Intelligent and Synergistic Regulation Method and System for Dissolved Oxygen in Haematococcus pluvialis Culture Medium

Abstract

This invention discloses a method and system for intelligent synergistic regulation of dissolved oxygen in Haematococcus pluvialis culture medium, relating to the field of bioengineering. The control decision layer dynamically switches between growth-promoting and stress-induced regulation modes based on biomass and physiological state data during the Haematococcus pluvialis culture stage, achieving synergistic regulation of dissolved oxygen in conjunction with light, pH, and carbon sources. This invention dynamically switches between growth-promoting and stress-induced modes: during the nutrient stage, biomass accumulation is promoted through sinusoidal pulse fluctuations of dissolved oxygen; during the stress stage, astaxanthin synthesis is induced through high-oxygen, low-oxygen cycles, and blue light regulation. Compared to traditional static regulation, this results in increased biomass and astaxanthin production. The introduction of a photo-oxygen synergistic factor k dynamically adjusts the dissolved oxygen fluctuation range, resolving the light-dissolved oxygen mismatch problem and improving photosynthetic efficiency. A stress intensity index SI is established, and stress parameters are adaptively optimized based on the astaxanthin synthesis rate to avoid cell damage caused by excessive stress, thus improving the stability of the astaxanthin synthesis rate during the stress stage.

Description

Technical Field

[0001] This invention relates to the field of bioengineering technology, and in particular to a method and system for intelligent and coordinated regulation of dissolved oxygen in Haematococcus pluvialis culture medium. Background Technology

[0002] In the large-scale cultivation of *Haematococcus pluvialis*, dissolved oxygen concentration in the culture medium is a core environmental factor regulating its growth, reproduction, and astaxanthin synthesis. During the vegetative growth stage, suitable dissolved oxygen levels promote photosynthesis and biomass accumulation in algal cells. During the stress-induced stage (the critical period for astaxanthin synthesis), precise dissolved oxygen fluctuations (hyperoxygen stress-hypoxia recovery cycle) can significantly activate the activity of astaxanthin synthesis-related enzymes, thereby increasing the yield of the target product. However, existing dissolved oxygen regulation technologies for *Haematococcus pluvialis* suffer from the following key bottlenecks, making it difficult to meet the demands for efficient, stable, and low-cost industrial cultivation:

[0003] Existing technologies mostly employ fixed parameter control strategies, maintaining a single dissolved oxygen range throughout the process, such as 50%-70% air saturation. This fails to dynamically switch control logic based on the physiological states of *Haematococcus pluvialis* at different growth stages (nutritional phase, stress phase), such as biomass and photosynthetic efficiency. For example, excessively high dissolved oxygen during the nutritious phase can easily lead to oxygen free radical damage, while insufficient dissolved oxygen during the stress phase cannot effectively induce astaxanthin synthesis, ultimately resulting in a double inefficiency in both biomass accumulation and target product synthesis.

[0004] Dissolved oxygen concentration is influenced by a combination of factors, including ventilation rate, gas ratio, light intensity, pH, and carbon source supply, and is strongly correlated with the photosynthetic efficiency and metabolic rate of *Rhodochophora*. Current technologies often regulate dissolved oxygen in isolation (e.g., maintaining dissolved oxygen levels solely by adjusting ventilation rate), neglecting the synergistic effects of dissolved oxygen, light intensity, pH, and carbon source. For example, if dissolved oxygen is not optimized synchronously when light intensity is increased, photosynthetic efficiency will decrease; blindly increasing dissolved oxygen when carbon sources are insufficient can exacerbate cellular respiration and lead to energy waste.

[0005] Astaxanthin synthesis depends on a combination of stresses, including high oxygen and blue light. However, the intensity of these stresses, such as high oxygen concentration, duration, and blue light ratio, must be precisely matched to the tolerance of algal cells: too low an intensity results in a slow astaxanthin synthesis rate, while too high an intensity can inhibit algal cell activity or even lead to death. Current technologies often use fixed stress parameters, such as maintaining high oxygen for 20 minutes. This fails to adaptively adjust the stress intensity based on the astaxanthin synthesis rate of the previous cycle, leading to large fluctuations in stress efficiency and unstable product yield.

[0006] In summary, there is an urgent need for an intelligent control method and system that can achieve stage-adaptive, multi-parameter collaborative, data-driven optimization, and experience reusability, in order to break through the efficiency bottleneck of industrialized cultivation. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention provides a method and system for intelligent and coordinated regulation of dissolved oxygen in Haematococcus pluvialis culture medium. The technical solution adopted is as follows:

[0008] The intelligent and coordinated regulation method of dissolved oxygen in Haematococcus pluvialis culture medium includes the following steps:

[0009] Step 1: Construct a closed-loop control system that includes a perception layer, an execution layer, a control decision layer, and a cloud platform;

[0010] Step 2: The control decision layer dynamically switches between growth-promoting regulation mode and stress-induced regulation mode based on biomass data and physiological state data of Haematococcus pluvialis culture stage, so as to achieve synergistic regulation of dissolved oxygen, light, pH and carbon source.

[0011] Step 3: The edge computing gateway synchronizes full control data to the cloud platform at set intervals. The cloud platform performs multi-objective optimization simulation based on the digital twin model. The optimization objectives include final biomass, astaxanthin production and energy consumption.

[0012] Step 4: The cloud platform knowledge base uses graph neural networks to construct a correlation map of Haematococcus pluvialis strains, environmental parameters, regulation strategies, and output effects;

[0013] Step 5: When starting a new batch of culture, based on the current strain and environmental characteristics, retrieve the Top N historical cases with the highest similarity from the spectrum, and send their optimal control parameter sequence as the initial set value to the edge controller.

[0014] Optionally, in step 2, when the optical density OD value measured by the online biomass monitoring probe is less than the OD threshold and the maximum photochemical efficiency of PSII measured by the photosynthetic efficiency sensor is greater than the photochemical threshold, the growth promotion regulation mode is triggered.

[0015] The growth-promoting regulation mode controls the three-way mixing valve and the micro-nano bubble generator to make the dissolved oxygen level fluctuate sinusoidally in a 4-6 hour cycle within the range of 40%-80% air saturation.

[0016] Optionally, a photo-oxygen synergy factor k can be introduced. ;

[0017] in This is the current light intensity. It is the optimal light intensity. This is the current dissolved oxygen level. It is the optimal dissolved oxygen value;

[0018] The system dynamically adjusts the sine wave parameters. When the K value deviates from 1 by more than 20%, the original dissolved oxygen fluctuation range of 40%-80% is converged to the current instantaneous dissolved oxygen value by 50%.

[0019] Optionally, the stress-induced modulation mode is triggered when any of the following conditions are met:

[0020] The OD value measured by the online biomass monitoring probe of the stress-induced regulation mode is greater than or equal to the OD threshold, and the maximum photochemical efficiency of PSII measured by the photosynthetic efficiency sensor is less than or equal to the photochemical threshold.

[0021] The growth promotion mode was continuously operated for 7 days or more;

[0022] The stress-induced regulation mode controls the three-way mixing valve to rapidly raise dissolved oxygen to 120%-150% air saturation and maintain it for 10-20 minutes, during which the blue light ratio of the LED dimmer is adjusted to 30%-50%; then, through strong ventilation, the dissolved oxygen is rapidly reduced to 20%-30% air saturation and maintained for 45-60 minutes.

[0023] Optionally, a stress intensity index (SI) can be established for the stress-induced regulation model;

[0024] ;

[0025] in It refers to the duration of hyperoxia. It is the maximum permissible hyperoxia time. It is a high oxygen concentration. It is the maximum permissible high oxygen concentration. It is a low oxygen concentration. This is the baseline dissolved oxygen concentration. It's the blue light ratio. It is the maximum permissible blue light ratio. , , and These are the weighting indices;

[0026] The system dynamically adjusts the SI value of the current cycle based on the astaxanthin synthesis rate of the previous cycle. If the astaxanthin synthesis rate increases, the SI value increases by 10% increments, and vice versa, thus achieving adaptive optimization of stress intensity.

[0027] Optionally, a soft measurement model of dissolved oxygen based on a long short-term memory network can be constructed, taking historical ventilation, light, pH, temperature, and biomass as inputs and dissolved oxygen value as output, to provide redundant measurement when the reliability of hardware sensors is low.

[0028] Optionally, in step 3, the cloud platform first preprocesses all the data synchronized by the edge gateway; then it constructs three types of models: physical mass transfer, Haematococcus pluvialis growth, metabolism, and energy consumption calculation; subsequently, it sets a multi-objective function to maximize biomass, maximize astaxanthin, and minimize energy consumption, and sets safety constraints for dissolved oxygen and pH processes; finally, it uses a multi-objective optimization algorithm to solve the problem, selects the optimal control parameters, and converts them into execution instructions to be synchronized to the edge gateway.

[0029] Optionally, the retrieval process of the graph neural network in step 4 is as follows: using the strain genotype, initial pH, temperature range and culture objective of the current culture task as query vectors, calculating the similarity with historical case nodes in the knowledge graph, and returning the control strategy of the case with the highest similarity.

[0030] Optionally, in step 5, a query vector for the current training task is first constructed, which includes: strain characteristics, initial environmental parameters, and training target weights. The same-dimensional vector of historical cases in the knowledge graph is also extracted. Then, candidate cases that meet the similarity criteria are selected through similarity calculation. The top N cases are selected in descending order of similarity. If the similarity is the same, the case that is closer to the target is selected first. Finally, the control parameters of the Top N cases are weighted and fused, and the current environmental deviation is fine-tuned to generate an initial parameter sequence, which is then sent to the edge controller and labeled with reference case information.

[0031] The intelligent collaborative regulation system for dissolved oxygen in Haematococcus pluvialis culture medium is used to realize the intelligent collaborative regulation method for dissolved oxygen in Haematococcus pluvialis culture medium. The system includes a sensing layer, an execution layer, a control decision layer, and a cloud platform.

[0032] The multi-parameter sensors in the sensing layer include a dissolved oxygen sensor, a pH sensor, a temperature sensor, a photosynthetic efficiency sensor, and an online biomass monitoring probe.

[0033] The execution layer includes a three-way mixing valve, a micro / nano bubble generator, an LED dimmer, and a CO2 injection valve;

[0034] The control decision layer establishes a dynamic model with ventilation rate, gas ratio, stirring speed, and light intensity as inputs and dissolved oxygen and pH as outputs. The MPC controller calculates the optimal control command through rolling optimization and drives the actuators in the execution layer to work together to track the dynamic dissolved oxygen setpoint curve.

[0035] The cloud platform constructs a virtual model that corresponds one-to-one with the physical cultivation system, receives real-time data from the perception layer, and performs simulation and prediction.

[0036] The cloud platform stores all historical data.

[0037] In summary, the present invention has at least one of the following beneficial technical effects:

[0038] This invention provides a method and system for intelligent and coordinated regulation of dissolved oxygen in Haematococcus pluvialis culture medium. By monitoring biomass and photosynthetic efficiency in real time, it dynamically switches between growth promotion mode and stress induction mode: during the nutrient period, biomass accumulation is promoted by dissolved oxygen sinusoidal pulse fluctuations, and during the stress period, astaxanthin synthesis is induced by high-oxygen-low-oxygen cycle + blue light regulation. Compared with traditional static regulation, biomass is increased by 15%-25% and astaxanthin production is increased by 20%-30%.

[0039] Introducing a photo-oxygen synergistic factor k to dynamically adjust the dissolved oxygen fluctuation range solves the problem of light-dissolved oxygen mismatch, improving photosynthetic efficiency by 10%-15%; establishing a stress intensity index SI and adaptively optimizing stress parameters based on astaxanthin synthesis rate avoids cell damage caused by excessive stress, improving the stability of astaxanthin synthesis rate during stress by more than 25%.

[0040] A soft measurement model for dissolved oxygen based on a long short-term memory network was constructed to provide redundant measurement data in the event of hardware sensor failure, ensuring uninterrupted regulation. Combined with real-time data fusion from multiple parameter sensors, the reliability of monitoring data was improved to over 95%, significantly reducing the risk of culture failure due to sensor problems.

[0041] The cloud platform is based on a digital twin model and uses a multi-objective optimization algorithm to solve for the optimal control parameters. While ensuring product yield, it reduces energy consumption by optimizing parameters such as ventilation and light ratio. Compared with traditional control, the energy consumption per unit of astaxanthin yield is reduced by 18%-25%, and the cost competitiveness of industrial cultivation is significantly improved.

[0042] The system achieves a real-time closed loop of perception, decision-making, and execution through the control decision-making layer. Combined with remote monitoring and optimization of the cloud platform, it reduces manual intervention. Data collaboration between edge computing and the cloud platform enables efficient processing of all data and strategy iteration, reducing the workload of operation and maintenance personnel and meeting the intelligent management needs of large-scale training. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating the intelligent and synergistic regulation method of dissolved oxygen in Haematococcus pluvialis culture medium according to the present invention.

[0044] Figure 2 This is a dynamic change curve of dissolved oxygen and key parameters during a 14-day culture period in a specific embodiment of the present invention. Detailed Implementation

[0045] The present invention will be further described in detail below with reference to the accompanying drawings.

[0046] This invention discloses a method and system for intelligent and coordinated regulation of dissolved oxygen in Haematococcus pluvialis culture medium.

[0047] Reference Figure 1 and Figure 2Example 1, a method for intelligent and coordinated regulation of dissolved oxygen in Haematococcus pluvialis culture medium, includes the following steps:

[0048] Step 1: Construct a closed-loop control system that includes a perception layer, an execution layer, a control decision layer, and a cloud platform;

[0049] Step 2: The control decision layer dynamically switches between growth-promoting regulation mode and stress-induced regulation mode based on biomass data and physiological state data of Haematococcus pluvialis culture stage, so as to achieve synergistic regulation of dissolved oxygen, light, pH and carbon source.

[0050] Step 3: The edge computing gateway synchronizes full control data to the cloud platform at set intervals. The cloud platform performs multi-objective optimization simulation based on the digital twin model. The optimization objectives include final biomass, astaxanthin production and energy consumption.

[0051] Step 4: The cloud platform knowledge base uses graph neural networks to construct a correlation map of Haematococcus pluvialis strains, environmental parameters, regulation strategies, and output effects;

[0052] Step 5: When starting a new batch of culture, based on the current strain and environmental characteristics, retrieve the Top N historical cases with the highest similarity from the spectrum, and send their optimal control parameter sequence as the initial set value to the edge controller.

[0053] By adopting the above technical solution, during the cultivation of *Haematococcus pluvialis*, dissolved oxygen changes interact in real time with environmental parameters, algal cell physiological states, and actuator actions. A closed-loop system is needed to achieve a complete chain of data acquisition, decision-making calculation, command execution, and effect feedback. The sensing layer is responsible for capturing key information of the cultivation system in real time, providing a data foundation for regulation; the execution layer, as the command delivery carrier, directly changes dissolved oxygen and related environmental conditions; the control decision layer undertakes local real-time decision-making functions, ensuring rapid regulatory response; and the cloud platform undertakes the tasks of massive data storage, in-depth optimization, and knowledge accumulation. These four layers are interconnected, forming a closed loop without data interruptions or decision delays, avoiding the problems of data disconnection and decision lag in traditional open-loop regulation, and laying a system foundation for subsequent collaborative regulation.

[0054] The life cycle of *Haematococcus pluvialis* consists of two core stages: vegetative growth and stress metabolism. The requirements for dissolved oxygen and related environmental factors differ fundamentally between these stages. During the vegetative growth stage, the focus is on promoting cell division and biomass accumulation. At this time, dissolved oxygen needs to be maintained within a range suitable for photosynthesis, and moderate fluctuations can enhance cellular metabolic activity. During the stress-induced stage, the focus is on activating enzyme systems related to astaxanthin synthesis. This requires drastic changes in dissolved oxygen combined with adjustments in light and carbon sources to generate stress signals.

[0055] The control decision layer uses biomass data to reflect the degree of algal cell accumulation and physiological state data to reflect cellular photosynthetic activity and metabolic direction. Based on this, it dynamically switches between growth-promoting and stress-induced regulatory modes to ensure precise matching between the regulatory logic and the culture stage. Meanwhile, dissolved oxygen is not an isolated environmental factor; it, along with light, affects photosynthetic efficiency, and along with pH and carbon source, it affects cellular respiration and carbon metabolism. Therefore, while regulating dissolved oxygen, light, pH, and carbon source are adjusted synergistically to avoid metabolic imbalances caused by single-factor regulation and improve regulatory efficiency.

[0056] Edge computing gateways are located close to the cultivation site, enabling real-time collection and preliminary processing of control data. They synchronize full data to the cloud platform at set intervals, ensuring timely data transmission while avoiding excessive network load caused by high-frequency data transmission. The cloud platform possesses powerful computing and modeling capabilities, constructing a virtual scenario consistent with the actual cultivation system based on a digital twin model. This virtual environment can simulate cultivation effects under different control parameters, eliminating the need for repeated trial and error in the actual system and reducing trial-and-error costs and cultivation risks.

[0057] The industrial cultivation of *Hydrocotyle vulgaris* needs to balance output efficiency and cost control. Maximizing biomass or astaxanthin yield alone may lead to a surge in energy consumption, while minimizing energy consumption alone will sacrifice output. Therefore, the cloud platform sets up a multi-objective function that includes final biomass, astaxanthin yield, and energy consumption. A multi-objective optimization algorithm is used to solve for the optimal control parameters that ensure high output while controlling low energy consumption. The optimized parameters are then converted into execution instructions and synchronized to the edge gateway, achieving a closed loop of data-simulation-optimization-execution, thus balancing output and cost.

[0058] The cultivation effect of *Haematococcus pluvialis* is influenced by multiple factors, including strain characteristics, environmental parameters, and regulatory strategies. These factors exhibit complex nonlinear relationships, and different strains have varying tolerances to environmental parameters. Furthermore, the same regulatory strategy can produce different results under different environmental parameters. Traditional databases struggle to effectively capture these complex relationships, while graph neural networks excel at processing structured data with inter-node relationships. They can treat *Haematococcus pluvialis* strains, environmental parameters, regulatory strategies, and yield results as nodes in a knowledge graph. Through algorithmic learning and quantification of the relationship strength between nodes, a structured knowledge system is formed.

[0059] This correlation map not only enables the orderly storage of historical culture data, but also transforms scattered data into reusable knowledge, providing data support for parameter setting of subsequent new batches of culture and avoiding knowledge waste caused by idle historical data.

[0060] When cultivating a new batch of Haematococcus pluvialis, if the characteristics of the current strain and the initial environmental features are similar to a certain type of historical case, the appropriate control parameters are also of high reference value. Based on this, a query vector containing the characteristics of the current strain, initial environmental parameters, and the weight of the cultivation target is first constructed. Then, the same-dimensional vectors of historical cases are extracted from the knowledge graph. The Top N historical cases most similar to the current cultivation task are selected through similarity calculation to ensure the relevance and reliability of the reference cases.

[0061] While the control parameters from different historical cases are valuable for reference, they need to be adjusted to account for deviations between the current and historical environments to avoid inaccurate control due to direct application. Therefore, the control parameters from the Top N cases are weighted and fused, then fine-tuned based on current environmental deviations to generate an initial parameter sequence suitable for the current culture task, which is then distributed to the edge controller. This significantly shortens the parameter tuning cycle for new batches of culture, reduces inefficiencies caused by unsuitable parameters in the initial stage, and, by labeling reference case information, facilitates subsequent traceability and parameter optimization, improving the stability and start-up speed of new batches of culture.

[0062] In Example 2, in step 2, when the optical density OD value measured by the online biomass monitoring probe is less than the OD threshold and the maximum photochemical efficiency of PSII measured by the photosynthetic efficiency sensor is greater than the photochemical threshold, the growth promotion regulation mode is triggered.

[0063] The growth-promoting regulation mode controls the three-way mixing valve and the micro-nano bubble generator to make the dissolved oxygen level fluctuate sinusoidally in a 4-6 hour cycle within the range of 40%-80% air saturation.

[0064] Example 3: Introducing a photo-oxidation synergistic factor k. ;

[0065] in This is the current light intensity. It is the optimal light intensity. This is the current dissolved oxygen level. It is the optimal dissolved oxygen value;

[0066] The system dynamically adjusts the sine wave parameters. When the K value deviates from 1 by more than 20%, the original dissolved oxygen fluctuation range of 40%-80% is converged to the current instantaneous dissolved oxygen value by 50%.

[0067] Example 4: The stress-induced modulation mode is triggered when any of the following conditions are met:

[0068] The OD value measured by the online biomass monitoring probe of the stress-induced regulation mode is greater than or equal to the OD threshold, and the maximum photochemical efficiency of PSII measured by the photosynthetic efficiency sensor is less than or equal to the photochemical threshold.

[0069] The growth promotion mode was continuously operated for 7 days or more;

[0070] The stress-induced regulation mode controls the three-way mixing valve to rapidly raise dissolved oxygen to 120%-150% air saturation and maintain it for 10-20 minutes, during which the blue light ratio of the LED dimmer is adjusted to 30%-50%; then, through strong ventilation, the dissolved oxygen is rapidly reduced to 20%-30% air saturation and maintained for 45-60 minutes.

[0071] Example 5: Establishing the stress intensity index SI in the stress-induced regulation model;

[0072] ;

[0073] in It refers to the duration of hyperoxia. It is the maximum permissible hyperoxia time. It is a high oxygen concentration. It is the maximum permissible high oxygen concentration. It is a low oxygen concentration. This is the baseline dissolved oxygen concentration. It's the blue light ratio. It is the maximum permissible blue light ratio. , , and These are the weighting indices;

[0074] The system dynamically adjusts the SI value of the current cycle based on the astaxanthin synthesis rate of the previous cycle. If the astaxanthin synthesis rate increases, the SI value increases by 10% increments, and vice versa, thus achieving adaptive optimization of stress intensity.

[0075] By adopting the above technical solution, the core objective of the vegetative growth stage of Haematococcus pluvialis is to promote algal cell division and proliferation and accumulate sufficient biomass. This stage requires meeting the dual conditions of biomass not reaching the threshold and sufficient photosynthetic activity. The triggering logic is designed based on these two key physiological state indicators.

[0076] The significance of optical density (OD) value: The OD value directly reflects the biomass concentration of Haematococcus pluvialis in the culture medium. If the OD value is less than the OD threshold, it indicates that the current number of algal cells has not yet reached the target accumulation amount of the vegetative growth stage, and growth still needs to be continuously promoted; if the OD value reaches the target, it means that the cells have the biomass basis to enter the subsequent metabolic stage.

[0077] The significance of the maximum photochemical efficiency of PSII: The maximum photochemical efficiency of PSII is the core indicator for measuring the activity of the photosynthetic system II of Haematococcus pluvialis. If its value is greater than the photochemical threshold, it indicates that the photosynthetic mechanism of algal cells is functioning well, can efficiently utilize light energy for photosynthesis and material synthesis, and has the physiological basis for rapid growth. If the value decreases, it indicates that the photosynthetic activity is declining, the cell growth potential is decreasing, and the direction of regulation needs to be adjusted.

[0078] When both indicators are met simultaneously, it indicates that the cells are in a critical period of vegetative growth characterized by low biomass and high photosynthetic activity, triggering a growth-promoting regulatory mode that can precisely match the cell's growth needs and avoid activating other regulatory modes at inappropriate stages, thus preventing resource waste.

[0079] Under the growth-promoting regulation mode, dissolved oxygen control adopts a method of 40%-80% air saturation and 4-6 hour periodic sinusoidal pulse fluctuation. The core logic revolves around adapting to the photosynthetic metabolism pattern and enhancing cell activity.

[0080] The principle of dissolved oxygen range selection: 40%-80% air saturation is the suitable dissolved oxygen range for the vegetative growth stage of Haematococcus pluvialis. Below 40% will lead to the obstruction of the photosynthetic electron transport chain and limit the efficiency of photosynthesis; above 80% may cause the accumulation of oxygen free radicals, causing oxidative damage to algal cells. This range can balance photosynthetic needs and cell protection.

[0081] Advantages of sinusoidal pulse oscillation: Compared with constant dissolved oxygen control, sinusoidal oscillation is closer to the dynamic changes of dissolved oxygen in the natural environment, and can continuously provide appropriate stimulation to algal cells. During the rising phase of dissolved oxygen, it can meet the oxygen demand when the photosynthetic rate of cells increases, while during the falling phase, it avoids metabolic inhibition caused by excessive oxygen concentration, promotes the maintenance of cell metabolic enzyme activity at a high level, and thus accelerates the rate of biomass accumulation.

[0082] The actuator-assisted control principle: The three-way mixing valve controls the oxygen content of the gas entering the culture medium by adjusting the mixing ratio of air, oxygen, and other gases; the micro / nano bubble generator produces tiny bubbles, increasing the contact area between the gas and the culture medium and improving dissolved oxygen mass transfer efficiency. Working together, these two components precisely control the dissolved oxygen concentration to fluctuate according to a set sine wave curve, avoiding excessive fluctuations or deviations from the target range, thus ensuring precise control.

[0083] The photosynthetic efficiency of *Rhodococcus pluvialis* depends on the synergistic matching of light and dissolved oxygen. Light provides energy for photosynthesis, while dissolved oxygen is an important intermediate product and environmental factor in photosynthetic metabolism. Both need to be maintained at suitable levels to maximize photosynthetic efficiency. The light-oxygen synergistic factor *k* is an indicator used to quantify the degree of matching between the current light-oxygen state and the optimal light-oxygen state.

[0084] When the k value is close to 1, it indicates that the current light intensity and dissolved oxygen value are close to their respective optimal values, the light-oxygen matching degree is high, and the photosynthetic system can operate efficiently.

[0085] When the k value deviates from 1, it means that the light and oxygen state is unbalanced (such as excessive light but insufficient dissolved oxygen, or excessive dissolved oxygen but insufficient light). If the original dissolved oxygen regulation logic is maintained at this time, it will aggravate the imbalance of photosynthetic metabolism and lead to a decrease in photosynthetic efficiency.

[0086] When the k value deviates from 1 by more than 20%, it indicates that the light-oxygen imbalance has reached a level requiring intervention. At this point, the original dissolved oxygen fluctuation range of 40%-80% is narrowed to the current instantaneous dissolved oxygen value by 50%. The core purpose is to quickly alleviate the light-oxygen imbalance and maintain stable photosynthetic efficiency.

[0087] The necessity of narrowing the fluctuation range: When there is an imbalance between light and oxygen, the already large fluctuations in dissolved oxygen may further amplify the imbalance. For example, when the light intensity is far below the optimum value (the k value is too small), the photosynthetic rate of algal cells decreases, and the consumption of dissolved oxygen decreases. If the upper limit of dissolved oxygen is still maintained at 80%, it will lead to the continuous accumulation of dissolved oxygen and cause oxygen damage. At this time, narrowing the fluctuation range can prevent dissolved oxygen from shifting to extreme values ​​and reduce the negative impact of the imbalance.

[0088] The design logic of a 50% convergence ratio is as follows: a 50% convergence ratio can significantly reduce the amplitude of dissolved oxygen fluctuations and quickly bring the dissolved oxygen state closer to the current suitable level, without completely eliminating dissolved oxygen fluctuations (avoiding metabolic adaptation problems caused by constant dissolved oxygen). It finds a balance between alleviating imbalance and maintaining cell activity stimulation, ensuring that photosynthetic efficiency can still be maintained within an acceptable range when there is a light-oxygen imbalance. After the light or other conditions are adjusted, the original fluctuation range can be gradually restored.

[0089] Astaxanthin synthesis in Haematococcus pluvialis requires a stress state where cell growth has essentially ceased and metabolism shifts to secondary metabolism. The triggering conditions are designed to precisely capture the transition signals at this stage, while a time safety net mechanism is included to ensure the orderly progress of the culture process.

[0090] The dual indicators of biomass and photosynthetic activity trigger the process: when the OD value is greater than or equal to the OD threshold, it indicates that the biomass has reached the target accumulation level and the cells have the material basis for synthesizing astaxanthin; when the maximum photochemical efficiency of PSII is less than or equal to the photochemical threshold, it indicates that the photosynthetic activity is declining and the cell growth potential is decreasing. At this time, metabolic resources can be tilted towards secondary metabolism (astaxanthin synthesis), which is in line with the physiological law of stage transition.

[0091] Time-based safety net trigger: The growth promotion mode runs continuously for 7 days, which is based on the normal cycle setting of the vegetative growth stage of Haematococcus pluvialis. Even if the biomass or photosynthetic activity index does not reach the threshold due to environmental fluctuations, the biomass accumulation has been basically completed in 7 days. If the growth promotion mode is maintained, it will lead to excessive cell proliferation and depletion of nutrients in the culture medium, which will affect the subsequent astaxanthin synthesis. Therefore, a time-based safety net is set to ensure that the stage transition is not delayed.

[0092] The stress-induced regulatory model constructs a complex stress environment through hyperoxia-hypoxia cycling and blue light regulation. The core of this model is the activation of astaxanthin synthesis-related enzyme systems (such as β-carotene ketolase) while simultaneously protecting cell viability.

[0093] The design principle of hyperoxia stress is to rapidly increase dissolved oxygen to 120%-150% of air saturation and maintain it for 10-20 minutes. High oxygen concentration will produce an appropriate amount of reactive oxygen species (ROS) in cells. ROS, as signaling molecules, can activate the regulatory pathway of astaxanthin synthesis. The maintenance time is controlled at 10-20 minutes because excessive hyperoxia will lead to excessive accumulation of ROS and cause cell damage. This duration can balance the generation of stress signals and cell protection.

[0094] The principle of blue light synergy: Blue light (30%-50% ratio) can regulate the expression of key genes for astaxanthin synthesis through photoreceptors (such as rhodopsin-like proteins), forming a synergistic effect with hyperoxic stress and significantly improving stress-induced efficiency; a blue light ratio of 30%-50% can effectively activate gene expression without causing damage to photosynthetic structures due to excessive blue light.

[0095] The principle of hypoxia recovery: After hyperoxia stress, the dissolved oxygen is rapidly reduced to 20%-30% of the air saturation through strong ventilation and maintained for 45-60 minutes. This is to allow cells to clear excess ROS in a hypoxic environment and restore metabolic balance. If hyperoxia continues, cells will lose the ability to synthesize astaxanthin due to oxidative damage. The hypoxia recovery phase can repair cell function and prepare for the next cycle of stress. Strong ventilation can ensure that dissolved oxygen drops rapidly and avoid the recovery phase being too long, which would affect the efficiency of stress.

[0096] The efficiency of astaxanthin synthesis in Haematococcus pluvialis is directly related to the intensity of stress, but a single stress factor (such as high oxygen concentration or blue light ratio) cannot fully reflect the stress effect. It is necessary to integrate multi-dimensional stress factors to construct a comprehensive index. The design of SI is based on the law that multiple factors synergistically affect the stress effect.

[0097] SI's factor selection logic: The duration and concentration of high oxygen directly determine the intensity of high oxygen stress; low oxygen concentration affects cell recovery efficiency and indirectly affects the stress cycle effect; the proportion of blue light is the core indicator of light stress. These four factors together constitute the key dimensions of the stress environment. By adjusting the influence of each factor through the weight index, the characterization accuracy of SI for the actual stress effect can be optimized according to the differences in the sensitivity of different Haematococcus pluvialis strains to stress.

[0098] The significance of setting maximum allowable values: The maximum allowable values ​​for high oxygen time, high oxygen concentration, and blue light ratio are safety thresholds set based on the tolerance limit of Haematococcus pluvialis cells. This avoids cell death caused by a single factor exceeding the threshold and ensures that the SI value fluctuates within the range of effective stress and cell safety.

[0099] Optimization of stress intensity requires feedback based on astaxanthin synthesis effects. The SI value of the current cycle is adjusted based on the astaxanthin synthesis rate of the previous cycle. The core is to achieve precise matching of cellular stress response characteristics.

[0100] Positive feedback adjustment (SI value increased by 10%): If the astaxanthin synthesis rate accelerated in the previous cycle, it indicates that the current stress intensity has not reached the cell's optimal response threshold. Appropriately increasing the SI value (such as increasing the high oxygen concentration or prolonging the high oxygen time) can further enhance the stress signal, activate more astaxanthin synthase systems, and increase the synthesis rate. A 10% increase can effectively strengthen the stress without causing the cell to suddenly become unable to adapt due to excessive increase.

[0101] Negative feedback adjustment (SI value reduction): If the astaxanthin synthesis rate slows down in the previous cycle, it may be due to excessive stress intensity (causing cell damage) or insufficient stress signal. In this case, reducing the SI value can adjust the stress intensity. If the intensity is too high, reduction can alleviate cell damage and restore synthesis capacity. If the intensity is insufficient, after reduction, further judgment can be made on whether fine-tuning is needed through observation of subsequent cycles to avoid blindly increasing the stress.

[0102] This adaptive mechanism of effect feedback and intensity adjustment can dynamically find the optimal stress intensity under the current culture conditions, avoiding the problems of insufficient or excessive stress caused by traditional fixed stress parameters, ensuring that the astaxanthin synthesis rate is maintained at a high level and improving the stability of product yield.

[0103] Example 6: A soft measurement model for dissolved oxygen based on a long short-term memory network is constructed. The model takes historical ventilation, light intensity, pH, temperature and biomass as inputs and dissolved oxygen values ​​as outputs, providing redundant measurements when the reliability of hardware sensors is low.

[0104] By adopting the above technical solution, the Long Short-Term Memory (LSTM) network possesses the ability to capture the correlation characteristics of time-series data, effectively handling the dynamic dependence between historical parameters and current dissolved oxygen values ​​during dissolved oxygen changes. Therefore, this network was chosen to construct a soft-sensing model for dissolved oxygen. Historical aeration, light intensity, pH, temperature, and biomass were selected as model inputs because these parameters all have direct or indirect effects on dissolved oxygen values. Aeration directly determines the efficiency of oxygen transfer to the culture medium; light intensity changes dissolved oxygen production by affecting the photosynthetic rate of *Haemaphysalis*; pH and temperature indirectly affect dissolved oxygen levels by influencing algal physiological activities and oxygen solubility; and biomass reflects the overall scale of algal oxygen consumption and production. Integrating these parameters comprehensively covers the key influencing factors of dissolved oxygen changes. The model output is set to the dissolved oxygen value. Through learning and training on historical data, the model acquires the ability to calculate dissolved oxygen values ​​based on the input parameters. When hardware dissolved oxygen sensors experience data drift, malfunctions, or other low-reliability issues, this soft-sensing model can replace or supplement the hardware measurement results, providing continuous and reliable redundant measurement data for dissolved oxygen regulation, avoiding regulation interruptions or deviations due to hardware problems.

[0105] In Example 7, in step 3, the cloud platform first preprocesses all the data synchronized by the edge gateway; then it constructs three types of models: physical mass transfer, Haematococcus pluvialis growth, metabolism, and energy consumption calculation; subsequently, it sets a multi-objective function to maximize biomass, maximize astaxanthin, and minimize energy consumption, and sets safety constraints for dissolved oxygen and pH processes; finally, it uses a multi-objective optimization algorithm to solve the problem, selects the comprehensive optimal control parameters, and converts them into execution instructions to be synchronized to the edge gateway.

[0106] By adopting the above technical solution, the cloud platform first performs preprocessing on the full amount of control data synchronized by the edge gateway. The purpose is to eliminate noise in the data, fill in missing values, and unify the data format to ensure the accuracy and consistency of the data, providing a reliable data foundation for subsequent modeling and optimization.

[0107] Three core models were then constructed. The physical mass transfer model was used to simulate the transfer of substances such as oxygen and carbon source in the culture medium system, accurately reflecting the exchange process between the liquid and gas phases. The Haematococcus pluvialis growth model and metabolic model were used to characterize the growth rate variation of Haematococcus pluvialis under different environmental conditions, as well as the metabolic mechanism of its internal substance transformation pathways, especially the synthesis of astaxanthin. The energy consumption calculation model was used to quantify the energy consumption of actuators such as three-way mixing valves and micro-nano bubble generators during operation, so as to achieve accurate calculation of energy consumption in the control process.

[0108] Subsequently, a multi-objective function was set, with maximizing the final biomass of Haematococcus pluvialis and maximizing astaxanthin production as the core output objectives, while minimizing the energy consumption of the entire regulation process as the cost control objective, forming a multi-dimensional optimization direction that takes into account both output and efficiency. On this basis, safety constraints on dissolved oxygen and pH processes were set to clarify the reasonable fluctuation range of both, so as to avoid the inhibition or death of Haematococcus pluvialis growth due to parameters exceeding the safety threshold.

[0109] Finally, a multi-objective optimization algorithm is used to solve the above function. Under the premise of satisfying safety constraints, the algorithm can coordinate the conflict relationship between different objectives, select the control parameters that have the best comprehensive performance in terms of biomass, astaxanthin production and energy consumption, and then convert these parameters into execution instructions that can be recognized by the execution layer and synchronize them to the edge gateway, so as to provide optimized control basis for the precise operation of subsequent actuators.

[0110] In Example 8, the retrieval process of the graph neural network in step 4 is as follows: the strain genotype, initial pH, temperature range and culture objective of the current culture task are used as query vectors, the similarity with historical case nodes in the knowledge graph is calculated, and the control strategy of the case with the highest similarity is returned.

[0111] By adopting the above technical solution, a query vector is constructed based on the strain genotype, initial pH, temperature range, and culture objectives of the current culture task. The similarity between this vector and historical case nodes in the knowledge graph is calculated using a graph neural network. Finally, the regulatory strategy corresponding to the case with the highest similarity is returned, thereby achieving the matching and retrieval of historical experience with the current task.

[0112] In Example 9, step 5 first constructs a query vector for the current training task, which includes: strain characteristics, initial environmental parameters, and training target weights, and extracts vectors of the same dimension from historical cases in the knowledge graph; then, candidate cases that meet the similarity criteria are selected through similarity calculation, and the top N cases are selected in descending order of similarity. If the similarity is the same, the case that is closer to the target is selected first; finally, the control parameters of the Top N cases are weighted and fused, and the current environmental deviation is fine-tuned to generate an initial parameter sequence, which is then sent to the edge controller and labeled with reference case information.

[0113] By adopting the above technical solution, a query vector is first constructed based on the strain characteristics, initial environmental parameters, and training target weights of the current training task. The same-dimensional vector of historical cases in the knowledge graph is extracted to ensure that the two have a comparable basis. Then, candidate cases that meet the target are screened through similarity calculation. The top N cases are selected in descending order of similarity. If the similarity is the same, the case that produces the closest result to the current target is selected first to ensure the relevance and practicality of the reference case. Finally, the control parameters of the top N cases are weighted and fused, and fine-tuned in combination with the current and historical environmental deviations to generate an initial parameter sequence adapted to the current task and send it to the edge controller. At the same time, the reference case information is marked for traceability, so as to realize the effective reuse of historical experience and the accurate adaptation to current needs.

[0114] Example 10: Intelligent and coordinated regulation system for dissolved oxygen in Haematococcus pluvialis culture medium, used to realize intelligent and coordinated regulation method of dissolved oxygen in Haematococcus pluvialis culture medium. The system includes a sensing layer, an execution layer, a control decision layer and a cloud platform.

[0115] The multi-parameter sensors in the sensing layer include a dissolved oxygen sensor, a pH sensor, a temperature sensor, a photosynthetic efficiency sensor, and an online biomass monitoring probe.

[0116] The execution layer includes a three-way mixing valve, a micro / nano bubble generator, an LED dimmer, and a CO2 injection valve;

[0117] The control decision layer establishes a dynamic model with ventilation rate, gas ratio, stirring speed, and light intensity as inputs and dissolved oxygen and pH as outputs. The MPC controller calculates the optimal control command through rolling optimization and drives the actuators in the execution layer to work together to track the dynamic dissolved oxygen setpoint curve.

[0118] The cloud platform constructs a virtual model that corresponds one-to-one with the physical cultivation system, receives real-time data from the perception layer, and performs simulation and prediction.

[0119] The cloud platform stores all historical data.

[0120] The following specific embodiments illustrate the implementation principle of the present invention:

[0121] A large-scale cultivation experiment of *Rhodotorula glutinis* was conducted in a 500L flat-plate photobioreactor to verify the feasibility and effectiveness of the intelligent synergistic regulation method and system for dissolved oxygen, as detailed below:

[0122] Cultivation scale: 500L flat-plate photobioreactor;

[0123] Haematococcus pluvialis strain: HA-01 strain;

[0124] Initial environmental parameters: initial pH 7.1±0.1, culture temperature 25±1℃, initial biomass (OD) 6800.1, initial light intensity 150 μmol·m - ²・s - ¹(White light, blue light accounts for 10%)

[0125] Culture cycle: 14 days (7 days of vegetative growth stage + 7 days of stress induction stage).

[0126] Control system setup:

[0127] Perception layer configuration:

[0128] Dissolved oxygen sensor: Hach LDOII type, measurement range 0-200% air saturation, accuracy ±0.1%;

[0129] pH sensor: Mettler Toledo InPro3250, measuring range 0-14, accuracy ±0.01;

[0130] Temperature sensor: PT100 platinum resistance thermometer, measuring range 0-50℃, accuracy ±0.1℃;

[0131] Photosynthetic efficiency sensor: WalzPAM-2500, which can monitor the maximum photochemical efficiency of PSII (Fv / Fm) in real time with an accuracy of ±0.01;

[0132] Online biomass monitoring probe: OceanOptics STS-VIS type based on near-infrared spectroscopy, real-time output of OD680 value, measurement range 0-3.0, accuracy ±0.02.

[0133] Execution layer configuration:

[0134] Three-way mixing valve: SMCVX3100 model, which can adjust the air-oxygen mixing ratio (0-100% oxygen ratio).

[0135] Micro / nano bubble generator: pore size 5-10μm, aeration efficiency ≥90%, air flow rate range 0.2-2.0vvm;

[0136] LED dimmer: Adjustable white / blue light ratio (0-100% blue light percentage), illuminance range 50-300 μmol·m⁻² - ²・s - ¹;

[0137] CO2 injection valve: Baode 2033 type, CO2 injection range 0-0.2vvm, response time ≤1s.

[0138] Configuration of the control decision-making level:

[0139] Hardware: Industrial-grade edge controller (Advantech UNO-2484G), sampling period of 10 minutes, communication protocol MQTT;

[0140] Dynamic model: A nonlinear dynamic model is established with ventilation rate, oxygen ratio, stirring rate, and light intensity as inputs, and dissolved oxygen and pH as outputs;

[0141] MPC controller: 2 hours in the prediction time domain and 1 hour in the control time domain. The rolling optimization objective is to minimize dissolved oxygen tracking error and smooth actuator actions.

[0142] Cloud platform configuration:

[0143] Data preprocessing: Outliers were removed using the 3σ criterion, and missing values ​​were filled using linear interpolation;

[0144] The digital twin model comprises three sub-models:

[0145] Physical mass transfer model (based on the two-membrane theory, simulating the efficiency of oxygen transfer between gas, liquid, and algal cells).

[0146] Haematococcus pluvialis growth model (Logistic equation, fitting biomass growth: X(t) = X max / (1+e^(k(τ-t))), X max =12g・L - ¹, k=0.3d - ¹);

[0147] Metabolic-energy model (Monod equation describes astaxanthin synthesis, energy model quantifies energy consumption of ventilation, light, and stirring: E=0.8Q+0.05I+0.02N, where Q is ventilation rate, I is light intensity, and N is stirring rate).

[0148] Multi-objective optimization: The NSGA-II algorithm was adopted, with the optimization objectives being maximum biomass X, maximum astaxanthin production A, and minimum energy consumption E. The constraints were: dissolved oxygen 20%-150% air saturation and pH 6.5-8.0.

[0149] Knowledge graph: Constructed based on graph neural network (GNN), nodes include strain genotype, initial pH, temperature range, regulatory strategy, biomass, astaxanthin production, and edge weights are the association strength (0-1).

[0150] Complete control process:

[0151] Phase 1: Vegetative Growth Phase (Days 1-7):

[0152] Mode Trigger: Set OD 680 Threshold = 1.5, Fv / Fm threshold = 0.75; Days 1-6, online monitoring showed OD 680 =0.3-1.4, Fv / Fm=0.76-0.82, satisfying OD<threshold and Fv / Fm>threshold, triggering the growth promotion regulation mode.

[0153] Dissolved oxygen regulation: Control the three-way mixing valve (oxygen ratio 5%-15%) and the micro-nano bubble generator to make the dissolved oxygen fluctuate in a sinusoidal pulse pattern with an air saturation of 40%-80% and a cycle of 5 hours (e.g. dissolved oxygen curve on day 3: 8:00 reaches 40% → 10:30 reaches 60% → 13:00 reaches 80% → 15:30 reaches 60% → 18:00 reaches 40%).

[0154] Photo-oxygen synergistic adjustment: Set the optimal light intensity = 200 μmol·m - ²・s - ¹, Optimal dissolved oxygen = 60% air saturation; on the 5th day, due to cloudy weather, the actual light intensity dropped to 120 μmol·m⁻¹. - ²・s - ¹, The current dissolved oxygen is 70%. The calculated photo-oxygen synergy factor k = (120 / 200) × (70 / 60) = 0.7, which deviates from 1 by more than 20%. The system automatically converges the dissolved oxygen fluctuation range to 70% by 50% (original amplitude 40% → new amplitude 20%). The new range is 60%-80% air saturation, avoiding excessive accumulation of dissolved oxygen that could lead to oxygen damage.

[0155] Phase 2: Stress Induction Phase (Days 7-14):

[0156] Mode Trigger: Day 7. The growth promotion mode has lasted for 7 days, and monitoring shows OD... 680 =1.6 (≥ threshold 1.5) and Fv / Fm=0.72 (≤ threshold 0.75), which meet the dual-indicator + time-based trigger condition, and switch to the stress-induced regulation mode.

[0157] Stress parameter control:

[0158] High-oxygen phase: The three-way mixing valve increases the oxygen content to 40%, and the dissolved oxygen reaches 130% air saturation within 10 minutes, maintaining this level for 15 minutes; simultaneously, the LED dimmer adjusts the blue light ratio to 40% (total illuminance 250 μmol·m⁻¹). - ²・s - ¹);

[0159] Low oxygen phase: Strong ventilation (ventilation volume increased to 1.5 vvm), dissolved oxygen drops to 25% air saturation within 5 minutes, and is maintained for 50 minutes;

[0160] Adaptive optimization of the stress intensity index (SI):

[0161] Weights were set as follows: α = 0.3 (hyperoxygenation time), β = 0.4 (hyperoxygenation concentration), γ = 0.15 (hypoxia concentration), δ = 0.15 (blue light ratio); maximum allowable hyperoxygenation time = 20 minutes, maximum allowable hyperoxygenation concentration = 150%, baseline dissolved oxygen = 60%, maximum allowable blue light ratio = 50%.

[0162] SI calculation for the first cycle on day 7: (15 / 20)×0.3+(130 / 150)×0.4+((60-25) / (60-20))×0.15+(40 / 50)×0.15=0.225+0.347+0.131+0.12=0.823;

[0163] On day 8, the astaxanthin synthesis rate of the previous cycle was monitored and found to be 1.0 mg·L⁻¹. - ¹・d - ¹(Compared to 0.8 mg / L on day 7) - ¹・d - ¹Accelerate), SI is adjusted to 0.905 with a 10% increase; simultaneously increase the high oxygen time to 17 minutes, high oxygen concentration to 135%, and blue light ratio to 42%, further activating the astaxanthin synthase system.

[0164] On day 10, the dissolved oxygen sensor experienced data drift due to culture medium contamination (dissolved oxygen was shown as 100%, but no abnormalities were observed in the algal solution). The system automatically activated the LSTM-based soft-sensing model for dissolved oxygen (inputs: aeration rate of 0.8 vvm and illumination of 220 μmol·m⁻² over the past 24 hours). - ²・s - ¹, pH 7.3, temperature 25℃, biomass OD380 = 1.8, dissolved oxygen output = 85%); use soft measurement data as the basis for regulation to avoid regulation deviations caused by hardware failure.

[0165] Cloud optimization: The edge gateway synchronizes all data to the cloud platform every 30 minutes. The cloud platform solves the multi-objective function using the NSGA-II algorithm, and outputs the optimal control parameters on day 12: ventilation rate 1.0 vvm, oxygen content 35%, blue light content 43%, and CO2 injection rate 0.08 vvm. After the instructions are synchronized to the edge controller, the astaxanthin synthesis rate increases to 1.2 mg·L⁻¹. - ¹・d - ¹, energy consumption is reduced by 8%.

[0166] New batch case reuse: Start a new batch of culture on day 15 (strain HA-01, initial pH 7.0, temperature 24-26℃, target astaxanthin yield ≥15mg·L). - ¹);

[0167] Construct the query vector: strain characteristics (HA-01 genotype), initial parameters (pH 7.0, temperature 24-26℃), target weights (astaxanthin 0.6, biomass 0.3, energy consumption 0.1).

[0168] Knowledge graph retrieval: After calculating similarity, the top 3 historical cases were selected (92%, 88%, 88% similarity), with the 3rd case showing an astaxanthin yield of 16.2 mg·L⁻¹. - ¹Closer to the target;

[0169] Parameter fusion and fine-tuning: The parameters of the top 3 cases are weighted and fused (weights 0.5, 0.25, 0.25). Combined with the fact that the current temperature is 1°C higher than the historical temperature, the ventilation rate is fine-tuned by +5% to generate the initial parameter sequence. After the initial parameters are issued, the start-up cycle of the new batch is shortened from the traditional 3 days to 1 day, and the initial dissolved oxygen control error is <5%.

[0170] After 14 days of cultivation, the measured indicators are as follows:

[0171] Biomass (dry weight): 10.2 g·L - ¹;

[0172] Astaxanthin production: 16.5 mg·L⁻¹ - ¹;

[0173] Energy consumption per unit of astaxanthin: 8.2 kWh / kg - ¹;

[0174] Astaxanthin yield fluctuation between batches: 7.2%;

[0175] Dissolved oxygen monitoring reliability: 96% (the soft measurement model seamlessly connects in the event of sensor failure).

[0176] Traditional methods for regulating dissolved oxygen in *Haematococcus pluvialis* are characterized by fixed static parameters, single-factor regulation, lack of data optimization, and no ability to reuse experience (e.g., maintaining dissolved oxygen at 50%-70% air saturation throughout the process, adjusting dissolved oxygen only through ventilation, without phased switching or cloud-based optimization). Table 1 shows a comparison of key performance indicators between this technical solution and traditional methods:

[0177] Table 1

[0178] Comparison Dimensions Traditional control methods This technical solution Increase / Decrease Amount Biomass (dry weight) after 14 days of culture <![CDATA[8.5g·L - ¹]]> <![CDATA[10.2g·L - ¹]]> An increase of 19.0% Astaxanthin yield after 14 days of cultivation <![CDATA[10.2mg·L - ¹]]> <![CDATA[16.5mg·L - ¹]]> An increase of 61.8% Energy consumption per unit of astaxanthin <![CDATA[12.5kWh·kg - ¹]]> <![CDATA[8.2kWh·kg - ¹]]> Reduced by 34.4% New batch start-up period 3 days 1 day Shortened by 66.7% Astaxanthin production fluctuates between batches 15.0% 7.2% Reduced by 52.0% Dissolved oxygen monitoring reliability (uninterrupted) 80% (Control interrupted in case of sensor failure) 96% (soft measurement model redundancy) Increase by 20 percentage points Mean photosynthetic efficiency (Fv / Fm) 0.68 0.74 An increase of 8.8% Peak rate of astaxanthin synthesis during stress period <![CDATA[0.7mg·L - ¹·d - ¹]]> <![CDATA[1.2mg·L - ¹·d - ¹]]> An increase of 71.4%

[0179] Biomass and astaxanthin production increased by 19.0% and 61.8% respectively, solving the problems of insufficient oxygen during the nutrient period and ineffective oxygen during the stress period under traditional static regulation.

[0180] The multi-objective optimization algorithm reduces energy consumption per unit of astaxanthin by 34.4%, avoiding energy waste caused by blind ventilation and lighting in traditional methods;

[0181] Reusing historical cases shortens the startup cycle by 66.7% and reduces batch fluctuations by 52.0%, solving the pain points of long parameter debugging and large batch differences in traditional methods.

[0182] The dissolved oxygen soft measurement model improves monitoring reliability to 96%, avoiding the risks of traditional methods that rely on hardware and are interrupted by failure.

[0183] In summary, this technical solution provides a feasible path for the efficient industrial cultivation of Haematococcus pluvialis by comprehensively overcoming the bottlenecks of traditional methods.

[0184] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent and synergistic regulation of dissolved oxygen in Haematococcus pluvialis culture medium, characterized in that, Includes the following steps: Step 1: Construct a closed-loop control system that includes a perception layer, an execution layer, a control decision layer, and a cloud platform; Step 2: The control decision layer dynamically switches between growth-promoting regulation mode and stress-induced regulation mode based on biomass data and physiological state data of Haematococcus pluvialis culture stage, so as to achieve synergistic regulation of dissolved oxygen, light, pH and carbon source. Step 3: The edge computing gateway synchronizes full control data to the cloud platform at set intervals. The cloud platform performs multi-objective optimization simulation based on the digital twin model. The optimization objectives include final biomass, astaxanthin production and energy consumption. Step 4: The cloud platform knowledge base uses graph neural networks to construct a correlation map of Haematococcus pluvialis strains, environmental parameters, regulation strategies, and output effects; Step 5: When a new batch of breeding is started, based on the current strain and environmental characteristics, the Top N historical cases with the highest similarity are retrieved from the spectrum, and their optimal control parameter sequences are sent to the edge controller as initial settings. In step 2, when the optical density OD value measured by the online biomass monitoring probe is less than the OD threshold and the maximum photochemical efficiency of PSII measured by the photosynthetic efficiency sensor is greater than the photochemical threshold, the growth promotion regulation mode is triggered. The growth promotion regulation mode controls the three-way mixing valve and the micro-nano bubble generator to make the dissolved oxygen level fluctuate sinusoidally in the range of 40%-80% air saturation with a period of 4-6 hours. Introducing the photo-oxygen synergistic factor k, ; in This is the current light intensity. It is the optimal light intensity. This is the current dissolved oxygen level. It is the optimal dissolved oxygen value; The system dynamically adjusts the sine wave parameters. When the K value deviates from 1 by more than 20%, the original dissolved oxygen fluctuation range of 40%-80% is converged to the current instantaneous dissolved oxygen value by 50%. The stress-induced modulation mode is triggered when any of the following conditions are met: The OD value measured by the online biomass monitoring probe of the stress-induced regulation mode is greater than or equal to the OD threshold, and the maximum photochemical efficiency of PSII measured by the photosynthetic efficiency sensor is less than or equal to the photochemical threshold. The growth promotion mode was continuously operated for 7 days or more; The stress-induced regulation mode controls the three-way mixing valve to rapidly raise dissolved oxygen to 120%-150% air saturation and maintain it for 10-20 minutes, during which the blue light ratio of the LED dimmer is adjusted to 30%-50%; then, through strong ventilation, the dissolved oxygen is rapidly reduced to 20%-30% air saturation and maintained for 45-60 minutes. Establish a stress intensity index (SI) for a stress-induced regulation model; ； in It refers to the duration of hyperoxia. It is the maximum permissible hyperoxia time. It is a high oxygen concentration. It is the maximum permissible high oxygen concentration. It is a low oxygen concentration. This is the baseline dissolved oxygen concentration. It's the blue light ratio. It is the maximum permissible blue light ratio. , , and These are the weighting indices; The system dynamically adjusts the SI value of the current cycle based on the astaxanthin synthesis rate of the previous cycle. If the astaxanthin synthesis rate increases, the SI value increases by 10% increments, and vice versa, thus achieving adaptive optimization of stress intensity.

2. The intelligent synergistic regulation method for dissolved oxygen in Haematococcus pluvialis culture medium according to claim 1, characterized in that, A soft measurement model for dissolved oxygen based on a long short-term memory network was constructed. The model takes historical ventilation, light intensity, pH, temperature, and biomass as inputs and dissolved oxygen values ​​as outputs, providing redundant measurements when the reliability of hardware sensors is low.

3. The intelligent synergistic regulation method for dissolved oxygen in Haematococcus pluvialis culture medium according to claim 2, characterized in that, In step 3, the cloud platform first preprocesses all the data synchronized by the edge gateway; then it constructs three types of models: physical mass transfer, Haematococcus pluvialis growth, metabolism, and energy consumption calculation; subsequently, it sets a multi-objective function to maximize biomass, maximize astaxanthin, and minimize energy consumption, and sets safety constraints for dissolved oxygen and pH processes; finally, it uses a multi-objective optimization algorithm to solve the problem, selects the comprehensive optimal control parameters, and converts them into execution instructions to be synchronized to the edge gateway.

4. The intelligent synergistic regulation method for dissolved oxygen in Haematococcus pluvialis culture medium according to claim 3, characterized in that, In step 4, the retrieval process of the graph neural network is as follows: the strain genotype, initial pH, temperature range and culture objective of the current culture task are used as query vectors, the similarity with historical case nodes in the knowledge graph is calculated, and the control strategy of the case with the highest similarity is returned.

5. The intelligent synergistic regulation method for dissolved oxygen in Haematococcus pluvialis culture medium according to claim 4, characterized in that, In step 5, a query vector for the current training task is first constructed, which includes strain characteristics, initial environmental parameters, and training target weights. The same-dimensional vectors of historical cases in the knowledge graph are also extracted. Then, candidate cases that meet the similarity criteria are selected through similarity calculation. The top N cases are selected in descending order of similarity. If the similarity is the same, the case that produces the closest result to the target is selected first. Finally, the control parameters of the Top N cases are weighted and fused, and the current environmental deviation is fine-tuned to generate an initial parameter sequence, which is then sent to the edge controller and labeled with reference case information.

6. A smart and coordinated regulation system for dissolved oxygen in Haematococcus pluvialis culture medium, characterized in that, To implement the intelligent and coordinated regulation method of dissolved oxygen in the Haematococcus pluvialis culture medium as described in claim 5, the system includes a sensing layer, an execution layer, a control and decision-making layer, and a cloud platform; The multi-parameter sensors in the sensing layer include a dissolved oxygen sensor, a pH sensor, a temperature sensor, a photosynthetic efficiency sensor, and an online biomass monitoring probe. The execution layer includes a three-way mixing valve, a micro / nano bubble generator, an LED dimmer, and a CO2 injection valve; The control decision layer establishes a dynamic model with ventilation rate, gas ratio, stirring speed, and light intensity as inputs and dissolved oxygen and pH as outputs. The MPC controller calculates the optimal control command through rolling optimization and drives the actuators in the execution layer to work together to track the dynamic dissolved oxygen setpoint curve. The cloud platform constructs a virtual model that corresponds one-to-one with the physical cultivation system, receives real-time data from the perception layer, and performs simulation and prediction. The cloud platform stores all historical data.

Images