Small-particle plant negative oxygen ion intelligent regulation method based on multi-parameter linkage

Abstract

The application discloses a small-particle-plant-negative-oxygen-ion intelligent regulation method based on multi-parameter linkage, which firstly collects a multi-dimensional parameter set including environmental parameters, plant physiological parameters and negative oxygen ion parameters in real time; then, the parameter set is input into a pre-trained multi-parameter linkage prediction model to predict the generation trend of small-particle negative oxygen ions; based on the prediction result, a multi-parameter linkage regulation instruction set is generated through a control algorithm with the goal of comprehensively optimizing small-particle concentration, proportion, energy consumption and stability; finally, the regulation is executed through the coordinated dynamic adjustment of plant growth environment, auxiliary negative oxygen ion generation conditions and plant physiological state, and closed-loop adaptive optimization is carried out based on the feedback result; the application realizes multi-parameter, intelligent and accurate coordinated regulation of the small-particle-plant-negative-oxygen-ion generation process, and significantly improves the generation efficiency, concentration stability and overall energy efficiency of small-particle negative oxygen ions.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology, and specifically discloses a method for intelligent regulation of negative oxygen ions in small-particle plants based on multi-parameter linkage. Background Technology

[0002] Negative oxygen ions are known as "air vitamins" and have multiple benefits for human health and the ecological environment, such as improving cardiopulmonary function, enhancing immunity, inhibiting bacteria and removing dust, and regulating mood. Among them, small-particle negative oxygen ions with high mobility and strong biological activity have become the focus of research in the field of environmental health because they can more easily penetrate the human blood-brain barrier and have a wider range of effects.

[0003] Currently, the main technical approaches for artificially generating negative oxygen ions include the following:

[0004] High-voltage ionization, waterfall / fountain method (Lenard effect), mineral radioactivity method, and plant release method; research has found that certain plants (such as conifers) can release negative oxygen ions from their leaf tips during their metabolism, especially during photosynthesis and transpiration; this method is ecological, safe, and produces no secondary pollution, and is considered the most ideal green source of negative oxygen ions; however, existing plant-based technologies have the following key bottlenecks:

[0005] Uncontrollable particle size: Existing research and technology mainly focus on the total concentration of negative oxygen ions, lacking active monitoring and targeted control methods for the particle size distribution of released ions, making it difficult to ensure the generation of a high proportion of small-particle-size negative oxygen ions;

[0006] The control parameters are singular and isolated: Traditional environmental control only considers the effects of a few environmental factors such as temperature, humidity, and light on plant growth, without deeply coupling these environmental parameters with the real-time physiological state of the plant (such as stomatal conductance, transpiration rate, and membrane potential), and without establishing a dynamic correlation model between these multi-source parameters and the particle size and concentration of negative oxygen ions. The control methods are crude, mostly open-loop or simple feedback control;

[0007] Lagging response and insufficient intelligence: Existing systems typically use threshold-based switching control, lacking the ability to predict future trends in ion concentration and thus failing to achieve "pre-adjustment." This results in a lag in system response to environmental fluctuations and poor stability of negative oxygen ion concentration.

[0008] Ignoring the dominant role of plant life: Many technical solutions treat plants as passive release sources, failing to consider the fundamental impact of their own physiological rhythms, health status, and stress on ion release efficiency, and lacking "synergistic" or "inducible" regulatory strategies based on plant physiological feedback.

[0009] In addition, the existing technology lacks a control system that can deeply integrate multi-source sensor data and realize intelligent predictive control and multi-objective collaborative optimization, resulting in a slow response, coarse regulation, high energy consumption and poor stability in the negative oxygen ion generation process, making it difficult to achieve long-term adaptive operation and precise regulation.

[0010] Therefore, it is necessary to invent a smart regulation method for small-particle-size plant negative oxygen ions based on multi-parameter linkage to solve the above problems. Summary of the Invention

[0011] To overcome the aforementioned deficiencies in the prior art, this invention provides a method for intelligent regulation of small-particle-size plant negative oxygen ions based on multi-parameter linkage. This method involves real-time acquisition of a multi-dimensional parameter set, including environmental parameters, plant physiological parameters, and negative oxygen ion parameters. The parameter set is then input into a pre-trained multi-parameter linkage prediction model to predict the generation trend of small-particle-size negative oxygen ions. Based on the prediction results, a multi-parameter linkage regulation instruction set is generated using a control algorithm aimed at comprehensively optimizing the concentration, proportion, energy consumption, and stability of small-particle-size ions. Finally, regulation is executed by synergistically and dynamically adjusting the plant growth environment, auxiliary negative oxygen ion generation conditions, and plant physiological state, and closed-loop adaptive optimization is performed based on feedback results, effectively solving the problems mentioned in the background art.

[0012] To achieve the above objectives, the present invention provides the following technical solution: a method for intelligent regulation of small-particle-size plant negative oxygen ions based on multi-parameter linkage, specifically including the following steps:

[0013] S1: Real-time acquisition of a multi-dimensional parameter set, which includes a subset of environmental parameters, a subset of plant physiological parameters, and a subset of negative oxygen ion parameters; the environmental parameter subset includes temperature, humidity, light intensity, carbon dioxide concentration, and air velocity;

[0014] S2: Input the multi-dimensional parameter set into the pre-trained multi-parameter linkage prediction model, and output the prediction results of the generation trend and target concentration of small-particle negative oxygen ions within a future set time period. The small-particle size refers to negative oxygen ions with an aerodynamic diameter of no more than 0.003 micrometers.

[0015] S3: Based on the prediction results, a multi-parameter linkage control instruction set is generated through an optimized control algorithm. The optimized control algorithm aims to achieve the comprehensive optimization of small-particle-size negative oxygen ion concentration, proportion, system energy consumption, and output stability.

[0016] S4: Execute the set of control instructions to synergistically intervene in the generation process of small-particle-size negative oxygen ions by dynamically adjusting the plant growth environment conditions, auxiliary negative oxygen ion generation conditions, and plant physiological regulation conditions.

[0017] S5: Based on the subset of negative oxygen ion parameters collected after intervention, calculate the regulation effect index, and feed the index back to the multi-parameter linkage prediction model in step S2 and the optimization control algorithm in step S3 to achieve closed-loop adaptive optimization.

[0018] The technical effects and advantages of this invention are as follows:

[0019] 1. By introducing precise monitoring methods such as ion mobility spectrometry, the key indicator of small particle size is incorporated into real-time feedback and optimization targets; combined with multi-parameter linkage prediction models and intelligent optimization algorithms, the system can dynamically coordinate the environment, plant physiology and auxiliary generation modules, thereby significantly increasing the absolute concentration of small-particle negative oxygen ions and their proportion in the total negative oxygen ions.

[0020] 2. It abandons the traditional single or open-loop control logic and establishes a complete intelligent closed loop of "perception-prediction-decision-execution-learning"; the multi-parameter linkage model can predict the changing trend in advance, enabling the system to intervene in advance and shorten the time to reach the target concentration; at the same time, based on the feedback of the regulation effect index and the online incremental learning of the model, the system can continuously adapt to environmental fluctuations and the plant's own growth changes, ensuring long-term operational stability (24-hour stability improved by about 25%), solving the pain points of fixed parameters and poor adaptability of traditional methods;

[0021] 3. It deeply couples the technologies of the three previously isolated fields of "environmental regulation," "plant physiology," and "physical generation." It not only regulates environmental factors such as temperature and humidity to adapt to plant growth, but also observes and actively regulates the plant's own "release intention" and metabolic state by monitoring signals such as leaf electrophysiology and transpiration rate. At the same time, it dynamically matches the parameters of the auxiliary high-voltage electric field (such as frequency and field strength) with the plant's release rhythm and environmental conditions, forming a positive superposition of biological and physical effects. This synergy of the multi-mode execution layer greatly improves the overall energy efficiency, achieving energy saving while enhancing the effect. Attached Figure Description

[0022] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating the overall steps of the present invention.

[0024] Figure 2 This is a schematic diagram of the collaborative intervention of the execution control instruction set of the present invention.

[0025] Figure 3 This is a flowchart of the closed-loop adaptive optimization process of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] This invention provides, for example Figure 1 The intelligent regulation method for small-particle-size plant negative oxygen ions based on multi-parameter linkage shown includes the following steps:

[0028] S1: Real-time acquisition of a multi-dimensional parameter set, which includes a subset of environmental parameters, a subset of plant physiological parameters, and a subset of negative oxygen ion parameters; the environmental parameter subset includes temperature, humidity, light intensity, carbon dioxide concentration, and air velocity;

[0029] Furthermore, in the above technical solution, the collection of multi-dimensional parameter sets specifically includes:

[0030] The subset of environmental parameters is collected synchronously through an environmental sensor grid deployed in the plant canopy space;

[0031] Furthermore, the environmental sensor grid is uniformly distributed in three dimensions, with a node spacing of no more than 0.5 meters. It is primarily positioned in the middle and upper airflow areas of the plant canopy to avoid direct shading by equipment or leaves. Temperature, humidity, and CO2 sensors are installed inside the canopy, 10–20 cm from representative leaves; the light sensor (PAR) is installed facing upwards at the top of the canopy to ensure its receiving angle aligns with the light-receiving surface of the plant canopy. All sensors are synchronized via a unified time-scaled protocol (such as PTP), with a time synchronization error of less than 100 milliseconds.

[0032] The plant physiological parameters are a subset of those parameters, which are obtained by collecting leaf electrophysiological signals using a contact leaf sensor and measuring the plant transpiration rate using a heat balance method or a stem flow meter method.

[0033] Furthermore, leaf electrophysiological signal acquisition employed a minimally invasive clamp-on silver-silver chloride electrode. A plant-safe conductive gel (such as 0.9% NaCl agar) was coated between the electrode and the lower surface of the leaf (avoiding the main vein). The electrode was cleaned and the contact position changed every 48 hours to prevent tissue damage from prolonged contact. The electrophysiological signal sampling frequency was 100 Hz, and after bandpass filtering of 0.1–100 Hz, the root mean square amplitude of the 5–30 Hz frequency band was extracted as the characteristic signal.

[0034] Transpiration rate is preferably measured using a heat balance stem flow meter (such as the Dynamax SGB series), installed in the middle of the main stem of the plant (5–10 mm in diameter), with an additional heat-insulating radiation protection sleeve to reduce the impact of ambient temperature fluctuations on measurement accuracy.

[0035] The particle size distribution of negative oxygen ions is measured by ion mobility spectrometry, and the concentration of negative oxygen ions is measured by capacitive ion collection method to determine the real-time concentration and proportion of small-particle negative oxygen ions, which constitute the negative oxygen ion parameter subset. The small-particle size refers to negative oxygen ions with an aerodynamic diameter of no more than 0.003 micrometers.

[0036] Furthermore, the ion mobility spectrometer (IMS) and capacitive ion collector require cross-calibration weekly. Calibration is performed under a zero-gas (filtered with activated carbon and molecular sieves) environment at constant temperature (25±1℃) and constant humidity (50±5%RH). 241 Am standard ion source. If the total concentration of negative oxygen ions measured by the two systems consistently deviates by more than 15%, the system will automatically diagnose and prioritize the particle size resolution data from the ion mobility spectrometer. The ion sampling flow rate is uniformly set at 1.0 L / min, and the sampling tube length does not exceed 2 meters to reduce ion recombination losses in the pipeline.

[0037] It should be further explained that, in the preferred technical solution of the invention, a distributed sensor network needs to be constructed to achieve synchronous and accurate acquisition of multi-dimensional parameters. The acquisition of a subset of environmental parameters is accomplished by deploying high-precision digital sensors in three dimensions within the plant canopy space: temperature and humidity are measured using integrated temperature and humidity sensors, networked via an I2C bus, with at least three nodes per cubic meter of space to ensure spatial gradient monitoring; light intensity and spectrum are measured using full-spectrum PAR sensors with cosine correction, installed vertically at the top, middle, and bottom of the canopy; carbon dioxide concentration is measured using non-dispersive infrared (NDIR) sensors, installed in the middle of the canopy where plant respiration is active; air velocity is measured at multiple points, including air inlets, between canopies, and at air outlets, using hot-wire micro-anemometers. All environmental sensors use a 5-second sampling period and are synchronized to the central controller via an RS-485 or LoRa wireless network.

[0038] Acquiring a subset of plant physiological parameters requires a combination of non-invasive and minimally invasive techniques. Leaf electrophysiological signals are transmitted via a silver-silver chloride electrode contacting the lower surface of the leaf (avoiding the main vein), connected to a high input impedance (>10). 12A bioelectric amplifier (Ω) is used to acquire raw signals, including membrane potential and action potential, and the characteristic frequency amplitude is extracted after bandpass filtering in the range of 0.1-100Hz. Transpiration rate measurement is preferably performed using a heat-balanced stem flow meter, with a heating ring installed on the main stem of the plant. Instantaneous transpiration flow is calculated by measuring the heating power and temperature gradient; alternatively, a high-precision weighing lysimeter can be used for indirect calibration. Physiological parameter sampling and negative oxygen ion parameter acquisition must be synchronized to analyze the instantaneous correlation.

[0039] The acquisition of a subset of negative oxygen ion parameters is the core step. A parallel-plate ion mobility spectrometer (IMS) is used to analyze the particle size distribution of negative oxygen ions in real time: air samples are passed through an ionization zone (using a parallel-plate ion mobility spectrometer) at a constant flow rate (typically 1.0 L / min). 63 After entering the migration region using a Ni soft β source, ions are separated according to their mobility in a superimposed DC and RF electric field, and finally, the particle size distribution is recorded by a Faraday cup detector. Small particle sizes (aerodynamic diameter ≤ 0.003 µm) correspond to the high mobility range, and their concentration and percentage of the total negative oxygen ion concentration are calculated by integrating specific peak areas. Simultaneously, a dual-cylinder capacitive ion collector (such as an AlphaLab Inc. product) is used to continuously monitor the total negative oxygen ion concentration as auxiliary verification. Both systems need to be periodically tested with a standard ion source (such as... 241 Cross-calibration is performed using Am to ensure particle size resolution and concentration accuracy.

[0040] S2: Input the multi-dimensional parameter set into the pre-trained multi-parameter linkage prediction model and output the prediction results of the generation trend and target concentration of small-particle negative oxygen ions within a future set time period;

[0041] Furthermore, in the above technical solution, the construction and prediction process of the multi-parameter linkage prediction model includes:

[0042] S201: Preprocess historical time-series data, which includes a subset of environmental parameters, a subset of plant physiological parameters, and a corresponding subset of negative oxygen ion parameters;

[0043] S202: Based on the preprocessed data, the correlation analysis method is used to determine the dynamic weight relationship between each input parameter and the generation rate of small-particle negative oxygen ions, and to screen out the core control parameters.

[0044] S203: Using the time-series data of the core regulation parameters as input and the concentration of small-particle negative oxygen ions as output, train a long short-term memory network model to learn the time-series dependencies between parameters.

[0045] S204: Input the core control parameters collected in real time into the trained long short-term memory network model to predict the change curve of small-particle negative oxygen ion concentration within a future set time period.

[0046] It should be further explained that in the preferred technical solution of the invention, model construction is divided into three stages: data preprocessing, feature selection, and model training and validation.

[0047] Data preprocessing stage (S201): Collect historical time-series data for at least three complete plant growth cycles, including all raw sensor readings. First, outlier removal is performed using the 3σ rule based on dynamic thresholds to eliminate wild values, and missing values ​​are filled using linear interpolation of adjacent points. Next, time alignment and resampling are performed, unifying signals with different sampling rates to a 1-minute interval using linear interpolation. Finally, normalization is performed, applying Min-Max normalization to the [0,1] interval for each parameter sequence to eliminate dimensional effects.

[0048] Feature screening stage (S202): Grey relational analysis is used to dynamically determine the correlation weight between each parameter and the generation rate of small-particle negative oxygen ions (calculated from concentration difference). Using the generation rate sequence as a reference sequence and the sequences of various environmental and physiological parameters as comparison sequences, the correlation coefficient at each time point is calculated, and the dynamic correlation degree is calculated through a sliding time window (e.g., 30 minutes). Parameters with a correlation degree consistently higher than 0.7 are selected as core control parameters, typically including photosynthetically active radiation intensity, relative humidity, stomatal conductance (derived from transpiration), canopy CO2 concentration, and air velocity.

[0049] Model Training Phase (S203): A Long Short-Term Memory (LSTM) network is used as the main component of the multi-parameter linked prediction model. The network structure is designed as follows: the input layer dimension equals the number of core control parameters; two LSTM hidden layers, each with 64 neurons, using the tanh activation function; followed by a Dropout layer (dropout rate 0.2) to prevent overfitting; the output layer is a fully connected layer, outputting the predicted sequence of small-particle negative oxygen ion concentration for the next 30 minutes (step size 5 minutes, 6 points in total). The loss function is the mean squared error (MSE), the optimizer is Adam, and the initial learning rate is set to 0.001. During training, the data is divided into a training set (70%), a validation set (15%), and a test set (15%) in chronological order, with early stopping criterion being that the loss on the validation set no longer decreases. The final model must satisfy the condition that the coefficient of determination R² between the predicted and measured values ​​on the test set is greater than 0.85.

[0050] Prediction Execution Phase (S204): The latest 30-minute sequence of core regulatory parameters (after standardization) is input into the trained LSTM model in real time, directly outputting the concentration prediction curve for the next 30 minutes. Simultaneously, the model outputs confidence interval estimates, providing uncertainty references for subsequent optimized control.

[0051] S3: Based on the prediction results, a multi-parameter linkage control instruction set is generated through an optimized control algorithm. The optimized control algorithm aims to achieve the comprehensive optimization of small-particle-size negative oxygen ion concentration, proportion, system energy consumption, and output stability.

[0052] Furthermore, in the above technical solution, the optimization control algorithm is a swarm intelligence-based optimization algorithm. It uses an iterative search to solve for a set of optimal controlled variable setpoints by taking into account a multi-objective optimization function that comprehensively considers the concentration of small-particle negative oxygen ions, the proportion of small-particle size, system energy consumption, and concentration stability as the evaluation criteria, thereby forming the control instruction set.

[0053] It should be further noted that, in the preferred embodiment of the present invention, the system employs a swarm intelligence optimization algorithm based on the non-dominated sorting genetic algorithm (NSGA-II) to solve multi-objective optimization problems.

[0054] First, a multi-objective optimization function is defined. The decision variables are a set of adjustable actuator setpoints, including: target temperature of the culture space T_set (range 18-28℃), target humidity RH_set (range 50-80%), artificial light source intensity P_light (range 0-1000 µmol / m² / s) and R:B light quality ratio, high-voltage pulse electric field intensity E_field (range 1-10 kV / cm), pulse frequency f_pulse (range 10-1000 Hz), and nutrient solution conductivity EC_set (range 1.0-3.0 mS / cm). The optimization objective function contains four components:

[0055] Concentration maximization target: f1 = -avg(C_pred), where C_pred is the predicted small particle size concentration.

[0056] The objective for maximizing the proportion of small particles is: f2 = -avg(R_pred), where R_pred is the predicted proportion of small particles.

[0057] Energy consumption minimization objective: f3 = P_total, which is the estimated total power of all actuators (temperature control, humidification, light source, high voltage power supply, pump).

[0058] Concentration stability target: f4=std(C_pred), which is the standard deviation of the predicted concentration sequence; the smaller the deviation, the more stable the concentration.

[0059] During algorithm initialization, a population of 100 is randomly generated, with each individual representing a set of decision variables. During iteration, the four objective function values ​​for each individual are evaluated by simulating actuator actions and invoking a multi-parameter linked prediction model. NSGA-II performs hierarchical selection of the population through fast non-dominated sorting and crowding calculation, and uses simulated binary crossover and polynomial mutation to generate offspring. The optimization process terminates after a preset maximum number of iterations (e.g., 50 generations). From the final non-dominated solution set (Pareto front), the decision-maker selects a compromise solution based on real-time preferences (e.g., prioritizing energy efficiency or concentration). The decision variable values ​​corresponding to this solution constitute the control instruction set for the current control cycle.

[0060] S4: Execute the set of control instructions to synergistically intervene in the generation process of small-particle-size negative oxygen ions by dynamically adjusting the plant growth environment conditions, auxiliary negative oxygen ion generation conditions, and plant physiological regulation conditions.

[0061] Furthermore, in the above technical solution, the dynamic adjustment of plant growth environment conditions includes: according to the control instructions, using fuzzy proportional-integral-derivative control logic to quickly and accurately adjust the temperature and humidity of the cultivation space, and adjusting the intensity and spectrum of the artificial light source according to the preset light quality and light cycle.

[0062] Furthermore, in the above technical solution, the dynamic adjustment of the auxiliary negative oxygen ion generation conditions includes: starting, stopping, or adjusting the parameters of a high-voltage pulse electric field according to the control command; the pulse width, frequency, and field strength of the high-voltage pulse electric field are dynamically matched and adjusted according to the real-time ambient humidity, air velocity, and predicted plant release rhythm, so as to assist in the generation of small-particle negative oxygen ions and inhibit ozone generation.

[0063] Furthermore, in the above technical solution, the dynamic adjustment of plant physiological regulation conditions includes: adjusting the conductivity and composition of the nutrient solution supplied to the plant according to the regulation instructions and real-time plant transpiration rate data; and triggering the spraying system to apply plant-derived biostimulants when the plant is identified to regulate the plant's secondary metabolic activities.

[0064] It should be further noted that, in the preferred embodiment of this invention, the collaborative intervention process is as follows: Figure 2 As shown, the detailed operation is as follows:

[0065] The control command set is issued to each execution subsystem for coordinated dynamic adjustment.

[0066] Plant growth environment condition regulation: A fuzzy adaptive PID controller is used to regulate temperature and humidity. Temperature control uses T_set as the target, and combines real-time temperature and rate of change to adjust PID parameters (Kp, Ki, Kd) online through fuzzy rules. The output signal controls the semiconductor cooling chip (heating / cooling) and the speed of the variable frequency fan. Humidity control uses RH_set as the target, and employs similar fuzzy PID logic to control the start and stop of the ultrasonic humidifier and the operation of the dehumidifying fan. Light regulation is based on P_light and light quality ratio in the command, and uses PWM dimming to drive programmable LED plant lights to precisely match the preset photoperiod (e.g., 16 hours of light / 8 hours of darkness) and spectral formula (e.g., increasing the red and blue light ratio during the photosynthetic period).

[0067] Assisted negative ion generation condition adjustment: The high-voltage pulsed electric field generator receives E_field, f_pulse, and pulse width (usually fixed at the microsecond level) commands. Its output parameters need to be matched with the environment in real time: when the ambient humidity is >70%, the field strength is automatically reduced to avoid air breakdown and arcing; when the air velocity is high, the pulse frequency is increased accordingly to maintain the ion generation intensity. Most importantly, it is synchronized with the predicted plant release rhythm: during the predicted peak period of natural plant release, the external electric field strength is appropriately reduced to avoid inhibiting the plant's own electrophysiological activity; during the trough period, the external field assistance is enhanced to form a synergistic effect of "peak shaving and valley filling," while ozone generation is suppressed by optimizing the pulse leading edge and polarity.

[0068] Furthermore, the specific method for dynamically matching the parameters of the high-voltage pulsed electric field with the plant's release rhythm is as follows: the system incorporates a plant daily release curve fitted based on a daily accumulated temperature model and historical release data. When the predicted peak release period (usually mid-day of the photoperiod) is in progress, the external electric field intensity E_field is reduced to 30%-50% of the baseline value; when the predicted trough release period (usually early-day of the dark period) is in progress, E_field is increased to 150%-200% of the baseline value. Simultaneously, when the real-time ambient humidity is >70%, the system automatically limits the upper limit of E_field to below 5kV / cm to prevent air breakdown from generating electric arcs and ozone.

[0069] Plant physiological regulation: The nutrient solution control system dynamically adjusts the A / B stock solution injection ratio and rate based on real-time transpiration rate and EC_set commands to maintain a stable rhizosphere environment. When the system detects that the plant has entered a rapid growth phase or flowering stage (determined through accumulated temperature model and image recognition) and the negative oxygen ion release efficiency is lower than expected, it triggers the plant-derived biostimulant spraying subsystem. This system pre-prepares a solution containing seaweed extract, humic acid, and other components, and sprays it onto the leaves above the canopy through precision atomizing nozzles. The spraying duration and dosage are determined by referring to tables based on the plant species and physiological stage, aiming to gently stimulate plant secondary metabolism and enhance the release of volatile organic compounds and negative oxygen ions.

[0070] It should be further noted that the spraying dosage is calculated using the formula Dose=k*LAI*(1+ΔET), where LAI is the leaf area index, ΔET is the rate of change of transpiration rate relative to the baseline, and k is a plant species-related constant (e.g., k=0.8mL / m² for conifers). Spraying should be carried out during the daytime, avoiding periods of high temperature.

[0071] S5: Based on the subset of negative oxygen ion parameters collected after intervention, calculate the regulation effect index, and feed the index back to the multi-parameter linkage prediction model in step S2 and the optimization control algorithm in step S3 to achieve closed-loop adaptive optimization.

[0072] Furthermore, in the above technical solution, the closed-loop adaptive optimization process is as follows: Figure 3 As shown, it specifically includes:

[0073] S501: After a control cycle is completed, obtain the actual concentration of small-particle-size negative oxygen ions and the total concentration;

[0074] S502: Calculate a comprehensive regulatory effect index based on the ratio of actual concentration to target concentration and the actual proportion of small-particle negative oxygen ions.

[0075] S503: Compare the regulation effect index with a preset threshold. If the index continues to be lower than the threshold or the environmental parameters fluctuate drastically, it is determined to be a model mismatch, and the online model update process is triggered.

[0076] It should be further explained that the criteria for determining "drastic fluctuations in environmental parameters" are: the rate of change of any core environmental parameter (temperature, humidity, light intensity) within a single control cycle exceeds twice the standard deviation of its normal fluctuation range (the normal fluctuation range is defined as the standard deviation of the rate of change of the parameter per minute in the past 24 hours), for example, a temperature change exceeding ±3℃ / minute, or a relative humidity change exceeding ±10% / minute, or a light intensity change exceeding ±200µmol / m² / s / minute.

[0077] S504: In the online model update process, the data from the latest few periods are used as incremental datasets to perform incremental learning on the multi-parameter linkage prediction model and update its network weight parameters.

[0078] It should be further noted that, in the preferred embodiment of this invention, the closed-loop adaptive optimization is executed in units of control cycles (configurable, default 1 minute). After each cycle, the system collects the actual negative oxygen ion parameters and calculates the regulation effect index η:

[0079] η=α*(C_act / C_target)+β*(R_act / R_target)-γ*(P_act / P_base);

[0080] Where C_act, R_act, and P_act are the measured values ​​of concentration, percentage, and cycle energy consumption, respectively; C_target and R_target are the set targets for concentration and percentage; P_base is the baseline energy consumption; and α, β, and γ are weighting coefficients (default 0.4, 0.4, 0.2). An η value close to 1 indicates ideal performance.

[0081] The system continuously monitors the η value and the variance of environmental parameter fluctuations. If η is below a threshold (e.g., 0.75) for five consecutive cycles, or if a key environmental parameter (e.g., temperature) fluctuates beyond a set range (e.g., ±3℃) within a single cycle, it is determined to be a model prediction mismatch, triggering an online incremental learning process. This process uses the data from the most recent two hours (pre-processed) as the incremental dataset, and retrains the LSTM model for a short period (usually 5-10 epochs) with a small learning rate (e.g., 0.0001), only fine-tuning the weights of the last two network layers, thereby quickly adapting to new operating conditions without forgetting existing knowledge.

[0082] Furthermore, in the above technical solution, the method also includes a model training and initialization step, which is performed during the system's first startup or periodic maintenance, specifically as follows:

[0083] Within typical environmental parameter variation ranges and plant growth cycles, the described regulation method is run but closed-loop feedback is not performed; only complete parameter data is collected.

[0084] Using the collected dataset, the parameter linkage prediction model is trained and cross-validated offline to obtain initial model parameters, which serve as the basis for step S204 and online incremental learning.

[0085] Furthermore, in the above technical solution, the overall control cycle of the method is configurable between 5 seconds and 5 minutes, and during the execution process, if any key sensor failure is detected, it automatically switches to a degraded control mode based on historical data and remaining valid parameters to maintain basic control functions.

[0086] Furthermore, the specific implementation of the degradation control mode is as follows: When a critical sensor (such as an ion mobility spectrometer) fails, the system first attempts to estimate the missing data based on redundant sensors (such as capacitive collectors) and a correlation model. If estimation is not possible, a historical data matching engine is activated, employing the K-nearest neighbor algorithm. Using currently valid environmental and physiological parameters as query conditions, the system searches for the 50 most similar operating condition records in the historical database, and takes the weighted average of their corresponding control commands as the execution basis for the current cycle. Simultaneously, the system records the operating data under degradation mode for subsequent fault analysis and model correction.

[0087] It should be further noted that initial deployment requires 1-2 weeks of model training and initialization. During this period, the system operates in "data acquisition mode," traversing typical environmental setpoint combinations and recording plant response data for offline training to obtain a reliable initial LSTM model. During operation, if any critical sensor (such as an ion mobility spectrometer or temperature and humidity sensor) is detected to be continuously malfunctioning or experiencing abnormal data for more than three cycles, the system automatically switches to degraded control mode. In this mode, the system utilizes optimal control commands under similar environmental and physiological conditions from a historical database, combined with readings from remaining effective sensors, for interpolation and inference to maintain basic negative oxygen ion generation while simultaneously issuing maintenance alarms to ensure system robustness. The overall control cycle can be flexibly configured between 5 seconds and 5 minutes according to network load and computing resources. This range is set based on the following considerations: the lower limit of 5 seconds ensures that the system can respond to rapid changes in sensor data (such as sudden changes in light intensity) and make pre-adjustments in a timely manner; the upper limit of 5 minutes matches the typical time constant of plant physiological responses (such as stomatal movement and transpiration regulation) and negative ion generation processes, avoiding frequent actuator actions and energy waste caused by excessive control frequency, while taking into account the real-time performance and stability of the system.

[0088] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent regulation of negative oxygen ions in small-particle plants based on multi-parameter linkage, characterized in that, Specifically, the following steps are included: S1: Real-time acquisition of a multi-dimensional parameter set, which includes a subset of environmental parameters, a subset of plant physiological parameters, and a subset of negative oxygen ion parameters; the environmental parameter subset includes temperature, humidity, light intensity, carbon dioxide concentration, and air velocity; S2: Input the multi-dimensional parameter set into the pre-trained multi-parameter linkage prediction model, and output the prediction results of the generation trend and target concentration of small-particle negative oxygen ions within a future set time period. The small-particle size refers to negative oxygen ions with an aerodynamic diameter of no more than 0.003 micrometers. S3: Based on the prediction results, a multi-parameter linkage control instruction set is generated through an optimized control algorithm. The optimized control algorithm aims to achieve the comprehensive optimization of small-particle-size negative oxygen ion concentration, proportion, system energy consumption, and output stability. S4: Execute the set of control instructions to synergistically intervene in the generation process of small-particle-size negative oxygen ions by dynamically adjusting the plant growth environment conditions, auxiliary negative oxygen ion generation conditions, and plant physiological regulation conditions. S5: Based on the subset of negative oxygen ion parameters collected after intervention, calculate the regulation effect index, and feed the index back to the multi-parameter linkage prediction model in step S2 and the optimization control algorithm in step S3 to achieve closed-loop adaptive optimization.

2. The intelligent regulation method for small-particle-size plant negative oxygen ions based on multi-parameter linkage according to claim 1, characterized in that: The collection of multi-dimensional parameter sets specifically includes: The subset of environmental parameters is collected synchronously through an environmental sensor grid deployed in the plant canopy space; The plant physiological parameters are a subset of those parameters, which are obtained by collecting leaf electrophysiological signals using a contact leaf sensor and measuring the plant transpiration rate using a heat balance method or a stem flow meter method. The particle size distribution of negative oxygen ions is measured by ion mobility spectrometry, and the concentration of negative oxygen ions is measured by capacitive ion collection method to determine the real-time concentration and proportion of small-particle negative oxygen ions, which constitute the negative oxygen ion parameter subset.

3. The intelligent regulation method for small-particle-size plant negative oxygen ions based on multi-parameter linkage according to claim 1, characterized in that: The construction and prediction process of the multi-parameter linkage prediction model includes: S201: Preprocess historical time-series data, which includes a subset of environmental parameters, a subset of plant physiological parameters, and a corresponding subset of negative oxygen ion parameters; S202: Based on the preprocessed data, the correlation analysis method is used to determine the dynamic weight relationship between each input parameter and the generation rate of small-particle negative oxygen ions, and to screen out the core control parameters. S203: Using the time-series data of the core regulation parameters as input and the concentration of small-particle negative oxygen ions as output, train a long short-term memory network model to learn the time-series dependencies between parameters. S204: Input the core control parameters collected in real time into the trained long short-term memory network model to predict the change curve of small-particle negative oxygen ion concentration within a future set time period.

4. The intelligent regulation method for small-particle-size plant negative oxygen ions based on multi-parameter linkage according to claim 1, characterized in that: The optimization control algorithm is a swarm intelligence-based optimization algorithm. It uses an iterative search and a multi-objective optimization function that comprehensively considers the concentration of small-particle negative oxygen ions, the proportion of small-particle size, system energy consumption, and concentration stability as the evaluation criteria to solve for a set of optimal controlled variable setpoints, thus forming the control instruction set.

5. The intelligent regulation method for small-particle-size plant negative oxygen ions based on multi-parameter linkage according to claim 1, characterized in that: The dynamic adjustment of plant growth environment conditions includes: according to the control instructions, using fuzzy proportional-integral-derivative control logic to quickly and accurately adjust the temperature and humidity of the cultivation space, and adjusting the intensity and spectrum of the artificial light source according to the preset light quality and light cycle.

6. The intelligent regulation method for small-particle-size plant negative oxygen ions based on multi-parameter linkage according to claim 1, characterized in that: The dynamic adjustment of the conditions for generating negative oxygen ions includes: starting, stopping, or adjusting the parameters of a high-voltage pulse electric field according to the control command; the pulse width, frequency, and field strength of the high-voltage pulse electric field are dynamically matched and adjusted according to the real-time ambient humidity, air velocity, and predicted plant release rhythm, so as to assist in the generation of small-particle negative oxygen ions and inhibit ozone production.

7. The intelligent regulation method for small-particle-size plant negative oxygen ions based on multi-parameter linkage according to claim 1, characterized in that: The dynamic regulation of plant physiological regulation conditions includes: adjusting the conductivity and composition of the nutrient solution supplied to the plant according to the regulation instructions and real-time plant transpiration rate data; and triggering the spraying system to apply plant-derived biostimulants when the plant is identified to regulate the plant's secondary metabolic activities when a specific physiological stage is detected.

8. The intelligent regulation method for small-particle-size plant negative oxygen ions based on multi-parameter linkage according to claim 1, characterized in that: The closed-loop adaptive optimization specifically includes: S501: After a control cycle is completed, obtain the actual concentration of small-particle-size negative oxygen ions and the total concentration; S502: Calculate a comprehensive regulatory effect index based on the ratio of actual concentration to target concentration and the actual proportion of small-particle negative oxygen ions. S503: Compare the regulation effect index with a preset threshold. If the index continues to be lower than the threshold or the environmental parameters fluctuate drastically, it is determined to be a model mismatch, and the online model update process is triggered. S504: In the online model update process, the data from the latest few periods are used as incremental datasets to perform incremental learning on the multi-parameter linkage prediction model and update its network weight parameters.

9. The intelligent regulation method for small-particle-size plant negative oxygen ions based on multi-parameter linkage according to claim 1, characterized in that: The method also includes a model training and initialization step, which is performed during the system's first startup or periodic maintenance, specifically as follows: Within typical environmental parameter variation ranges and plant growth cycles, the described regulation method is run but closed-loop feedback is not performed; only complete parameter data is collected. Using the collected dataset, the parameter linkage prediction model is trained and cross-validated offline to obtain initial model parameters, which serve as the basis for step S204 and online incremental learning.

10. The intelligent regulation method for small-particle-size plant negative oxygen ions based on multi-parameter linkage according to claim 1, characterized in that: The overall control cycle of the method is configurable between 5 seconds and 5 minutes. During execution, if any key sensor failure is detected, it automatically switches to a degraded control mode based on historical data and remaining valid parameters to maintain basic control functions.

Images