Sugar-free tissue culture system for industrialization of rare medicinal plants

Abstract

The invention discloses a sugar-free tissue culture system for industrialization of rare medicinal plants, and relates to the technical field of plant tissue culture and intelligent bioreactor control. Comprising a temporary immersion culture unit, a carbon dioxide supply and uniform mixing unit, a liquid phase buffer adding unit, an environment regulation and control unit, a physiological state acquisition unit, a multi-point carbon dioxide sensor array, an edge controller, a special risk constraint large model controller and a block chain-based credible strategy consistency network. According to the system, a sugar-free culture medium is used as a culture solution, a temporary immersion rhythm formed by an immersion phase and a ventilation phase is adopted, a carbon supply window is set at the ventilation phase, and carbon dioxide injection is forbidden at the immersion phase; a uniformity threshold value, a rise time threshold value, a deviation correction and amplitude limiting rule, a carbon supply curve parameter set and a liquid phase buffer parameter set bound with the carbon supply curve parameter set are generated in the large model, and a risk upper bound quantile and a strategy admission threshold value are calculated based on a rolling calibration set to achieve admission; the strategy packet is executed after being subjected to block chain consensus release and edge verification, correction, amplitude limiting and degradation are triggered in a window according to criteria, and a calibration set is written back, so that cross-base consistency triggering calibration and recovery are realized.

Description

Technical Field

[0001] This invention relates to the field of plant tissue culture and intelligent bioreactor control technology, specifically a sugar-free tissue culture system for the industrialization of rare medicinal plants. Background Technology

[0002] The rapid propagation and robust seedling production of rare medicinal plants generally rely on tissue culture technology. Temporary immersion culture systems are used for propagation and large-scale expansion due to their high efficiency in liquid-gas exchange and mass transfer. In current industrial production, most culture systems still rely on sugar-containing media and external carbon sources to maintain growth. Although they can achieve certain propagation effects, they are prone to problems such as high risk of contamination, large fluctuations in the physiological state of seedlings, and poor batch consistency when used in mass production, long-term operation, and cross-site replication. At the same time, sugar-containing culture brings contradictions such as sensitivity of culture medium components, rapid spread of contamination, and high subsequent acclimatization costs, making the industrialization process highly dependent on personnel experience and on-site parameter adjustment.

[0003] To reduce pollution risks and enhance the photosynthetic autotrophic capacity of seedlings, sugar-free or low-sugar culture approaches have emerged in recent years. These approaches typically promote photosynthetic carbon supply by increasing ventilation, supplementing carbon dioxide, and increasing light intensity. However, this approach still has significant shortcomings in engineering implementation: the carbon dioxide supply process often employs a single-point concentration closed-loop or fixed valve control strategy, which fails to reflect the spatial unevenness and dynamic lag of the gas phase distribution within the culture chamber, easily leading to local concentration shocks and uncontrollable carbon supply rise. Carbon dioxide injection and liquid buffer addition are often set separately, lacking strong binding constraints between carbon supply curves and buffer parameters, making it difficult to suppress pH drift and ionic strength fluctuations, which are more sensitive under sugar-free conditions, in a timely manner. Equipment differences, sensor drift, and environmental disturbances at different sites can cause continuous deviations in operating condition distribution. Existing systems mainly rely on recording or manual calibration, lacking verifiable strategy access, mandatory execution, and cross-site consistency triggering mechanisms, resulting in the problem of non-replicable industrialization of different results with the same formula.

[0004] Existing technologies achieve tissue culture production through temporary immersion cycle control, carbon dioxide ventilation regulation, light and temperature environment control, and culture medium buffer preparation. However, there are still certain limitations, such as the carbon supply window and cycle are not tightly coupled, the uniformity of gas phase carbon supply space lacks quantitative criteria, the carbon supply and liquid phase buffer lack integrated linkage, the strategy generation and execution lack verifiable constraints, and the cross-site consistency deviation lacks automatic trigger calibration and recovery paths.

[0005] Therefore, there is an urgent need for a sugar-free tissue culture system for the industrialization of rare medicinal plants that can control the particle size of carbon supply under sugar-free culture medium conditions, form uniformity and rise time criteria based on multi-point carbon dioxide sampling, strongly bind the carbon supply curve with liquid phase buffer parameters, and introduce risk constraint strategy access, blockchain consensus enforcement, and cross-site consistency trigger calibration to achieve stable operation, reduce pollution, improve the physiological consistency of seedlings, and support large-scale replication at multiple sites. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and propose a sugar-free tissue culture system for the industrialization of rare medicinal plants, so as to solve the above-mentioned problems.

[0007] The objective of this invention is achieved through the following technical solution: a sugar-free tissue culture system for the industrialization of rare medicinal plants, comprising a temporary immersion culture unit, a carbon dioxide supply and mixing unit, a liquid buffer addition unit, an environmental control unit, a physiological state acquisition unit, a multi-point carbon dioxide sensor array, an edge controller, a dedicated risk-constrained large model controller, and a blockchain-based trusted policy consistency network. The temporary immersion culture unit uses sugar-free culture medium as the culture medium and is set to a temporary immersion cycle consisting of an immersion phase and a ventilation phase. A carbon supply window is set within the ventilation phase, and the start and end times of the carbon supply window and the ventilation phase are phase-locked so that carbon dioxide injection is only allowed to be performed within the carbon supply window and is prohibited from being performed during the immersion phase. The carbon dioxide supply and mixing unit is connected to the temporary immersion culture unit, and performs carbon dioxide injection according to the carbon supply curve parameter set within the carbon supply window and performs mixing airflow organization switching according to the correction action set. A multi-point carbon dioxide sensor array is arranged in the cavity of the temporary immersion culture unit, and the carbon dioxide concentration at multiple points is collected simultaneously within the carbon supply window to form a multi-point sampling sequence; The liquid phase buffer dosing unit is connected to the temporary immersion culture unit and includes a liquid phase state monitoring channel. It performs buffer component dosing according to the liquid phase buffer parameter set and ion balancing according to the ion balancing constraint identifier. The liquid phase state characteristics are output by the liquid phase state monitoring channel. The environmental control unit sets the light intensity and temperature for the temporary immersion culture unit. The physiological state acquisition unit generates a physiological state vector summary at the end of the carbon supply window; The edge controller maintains a rolling calibration set; The dedicated risk-constrained large model controller receives temporary immersion beat identifiers, multi-point sampling sequences, liquid phase state characteristics, physiological state vector summaries, and cross-site consistency statistics issued by a blockchain-based trusted policy consistency network. It generates candidate control strategies based on a rolling calibration set and generates uniformity thresholds, rise time thresholds, correction action sets, and amplitude limiting rule sets before each carbon supply window begins. Simultaneously, it generates a set of carbon supply curve parameters and a set of liquid phase buffer parameters that are bound one-to-one with the set of carbon supply curve parameters. It calculates the risk upper bound quantile value based on the rolling calibration set and compares the risk upper bound quantile value with the policy admission threshold as the policy admission condition. If the policy admission condition is not met, the generation of non-degraded policy packages is prohibited. If the policy admission condition is met, the candidate control strategies are compiled into a policy package and the uniformity threshold, rise time threshold, correction action set, amplitude limiting rule set, carbon supply curve parameter set, liquid phase buffer parameter set, degradation condition, risk upper bound quantile value, and policy admission threshold are written into the policy package. A blockchain-based trusted policy consensus network is used to publish policy packages in consensus, distribute cross-base consensus statistics, and output the consensus trigger status. Before the carbon supply window begins, the edge controller verifies the strategy package hash value and the calibration summary hash value. If the verification is successful and the consistency trigger status is not triggered, the strategy package is executed. Within the carbon supply window, the uniformity index and rise time index are calculated from the multi-point sampling sequence. The uniformity index is compared with the uniformity threshold and the rise time index is compared with the rise time threshold. If either comparison is not satisfied, the carbon supply curve is limited and mixed to correct the deviation according to the set of correction actions and the set of limiting rules. When the degradation condition is met, the degraded carbon supply mode is executed. The uniformity index, rise time index, liquid phase state characteristics, physiological state vector summary, cross-site consistency statistics and strategy package hash value are written into the rolling calibration set. When the edge controller fails verification or the consistency trigger state is triggered, it refuses to execute the non-degraded policy package and executes the conservative mode, so that the carbon supply window criterion generation, carbon supply window availability verification, carbon supply curve binding with liquid phase buffer, risk upper bound quantile value admission, policy package blockchain consensus release, policy package verification execution, consistency trigger and rolling calibration set update constitute a control chain that must be closed.

[0008] The policy packet hash value is the hash value obtained by performing a hash operation on the policy packet payload that does not contain the policy packet hash value field, and is written into the policy packet for verification; the calibration summary hash value is the hash value obtained by performing a hash operation on the window record summary of the rolling calibration set, and is written into the policy packet for verification; the uniformity index is determined by the ratio of the difference between the maximum and minimum carbon dioxide concentrations in the multi-point sampling sequence at the same sampling time to the average carbon dioxide concentration in the multi-point sampling sequence.

[0009] The rise time index is determined by the length of time from the start of the carbon supply window to the time when the average carbon dioxide concentration in the chamber corresponding to the multi-point sampling sequence enters the target platform concentration tolerance band and remains stable. The target platform concentration, the width of the target platform concentration tolerance band, and the duration of stability are given by the set of parameters of the carbon supply curve.

[0010] The carbon dioxide supply and mixing unit includes a proportional control valve, a carbon dioxide injection pipeline, a mixing fan, and a switchable duct structure. The edge controller controls the speed of the mixing fan and the duct state of the switchable duct structure according to the correction action set, and controls the opening of the proportional control valve according to the carbon supply curve parameter set, so as to perform the linkage of carbon dioxide injection and mixing correction within the carbon supply window.

[0011] The set of carbon supply curve parameters includes target platform concentration, target platform concentration tolerance band width, duration of continuous stability, range of rising slope, duration of carbon supply window, pulse duty cycle, and upper limit of cumulative injection volume within the permeable phase; the set of limiting rules includes reducing target platform concentration, reducing upper limit of rising slope, shortening duration of carbon supply window, and reducing upper limit of cumulative injection volume within the permeable phase.

[0012] The set of liquid phase buffer parameters includes the amount of buffer component added, the timing of buffer component addition, and the ion balancing constraint identifier. The edge controller triggers the addition of buffer component compensation based on the liquid phase state characteristics and degradation conditions output by the liquid phase state monitoring channel, and simultaneously triggers the set of limiting rules to suppress the accumulation of liquid phase drift.

[0013] The dedicated risk constraint large model controller has a built-in variety stage map; the variety stage map outputs the value boundaries of the carbon supply curve parameter set, the value boundaries of the liquid phase buffer parameter set, and the value boundaries of the light intensity setting and temperature setting, using variety identifier and growth stage identifier as indexes; the dedicated risk constraint large model controller generates a strategy package that meets the strategy admission conditions within the value boundaries.

[0014] The rolling calibration set consists of execution records at the carbon window scale, and each execution record includes a uniformity index, a rise time index, liquid phase state characteristics, a physiological state vector summary, a cross-base consistency statistic, and a strategy package hash value. The dedicated risk-constrained large model controller generates stability loss, physiological loss, and consistency loss from the execution records, and uses the maximum value among the three as the overall loss, and calculates the risk upper bound quantile value based on the overall loss.

[0015] The blockchain-based trusted strategy consistency network calculates the consistency deviation based on the physiological state vector summary of the same variety and growth stage and the cross-site consistency statistics. When the consistency deviation exceeds the consistency threshold, the consistency trigger state is set to trigger and a calibration task strategy package is released. The calibration task strategy package includes sensor calibration instructions, air duct self-check instructions and conservative mode execution time. The upper risk quantile value meets the policy admission conditions as the recovery condition for allowing the release of non-degraded strategy packages.

[0016] The conservative mode includes a pulsed carbon supply that forces a low target plateau concentration and a maximum cumulative injection volume during the low permeability phase, and simultaneously executes a set of liquid-phase buffer parameters that are one-to-one with the pulsed carbon supply.

[0017] The beneficial effects of this invention are: Using a temporary immersion cycle as the sole time base, the immersion phase and the aeration phase are clearly separated. Carbon dioxide injection is only permitted within the carbon supply window of the aeration phase, and carbon dioxide injection is prohibited during the immersion phase. This fundamentally reduces pH drift and ionic strength fluctuations caused by non-target dissolution of carbon dioxide in the liquid phase, lowers the probability of contamination and the risk of contamination spread due to fluctuations in the liquid phase environment under sugar-free conditions, and reduces the risk of physiological stress, thus transforming sugar-free culture from feasible to stable operation. Based on this, the invention uses a multi-point carbon dioxide sensor array to synchronously sample within the carbon supply window, forming a multi-point sampling sequence. The uniformity index is calculated using the maximum, minimum, and average concentrations at the same sampling time, and the rise time index is calculated using the length of time it takes for the average carbon dioxide concentration in the chamber to enter the target platform concentration tolerance zone and remain stable. This elevates the gas phase carbon supply process from a traditional single-point concentration setting to a verifiable object with dual criteria of spatial uniformity and dynamic accessibility, avoiding uneven photosynthetic carbon source supply caused by local high concentrations, flow field dead zones, and delayed rise, thereby more stably maintaining the gas phase environment required for seedling carbon assimilation under sugar-free conditions.

[0018] The carbon supply curve parameter set and the liquid phase buffer parameter set are designed to be bound one-to-one and written into the strategy package. This ensures that any adjustment to the target plateau concentration, rise slope range, carbon supply window duration, pulse duty cycle, and the upper limit of cumulative injection volume within the permeability phase of the carbon supply curve must simultaneously cause corresponding adjustments to the buffer component dosage, addition time, and ion balancing constraint flags. The edge controller only executes the strategy package containing the binding relationship, thereby solidifying gas phase carbon supply and liquid phase stability into the same strategy object. This avoids the accumulation of pH drift, osmotic pressure fluctuations, and unreproducible culture environment caused by field personnel or equipment independently adjusting the gas and liquid phases at different sites. After the liquid phase state monitoring channel outputs the liquid phase state characteristics, the edge controller synchronously triggers the addition of buffer component compensation and triggers the limit rule set when the liquid phase state characteristics approach or reach the degradation condition. The limit rule set achieves carbon supply convergence by reducing the target platform concentration, reducing the upper limit of the rise slope, shortening the duration of the carbon supply window, and reducing the upper limit of the cumulative injection amount within the permeable phase. The compensation addition and carbon supply convergence occur in the same direction, so that liquid phase drift suppression no longer depends on experience-based treatment but forms an executable closed-loop action chain, which significantly reduces the batch fluctuation and abnormal diffusion probability of sugar-free tissue culture.

[0019] A dedicated risk-constrained large-scale model controller serves as the sole center for strategy generation and admission. It endogenously generates uniformity thresholds, rise time thresholds, correction action sets, amplitude limiting rule sets, carbon supply curve parameter sets, and their associated liquid phase buffer parameter sets. Based on a rolling calibration set, the upper risk bound quantile is calculated and compared with the strategy admission threshold as the strategy admission condition. If the policy admission condition is not met, the generation of non-degraded strategy packages is prohibited; only degraded strategy packages are allowed. This mechanism elevates strategy selection from predictive or optimization output to risk-constrained decision-making with admission evidence, effectively suppressing the accumulation and amplification of tail failure events under a few extreme conditions in long-term multi-site operation. The rolling calibration set consists of execution records at the carbon supply window scale. These records include uniformity indices, rise time indices, liquid phase state characteristics, physiological state vector summaries, cross-site consistency statistics, and strategy package hash values. This ensures that model updates and admission decisions are based on real closed-loop data, avoiding control failures caused by the disconnect between offline models and field distribution, thereby improving the stability and interpretability of strategy iteration.

[0020] The strategy package is published through a blockchain-based trusted strategy consensus network, and cross-site consensus statistics are distributed and consensus trigger status is output. Before the carbon supply window begins, the edge controller verifies the hash value of the strategy package and the hash value of the calibration summary. The strategy package is executed only when the verification is successful and the consensus trigger status is not triggered. If the verification fails or the consensus trigger status is triggered, the non-degraded strategy package is rejected and the conservative mode is executed. The conservative mode maintains the safety boundary by pulsed carbon supply with the upper limit of cumulative injection volume in the low target platform concentration and low permeability phase, and simultaneously executes the set of liquid phase buffer parameters that are bound one-to-one with the pulsed carbon supply. This prevents temporary on-site parameter changes from destroying calibration assumptions and access evidence, so that risk access - trusted execution - data write-back form a closed loop. Furthermore, the Trusted Strategy Consistency Network calculates the consistency deviation based on the physiological state vector summary of the same variety and growth stage and the cross-site consistency statistics. When the consistency deviation exceeds the consistency threshold, the consistency trigger state is set to trigger and a calibration task strategy package is released. The calibration task strategy package includes sensor calibration instructions, air duct self-check instructions and conservative mode execution time. The risk upper bound quantile value meets the strategy access condition as the recovery condition for allowing the release of non-degradation strategy packages. Thus, when distribution deviation occurs across sites, it automatically enters the trigger-calibration-reconstruction-recovery governance process, avoiding the gradual uncontrollable differentiation path of different sites due to equipment aging, flow field changes, sensor drift or environmental differences. This transforms industrial replication from relying on personnel experience to relying on system mechanisms.

[0021] Using the carbon supply window as the unified control granularity, a continuous data processing chain is formed within the window, consisting of acquisition, calculation, comparison, correction, amplitude limiting, compensation, and writing. The strategy package fields fully cover thresholds, rules, parameter sets, degradation conditions, and verification summaries. The system has a clear engineering implementation path, facilitating standardized deployment, standardized debugging, and standardized operation and maintenance. At the same time, the variety stage map outputs the carbon supply curve parameter set, liquid phase buffer parameter set, and value boundaries of light intensity and temperature settings using variety identifiers and growth stage identifiers. This allows the same system to quickly switch parameter domains in different rare medicinal plants such as Bletilla striata, Dendrobium officinale, and Lilium lancifolium at different growth stages without exceeding physiological and equipment safety boundaries, improving the efficiency of variety expansion and reducing trial and error costs. Attached Figure Description

[0022] Figure 1 The process of this invention Figure 1 ; Figure 2 The process of this invention Figure 2 ; Figure 3 The process of this invention Figure 3 . Detailed Implementation

[0023] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Example 1 like Figure 1 As shown, this embodiment discloses a basic implementation method of a sugar-free tissue culture system for the industrialization of rare medicinal plants. The system is used to complete the propagation of seedlings under sugar-free culture medium conditions with a temporary immersion rhythm. It uses a carbon supply window as the control particle size formation strategy generation - strategy consensus release - strategy verification and execution - window criterion verification - correction and limit degradation - data write-back calibration to ensure that the carbon dioxide supply process is stably realized in terms of spatial uniformity, controllability of the rise process, liquid phase stability and anomaly convergence.

[0025] The system consists of a temporary immersion culture unit, a carbon dioxide supply and mixing unit, a liquid buffer dosing unit, an environmental control unit, a physiological state acquisition unit, a multi-point carbon dioxide sensor array, an edge controller, a dedicated risk-constrained large-scale model controller, and a blockchain-based trusted policy consensus network. The temporary immersion culture unit includes a culture chamber, a culture rack, and a cycle execution mechanism. The cycle execution mechanism decomposes the culture process into an immersion phase and a ventilation phase and executes them cyclically according to a preset period. The culture chamber is connected to the liquid supply circuit to achieve immersion of the culture medium during the immersion phase and drainage and reflux during the ventilation phase. The culture medium uses a sugar-free medium with a sucrose concentration set to 0 g / L. The sugar-free medium uses MS-based inorganic salts and vitamin components. The culture medium, after being prepared with plant growth regulators and sterilized, is introduced into the culture medium storage tank. In this embodiment, the plant growth regulator components are set as a combination of 6-benzylaminopurine and naphthaleneacetic acid, with mass concentrations set at 0.5 mg / L and 0.1 mg / L, respectively, to meet the differentiation and proliferation requirements of rare medicinal plant tissue culture. The initial pH of the culture medium is set at 5.7 and maintained by the liquid phase buffer addition unit and ion balancing constraint. The environmental control unit sets the light intensity and temperature of the culture chamber. The light intensity is set at 80 μmol·m⁻²·s⁻¹ and the photoperiod is set at 16 hours of light and 8 hours of darkness. The temperature is set at 25℃ and the relative humidity in the chamber is controlled in a closed loop to suppress leaf water loss in the ventilated phase.

[0026] The temporary immersion culture unit sets a carbon supply window within the ventilation phase and locks the carbon supply window with the start and end times of the ventilation phase. At the start of the ventilation phase, the edge controller only allows the execution of carbon dioxide injection commands within the carbon supply window and prohibits carbon dioxide injection during the immersion phase. During the ventilation phase, except for the carbon supply window, only circulating ventilation and mixing pretreatment are allowed to stabilize the chamber boundary layer. The carbon dioxide supply and mixing unit includes a proportional control valve, a carbon dioxide injection pipeline, a mixing fan, and a switchable air duct structure. The proportional control valve and carbon dioxide injection pipeline are used to inject carbon dioxide gas into the culture chamber according to the time sequence given by the carbon supply curve parameter set. The mixing fan and the switchable air duct structure are used to change the airflow organization within the chamber to reduce the spatial distribution difference of carbon dioxide and shorten the time to reach the target platform concentration.

[0027] A multi-point carbon dioxide sensor array is arranged within the culture chamber and simultaneously collects carbon dioxide concentrations at multiple points within the carbon supply window to form a multi-point sampling sequence. In this embodiment, the sensor array is arranged at four points and outputs synchronous sampling data at a sampling frequency of 1Hz. The edge controller performs timestamp alignment and outlier removal on the multi-point sampling sequence within the carbon supply window and calculates the average carbon dioxide concentration and spatial dispersion of the chamber. The spatial dispersion is quantified by a uniformity index, which is calculated using the following formula: in, Indicates the first sampling time at the same sampling moment Carbon dioxide concentration (unit: ppm) of each sensor. This represents the average concentration at four points at the same sampling time (unit: ppm). The uniformity index (dimensionless) has a value range of [value range missing]. The smaller the value, the more uniform the spatial distribution. The edge controller uses the comparison between the uniformity index and the uniformity threshold in the strategy package as a necessary condition for verifying the availability of the carbon supply window and to trigger mixing correction and carbon supply limiting.

[0028] To ensure that timestamp alignment and outlier removal have a judgment standard that can be directly implemented in engineering, the edge controller in this embodiment executes the following synchronization and removal rules for each sampling period: Synchronous alignment: for each sampling period ( The edge controller uses the window start time. With sampling period Constructing target sampling time (In this embodiment) Only receive timestamps from each sensor. The sampled value is used as the input for this period, where To ensure alignment tolerance (unit: seconds), in this embodiment... ).

[0029] Outlier removal: For each aligned sensor sample value The edge controller simultaneously performs dual-threshold culling based on both slope threshold and multi-point consistency threshold: in, The upper limit of the allowable rate of change of carbon dioxide concentration at a single point (unit: ppm / s, which can be set to [value] in this embodiment). ), This represents the median of a four-point sample (unit: ppm). The consistency tolerance coefficient (dimensionless, which can be set to 0.25 in this embodiment) is given by the formula. This is used to avoid false rejection due to an excessively small denominator at low concentrations. If any threshold is not met... Those marked as invalid are not included in the calculation of uniformity index and average concentration for this period.

[0030] Data quality indicator: The edge controller uses the number of valid sensors in this cycle. Constructing data quality labels : in, .when At that time, the edge controller uses the arithmetic mean of the effective points as... ;when At that time, the edge controller maintains the proportional control valve opening and the fan control output unchanged from the previous effective sampling period and... A summary of the window records written to the rolling calibration set is used to penalize windows with unverifiable data in subsequent policy admissions.

[0031] The liquid phase buffer dosing unit is connected to the temporary immersion culture unit and includes a buffer component dosing mechanism, an ion balancing actuator, and a liquid phase state monitoring channel. In this embodiment, the liquid phase state monitoring channel uses online pH and conductivity data acquisition and outputs liquid phase state characteristics. The liquid phase buffer parameter set in this embodiment includes the buffer component dosing amount, the buffer component dosing time, and ion balancing constraint indicators. The buffer component dosing mechanism adds the buffer component to the culture medium loop according to the dosing amount and dosing time specified in the liquid phase buffer parameter set. In this embodiment, the buffer component is set to sodium bicarbonate and diphosphate. The combination of hydrogen and potassium, and the dosage of both, is generated by a dedicated risk-constrained large model controller based on the one-to-one binding relationship of the carbon supply curve parameter set. The ion balancing actuator compensates for the ratio of cations and anions according to the ion balancing constraint identifier to control conductivity drift and suppress osmotic pressure fluctuations in the culture medium. The edge controller continuously receives the liquid phase state characteristics output by the liquid phase state monitoring channel within the carbon supply window and determines the liquid phase drift. When the liquid phase state characteristics meet the degradation conditions in the strategy package, the buffer component compensation dosage and carbon supply curve limiting are triggered simultaneously to suppress the accumulation of liquid phase drift.

[0032] To ensure a clear calculation method for the duration of pH drift exceeding the threshold, the edge controller in this embodiment solidifies the pH drift determination within the carbon supply window into a dual measure of amplitude and duration ratio. The pH change within the window and the drift duration ratio are determined according to the following formula: in, The start time of the carbon supply window. The end time of the carbon supply window. Carbon supply window duration (in seconds). The number of sampling points in the window (dimensionless). The threshold value for pH drift amplitude is dimensionless; in this embodiment, it can be taken as 0.2. This is an indicator function. This represents the percentage of time within the window that exceeds the threshold (dimensionless, range of values). This is used to determine when the duration of a condition exceeding a threshold reaches one-third of the window duration.

[0033] At the end of the carbon supply window, the physiological state acquisition unit acquires images of the seedlings and generates a physiological state vector summary. In this embodiment, the physiological state vector summary includes leaf area estimates, leaf color index, and morphological compactness index, along with timestamps and culture chamber identifiers. The edge controller writes the physiological state vector summary, along with the uniformity index, rise time index, and liquid phase state characteristics within the carbon supply window, into a rolling calibration set. In this embodiment, the rolling calibration set uses a window-scale sequential append storage method and saves the execution records of the most recent 120 carbon supply windows for subsequent policy admission calculations and policy iterations.

[0034] The blockchain-based trusted policy consensus network is used to publish policy packages in consensus and to send cross-site consensus statistics to the edge controller and output the consensus trigger state. In this embodiment, the cross-site consensus statistics use the deviation of the physiological state vector summary distribution of the same species and growth stage as input. The data is calculated by the trusted policy consensus network on-chain or off-chain and then sent to the edge controller. The edge controller reads the consensus trigger state at the beginning of each ventilation phase and uses the consensus trigger state as a pre-threshold for whether to allow the execution of non-degradation policy packages.

[0035] To achieve a clear calculation method for cross-site consistency statistics, the Trusted Policy Consistency Network in this embodiment can output cross-site consistency statistics in the form of window summary mean deviation. Specifically, this refers to: the base The recent One carbon supply window (in this embodiment) Summary of Internal Physiological State Vectors (In this embodiment) Find the mean. and the reference average for network maintenance. and reference covariance matrix Constructing statistics: in, This is a cross-site consistency statistic (dimensionless). , For reference average, For reference covariance matrix, superscript Indicates transpose. The Trusted Policy Consistency Network will... Compare with a consistency threshold to generate a consistency-triggered state, and As a cross-site consistency statistic field, it is sent to the edge controller and written into the rolling calibration set.

[0036] A dedicated risk-constrained large-scale model controller is deployed on a cloud computing node and maintains an encrypted communication connection with the edge controller. The controller receives temporary immersion beat identifiers, multi-point sampling sequences, liquid phase state characteristics, physiological state vector summaries, and cross-site consistency statistics issued by a trusted policy consistency network. Based on a rolling calibration set, it generates candidate control strategies. Before each carbon supply window begins, it generates uniformity thresholds and rise time thresholds, as well as a set of correction actions and a set of limiting rules. Simultaneously, it generates a set of carbon supply curve parameters and a set of liquid phase buffer parameters that are bound one-to-one with the carbon supply curve parameter set. In this embodiment, the carbon supply curve parameter set includes at least the target platform concentration, target... The platform concentration tolerance band width, the duration of continuous stability, the range of rising slope, the duration of carbon supply window, the pulse duty cycle, and the upper limit of cumulative injection volume within the permeable phase are all included in this embodiment. The limit rule set includes at least reducing the target platform concentration, reducing the upper limit of rising slope, shortening the duration of carbon supply window, and reducing the upper limit of cumulative injection volume within the permeable phase. The correction action set includes at least the duct state switching command and the fan speed setting command. The degradation condition is defined in this embodiment as the number of times the uniformity index continuously exceeds the uniformity threshold reaches 3 times or the duration of pH drift characterized by liquid phase state characteristics exceeding 0.2 reaches one-third of the duration of carbon supply window.

[0037] The dedicated risk-constrained large model controller calculates the risk upper bound quantile value based on the rolling calibration set and compares the risk upper bound quantile value with the policy admission threshold as the policy admission condition. In this embodiment, the risk upper bound quantile value is calculated using the empirical quantile of the overall loss at the window scale, and the overall loss is taken as the maximum value of stability loss, physiological loss and consistency loss. A loss sequence is formed for the most recent 120 carbon supply windows. The risk upper bound quantile value is calculated according to the following formula: in, For the first Total loss per carbon supply window (dimensionless). , For quantile parameters, This indicates empirical quantile operations. This is the upper bound quantile value of the risk (dimensionless). The policy admission threshold is set to [value] in this embodiment. (Dimensionless) and written to the policy package. When When the dedicated risk-constrained large model controller prohibits the generation of non-degradation policy packages and only generates degradation policy packages to control tail failure risk; when At that time, the dedicated risk-constrained large model controller compiles the candidate control strategies into a strategy package and writes in the uniformity threshold, rise time threshold, set of correction actions, set of amplitude limiting rules, set of carbon supply curve parameters, set of liquid phase buffer parameters, degradation conditions, strategy admission threshold and risk upper bound quantile value.

[0038] After the policy package is generated, the dedicated risk-constrained large model controller calculates the policy package hash value and the calibration digest hash value, and publishes them through consensus via a trusted policy consensus network. In this embodiment, the policy package hash value is calculated by a hash function on the policy package payload that does not contain the policy package hash value field and the calibration digest hash value field, and satisfies the following expression: in, This is the policy package hash value (a 256-bit hexadecimal string). The strategy package payload includes uniformity threshold, rise time threshold, set of corrective actions, set of limiting rules, set of carbon supply curve parameters, set of liquid phase buffer parameters, degradation conditions, risk upper bound quantile, and strategy admission threshold. The hash function is SHA-256, which outputs a fixed-length hash value for verification. The calibration digest hash value is calculated from the window record digests of the rolling calibration set using the same hash function and is used to prove the binding relationship between policy admission evidence and execution data.

[0039] Before the start of the ventilation phase, the edge controller reads the latest policy package from the trusted policy consistency network and verifies the policy package hash value and the calibration summary hash value. If the verification is successful and the consistency trigger status is not triggered, it enters the carbon supply window and executes the policy package. If the verification fails or the consistency trigger status is triggered, it refuses to execute the non-degraded policy package and executes the conservative mode. In this embodiment, the conservative mode includes forcibly adopting pulsed carbon supply with a low target platform concentration and a cumulative injection limit within the low ventilation phase, and simultaneously executing the set of liquid phase buffer parameters that are bound one-to-one with the pulsed carbon supply. The low target platform concentration is set to 900 ppm and the cumulative injection limit within the low ventilation phase is set to 50% of the conventional policy package to ensure that the controllable carbon supply boundary is maintained under abnormal conditions.

[0040] Within the carbon supply window of the execution strategy package, the edge controller reads the multi-point sampling sequence at a frequency of 1Hz and calculates the uniformity index, the average carbon dioxide concentration in the cavity, and the rise time index in real time. The edge controller controls the proportional control valve opening according to the carbon supply curve parameter set to form an injection flow sequence and controls the mixing fan speed and duct status according to the correction action set to form a mixing sequence. The edge controller compares the uniformity index with the uniformity threshold and the rise time index with the rise time threshold. If either comparison fails, the correction action set is immediately executed, and the amplitude limiting rule set is triggered to adjust the target platform concentration, the upper limit of the rise slope, the duration of the carbon supply window, and the upper limit of the cumulative injection volume within the permeable phase in the next carbon supply curve parameter set. When the control cycle takes effect, if continuous comparisons fail to meet the degradation conditions, the edge controller switches to the degradation carbon supply mode and forces the use of the carbon supply curve parameter set and liquid phase buffer parameter set in the degradation strategy package to achieve rapid convergence. After the carbon supply window ends, the edge controller calls the physiological state acquisition unit to generate a physiological state vector summary, which is then appended to the rolling calibration set along with the evenness index, rise time index, liquid phase state characteristics, cross-site consistency statistics, and strategy package hash value. The window record summary of the rolling calibration set is then reported to the dedicated risk constraint large model controller as the data source for the next strategy admission calculation, thereby enabling a continuous closure of data source, data processing, strategy output, execution verification, and data write-back on the carbon supply window scale.

[0041] Example 2 like Figure 1 and Figure 2 As shown, based on the system structure and closed control chain operation mode of Example 1, this example further solidifies the quantitative calculation process for carbon supply window availability verification within the carbon supply window, and establishes a synchronous triggering relationship between the liquid phase state characteristics output by the liquid phase state monitoring channel and the buffer component compensation addition and carbon supply curve limiting, so that the uniformity index, rise time index and liquid phase drift suppression form a continuous data processing link within the same carbon supply window scale, avoiding the window criterion from only staying at the monitoring layer and failing to form an executable control action.

[0042] In this embodiment, the multi-point carbon dioxide sensor array includes four carbon dioxide sensors arranged on the inlet side, exhaust side, and two diagonal positions of the temporary immersion culture unit cavity. The four carbon dioxide sensors synchronously output carbon dioxide concentration values ​​with a unified clock timestamp within the carbon supply window, using the same sampling period. After the carbon supply window begins, the edge controller establishes a window-level data buffer queue and performs synchronous verification on the four carbon dioxide concentration values ​​for each sampling period. If any sensor lacks data in a sampling period, the edge controller marks that sampling period as invalid and maintains the control output of the previous valid sampling period unchanged, thereby ensuring that the input data sources for the uniformity index and rise time index are continuous and traceable. In each valid sampling period, the edge controller calculates the average carbon dioxide concentration of the cavity based on the multi-point sampling sequence and calculates the uniformity index, which is determined by the following formula: in, This represents the maximum carbon dioxide concentration (in ppm) at four points at the same sampling time. This represents the minimum carbon dioxide concentration (in ppm) at four points at the same sampling time. This represents the average carbon dioxide concentration at four points at the same sampling time (unit: ppm). The uniformity index (dimensionless) has a value range of [value range missing]. The edge controller compares the uniformity index with the uniformity threshold written in the strategy package and forms a uniformity judgment result, which serves as the direct basis for mixing correction and amplitude limiting triggering. When the uniformity judgment result is not met, the edge controller simultaneously issues a duct state switching command and a mixing fan speed setting command according to the correction action set, switching the airflow organization from unidirectional flow to countercurrent flow and increasing the fan speed to the correction speed limited by the strategy package. At the same time, the edge controller performs convergence adjustment on the carbon supply curve parameter set according to the amplitude limiting rule set, so that the proportional control valve opening sequence meets the requirements in subsequent control cycles. The dual constraints of reducing the upper limit of the upward slope and reducing the target platform concentration ensure that rapid high-amplitude carbon supply does not cause local concentration shocks when the spatial uniformity is insufficient. When the number of consecutive times that the uniformity judgment result is not met reaches the degradation condition written in the strategy package, the edge controller switches to the degradation carbon supply mode and replaces the carbon supply curve parameter set with the pulse carbon supply parameters corresponding to the low target platform concentration and the upper limit of cumulative injection in the low permeability phase in the degradation carbon supply mode. This closes the uniformity index - correction action set - amplitude limiting rule set - degradation condition into an instant execution chain within the window.

[0043] In this embodiment, the rise time metric is calculated using a method that determines whether the target platform concentration tolerance band is reached and remains stable. The edge controller records the start time of the carbon supply window as [missing information]. The average carbon dioxide concentration in the cavity was calculated for each valid sampling period. The set of carbon curve parameters is written into the target platform concentration. Target platform concentration tolerance band width and maintain a stable duration The edge controller searches for the earliest entry time that satisfies the stability condition within the carbon supply window. The rise time index is determined at that moment, and the rise time index is determined by the following formula: in, Ascent time indicator (unit: seconds). The start time of the carbon supply window. The earliest moment to enter the stable range. Target platform concentration (unit: ppm). The target platform concentration tolerance bandwidth (unit: ppm). To maintain a stable duration (in seconds), the edge controller compares the rise time metric with the rise time threshold written in the strategy package and forms a rise time determination result. When the rise time determination result is not met, the edge controller considers this non-compliance as an uncontrollable dynamic process of carbon supply and executes a reduction in the target platform concentration and a shortening of the carbon supply window duration according to the limit rule set. At the same time, it maintains the airflow state of the correction action set to reduce hysteresis caused by dead zones in the flow field, and writes the event of the rise time non-compliance into the rolling calibration set so that the dedicated risk constraint large model controller can reduce the allowable rise slope range and the upper limit of the cumulative injection amount within the permeable phase in subsequent strategy admissions. When the rise time determination result is continuously non-compliant and reaches the degradation condition, the edge controller and the uniformity determination link switch to the degradation carbon supply mode in unison, so that the carbon supply curve changes from platform carbon supply to pulse carbon supply and shortens the pulse duty cycle to quickly return to the verifiable dynamic range in the next window.

[0044] In order to make The search process adapts to discrete sampling and avoids ambiguity introduced by continuous-time universal quantifiers in engineering implementation. In this embodiment, the edge controller uses a continuous on-band counter to achieve stable holding determination: let the sampling period be... (In this embodiment) ), and define a stable counting threshold. For each valid sampling period Defined in the case of an indicator And define a continuous counter. : When it first appears At that time, the edge controller will solidify into Thus, a discrete realizable rise time index is obtained. .in, , For the indicator quantity (dimensionless). For continuous use of a tape counter (dimensionless). This is for rounding up.

[0045] Furthermore, to ensure that window availability verification not only relies on a single instantaneous criterion but also possesses a traceable window-level summary, the edge controller in this embodiment can construct a carbon supply window availability score. As an auxiliary feature for writing to the rolling calibration set: in, Uniformity index within the carbon supply window The 95th percentile value (dimensionless, used to suppress the interference of single-point peaks on window evaluation). The uniformity threshold (dimensionless) to be written to the policy package. The rise time threshold (in seconds) for writing the policy package. Data quality indicators within the window Time average (dimensionless, range of values) ), The weighting coefficients are satisfied. (Dimensionless, in this embodiment, it can be taken as...) ). The range of values ​​is The larger the value, the more verifiable and controllable the window is. This score can be used in dedicated risk-constrained large model controllers to give higher weight to windows with data credibility, dynamic reachability and spatial uniformity in policy admission calculation.

[0046] In this embodiment, liquid phase drift suppression is achieved through a linkage triggering mechanism established by the liquid phase state characteristics output by the liquid phase state monitoring channel, buffer component compensation addition, and amplitude limiting rule set. The liquid phase state monitoring channel continuously outputs pH and conductivity and aligns them with the carbon supply window timestamp. The edge controller constructs liquid phase state characteristics within the carbon supply window. The liquid phase state characteristics include at least the pH value at the beginning of the window, the pH value at the end of the window, and the pH change within the window. The pH change is used as the core quantity for liquid phase drift determination. The pH change is determined according to the following formula: in, The end time of the carbon supply window. The edge controller will measure the pH change (dimensionless) within the window. The liquid phase drift determination result is formed by comparing the threshold values ​​related to the degradation conditions written in the strategy package. When the liquid phase drift determination result indicates that the liquid phase state characteristics are close to or meet the degradation conditions, the edge controller triggers the addition of buffer component compensation and simultaneously triggers the limit rule set. The execution object of the buffer component compensation is given by the liquid phase buffer parameter set. In this embodiment, the liquid phase buffer parameter set includes the amount of buffer component added, the timing of the buffer component addition, and the ion balancing constraint identifier. After triggering, the edge controller locks the addition timing within the limit effect period of the carbon supply window and executes the addition of buffer component according to the set value of the liquid phase buffer parameter set. At the same time, it calls the ion balancing execution mechanism to adjust the conductivity offset according to the ion balancing constraint identifier. Compensation is performed to avoid osmotic pressure abrupt changes introduced by buffer compensation. When the limit rule set is triggered synchronously, the edge controller lowers the upper limit of the cumulative injection amount in the permeable phase of the carbon supply curve parameter set and lowers the target platform concentration, so that the compensation addition on the liquid phase side and the carbon supply convergence on the gas phase side occur simultaneously, avoiding drift accumulation caused by only increasing the buffer without limiting the carbon supply or slow recovery caused by only limiting the carbon supply without compensating the buffer. When the liquid phase drift judgment result reaches the degradation condition written in the strategy package, the edge controller switches to the degradation carbon supply mode and synchronously switches to the liquid phase buffer parameter set that is bound one-to-one with the degradation carbon supply mode, so as to ensure that the liquid phase is still in a controlled buffer state when running at a low target platform concentration and a low upper limit of the cumulative injection amount in the permeable phase.

[0047] Within the carbon supply window, the edge controller appends and records the phased calculation status of four points of carbon dioxide concentration, uniformity index, average carbon dioxide concentration in the cavity, rise time index, liquid phase state characteristics, and the execution results of the triggered correction action set and amplitude limiting rule set for each effective sampling cycle in a window-level appending manner. At the end of the carbon supply window, it records the maximum value of the uniformity index sequence, rise time index, and other parameters. The results of buffer compensation addition, along with the physiological state vector summary, cross-base consistency statistics, and strategy package hash value, are written into the rolling calibration set. In this embodiment, the content of the rolling calibration set is kept fixed in terms of fields and appended in the order of carbon supply windows. This allows the dedicated risk-constrained large model controller to directly read the uniformity index and rise time index judgment results, liquid phase drift judgment results, and compensation addition results of the nearest window before the start of the next carbon supply window. This ensures the effective convergence of the carbon supply curve parameter set and the liquid phase buffer parameter set when generating the next strategy package, maintaining the one-to-one binding relationship between the two. Ultimately, this achieves continuous closed-loop execution of multi-point sampling sequence—uniformity index—rise time index—liquid phase state characteristics—buffer compensation addition—amplitude limit rule set—rolling calibration set writing.

[0048] Example 3 like Figures 1 to 3As shown, based on the temporary immersion cycle, carbon supply window phase locking, strategy package blockchain consensus release and edge verification execution mechanism of the sugar-free tissue culture system in Example 1, and the linkage of uniformity index, rise time index and liquid phase drift suppression in Example 2, this example further provides a collaborative implementation method of a dedicated risk constraint large model controller and a blockchain-based trusted policy consistency network in a cross-site industrial deployment scenario. This enables the dedicated risk constraint large model controller to have a feasible domain hard constraint source when generating strategy packages, provides a window-scale real data basis for the calculation of risk upper bound quantile values, and forms a repeatable engineering closed loop for the generation of consistency trigger state, the release of calibration task strategy packages, and the determination of allowing the release of non-degraded strategy packages. This supports the distribution drift convergence and policy adaptive update in the long-term operation of multiple sites.

[0049] The dedicated risk-constrained large-scale model controller has a built-in variety stage map. This map uses variety and growth stage identifiers as indexes, and solidifies the value boundaries of the output carbon supply curve parameter set, the liquid phase buffer parameter set, and the light intensity and temperature settings. In this embodiment, the variety identifiers are set as Bletilla striata, Dendrobium officinale, and Lilium brownii. The growth stage identifiers are set as proliferation stage, seedling vigor stage, and pre-acclimatization stage. The variety stage map is stored in the dedicated risk-constrained large-scale model controller in the form of table entries. The parameter library is managed by version number. Each entry outputs the lower and upper bounds of each parameter as key-value pairs, along with the policy admission threshold and consistency threshold bound to that entry. The parameter set for the carbon supply curve includes at least the lower and upper bounds of the target platform concentration, the lower and upper bounds of the rising slope range, the lower and upper bounds of the carbon supply window duration, the lower and upper bounds of the pulse duty cycle, the lower and upper bounds of the upper limit of the cumulative injection volume within the permeable phase, the lower and upper bounds of the target platform concentration tolerance band width, and the lower and upper bounds of the duration for which stability is maintained. Liquid phase buffer parameters... The value boundaries of the set of parameters include at least the lower and upper bounds of the buffer component dosage, the lower and upper bounds of the buffer component dosage time, and the allowed set of ion balancing constraint identifiers. The value boundaries of the light intensity setting and temperature setting include at least the lower and upper bounds of the light intensity and the lower and upper bounds of the temperature. Before each carbon supply window begins, the dedicated risk constraint large model controller receives the variety identifier and growth stage identifier reported by the edge controller and reads the corresponding value boundaries from the variety stage map. Then, it generates candidate control strategies within the value boundaries and performs boundary verification on the candidate control strategies before compiling the strategy package. If the boundary verification is satisfied, it is allowed to enter the risk upper bound quantile value admission calculation. If the boundary verification is not satisfied, the candidate control strategy is projected to the nearest boundary and the liquid phase buffer parameter set is recalculated to maintain the one-to-one binding relationship between the carbon supply curve parameter set and the liquid phase buffer parameter set. Then, it enters the risk upper bound quantile value admission calculation, so that the feasible domain constraint source of the strategy package is fixed and traceable, avoiding uncontrollable risks introduced by the model's spontaneous exploration output of parameters that exceed the equipment capacity or physiological safety range.

[0050] In this embodiment, the rolling calibration set is constructed using execution records at the carbon supply window scale and adopts a multi-site unified field specification. Execution records are generated and reported by the edge controller after the carbon supply window ends. Each execution record includes at least the uniformity index, rise time index, liquid phase state characteristics, physiological state vector summary, cross-site consistency statistics, and strategy package hash value. The calculation process for the uniformity index and rise time index is consistent with that in Embodiment 2. The liquid phase state characteristics are output from the liquid phase state monitoring channel, and in this embodiment, the window start pH, window end pH, and pH change within the window are used as key stability features. The physiological state vector summary is output from the physiological state acquisition unit and in this embodiment, it is composed of leaf area estimation, leaf color index, and morphological compactness index. The cross-site consistency statistics are constructed using a blockchain-based... The trusted policy consistency network aggregates data on the same variety and growth stage, distributes it to each base, and aligns it with the carbon supply window timestamp. To ensure data processing continuity, the edge controller uses the policy package hash value as the primary key when generating execution records and writes the execution records to the local rolling calibration set cache. Simultaneously, the execution records are uploaded to the data interface of the dedicated risk constraint large-scale model controller and the data interface of the blockchain-based trusted policy consistency network. The dedicated risk constraint large-scale model controller performs deduplication, sequence verification, and missing item verification on the reported execution records. Deduplication is determined by combining the policy package hash value and the carbon supply window timestamp; sequence verification is determined by the temporary immersion beat identifier and the carbon supply window sequence number; and missing item verification is determined by field integrity. Execution records that pass the verification are added to the rolling calibration set, which is maintained with a fixed window length. 1 record In this embodiment, it is set to 120.

[0051] The dedicated risk-constrained large-scale model controller generates stability loss, physiological loss, and consistency loss based on the execution records of the rolling calibration set. The maximum value among these three is used as the overall loss, and the risk upper bound quantile is calculated based on the overall loss. The stability loss characterizes the spatial and dynamic controllability of the carbon supply window and the risk of liquid phase drift. In this embodiment, the stability loss is obtained by fusing the degree of threshold exceedance of the uniformity index, the degree of threshold exceedance of the rise time index, and the degree of threshold exceedance of the pH change, and normalized to the 0-1 range. The physiological loss characterizes the degree to which the physiological state of the seedlings deviates from the target range and is obtained by fusing the growth rate of the leaf area estimate, the deviation of the leaf color index, and the deviation of morphological compactness, and normalized to the 0-1 range. The consistency loss characterizes the risk corresponding to the cross-site consistency deviation and is obtained by mapping the cross-site consistency statistics issued by the blockchain-based trusted policy consistency network and normalized to the 0-1 range. The overall loss is determined according to the following formula: in, For the first Stability loss corresponding to each execution record (dimensionless, range of values) ), For the first Physiological loss corresponding to each execution record (dimensionless, range of values) ), For the first Consistency loss corresponding to each execution record (dimensionless, range of values) ).

[0052] To ensure that the construction of the three types of losses has a reproducible implementation method and avoids the use of the general textbook simple mean square error expression, this embodiment presents a construction method for the final stage weighted overthreshold degree-interval deviation degree-credible weight consistency deviation loss oriented towards the carbon supply window control characteristics, as follows.

[0053] Stability loss : Regarding the first Uniformity sequence within a carbon supply window (Sampling period) ), window rise time pH change within the window Construct the weighted overthreshold at the end of the segment: ; ; ; in, The uniformity threshold (dimensionless) to be written to the policy package. The rise time threshold (in seconds) for writing the policy package. The threshold value for pH drift amplitude is dimensionless; in this embodiment, it can be taken as 0.2. pH drift duration ratio within the window (dimensionless, range of values) ), The number of sampling points in the window (dimensionless). The weighting coefficient for the final segment is defined as follows: (Dimensionless, this embodiment can be taken as) This is used to emphasize the risk of the window segment exceeding the threshold and not having enough time to correct the deviation. For numerical stability terms (which can be taken in this embodiment) ); ; ; To merge weights and satisfy (Dimensionless, this embodiment can be taken as) ); The amplification factor is the continuous ratio (dimensionless, and can be 0.5 in this embodiment).

[0054] Physiological loss The dedicated risk-constrained large-scale model controller reads the target intervals of each component of the physiological state vector summary from the variety stage map. ( ), and for the first Summary of physiological state vectors of each window Construction interval deviation: ; ; in, A summary of leaf area estimates or their growth rates. Leaf color index (dimensionless). It is an index of morphological compactness (dimensionless). These are the lower and upper bounds (the units are the same as the corresponding components). Weight and satisfy (Dimensionless, this embodiment can be taken as) ). It is a bounded mapping, ensuring that a single anomaly will not be amplified unbounded in numerical terms.

[0055] Consistency loss Trusted policy consistency network for base Calculate consistency deviation Cross-site consensus statistics will be distributed; to reflect the impact of on-chain consensus on data credibility, the network can also distribute credibility weights. (Dimensionless, indicating that the base was recently...) The combined score of the record verification pass rate and sensor calibration pass rate within the window is used to construct the consistency loss: in, The consistency threshold (dimensionless) for the output of the variety stage map. The definition is consistent with the Mahalanobis distance described above in this embodiment.

[0056] Through the above construction All were solidified in Within this range, and directly corresponding to the correctability, verifiability, and cross-site replicability of the closed-loop control of the carbon supply window, it serves as the subsequent risk upper bound quantile. The computation provides a repeatable data foundation.

[0057] The upper bound quantile of risk is determined by the following formula: in, The upper bound quantile value of risk (dimensionless, range of values) ), For quantile parameters, This indicates empirical quantile operations. The window length for the rolling calibration set. A dedicated risk-constrained large model controller will... The strategy admission threshold is compared with the output of the variety stage map to determine whether to allow the generation of non-downgraded strategy packages and write the risk upper bound quantile and strategy admission threshold fields into the strategy package accordingly. This ensures that the strategy admission evidence comes from the window-scale closed-loop execution data and is consistent with the carbon supply window criteria, liquid phase characteristics, physiological state vector summaries and cross-site consistency statistics.

[0058] In this embodiment, the blockchain-based trusted policy consistency network is responsible for calculating, triggering, and calibrating cross-site consistency deviation, as well as publishing policy packages. The network constructs a consistency deviation using physiological state vector summaries of the same species and growth stage, along with cross-site consistency statistics, and compares this deviation with a consistency threshold. The consistency deviation is defined using verifiable statistical distance, specifically represented in this embodiment by Mahalanobis distance (the deviation from the mean vector). The trusted policy consistency network uses the nearest... The physiological state vector summaries for each carbon supply window are used to form a mean vector at each site, and the degree of consistency deviation is calculated using a reference covariance matrix. In this embodiment, the value is set to 30, and the consistency deviation is determined according to the following formula: in, This represents the degree of consistency deviation (dimensionless). As a base The mean of physiological state vectors of the same variety and at the same growth stage ( In this embodiment, the physiological state vector dimension is used. ), The reference mean maintained by the trusted policy consistency network. For reference covariance matrix, Its inverse matrix, superscript This represents the matrix transpose. The Trusted Policy Consistency Network will... Compare with the consistency threshold and in When the consistency threshold is exceeded, the consistency trigger state is set to trigger and a calibration task strategy package is issued. In addition to the complete fields of the strategy package in Example 1, the calibration task strategy package in this example additionally includes sensor calibration instructions, duct self-test instructions, and conservative mode execution duration. The sensor calibration instructions include at least the target values ​​and durations of zero-point calibration and span calibration of the carbon dioxide sensor, as well as the pH calibration points and calibration sequence of the liquid phase state monitoring channel. The duct self-test instructions include at least the duct switching action sequence, the fan speed step sequence, and the self-test pass judgment condition. The conservative mode execution duration in this example is set to 12 consecutive carbon supply windows and counted by the carbon supply window sequence number. After receiving the calibration task strategy package, the edge controller refuses to execute the non-degraded strategy package during the period when the consistency trigger state is trigger and enters the conservative mode according to the calibration task strategy package. It completes the sensor calibration and duct self-test according to the instructions. At the same time, it continuously generates execution records during the conservative mode execution duration and writes the execution records into the rolling calibration set, so that the data formed during the calibration phase has the validity to be used for the recalculation of the risk upper bound quantile value.

[0059] In this embodiment, the recovery conditions for allowing the release of non-degraded policy packages, as written in the calibration task policy package, are based on the upper risk quantile value meeting the policy admission criteria. During the execution of the calibration task policy package, the trusted policy consistency network continuously receives execution records reported by each base, calculates the consistency deviation, and updates the cross-base consistency statistics. When the consistency deviation... Falling below the consistency threshold and continuing to meet the continuity requirement During a carbon supply window, the trusted policy consensus network sets the consensus trigger state to non-triggered. In this embodiment, it is set to 10; based on the consistency trigger state being set to untriggered, when consecutive Each carbon supply window meets When the policy admission threshold is reached, the trusted policy consistency network allows the release of non-degraded policy packages. This means that the conditions for resolving the consistency trigger state and allowing the release of non-degraded policy packages logically correspond to consistency decline and risk admission recovery, respectively, and are not contradictory. The dedicated risk-constrained large-scale model controller, after recovering from a triggered to a non-triggered state, maintains the boundary constraints of the product stage map and continues to generate candidate control strategies using a rolling calibration set and performs risk upper bound quantile value admission. This creates a continuous closed loop of cross-base consistency deviation triggering, forced calibration, data reconstruction, and risk admission recovery, with the input data sources, processing procedures, and output results all aligned on the carbon supply window scale.

[0060] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be modified within the scope of the concept described herein by means of the above teachings or the technology or knowledge in related fields.

Claims

1. A sugar-free tissue culture system for the industrialization of rare medicinal plants, characterized in that, It includes a temporary immersion culture unit, a carbon dioxide supply and mixing unit, a liquid phase buffer dosing unit, an environmental control unit, a physiological state acquisition unit, a multi-point carbon dioxide sensor array, an edge controller, a dedicated risk-constrained large model controller, and a blockchain-based trusted policy consistency network. The temporary immersion culture unit uses sugar-free culture medium as the culture medium and is set to a temporary immersion cycle consisting of an immersion phase and a ventilation phase. A carbon supply window is set within the ventilation phase, and the start and end times of the carbon supply window and the ventilation phase are phase-locked so that carbon dioxide injection is only allowed to be performed within the carbon supply window and is prohibited from being performed during the immersion phase. The carbon dioxide supply and mixing unit is connected to the temporary immersion culture unit, and performs carbon dioxide injection according to the carbon supply curve parameter set and performs mixing airflow organization switching according to the correction action set within the carbon supply window. The multi-point carbon dioxide sensor array is arranged in the cavity of the temporary immersion culture unit, and the carbon dioxide concentration at multiple points is collected simultaneously within the carbon supply window to form a multi-point sampling sequence. The liquid phase buffer dosing unit is connected to the temporary immersion culture unit and includes a liquid phase state monitoring channel. It performs buffer component dosing according to the liquid phase buffer parameter set and ion balancing according to the ion balancing constraint identifier. The liquid phase state characteristics are output by the liquid phase state monitoring channel. The environmental control unit sets the light intensity and temperature for the temporary immersion culture unit. The physiological state acquisition unit generates a physiological state vector summary at the end of the carbon supply window. The edge controller maintains a rolling calibration set; The dedicated risk-constrained large model controller receives the temporary immersion beat identifier, the multi-point sampling sequence, the liquid phase state characteristics, the physiological state vector summary, and the cross-site consistency statistics issued by the blockchain-based trusted policy consistency network. It generates candidate control strategies based on a rolling calibration set and generates uniformity thresholds and rise time thresholds, a set of corrective actions and a set of limiting rules before each carbon supply window begins. Simultaneously, it generates the set of carbon supply curve parameters and a set of liquid phase buffer parameters that are bound one-to-one with the set of carbon supply curve parameters. Based on the rolling calibration set, it calculates the risk upper bound quantile and compares it with the policy admission threshold as a policy admission condition. If the policy admission condition is not met, it prohibits the generation of non-degradation policy packages. If the policy admission condition is met, it compiles the candidate control strategies into a policy package and writes the uniformity threshold, rise time threshold, set of corrective actions, set of limiting rules, set of carbon supply curve parameters, set of liquid phase buffer parameters, degradation condition, risk upper bound quantile, and policy admission threshold into the policy package. The blockchain-based trusted policy consensus network is used to publish the policy package in consensus, distribute the cross-base consensus statistics, and output the consensus trigger status. The edge controller verifies the strategy package hash value and the calibration summary hash value before the carbon supply window begins. If the verification is successful and the consistency trigger state is not triggered, the strategy package is executed. Within the carbon supply window, the uniformity index and rise time index are calculated from the multi-point sampling sequence. The uniformity index is compared with the uniformity threshold, and the rise time index is compared with the rise time threshold. If either comparison fails, carbon supply curve limiting and mixing correction are performed according to the correction action set and the limiting rule set. A degraded carbon supply mode is executed when the degradation condition is met. The uniformity index, rise time index, liquid phase state characteristics, physiological state vector summary, cross-site consistency statistics, and strategy package hash value are written into the rolling calibration set. When the verification fails or the consistency trigger state is triggered, the edge controller refuses to execute the non-degraded strategy package and executes the conservative mode, so that the carbon supply window criterion generation, carbon supply window availability verification, carbon supply curve binding with liquid phase buffer, risk upper bound quantile value admission, strategy package blockchain consensus release, strategy package verification execution, consistency trigger and rolling calibration set update constitute a control chain that must be closed.

2. The sugar-free tissue culture system for the industrialization of rare medicinal plants according to claim 1, characterized in that: The policy package hash value is a hash value obtained by performing a hash operation on the policy package payload that does not contain the policy package hash value field, and is written into the policy package for verification; the calibration summary hash value is a hash value obtained by performing a hash operation on the window record summary of the rolling calibration set, and is written into the policy package for verification; the uniformity index is determined by the ratio of the difference between the maximum and minimum carbon dioxide concentrations in the multi-point sampling sequence at the same sampling time to the average carbon dioxide concentration in the multi-point sampling sequence.

3. The sugar-free tissue culture system for the industrialization of rare medicinal plants according to claim 2, characterized in that: The rise time index is determined by the length of time from the start of the carbon supply window to the time when the average carbon dioxide concentration in the cavity corresponding to the multi-point sampling sequence enters the target platform concentration tolerance band and remains stable; the target platform concentration, the width of the target platform concentration tolerance band, and the duration of stability are given by the set of parameters of the carbon supply curve.

4. The sugar-free tissue culture system for the industrialization of rare medicinal plants according to claim 3, characterized in that: The carbon dioxide supply and mixing unit includes a proportional control valve, a carbon dioxide injection pipeline, a mixing fan, and a switchable air duct structure. The edge controller controls the speed of the mixing fan and the air duct state of the switchable air duct structure according to the correction action set, and controls the opening of the proportional control valve according to the carbon supply curve parameter set, so as to perform the linkage of carbon dioxide injection and mixing correction within the carbon supply window.

5. The sugar-free tissue culture system for the industrialization of rare medicinal plants according to claim 4, characterized in that: The set of carbon supply curve parameters includes target platform concentration, target platform concentration tolerance band width, duration of continuous stability, range of rising slope, duration of carbon supply window, pulse duty cycle and upper limit of cumulative injection volume within the permeability phase. The set of limiting rules includes reducing the target platform concentration, reducing the upper limit of the rise slope, shortening the duration of the carbon supply window, and reducing the upper limit of the cumulative injection volume within the ventilated phase.

6. The sugar-free tissue culture system for the industrialization of rare medicinal plants according to claim 5, characterized in that: The set of liquid phase buffer parameters includes the amount of buffer component added, the timing of buffer component addition, and the ion balancing constraint identifier; the edge controller triggers the addition of buffer component compensation based on the liquid phase state characteristics output by the liquid phase state monitoring channel and the degradation conditions, and simultaneously triggers the set of limiting rules to suppress the accumulation of liquid phase drift.

7. The sugar-free tissue culture system for the industrialization of rare medicinal plants according to claim 6, characterized in that: The dedicated risk constraint large model controller has a built-in variety stage map; the variety stage map outputs the value boundaries of the carbon supply curve parameter set, the value boundaries of the liquid phase buffer parameter set, and the value boundaries of the light intensity setting and the temperature setting, using variety identifier and growth stage identifier as indexes; the dedicated risk constraint large model controller generates the strategy package that satisfies the strategy admission conditions within the value boundaries.

8. The sugar-free tissue culture system for the industrialization of rare medicinal plants according to claim 7, characterized in that: The rolling calibration set consists of execution records at the carbon window scale, and each execution record includes a uniformity index, a rise time index, liquid phase state characteristics, a physiological state vector summary, the cross-base consistency statistic, and the strategy package hash value; the dedicated risk-constrained large model controller generates stability loss, physiological loss, and consistency loss from the execution records, and uses the maximum value among the three as the overall loss, and calculates the risk upper bound quantile value based on the overall loss.

9. The sugar-free tissue culture system for the industrialization of rare medicinal plants according to claim 8, characterized in that: The blockchain-based trusted strategy consistency network calculates the consistency deviation based on the physiological state vector summary of the same variety and growth stage and the cross-site consistency statistics. When the consistency deviation exceeds the consistency threshold, the consistency trigger state is set to trigger and a calibration task strategy package is released. The calibration task strategy package includes sensor calibration instructions, air duct self-check instructions and conservative mode execution time. The upper risk quantile value satisfies the strategy admission condition as the recovery condition for allowing the release of a non-degradation strategy package.

10. The sugar-free tissue culture system for the industrialization of rare medicinal plants according to claim 9, characterized in that: The conservative mode includes a pulsed carbon supply that forces a low target platform concentration and a maximum cumulative injection volume within the low permeability phase, and simultaneously executes the set of liquid-phase buffer parameters that are one-to-one bound to the pulsed carbon supply.

Images