, and the cuticle peeling characteristics are extracted. For rigidification behavior, the insect body movement speed is tracked to be less than 0.1mm / s and lasts for more than 2 hours, which is determined to be a precursor to rigidification; when the RGB mean is greater than 150 and the variance is less than 10, the molting state is marked. The body wall of the larvae is thinnest within 48 hours after molting, which is the optimal window period for mycelial infection. Then, the trigger triggers the bacterial solution spray instruction to promote mycelial infection.

[0089] like Figure 3 As shown, the dynamic environment map is constructed:

[0090] Unify the sensor data into a spatiotemporal coordinate system. Specifically:

[0091] The cultivation chamber was divided into a three-dimensional grid with a length × width × height of 20 cm × 20 cm × 10 cm. Spatial gridding was performed to determine the specific location of each sensor in space, providing a basic unit for subsequent spatial data processing. All sensors were time-synchronized using the NTP protocol, with an error of less than 10ms, ensuring the consistency of the collected data in the time dimension and providing an accurate time reference for dynamic environmental monitoring.

[0092] The temperature and humidity fields are reconstructed based on the Kriging interpolation method to generate a three-dimensional thermal map: Among them, Z(s0) is the predicted value of the unsampled point, λ i is the weight coefficient, Z(s i ) is the actual value measured by the sensor;

[0093] The larval behavioral coordinates were superimposed on the environmental heat map to establish a location-environmental parameter-behavioral state association matrix.

[0094] Edge computing uses NVIDIA Jetson Xavier NX modules for computing.

[0095] Step S1 integrates a variety of high-precision sensors to comprehensively collect data on temperature, humidity, gas concentration, light intensity, and larval behavior. This data is then integrated using edge computing to generate a dynamic environmental map, providing a real-time, accurate monitoring foundation for the Cordyceps sinensis's growth environment. Building on this foundation, Step S2 further utilizes this dynamic environmental map, combined with historical meteorological data and host-strain interaction rules, to construct a digital twin model of the plateau ecosystem. This model generates a phased control sequence, enabling precise regulation of the Cordyceps sinensis growth environment.

[0096] In step S2, the historical meteorological data includes: historical day and night temperature difference and seasonal humidity curve of the plateau meteorological station;

[0097] Specific content of host-strain interaction rules:

[0098] The formation of Cordyceps sinensis is essentially a cross-species symbiotic process between the fungus, Cordyceps sinensis, and the host insect, the bat moth larvae. The interaction rules must strictly match the biological rhythms of both parties:

[0099] There are four growth stages:

[0100] Low-temperature diapause: Bat moth larvae enter a diapause state when exposed to sustained low temperatures. Their metabolic rate drops to 5% of that in the wild, and the chitin structure of their body walls relaxes, creating a window for hyphae to penetrate. During this process, when the temperature fluctuates by more than 1°C, the larvae's immune proteins activate, leading to infection failure. Environmental requirements include a temperature of 8–10°C and a CO2 concentration of ≤1000ppm to avoid interfering with the larvae's dormant metabolism.

[0101] During the mycelial infection phase, a high humidity environment induces the Cordyceps sinensis to secrete chitinase, which degrades the larval body wall and suppresses the host immune response. If the humidity is too low, the mycelium will dehydrate and become inactive, while if it is too high, the larvae will suffocate. Environmental requirements include: 92% ± 2% humidity, 1800–2000 ppm CO2 to stimulate the secretion of chitinase, and 18-21% O2 to ensure basal metabolism in the larvae. If the CO2 exceeds 2200 ppm, an exhaust command is triggered. At this time, spray induction is required to promote mycelial penetration through the body wall.

[0102] Sclerotium formation stage: larvae are rigid, movement speed is less than 0.1mm / s, hyphae proliferate inside the insect body, and environmental requirements are: O2 is increased to ≥20%, humidity is 90%±1%, and CO2 is reduced to 1200-1500ppm to avoid inhibition of sclerotium development;

[0103] Fruiting body induction period: The insect head swells, and the Cordyceps fungus HSP90 cold-resistant gene needs to be activated through a low-temperature pulse. Then the temperature is stepped up to simulate the snowmelt period on the plateau to trigger the differentiation of fruiting body primordia. The environmental requirements are: day and night temperature difference ≥8℃, PAR>200μmol, daytime CO2 is controlled at 1500-1700ppm to promote photosynthetic metabolism, and CO2 is allowed to rise to 2000ppm at night to simulate respiratory accumulation at night on the plateau.

[0104] In step S2, a digital twin model of plateau ecology is constructed:

[0105] The input data are: gridded spatiotemporal data of dynamic environmental maps, historical meteorological data, and interaction rules;

[0106] The radial basis function neural network was used to fit the environmental-biological response surface and establish parameter mapping.

[0107] Among them, y k represents the biological state, which is the biological response variable expected to be output by the model, with a value range of [0, 1]; x is the environmental vector, c i The historical optimal environmental center point is taken from the plateau environmental parameter values, corresponding to the sample points with representative environmental conditions in different growth stages or different production areas; ki is the weight coefficient, which indicates the influence of the i-th basis function on the k-th biological state output; φ(r) is the radial basis function, and the Gaussian kernel function is used here; ε is the Gaussian kernel width parameter, which is used to control the shape and range of the basis function.

[0108] Integrate real-time sensor data with historical meteorological characteristics to reconstruct the spatiotemporal environmental field. in, represents the reconstructed spatiotemporal environmental field, which is the estimated value of the environmental variable at the spatial position (x, y, z) and time t that integrates real-time and historical data. Its unit depends on the specific environmental variable. α is the real-time data weight, which reflects the proportion of real-time sensor data in the reconstructed environmental field and emphasizes the dominant role of current monitoring data on the environmental state. E sen Real-time sensor data provides observation values ​​of environmental variables, which come directly from the sensor network in the cultivation cabin and represent the actual measured values ​​of the environment at the current moment; E his (t sea ) is the historical meteorological data corresponding to the current seasonal phase t sea The environment variable value of t sea It represents the seasonal phase of the simulated plateau and is used to index the characteristics of different seasons in historical meteorological data, so that the reconstructed environmental field can reflect the impact of the seasonal laws of the plateau climate on the growth environment of Cordyceps sinensis.

[0109] The digital twin outputs the four-dimensional environment of the virtual cultivation cabin.

[0110] Spatial Dimension: Through a three-dimensional spatial grid display, the cultivation chamber space is divided into a 20cm×20cm×20cm cube grid. Each grid point corresponds to a spatial position (x, y, z). At each grid point, estimated values ​​of environmental parameters such as temperature, humidity, and CO2 / O2 gas concentration are output;

[0111] Temporal dimension: Using time series forecasting, we predict the environment for the next 72 hours, with a time step of 5 minutes. Starting from the current time t0, the forecast time is t0 + Δt. The forecast value is based on the environmental-biological response model and the reconstruction of the spatiotemporal environmental field, combined with the time series patterns in historical meteorological data, and obtained using the ARIMA time series forecasting algorithm. This provides forward-looking information on the temporal dimension for the dynamic regulation of the Cordyceps sinensis growth environment.

[0112] Biological response dimension: obtained through biological state probability, outputs the probability values ​​of two biological states: infection success rate and fruiting body differentiation probability. The probability values ​​are calculated based on the environmental-biological response surface fitting model. Based on current and predicted environmental parameters, it reflects the possibility of the key biological processes of Cordyceps sinensis host-strain interaction in a specific environment, providing a decision-making basis for the precise control of biological processes.

[0113] Step S2 constructs a digital twin model of plateau ecology, deeply integrates historical meteorological data with host-strain interaction rules, and accurately generates a phased control sequence, including low-temperature diapause period → mycelium infection period → sclerotium formation period → fruiting body induction period. This solves the problem that the static environment in traditional cultivation cannot simulate the day and night temperature difference, seasonal humidity changes, and fungus-insect symbiotic rhythm in the plateau, laying a dynamic environmental benchmark for subsequent coordinated control.

[0114] In step S3, the temperature and humidity coupling effect is dynamically corrected by a feedforward-feedback composite control algorithm;

[0115] When the temperature rises by 1°C, the saturated humidity increases by 4.5%. The compensation value is calculated in real time: ΔH=m×(T 目标 -T 当前 ), where ΔH is the humidity compensation, T is the temperature, and m is the temperature coefficient.

[0116] Feedback control uses fuzzy PID to adjust the actuator and uses triangular membership function to fuzzify the error and error change rate, output the control quantity, and the membership function of error and error change rate is: Where x is the input value, is the temperature / humidity error, c is the center point of the membership function, and w is the width of the membership function;

[0117] The temperature / humidity error (e) and error change rate (ec) were fuzzified using triangular membership functions and mapped into seven linguistic variables: negative large (NB), negative medium (NM), negative small (NS), zero (ZO), positive small (PS), positive medium (PM), and positive large (PB). A fuzzy rule base was established based on 49 sets of working condition experimental data.

[0118] The final output of the PID control quantity is Among them, e(t) is the current error, K p , K i , K d The proportional gain coefficient, integral gain coefficient, and differential gain coefficient are used. The proportional gain coefficient quickly responds to the current error and increases the heating power when the temperature deviates. The integral gain coefficient eliminates historical accumulated errors and delays mycelial infection caused by continuous low temperatures. The differential gain coefficient predicts future change trends and starts heating in advance when the temperature drops sharply to suppress fluctuations. is the sum of historical errors, compensating for long-term environmental drift; is the error rate of change, because a temperature change rate > 0.1℃ / min will activate the larval immune system.

[0119] Specifically, when a temperature error of e(t) = +0.5°C is detected and the actual temperature is lower than the target, a heating command is immediately output. If the low temperature has lasted for 2 hours, an additional +15% power compensation is applied. If the temperature is decreasing at a rate of 0.2°C / min, the power is increased by +20% in advance. Otherwise, a cooling command is issued to reduce the power.

[0120] The actuator linkage avoidance strategy is as follows: when the temperature needs to be increased, the heater is started immediately, and the humidification operation is performed after a delay of 120 seconds to prevent the humidity from dropping suddenly due to heating and triggering the humidifier action; when the temperature needs to be lowered, the cooler is started immediately and the humidification operation is performed immediately, and the humidity changes smoothly during the cooling process.

[0121] Humidity control execution: After the temperature stabilizes, the humidifier is driven according to the extreme compensation humidity value.

[0122] Step S3 solves the problem of excessive temperature and humidity fluctuations caused by equipment oscillation in traditional cultivation through a feedforward-feedback composite control algorithm. At the same time, through the actuator linkage avoidance strategy, heating / cooling priority, and humidity control delay of 120 seconds, energy consumption is reduced, providing a precise and stable environmental foundation for host-strain interaction.

[0123] In step S4, a 4K camera is used to capture the larvae's surface characteristics in real time: the transparency of the body wall after molting and the frequency of activity; the YOLOv7 model is used to identify key biomarkers;

[0124] AI larvae behavior analysis:

[0125] Molting is marked by the peeling of the stratum corneum, resulting in surface transparency. Among them, I tra I is the reflected light intensity of the transparent area of ​​the insect body in the near-infrared band, reflecting the degree of chitin degradation; op P is the reflected light intensity of the opaque area of ​​the insect body in the visible light band, reflecting the pigment deposition state of the body wall; 阈值 is the set molting threshold, when the ratio exceeds it, the mycelial infection process control is triggered;

[0126] The precursor to rigidity is when the movement speed is continuously less than 0.1 mm / s and the body color RGB mean is greater than 150 and the variance is less than 10;

[0127] Infection window determination: The golden infection period is within 48 hours after molting, when the body wall is thinnest and the mycelium penetration resistance is reduced. The bacterial solution spray command is automatically triggered, with a spray volume of 0.5 ml / larva and a droplet size of ≤20 μm.

[0128] Near-infrared spectroscopy continuously monitors mycelial infection at wavelengths of 1200-2500nm. Near-infrared spectroscopy penetrates the insect body and detects changes in the intensity of the chitin C-H bond absorption peak. Infection progress data is transmitted back to the S3 gas control module in real time to dynamically optimize CO2 concentration.

[0129] When the temperature rises, the larvae's respiration increases, leading to accelerated CO2 accumulation. Infrared spectroscopy (900-1700 nm) monitors the chitin degradation of the insect body. When the degradation rate is >70%, it indicates that the hyphae have successfully colonized. According to the formula ΔCO2=k(T 目标 -T 当前 ), k is the respiration coefficient; an independent fuzzy PID controller is used for CO2 / O2, and the avoidance strategy is as follows: the ventilation fan is temporarily turned off when the humidifier is spraying to prevent a sudden drop in humidity, and ventilation is restored according to the CO2 compensation curve after the spraying is completed.

[0130] Step S4 uses AI vision to accurately identify larval molting and ossification, monitors the progress of mycelial infection using near-infrared spectroscopy, and dynamically adjusts CO2 / O2 concentrations and spray volume, enabling intelligent management of the infection process and significantly improving the success rate of mycelial infection. Furthermore, step S5 combines dynamic environmental maps with insect monitoring data, employing an LSTM model to predict Cordyceps growth risks and proactively intervene in abnormal environments to ensure stable growth and the accumulation of active ingredients.

[0131] In step S5, a growth-environment correlation model is constructed:

[0132] Input layer:

[0133] Dynamic environmental maps: real-time temperature and humidity gradients, CO2 / O2 concentrations, and light intensity (PAR / UV);

[0134] Insect monitoring data: larval infestation progress, activity frequency, and body color changes;

[0135] Actuator history status: heater duty cycle, spray times, ventilation fan speed;

[0136] The model architecture is a two-layer LSTM network with a hidden layer dimension of 128 to capture spatiotemporal dependencies;

[0137] The fully connected layer outputs the risk probability P r ;

[0138] The loss function uses weighted cross entropy loss to strengthen the weight of high-risk samples. The total loss value is Among them, w i is the sample weight, N is the total sample size, y i is the true risk label, indicating the actual risk status of the i-th sample, To predict the risk probability.

[0139] When the predicted risk exceeds the threshold, the system automatically generates control instructions and sends them to the executor through the edge computing gateway.

[0140] like Figure 4 As shown, when the predicted risk is within 0-0.3 and the parameters are within the tolerance range, the current control sequence is maintained; if it exceeds 0.3, an alarm is issued, even if local adjustments are made through PID control and timely adjustments are made. If it exceeds 0.7, it is a high-risk stage and the current sequence needs to be interrupted, the emergency protocol is executed, and an alarm is sent to the administrator.

[0141] Example systems:

[0142] The intelligent bionic cultivation control system for Cordyceps sinensis based on the Internet of Things includes:

[0143] A sensor network module is used to collect data on Cordyceps sinensis substrate temperature and humidity, gas concentration, light intensity, and larval behavior;

[0144] Edge computing gateway module, used to fuse sensor data to generate dynamic environment maps;

[0145] A plateau ecological digital twin model construction module is used to combine dynamic environmental maps, historical meteorological data, and host-strain interaction rules to generate a staged control time series;

[0146] Feedforward-feedback composite control algorithm module is used to dynamically correct the temperature and humidity coupling effect to achieve precise environmental control;

[0147] AI visual recognition and near-infrared spectroscopy monitoring modules are used to identify the larval infection window and monitor hyphae infection dynamics;

[0148] LSTM risk prediction and intervention module, used to build an LSTM model to predict growth risks and proactively intervene in abnormal environments;

[0149] The actuator linkage control module is used to adjust actuators such as heaters, humidifiers, coolers, ventilation fans, etc. according to control instructions to achieve precise environmental control.

[0150] The sensor network module collects environmental and biological data in real time, and generates a dynamic environmental map through the edge computing gateway module; the plateau ecological digital twin model construction module generates control timing based on the map and historical data, and guides the feedforward-feedback composite control algorithm module to optimize temperature and humidity; the AI ​​visual recognition and near-infrared spectrum monitoring module uses image and spectral analysis to link the actuator linkage control module to adjust the spray and gas concentration; the LSTM risk prediction and intervention module continuously monitors the environment and insect body status, triggers abnormal intervention instructions, and ensures stable operation of the system.

[0151] The above description is based on the ideal embodiment of the present invention. Based on the above description, relevant personnel can make various changes and modifications without departing from the technical scope of this invention. The technical scope of this invention is not limited to the content of the specification and must be determined according to the scope of the claims.

Claims

1. An intelligent bionic cultivation control method for Cordyceps sinensis based on the Internet of Things, characterized in that: Including steps: S1. Collect data on Cordyceps sinensis substrate temperature and humidity, gas concentration, light intensity, and larval behavior through a sensor network, and generate a dynamic environmental map through edge computing gateway integration. S2. Based on the dynamic environmental map, combined with historical meteorological data and host-strain interaction rules, a digital twin model of plateau ecology is constructed to generate a staged control time series; S3, based on the staged control sequence, temperature and humidity coupling compensation is performed through the feedforward-feedback composite algorithm, and the actuator linkage is adjusted through fuzzy PID; S4. Use AI vision to identify the larval infection window and accurately spray, and combine near-infrared spectroscopy to monitor mycelial infection and dynamically adjust CO2 concentration; S5. Combining dynamic environmental maps and insect monitoring data, we build an LSTM model to predict growth risks and proactively intervene in abnormal environments.

2. The method for controlling the intelligent bionic cultivation of Cordyceps sinensis based on the Internet of Things according to claim 1, characterized in that: The host-strain interaction rules include four growth stages, and their environmental requirements include: Low-temperature diapause: The temperature is maintained at 8–10°C and the CO2 concentration is maintained at ≤1000ppm to induce the bat moth larvae to enter diapause; Mycelial infection period: humidity 92% ± 2%, CO2 1800–2000 ppm, O2 maintained at 18-21%, triggering the mycelium to secrete chitinase to infect the larval body wall; Sclerotium formation period: O2 is increased to ≥20%, humidity is 90%±1%, and CO2 is reduced to 1200-1500ppm to promote sclerotium development; Fruiting body induction period: day and night temperature difference ≥ 8℃, PAR>200μmol, daytime CO2 controlled at 1500-1700ppm, nighttime CO2 ≤ 2000ppm, simulating nocturnal respiratory accumulation in the plateau.

3. The method for controlling the intelligent bionic cultivation of Cordyceps sinensis based on the Internet of Things according to claim 1, characterized in that: The construction of the dynamic environment map includes: The cultivation chamber space is divided into a 20cm×20cm×10cm three-dimensional grid. Sensor data is synchronized using the NTP protocol with an error of <10ms. The temperature and humidity fields are reconstructed using the Kriging interpolation algorithm: Among them, Z(s0) is the predicted value of the unsampled point, λ i is the weight coefficient, Z(s i ) is the actual value measured by the sensor; The larval behavior coordinates were superimposed on the environmental heat map to form a location-environmental parameter-behavioral state association matrix.

4. The method for controlling the intelligent bionic cultivation of Cordyceps sinensis based on the Internet of Things according to claim 1, characterized in that: The actuator linkage avoidance strategy includes: Heating operation: Start the heater immediately and delay humidification for 120 seconds; Cooling operation: start the refrigerator immediately and humidify simultaneously; Spray operation: Turn off the ventilation fan to prevent a sudden drop in humidity. After spraying, resume ventilation according to the CO2 compensation curve.

5. The method for controlling the intelligent bionic cultivation of Cordyceps sinensis based on the Internet of Things according to claim 1, characterized in that: The plateau ecological digital twin model is constructed using a radial basis function neural network: Input the grid data of the dynamic environmental map, historical meteorological data and host-strain interaction rules, and output the four-dimensional environmental field: Spatial dimension: Through a three-dimensional spatial grid display, the cultivation chamber space is divided into a 20cm×20cm×20cm cube grid. The spatial position corresponding to each grid point is the estimated value of environmental parameters such as temperature, humidity, CO2 / O2 gas concentration, etc. output at each grid point; Time dimension: Through time series forecasting, the environment is predicted for the next 72 hours with a time step of 5 minutes. Combined with the time series patterns in historical meteorological data, the ARIMA time series forecasting algorithm is used to provide forward-looking information in the time dimension for the dynamic regulation of the Cordyceps sinensis growth environment; Biological response dimension: obtained through the biological state probability, outputs the probability values ​​of two biological states: infection success rate and fruiting body differentiation probability. The probability values ​​are calculated based on the environmental-biological response surface fitting model.

6. The method for controlling the intelligent bionic cultivation of Cordyceps sinensis based on the Internet of Things according to claim 1, characterized in that: The feedforward-feedback composite algorithm in step S3 dynamically corrects the temperature and humidity coupling effect; Feedforward compensation: Real-time calculation of compensation value ΔH=m×(T 目标 -T 当前 ); where ΔH is the humidity compensation, T is the temperature, and m is the temperature coefficient; Feedback control: Fuzzy PID is used to adjust the actuator. The fuzzy rule base is established based on 49 groups of working condition experiments. The input is temperature / humidity error and error change rate, and the output is PID control quantity. Among them, e(t) is the current error, K p , K i , K d are the proportional gain coefficient, integral gain coefficient and differential gain coefficient; is the sum of historical errors, compensating for long-term environmental drift; is the error change rate.

7. The method for controlling the intelligent bionic cultivation of Cordyceps sinensis based on the Internet of Things according to claim 1, characterized in that: In step S4, the larvae's surface characteristics are captured in real time: the transparency of the body wall after molting and the frequency of activity; the YOLOv7 model is used to identify key biomarkers; and AI larval behavior analysis is used to: Molting is marked by the peeling of the stratum corneum, resulting in surface transparency. Among them, I tra I is the reflected light intensity of the transparent area of ​​the insect body in the near-infrared band, reflecting the degree of chitin degradation; op P is the reflected light intensity of the opaque area of ​​the insect body in the visible light band, reflecting the pigment deposition state of the body wall; 阈值 is the set molting threshold. When the ratio exceeds this value, the mycelial infection process control is triggered. The precursor to rigidification is when the movement speed is continuously less than 0.1 mm / s and the body color RGB mean is greater than 150 and the variance is less than 10, which is considered to be the molting state. The bacterial solution spray command was triggered within 48 hours after molting, with a spray volume of 0.5 ml / larva and a droplet size of ≤20 μm.

8. The method for controlling the intelligent bionic cultivation of Cordyceps sinensis based on the Internet of Things according to claim 1, characterized in that: Dynamic adjustment of CO2 includes: infrared spectroscopy monitoring of chitin degradation in insects, when degradation is greater than 70%, it is marked that mycelium has successfully colonized, and respiratory compensation formula ΔCO2=k(T 目标 -T 当前 ) adjusts the CO2 concentration, and k is the respiration coefficient.

9. The method for controlling the intelligent bionic cultivation of Cordyceps sinensis based on the Internet of Things according to claim 1, characterized in that: The LSTM model in step S5 includes: Input layer: dynamic environment map data, insect infestation progress, and actuator historical status; Network architecture: Two-layer LSTM hidden layer, fully connected layer outputs risk probability P r ; The loss function uses weighted cross entropy loss to strengthen the weight of high-risk samples. The total loss value is Among them, w i is the sample weight, N is the total sample size, y i is the true risk label, indicating the actual risk status of the i-th sample, To predict the probability of risk; When the predicted risk is less than 0.3, the current control sequence is maintained; if it exceeds 0.3, an alarm is issued, that is, local adjustments are made through PID control and timely adjustments are made. If it exceeds 0.7, it is a high-risk stage and the current sequence needs to be interrupted, the emergency protocol is executed, and an alarm is sent to the administrator.

10. An intelligent bionic cultivation control system for Cordyceps sinensis based on the Internet of Things, based on the intelligent bionic cultivation control method for Cordyceps sinensis based on the Internet of Things according to any one of claims 1 to 9, characterized in that: include: A sensor network module is used to collect data on Cordyceps sinensis substrate temperature and humidity, gas concentration, light intensity, and larval behavior; Edge computing gateway module, used to fuse sensor data to generate dynamic environment maps; A plateau ecological digital twin model construction module is used to combine dynamic environmental maps, historical meteorological data, and host-strain interaction rules to generate a staged control time series; Feedforward-feedback composite control algorithm module is used to dynamically correct the temperature and humidity coupling effect to achieve precise environmental control; AI visual recognition and near-infrared spectroscopy monitoring modules are used to identify the larval infection window and monitor hyphae infection dynamics; LSTM risk prediction and intervention module, used to build an LSTM model to predict growth risks and proactively intervene in abnormal environments; The actuator linkage control module is used to adjust actuators such as heaters, humidifiers, coolers, ventilation fans, etc. according to control instructions to achieve precise environmental control.

Images