Self-adaptive greenhouse environment control system

Abstract

The invention relates to the technical field of agricultural engineering, in particular to a self-adaptive greenhouse environment control system which comprises a sensing layer, a decision-making layer, an execution layer and an optimization layer, and the sensing layer, the decision-making layer, the execution layer and the optimization layer achieve inter-layer data penetrating interaction and self-adaptive resource scheduling through a cross-layer dynamic cooperation mechanism. The sensing layer is used for collecting greenhouse environment parameters, crop physiological parameters and greenhouse physical structure parameters; the decision-making layer is used for carrying out collaborative decision-making through a physical model module and a self-adaptive MPC algorithm module based on the data collected by the sensing layer, and generating a control instruction; the execution layer is used for operating greenhouse environment regulation and control equipment according to the control instruction of the decision layer; and the optimization layer is used for dynamically adjusting the optimal combination of the equipment based on a preset regulation and control target. Through coupling of the cross-layer collaborative architecture, the physical model and the self-adaptive MPC algorithm, the greenhouse environment regulation and control precision is improved, the energy consumption is reduced, the per unit yield of crops is increased, and the labor and equipment cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of agricultural engineering technology, and in particular to an adaptive greenhouse environment control system based on physical models and model predictive control. Background Technology

[0002] Domestic control technology is still mainly based on traditional PID and simple distributed control, which is limited to single-factor regulation and cannot handle multi-variable coupling problems such as temperature, humidity and CO2 concentration.

[0003] Existing technologies suffer from four major bottlenecks: First, strong coupling of multiple variables leads to control failure. Parameters such as temperature, humidity, CO2 concentration, and light intensity are mutually constrained (e.g., ventilation and cooling simultaneously cause a sudden drop in humidity and CO2 loss). Traditional single-factor control logic often causes system oscillations, requiring frequent manual intervention for calibration. Second, parameters lack time-varying adaptability. Dynamic changes such as crop growth and equipment aging cannot be captured in real time, and fixed parameter models cause control accuracy to decline significantly with the growth stage. Third, energy consumption management is crude. There is a lack of meteorological forecasting and energy optimization mechanisms. Heating and supplemental lighting equipment are mostly started and stopped according to fixed thresholds, failing to fully utilize natural resources and energy dispatch strategies, resulting in a high proportion of energy costs. Fourth, crop physiological adaptability is insufficient. Key physiological indicators such as photosynthesis and transpiration are not linked, and environmental parameters are not dynamically adjusted according to crop growth stages, leading to impaired photosynthetic efficiency and reduced yield.

[0004] Invention Patent CN202210087874: A greenhouse environment energy-saving control method based on second-order Volterra model prediction. This method uses a data-driven model such as second-order Volterra series, which simplifies parameter identification but lacks physical interpretability, cannot cope with sudden changes in the greenhouse environment, and has insufficient model robustness. Invention Patent CN202410347074: A greenhouse environment multi-factor coordinated control algorithm based on fuzzy inference. This algorithm handles multi-variable coupling through fuzzy inference, but the control logic relies on empirical rules and does not incorporate a physical model or MPC algorithm. It cannot dynamically adapt to changes in environmental demands caused by crop growth and does not integrate meteorological forecasting capabilities, resulting in a lack of foresight in environmental regulation. Invention Patent CN202110048899: A blueberry greenhouse light and temperature coordination optimization method based on NSGA-II. This method uses the NSGA-II algorithm to optimize light and temperature coordination in blueberry greenhouses, focusing only on light and temperature variables. It does not address the coupling of multiple variables such as CO2 and humidity and does not incorporate a physical model. Therefore, the control logic is disconnected from the dynamic growth needs of the crops. Summary of the Invention

[0005] To address the shortcomings of existing technologies, such as multivariate coupling control failure, lack of time-varying parameter adaptability, crude energy consumption control, and insufficient crop physiological adaptability, as well as the lack of physical model support and multi-objective collaborative optimization mechanism, which prevent the achievement of precise, energy-saving, and dynamic control of the greenhouse environment that meets the needs of crop growth, this invention proposes an adaptive greenhouse environment control system. Through a cross-layer collaborative architecture, coupling of physical models and adaptive MPC algorithms, this system achieves improved greenhouse environment control precision, reduced energy consumption, increased crop yield, and reduced labor and equipment costs.

[0006] To achieve the above objectives, the technical solution adopted is:

[0007] This invention provides an adaptive greenhouse environment control system, comprising a sensing layer, a decision-making layer, an execution layer, and an optimization layer. The sensing layer, decision-making layer, execution layer, and optimization layer achieve inter-layer data penetration interaction and adaptive resource scheduling through a cross-layer dynamic collaboration mechanism.

[0008] The sensing layer is used to collect greenhouse environmental parameters, crop physiological parameters, and greenhouse physical structure parameters.

[0009] The decision layer is used to make collaborative decisions based on the data collected by the perception layer, through the physical model module and the adaptive MPC algorithm module, to generate control commands.

[0010] The execution layer is used to operate greenhouse environmental control equipment according to the control instructions of the decision-making layer;

[0011] The optimization layer is used to dynamically adjust the optimal combination of equipment based on preset control targets.

[0012] According to the adaptive greenhouse environment control system of the present invention, the sensing layer adopts a distributed heterogeneous sensor network to collect environmental information such as air temperature and humidity, light intensity, and carbon dioxide concentration, as well as crop physiological signals such as leaf area index, single leaf area, and transpiration, and basic physical parameters such as crop variety, greenhouse area, and specific heat capacity of the enclosure structure. The multi-source data is then aligned in time series and optimized by spatial interpolation through a spatiotemporal fusion algorithm.

[0013] According to the adaptive greenhouse environment control system of the present invention, the physical model module of the decision layer is further constructed based on the principles of energy balance and mass balance, and specifically includes:

[0014] The overall heat balance equation is expressed as: dQ total / dt=Q in -Q out +Q trans , where dQ total / dt represents the rate of change of the total heat inside the greenhouse over time, and the heat input term Q. in Including solar radiation heat Q solarHeat output of heating equipment (Q) heat Q, heat production from crop photosynthesis photo Heat generated during equipment operation Q metaβ Heat output term Q out Including ventilation and heat dissipation Q vent Building envelope heat dissipation Q loss Soil thermal conductivity Q transp And evaporative heat dissipation Q evap The heat conduction term Q trans Including air and equipment / crop conduction Q air-oβj and air-to-soil conduction Q air-soil ;

[0015] The water vapor mass balance equation is expressed as: dW total / dt=G in -G out Among them, water vapor input item G in Including crop transpiration and soil evaporation G evap and humidification equipment input G humid Water vapor output item G out Including the amount of water vapor removed by ventilation (G) vent and the amount of water vapor removed by the dehumidification equipment G dehumid ;

[0016] The carbon dioxide mass balance equation is expressed as: dC total / dt=C in -C out -C consume Among them, carbon dioxide input item C in Including carbon dioxide fertilizer input C fert Carbon dioxide output C out Including the amount of carbon dioxide C removed by ventilation vent Carbon dioxide consumption item C consume C is the amount consumed by crop photosynthesis. photo .

[0017] According to the adaptive greenhouse environment control system of the present invention, the adaptive MPC algorithm module of the decision layer further constructs a discrete state equation for predicting the environmental state at future times based on the total heat balance equation, the water vapor mass balance equation, and the carbon dioxide mass balance equation, including:

[0018] The equation of state for predicting the temperature is constructed based on the overall heat balance equation and takes the following form:

[0019]

[0020] Where T(k+1) is the predicted indoor temperature at time k+1, T(k) is the indoor temperature at time k, and Q in(k), Q out (k), Q trans (k) represents the total heat input, total heat output, and heat conduction at time k, m 空 For air quality, c 空 Here, Δt represents the specific heat capacity of air, and Δt represents the time period.

[0021] The state equation for predicting humidity is constructed based on the water vapor mass balance equation, and takes the following form:

[0022]

[0023] Where d(k+1) is the predicted indoor absolute humidity at time k+1, d(k) is the indoor absolute humidity at time k, and G in (k), G out (k) represents the total water vapor input and total water vapor output at time k, respectively, and V0 is the greenhouse volume;

[0024] The equation of state for predicting carbon dioxide concentration is constructed based on the carbon dioxide mass balance equation and takes the following form:

[0025]

[0026] Where c(k+1) is the predicted indoor CO2 concentration at time k+1, c(k) is the indoor CO2 concentration at time k, and C in (k), C out (k), C 消耗 (k) represents the total CO2 input, total CO2 output, and total CO2 consumed by crop photosynthesis at time k.

[0027] According to the adaptive greenhouse environment control system of the present invention, the adaptive MPC algorithm module further includes a prediction layer, an optimization layer, and a correction layer. The prediction layer outputs future temperature, absolute humidity, and CO2 concentration sequences based on discrete state equations T(k+1), d(k+1), and c(k+1), respectively. The optimization layer uses crop physiological needs as constraints to continuously solve for the sequence of operations that minimize the objective function. The correction layer receives the actual state deviations fed back by the sensing layer in real time and dynamically adjusts the physical model parameters and optimization target weights.

[0028] According to the adaptive greenhouse environment control system of the present invention, the optimization objective of the adaptive MPC algorithm module is further defined as "temperature tracking accuracy + energy consumption optimization" as dual objectives, and the objective function is:

[0029]

[0030] Where J is the loss function, ω T For temperature tracking weights, T 目标Let T(k+i|k) be the target temperature value required for crop growth in the greenhouse environment, and let T(k+i|k) represent the greenhouse indoor temperature value predicted at time k+i. N is the prediction time domain length, and ω is the value of the target temperature value. j Let μ be the energy consumption weight for the j-th type of device. j (k+i-1) represents the operation quantity of the j-th type of equipment at time k+i-1, and M represents the number of equipment types involved in greenhouse environment control.

[0031] According to the adaptive greenhouse environment control system of the present invention, the optimization objective of the adaptive MPC algorithm module is a dual objective of "humidity tracking accuracy + CO2 tracking accuracy", and the objective function is:

[0032] J=Σω d ·(d 目标 -d(k+i|k)) 2 +Σω c ·(c 目标 -c(k+i|k))2+ΣΣω j ·u j (k+i-1) 2

[0033] Where, ω d Indicates the humidity tracking weight, d 目标 d(k+i|k) represents the target humidity value required for crop growth in a greenhouse environment, d(k+i|k) represents the indoor humidity value predicted at time k+i, and ω represents the humidity value at time k+i. c c represents the CO2 concentration tracking weight. 目标 c(k+i|k) represents the target CO2 concentration required for crop photosynthesis, c(k+i|k) represents the indoor CO2 concentration predicted at time k+i, and ω represents the target CO2 concentration required for crop photosynthesis. j Let u be the energy consumption weight for the j-th type of equipment. j (k+i-1) represents the operation quantity of the j-th type of device at time k+i-1.

[0034] According to the adaptive greenhouse environment control system of the present invention, the execution layer further includes multiple device clusters, which are functionally divided into temperature regulation clusters, humidity regulation clusters, light regulation clusters and carbon dioxide regulation clusters.

[0035] According to the adaptive greenhouse environment control system of the present invention, the optimization layer further constructs a three-dimensional evaluation model based on environmental deviation, energy consumption cost and crop sensitivity, and solves the optimal equipment combination through particle swarm optimization algorithm.

[0036] According to the adaptive greenhouse environment control system of the present invention, the three-dimensional evaluation model is further expressed as: Priority=α·ΔEnv+β·Cost+γ·Sens, where ΔEnv is the environmental deviation degree, Cost is the energy consumption cost, Sens is the crop sensitivity, and α, β, and γ are weighting coefficients.

[0037] The beneficial effects achieved by adopting the above technical solution are:

[0038] 1. Cross-layer collaborative architecture innovation improves data processing efficiency and system adaptability.

[0039] By innovating "data penetration and interaction between the decision-making layer and the perception layer" (such as the MPC algorithm dynamically adjusting the sensor sampling frequency) and "dynamic load distribution between the edge and the cloud" (short-term prediction relies on the edge, and long-term growth models rely on the cloud), the limitations of the traditional unidirectional data flow in the Internet of Things are broken through, solving the problems of full data waste and the contradiction between computing power and latency. The data preprocessing accuracy is improved by 40%, while being compatible with different types of greenhouses and multiple crop scenarios. The installation and commissioning cycle is shortened by 50%, and the industrial adaptability is significantly enhanced.

[0040] 2. Closed-loop coupling of physical model and adaptive MPC algorithm enhances control stability and accuracy.

[0041] The physical model module constructs an energy / mass balance physical model, quantifies the multivariate coupling mechanism of temperature, humidity, and CO2, and combines it with the adaptive MPC algorithm's three-level architecture of "prediction-optimization-correction" to correct time-varying parameters such as leaf area index and transpiration coefficient in real time. Compared with existing technologies such as data-driven and fuzzy inference, the control stability is improved by 60% when the environment changes abruptly, and the control deviations of temperature, humidity, and CO2 concentration are reduced to ±1.2℃, ±5%, and ±50ppm, respectively, accurately matching the dynamic environmental requirements of crop growth.

[0042] 3. Dynamic priority scheduling driven by crop physiological signals optimizes equipment response and synergistic effects.

[0043] Abandoning the traditional scheduling logic of "fixed threshold" and "empirical rules", this method uses physiological parameters such as crop leaf water potential (e.g., irrigation priority is triggered when tomato fruiting period < -0.8MPa) and stomatal conductance as core criteria. It integrates environmental deviation and energy consumption cost to construct a three-dimensional objective function, which solves the defect of "only looking at environmental parameters and ignoring crop needs". This reduces the response time to actual crop needs from 20 minutes to 8 minutes, and the deviation of temperature and humidity coordinated control from ±15% to ±9.8%.

[0044] 4. Multi-variable collaborative optimization logic improves photosynthetic efficiency and energy saving effect.

[0045] By establishing a priority regulation and variable linkage mechanism for "temperature → humidity → light → CO2", we can overcome the limitations of single-factor optimization and achieve coordinated control such as "evaporative cooling + spray humidification" under high temperature and low humidity and "precise CO2 supplementation" when light intensity meets the standard. This results in a 35% increase in photosynthetic efficiency, a 15%-20% increase in crop yield, and a 40% reduction in disease incidence. At the same time, through energy optimization scheduling and multi-equipment collaboration to reduce redundancy, energy consumption is reduced by 25%-35%, and the frequency of manual intervention and equipment costs are significantly reduced. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. The drawings are merely illustrative of some embodiments of the present invention and are not intended to limit the scope of the present invention to all embodiments.

[0047] Figure 1 This is a structural diagram of the adaptive greenhouse environment control system according to an embodiment of the present invention. Detailed Implementation

[0048] The exemplary solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art.

[0049] like Figure 1 As shown in the figure, this embodiment discloses an adaptive greenhouse environment control system including a sensing layer, a decision-making layer, an execution layer, and an optimization layer. The sensing layer, decision-making layer, execution layer, and optimization layer achieve inter-layer data penetration interaction and adaptive resource scheduling through a cross-layer dynamic collaboration mechanism. The structure of each layer is explained in detail below.

[0050] (1) Sensing layer: used to collect greenhouse environmental parameters, crop physiological parameters and greenhouse physical structure parameters.

[0051] The perception layer adopts a distributed heterogeneous sensor network and innovatively constructs a three-in-one dual-mode perception system of "environment-crop-physical parameters". It collects environmental information such as air temperature and humidity, light intensity, and carbon dioxide concentration; crop physiological signals such as leaf area index, single leaf area, and transpiration; and basic physical parameters such as crop variety, plant growth, greenhouse area, and specific heat capacity of the enclosure structure. It also completes the time series alignment and spatial interpolation optimization of multi-source data through spatiotemporal fusion algorithm, which solves the problem of the one-sidedness of traditional single sensor data.

[0052] The spatiotemporal fusion algorithm of the sensing layer includes timestamp alignment and spatial interpolation processing. The timestamp alignment error is less than 50ms. The spatial interpolation uses the Kriging interpolation algorithm to correct the uneven distribution of sensors, reducing the temperature field distribution error after calibration from ±2.5℃ to ±1.5℃.

[0053] (2) Decision layer: Based on the data collected by the perception layer, the physical model module and the adaptive MPC algorithm module make collaborative decisions to generate control commands.

[0054] The decision-making level deploys edge computing nodes and adopts a "dual-core collaborative" edge computing architecture, which includes a lightweight physical model module and an adaptive MPC algorithm module. The two are tightly coupled and interact through a high-speed internal bus.

[0055] The physical model module of the decision-making layer is built based on the principles of energy balance and mass balance, and specifically includes:

[0056] ① Overall heat balance equation

[0057] The change in total heat inside the greenhouse follows the law of conservation of energy, and its dynamic equilibrium equation is:

[0058] dQ total / dt=Q in -Q out +Q trans

[0059] Among them, dQ total / dt represents the rate of change of the total heat inside the greenhouse over time, Q in Q represents the total heat input into the greenhouse per unit time. out Q represents the total heat output of the greenhouse per unit time. trans This represents the amount of heat transfer between different areas inside the greenhouse per unit time; the rate of change of greenhouse air temperature can be derived from the overall heat balance equation.

[0060] dT / dt=(dQ total / dt) / (m air ×c air )

[0061] Where, m air c represents the total mass of air inside the greenhouse. air Let T be the specific heat capacity of air and T be the greenhouse air temperature. To accurately calculate temperature changes, it is necessary to further quantify the specific components of heat input, output, and conduction. The following details each component.

[0062] 1. Caloric Input (Q) in )

[0063] Q in =Q solar +Q heat +Q photo +Q metaβ

[0064] Solar radiation heat Q solar Q solar=I×A×τ×α, where I is the solar radiation intensity, A is the greenhouse light-receiving area, τ is the light transmittance of the covering material, and α is the indoor surface absorptivity.

[0065] Heat of heating equipment Q heat Q heat =Σ(P i ×η i In the formula, P i η is the rated power of heating equipment of class i (such as electric heaters, hot air blowers). i This refers to the heat conversion efficiency of the equipment.

[0066] Crop photosynthesis heat production Q photo Q photo =LAI×r p ×ΔH, where LAI is the leaf area index, r p ΔH represents the photosynthetic heat production rate per unit leaf area, and ΔH represents the enthalpy change of photosynthesis.

[0067] Heat generated during equipment operation Q metaβ Q metaβ =Σ(P j ×(1-η j In the formula, P j For the power of the j-th type of power-consuming equipment (such as fans and water pumps), η j The electromechanical conversion efficiency of the equipment (the conversion of non-heat output parts into heat).

[0068] 2. Heat output term (Q) out )

[0069] Q out =Q vent +Q loss +Q transp +Q evap

[0070] Ventilation and heat dissipation Q vent Q vent =ρ air ×c air ×V×(T in -T out ), where ρ air V is the air density, T is the ventilation volumetric flow rate, and V is the air density. in Indoor temperature, T out This refers to the outdoor temperature.

[0071] Building envelope heat dissipation Q loss Q loss =Σ(k i ×A i ×(T in -T out In the formula, ki Let A be the heat transfer coefficient of the i-th type of building envelope (walls, roof, floor). i This represents the area of ​​the corresponding structure.

[0072] Soil thermal conductivity Q transp Q transp =k soil ×A soil ×(T in -T soil In the formula, k soil Let A be the soil heat transfer coefficient. soil T represents the greenhouse floor area. soil This refers to soil temperature.

[0073] Evaporative cooling Q evap Q evap =L×E, where L is the latent heat of vaporization of water and E is the total evaporation in the greenhouse (including crop transpiration and surface evaporation).

[0074] 3. Heat conduction term (Q) trans )

[0075] Q trans =Q air-oβj +Q air-soil

[0076] Air and equipment / crop conduction Q air-oβj Q air-oβj =h×A oβj ×(T in -T oβj In the formula, h is the convective heat transfer coefficient, and A oβ T represents the total surface area of ​​the equipment / crop. oβ This refers to the surface temperature of the equipment / crop.

[0077] Air-to-soil conduction Q air-soil Q air-soil =h s ×A soil ×(T in -T soil In the formula, h s The heat transfer coefficient at the air-soil interface.

[0078] It should be noted that the above is the basic equation. Depending on the specific conditions of each greenhouse, such as the cooling and heating equipment within the greenhouse, terms may be added or removed from the equation. If heating equipment is present, add a heating term: Q 加热 =P 加 ·t·η 加 The presence of shading devices increases the impact of solar radiation: Q 太阳 =I·A 顶 ·τ 遮 ·α吸 .

[0079] The above equations can quantify the impact of each device's operation on the total heat of the greenhouse, providing a physical model basis for subsequent MPC optimization of temperature control.

[0080] ② Water vapor mass balance equation

[0081] The water vapor mass balance equation is expressed as follows:

[0082] dW total / dt=G in -G out

[0083] Among them, water vapor input item G in Including crop transpiration and soil evaporation G evap and humidification equipment input G humid Water vapor output item G out Including the amount of water vapor removed by ventilation (G) vent and the amount of water vapor removed by the dehumidification equipment G dehumid .

[0084] 1. Water vapor input item (G) in )

[0085] Crop transpiration and soil evaporation G evap :G evap (k)=k 植 ·LAI·(d 饱 -d(k))·Δt+k 土 ·A 土 ·(θ 土 -θc)·Δt, where kplant is the plant transpiration coefficient, LAI is the leaf area index, and dsaturated is the saturated humidity at the current temperature (kg / m2). 3 ), k 土 A is the soil evaporation coefficient. 土 For soil area, θ 土 For soil moisture, θ 临 This is the critical humidity level (below this value, evaporation will not occur).

[0086] Humidifier input G humid (Water vapor generated by spray / wet curtain): G humid (k)=q 雾 (γ)·ρ 水 ·η 雾 ·Δt, where q 雾 (γ) represents the spray flow rate (m³ / s). 3 / h, γ is the valve opening), ρ 水 The density of water (1000 kg / m³) 3 ), η 雾For atomization efficiency.

[0087] 2. Water vapor output (G) out )

[0088] The amount of water vapor removed by ventilation, G vent :G vent (k)=V 风 (β)·(d(k)-d 外 (k))·Δt / 3600, where V 风 (β) represents the fan air volume (m³ / s). 3 / h), d 外 (k) represents the absolute outdoor humidity.

[0089] The amount of water vapor removed by the dehumidifier G dehumid :G dehumid (k)=P 除 ·ε·Δt, where P 除 ε represents the dehumidifier power (kW) and ε represents the dehumidification efficiency (kg / (kW·h)).

[0090] ③ Carbon dioxide mass balance equation

[0091] The carbon dioxide mass balance equation is expressed as: dC total / dt=C in -C out -C consume

[0092] Among them, carbon dioxide input item C in Including carbon dioxide fertilizer input C fert Carbon dioxide output C out Including the amount of carbon dioxide C removed by ventilation vent Carbon dioxide consumption item C consume C is the amount consumed by crop photosynthesis. photo .

[0093] 1. Carbon dioxide input (C in )

[0094] Carbon dioxide fertilizer input C fert : In the formula, CO2 flow rate (m 3 / h, where δ is the valve opening degree), CO2 density (1.977 kg / m³) 3 ).

[0095] 2. Carbon dioxide output (C out )

[0096] The amount of carbon dioxide C removed by ventilation vent :Cvent (k)=V 风 (β)·(c(k)-c 外 (k))·Δt / 3600, where cout (k) is the outdoor CO2 concentration (approximately 0.6 kg / m³). 3 ).

[0097] 3. Carbon dioxide consumption (C consume )

[0098] Crop photosynthetic consumption C photo :C photo (k)=k 光 ·I(k)·LAI·(c(k)-c 补偿 )·Δt, where k 光 I(k) is the photosynthetic coefficient, I(k) is the light intensity, and c is the light intensity. 补偿 This is the concentration at the compensation point (no consumption occurs below this value).

[0099] The adaptive MPC algorithm module of the decision layer constructs discrete state equations for predicting the environmental state at future moments, based on the total heat balance equation, water vapor mass balance equation, and carbon dioxide mass balance equation. These equations include:

[0100] ① The equation of state for predicting temperature is constructed based on the overall heat balance equation:

[0101]

[0102] Where T(k+1) is the predicted indoor temperature at time k+1, T(k) is the indoor temperature at time k, and Q in (k), Q out (k), Q trasi (k) represents the total heat input, total heat output, and heat conduction at time k, m 空 For air quality, c 空 Δt represents the specific heat capacity of air, and Δt represents the time period (e.g., 300s).

[0103] ②Predicted humidity (absolute humidity, unit kg / m³) 3 The state equation is constructed based on the water vapor mass balance equation:

[0104]

[0105] Where d(k+1) is the predicted indoor absolute humidity at time k+1, d(k) is the indoor absolute humidity at time k, and G in (k), G out (k) represents the total water vapor input and output (kg) at time k, respectively, and V0 represents the greenhouse volume (m³). 3 ).

[0106] ③ The equation of state for predicting carbon dioxide concentration is constructed based on the carbon dioxide mass balance equation:

[0107]

[0108] Where c(k+1) is the predicted indoor CO2 concentration at time k+1, c(k) is the indoor CO2 concentration at time k, and C in (k), C out (k), C 消耗 (k) represents the total CO2 input, total CO2 output, and total CO2 consumed by crop photosynthesis at time k.

[0109] The adaptive MPC algorithm module comprises a prediction layer, an optimization layer, and a correction layer. The prediction layer outputs future temperature, absolute humidity, and CO2 concentration sequences based on the discrete state equations T(k+1), d(k+1), and c(k+1), respectively. The optimization layer uses crop physiological needs as constraints to continuously solve for the sequence of operations that minimizes the objective function. The correction layer receives real-time feedback from the sensing layer regarding actual state deviations and dynamically adjusts the physical model parameters and optimization target weights to achieve online compensation for model errors. The entire rolling optimization control process executed by the adaptive MPC algorithm module includes the following steps:

[0110] (a) Prediction layer performs prediction: at time k in each control cycle, based on the discrete state equations from the physical model module and the current system state measurements, predicts the system state sequence for the next N time moments, where the system state includes at least one of temperature, humidity and carbon dioxide concentration.

[0111] (b) Rolling optimization in the optimization layer: A finite-time open-loop optimization problem is solved with crop physiological requirements as constraints. The objective function J of the optimization problem simultaneously considers system state tracking error and equipment operation energy consumption, and takes the following form:

[0112]

[0113] Among them, Y 目标 Y(k+i|k) represents the target state required for crop growth in a greenhouse environment, and is the predicted state at time k+i based on the state at time k. j For the operation quantity of the j-th type of equipment, ω i · and ω j The weighting coefficients are dynamically adjusted. By solving this objective function, a set of optimal equipment operation quantities [u*(k),u*(k+1),...,u*(k+i-1)] in the future time domain is obtained.

[0114] The system achieves collaborative optimization through a variable linkage mechanism:

[0115] Temperature and humidity are linked: when the temperature is high, ventilation and cooling will reduce humidity, so it is necessary to spray moisture to replenish humidity at the same time (for example, when tomatoes are hot and low in humidity, wet curtains are used to cool them down and spray moisture to increase humidity, which controls the temperature and prevents blossom-end rot); when the temperature is low, sealing and insulation will increase humidity, so it is necessary to dehumidify (for example, when cucumbers are cold and high in humidity, heating is used to raise the temperature and dehumidifiers are used to reduce humidity, which reduces downy mildew).

[0116] Light and carbon dioxide are linked: light intensity is "energy," and CO2 is "raw material," and the two must be matched: when light intensity is less than the light compensation point, supplementing CO2 is ineffective (photosynthesis < respiration); when light intensity reaches the light saturation point, insufficient CO2 will become a bottleneck (e.g., for tomatoes with a light intensity of 500 μmol / m²). 2 When CO2 increases from 400ppm to 800ppm per second, the photosynthetic rate can be increased by 40%.

[0117] Light and temperature are linked: strong sunlight is accompanied by increased temperature, requiring simultaneous shading and cooling (e.g., at midday in summer, shade nets should be used to reduce light intensity to 500 μmol / m²). 2 At the same time, the wet curtain controls the temperature below 28℃ to avoid high temperature exacerbating light suppression.

[0118] (c) Execution layer implements operation: The first element in the optimal equipment operation sequence, i.e. the instantaneous operation u*(k) (e.g., α*=0.4 (shading curtain open 40%), β*=3 (fan at 3rd speed)), is sent to the corresponding greenhouse environment control equipment for execution.

[0119] (d) Closed-loop feedback of the correction layer: At the next sampling time k+1, a new system state measurement value Y(k+1) is obtained and compared with the predicted value Y(k+1|k) of the previous period; if the deviation exceeds the preset threshold, the adaptive weights and constraints are dynamically corrected, and the objective function weights are dynamically adjusted: if the predicted temperature deviates from the target T by more than 2℃, the temperature tracking weight ω is increased. T (e.g., from 0.7 to 0.8); if energy consumption exceeds the historical average by 15%, increase the equipment energy consumption weight ω. j (like a fan) j Increase by 10%), correct equipment constraints: adjust the range of operating quantities according to the real-time status of the equipment (such as the upper limit of the flow rate of the wet curtain pump) (such as reducing the maximum opening of γ from 1 to 0.8); and use the corrected model for the next round of prediction in step (a).

[0120] Repeat steps (a) to (d) to achieve closed-loop feedback rolling optimization control based on model prediction.

[0121] The optimization objective of the adaptive MPC algorithm module is twofold: "temperature tracking accuracy + energy consumption optimization". The objective function is as follows:

[0122]

[0123] Where J is the loss function, ω T For temperature tracking weights, T 目标 Let T(k+i|k) be the target temperature value required for crop growth in the greenhouse environment, and let T(k+i|k) represent the greenhouse indoor temperature value predicted at time k+i. N is the prediction time domain length, and ω is the value of the target temperature value. j Let μ be the energy consumption weight for the j-th type of device. j (k+i-1) represents the operation quantity of the j-th type of equipment at time k+i-1, and M represents the number of equipment types involved in greenhouse environment control.

[0124] Constraints: Equipment operation constraints: 0≤α≤1 (shade curtain opening), β∈{1,2,3,4,5} (fan speed), 0≤γ≤1 (evaporative cooling pad valve), 0≤δ≤1 (heating power); Temperature safety constraints: T min ≤T(k+i|k)≤T max (e.g., the ideal growing temperature for tomatoes is 10-35℃).

[0125] When the optimization objective is "humidity tracking accuracy + CO2 tracking accuracy", the objective function is adjusted to:

[0126] J=Σω d ·(d 目标 -d(k+i|k)) 2 +Σω c ·(c 目标 -c(k+i|k))2+ΣΣω j ·u j (k+i-1) 2

[0127] Where, ω d Indicates the humidity tracking weight, d 目标 d(k+i|k) represents the target humidity value required for crop growth in a greenhouse environment, d(k+i|k) represents the indoor humidity value predicted at time k+i, and ω represents the humidity value at time k+i. c c represents the CO2 concentration tracking weight. 目标 c(k+i|k) represents the target CO2 concentration required for crop photosynthesis, c(k+i|k) represents the indoor CO2 concentration predicted at time k+i, and ω represents the target CO2 concentration required for crop photosynthesis. j Let u be the energy consumption weight for the j-th type of equipment. j (k+i-1) represents the operation quantity of the j-th type of device at time k+i-1.

[0128] Constraints: Humidity: d min ≤d(k+i|k)≤d max (e.g., tomato growth humidity 50%-80%); CO2:c min ≤c(k+i|k)≤c max(e.g., 800-1500 kg / m) 3 The equipment operation is the same as before (constraints on fans, valves, etc. remain unchanged).

[0129] Execution layer: Used to operate greenhouse environment control equipment according to the control instructions of the decision-making layer.

[0130] The execution layer includes multiple equipment clusters, which are divided into temperature control clusters (heating units, water curtains, variable frequency fans), humidity control clusters (humidifiers, dehumidifiers, ventilation valves), light control clusters (LED supplemental lighting, shading nets) and CO2 control clusters (gas cylinder release devices, circulating fans).

[0131] Optimization layer

[0132] The traditional fixed order of environmental variable control priorities, namely "temperature → humidity → light → carbon dioxide", is only applicable to normal scenarios where crops are not under stress. However, when crops encounter survival-level stress (such as severe water shortage), physiological signals trigger a "priority override mechanism", which strengthens the logic of "ensuring crop survival first, then optimizing environmental parameters".

[0133] Regulation should follow the priority order of "temperature → humidity → light → carbon dioxide," ensuring the basic conditions for crop metabolism first, and then optimizing photosynthetic efficiency.

[0134] 1. Temperature Priority: Locking in the Metabolic "Baseline"

[0135] Temperature directly determines enzyme activity (such as photosynthetic enzymes and respiratory enzymes) and cell function, and is a prerequisite for all physiological activities. Temperature must first be controlled within the optimal range for the crop (e.g., daytime temperature 25-28℃ and nighttime temperature 16-20℃ during the tomato fruiting period). If the temperature is below the lower limit (e.g., tomato <10℃), heating equipment (electric heating, fuel-fired hot air blower) should be activated, along with heat-insulating curtains to reduce heat loss. If the temperature is above the upper limit (e.g., tomato >30℃), priority should be given to using wet curtains + fans for cooling (evaporative heat absorption), and in extreme high temperatures, shade nets should be added (to reduce radiant heat).

[0136] 2. Humidity secondary: Balancing transpiration and disease risk

[0137] Once the temperature is suitable, adjust the humidity to the safe range for crops (e.g., 50%-70% for tomatoes, 60%-80% for cucumbers), avoiding two extremes: high humidity (e.g., >85% for tomatoes): turn on the dehumidifier or force ventilation (3-5 air changes per hour) to reduce the risk of gray mold and downy mildew; low humidity (e.g., <50% for tomatoes): increase humidity through a spray system (droplet diameter 50-100μm, avoid condensation on leaves) to ensure the transport of mineral elements such as calcium and boron (dependent on transpiration pull).

[0138] 3. Light adaptation: Matching the "energy requirements" of photosynthesis

[0139] Light is the energy source for photosynthesis and needs to be controlled "above the light compensation point and below the light inhibition threshold": insufficient light intensity (e.g., tomato <200 μmol / m²) 2 / s): Turn on the LED fill light (red to blue light 6:4) and fill to the light saturation point (300-500μmol / m²). 2 Avoid excessive light intensity (e.g., tomato > 800 μmol / m²) to prevent excessive growth; avoid excessive light intensity (e.g., tomato > 800 μmol / m²). 2 / s): Use shade nets (50%-70% light transmittance) to reduce light and prevent chloroplast damage (photoinhibition) and fruit sunburn.

[0140] 4. Carbon dioxide replenishment: Enhancing the efficiency of photosynthetic raw materials.

[0141] As a raw material for the dark reaction of photosynthesis, it should only be supplemented when other conditions are suitable (temperature and light intensity meet requirements) to avoid waste: if the concentration is insufficient (e.g., tomatoes <600ppm): increase it to 800-1000ppm using a CO2 generator (during the fruiting period), and only during the day when the light intensity is >200μmol / m². 2 Start when the concentration is too high (e.g., >1500ppm): Stop replenishment and ventilate appropriately to avoid stomata closing and hindering evaporation.

[0142] This scheme's dynamic prioritization mechanism breaks through the traditional "fixed threshold scheduling" or "empirical rule reasoning." Its core innovation lies in real-time priority generation and multi-objective optimization based on crop physiological signals, specifically including:

[0143] Priority-driven physiological signal mechanism: Abandoning the conventional fixed threshold logic of "activating water curtain when temperature > 30℃", crop physiological parameters (such as leaf water potential and stomatal conductance) are used as the first priority criterion. For example, when the leaf water potential of tomatoes during the fruiting period is < -0.8MPa (indicating water shortage stress), even if the temperature does not reach the threshold, the priority of irrigation equipment is automatically increased to the highest level (surpassing ventilation equipment), which solves the shortcomings of the traditional "only looking at environmental parameters and ignoring the actual needs of crops". Experiments show that when irrigation equipment has the highest priority, the system balances short-term environmental fluctuations through the following logic: if the actual indoor temperature (31℃) is slightly higher than the upper limit of the "suitable temperature range" for crop growth (28℃), irrigation is started first (10-15 minutes) to alleviate water shortage, during which ventilation is suspended (to avoid aggravating transpiration). After irrigation, the leaf water potential recovers to >-0.6MPa (escaping stress), and immediately and automatically returns to the normal priority of "temperature > humidity...". Temperature deviation is quickly corrected through ventilation and other equipment. Experimental data show that under this mechanism, the total time for crops to recover from water shortage stress is shortened by 60% (from the normal 20 minutes to 8 minutes), and short-term temperature fluctuations (within 15 minutes) do not have a significant impact on yield.

[0144] A multi-objective optimization model for conflict resolution: The optimization layer constructs a three-dimensional evaluation model based on environmental deviation, energy consumption cost, and crop sensitivity, and solves for the optimal equipment combination using the particle swarm optimization algorithm; The expression of the three-dimensional evaluation model is: Priority=α·ΔEnv+β·Cost+γ·Sens, where ΔEnv is the environmental deviation, Cost is the energy consumption cost, Sens is the crop sensitivity, and α, β, and γ are weight coefficients.

[0145] When environmental parameters deviate significantly from the target: for example, a sudden temperature spike exceeding 35°C, far exceeding the crop's tolerance range. In this case, it is imperative to quickly restore a safe environment at all costs, significantly increasing α. During periods of high energy prices: for example, when implementing peak-valley electricity pricing, β should be increased during peak electricity price periods to encourage the system to use passive control methods such as natural ventilation and shading as much as possible, avoiding the activation of high-power heating / cooling equipment. When physiological stress is detected in the crop: this is the most critical scenario; for example, if sensors detect excessively low leaf water potential (drought stress) or excessively high canopy temperature (heat stress), γ should be increased immediately. In this case, the system will place the crop's "sensitivity" as the absolute priority. In this scheme, crop sensitivity reflects its first priority through setting rules and triggering emergency responses; temperature is the most important daily control dimension under these rules; and energy is the resource that needs to be optimized as much as possible after the former two are guaranteed.

[0146] In normal circumstances (without sudden constraints): the optimization layer only fine-tunes the operational amount of the MPC output (e.g., reducing the CO2 generator power from 30% to 25%, while keeping the equipment unchanged); in extreme constraint scenarios: whether it is peak-valley electricity pricing (economic constraints) or crop stress (physiological constraints), the optimization layer's adjustments always revolve around "unchanged core objectives and dynamic path adaptation": under peak-valley electricity pricing, the objective is still "stable CO2 concentration", but it is adjusted from "continuous generator operation" to "intermittent operation + enhanced mixing by fans" (the operational amount and equipment operation mode change, while the equipment type may remain the same); under crop water shortage stress: the objective switches from "temperature priority" to "irrigation priority", and ventilation equipment may be suspended (to reduce transpiration), while only irrigation equipment is kept running at full load (the equipment combination is temporarily simplified, and the operational amount is concentrated).

[0147] The adaptive MPC algorithm module of the decision layer of this scheme outputs an "unconstrained theoretical solution" (an ideal match between equipment combination and operation quantity); the optimization layer outputs a "constrained feasible solution" (which may adjust the operation quantity or change the equipment combination, but always serves the global optimization goal of "environment-energy consumption-crop"). This closed loop of theoretical design and actual adaptation is the core of the system's robustness - ensuring the optimization direction and coping with complex real-world scenarios.

[0148] The present invention has the following advantages: 1. Significantly reduced energy consumption: Compared with the traditional PID control in China, this system reduces energy consumption by 25%-35% (the annual power consumption of a 3000㎡ tomato greenhouse is reduced from 120,000 kWh to 80,000 kWh) through "weather forecasting + energy optimization scheduling" (such as prioritizing natural ventilation over heating) and "multi-equipment collaborative redundancy reduction" (such as water curtain + fan combination replacing individual dehumidification at high temperatures). 2. Improved Control Precision and Crop Adaptability: Environmental parameter control precision: Temperature deviation ±1.2℃ (traditional PID ±2.5℃), humidity deviation ±5% (traditional ±10%), CO2 concentration deviation ±50ppm (traditional ±100ppm); Crop adaptability: Corresponding to photosynthetic / transpiration physiological indicators, parameters are dynamically adjusted according to growth stages (e.g., 70%-80% humidity during seedling stage, 50%-70% during fruiting stage), resulting in a 15%-20% increase in yield (tomato yield per mu increased from 8000kg to 9200-9600kg), and a 40% reduction in disease incidence (gray mold incidence decreased from 15% to 9%). 3. Reduced Labor and Equipment Costs: Frequency of manual intervention: Reduced from 3-5 calibrations per day to 1-2 times per month, saving 60% in annual labor costs; Equipment costs: Hardware costs (distributed sensors, edge nodes) of this system are reduced by 40%-50% (the cost of a 5000㎡ greenhouse system is reduced from 500,000 yuan to less than 300,000 yuan). 4. Strong industrial adaptability: It is compatible with different scenarios such as solar greenhouses and intelligent glass greenhouses, and supports the regulation of multiple crops such as tomatoes, cucumbers, and blueberries. It does not require large-scale modification of the existing greenhouse structure, and the installation and commissioning cycle is shortened by 50% (from 20 days to 10 days for a 3000㎡ greenhouse).

[0149] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An adaptive greenhouse environment control system, characterized in that, The system comprises a perception layer, a decision layer, an execution layer and an optimization layer, and the perception layer, the decision layer, the execution layer and the optimization layer realize the interlayer data penetrating interaction and adaptive resource scheduling through a cross-layer dynamic coordination mechanism. The perception layer is used for collecting greenhouse environment parameters, crop physiological parameters and greenhouse physical structure parameters. The decision layer is used for making a collaborative decision through a physical model module and an adaptive MPC algorithm module based on the data collected by the perception layer, and generating a control instruction. The execution layer is used for operating a greenhouse environment regulation device according to the control instruction of the decision layer. The optimization layer is used for dynamically adjusting an optimal combination of devices based on a preset regulation target.

2. The adaptive greenhouse environment control system of claim 1, wherein, The perception layer adopts a distributed heterogeneous sensing network and is used for collecting air temperature and humidity, light intensity, carbon dioxide concentration environment information, leaf area index, single leaf area, transpiration crop physiological signals, and crop variety, greenhouse area and enclosure structure specific heat capacity basic physical parameters, and performing time series alignment and spatial interpolation optimization on multi-source data through a time-space fusion algorithm.

3. The adaptive greenhouse environmental control system of claim 1, wherein, The physical model module of the decision layer is constructed based on the principles of energy balance and mass balance, and specifically comprises: The total heat balance equation is expressed as: dQ total / dt = Q in - Q out + Q trans , where dQ total / dt represents the rate of change of the total heat inside the greenhouse over time, the heat input term Q in includes solar radiation heat Q solar , heating device heat Q heat , crop photosynthesis heat Q photo , and device operation heat Q metaβ , the heat output term Q out includes ventilation heat dissipation Q vent , enclosure heat dissipation Q loss , soil heat conduction Q transp , and evaporation heat dissipation Q evap , and the heat conduction term Q trans includes air and device / crop conduction Q air-oβj and air and soil conduction Q air-soil ; The water vapor mass balance equation is expressed as: dW total / dt = G in -G out , wherein the water vapor input term G in includes the crop transpiration and soil evaporation G evap and the humidification equipment input G humid , the water vapor output term G out includes the water vapor carried away by ventilation G vent and the water vapor removed by the dehumidification equipment G dehumid ; The mass balance equation for carbon dioxide is expressed as: dC total / dt = C in -C out -C consume where C in includes the input of carbon dioxide C fert , the output of carbon dioxide C out includes the amount of carbon dioxide C vent , the consumption of carbon dioxide C consume is the amount of carbon dioxide consumed by photosynthesis of the crop C photo .

4. The adaptive greenhouse environment control system of claim 3, wherein, The adaptive MPC algorithm module of the decision layer is based on a total heat balance equation, a water vapor mass balance equation and a carbon dioxide mass balance equation, and constructs a discrete state equation for predicting the environment state at a future time, which comprises: A state equation for predicting temperature is constructed based on the total heat balance equation and has a form of: where T(k+1) is the indoor predicted temperature at k+1 time, T(k) is the indoor temperature at k time, Q in (k), Q out (k), Q trans (k) is the total heat input, total heat output and heat conduction at k time, m 空 is the air mass, c 空 is the air specific heat capacity, and Δt is the time period. A state equation for predicting humidity is constructed based on the water vapor mass balance equation and has a form of: wherein d(k+1) is the predicted indoor absolute humidity at k+1 time, d(k) is the indoor absolute humidity at k time, G in (k), G out (k) are the total water vapor input and output at k time, respectively, and V0 is the greenhouse volume. A state equation for predicting carbon dioxide concentration is constructed based on the carbon dioxide mass balance equation and has a form of: Wherein, c(k+1) is the indoor predicted CO2 concentration at k+1 time, c(k) is the indoor CO2 concentration at k time, C in (k), C out (k), C 消耗 (k) are the total amount of CO2 input, the total amount of CO2 output and the total amount of CO2 consumed by crop photosynthesis at k time, respectively.

5. The adaptive greenhouse environment control system of claim 4, wherein, The adaptive MPC algorithm module comprises a prediction layer, an optimization layer and a correction layer, the prediction layer outputs future temperature, absolute humidity and CO2 concentration sequences based on the discrete state equations T(k+1), d(k+1) and c(k+1) respectively, the optimization layer solves a sequence of operation amounts that minimizes an objective function under the constraint condition of crop physiological requirements, and the correction layer dynamically adjusts physical model parameters and optimization target weights in real time according to the actual state deviation fed back by the perception layer.

6. The adaptive greenhouse environment control system of claim 5, wherein, The optimization target of the adaptive MPC algorithm module has a double target of "temperature tracking accuracy + energy consumption optimization", and an objective function is: where J is a loss function, ω T is a temperature tracking weight, T 目标 is a target temperature value required for crop growth in the greenhouse environment, T(k+i|k) represents a greenhouse indoor temperature value predicted at the k+i time from the k time, N is a prediction time domain length, ω j is a weight of energy consumption of the jth type of equipment, μ j (k+i-1) is an operating amount of the jth type of equipment at the k+i-1 time, and M represents the number of types of equipment participating in the regulation of the greenhouse environment.

7. The adaptive greenhouse environmental control system of claim 5, wherein, The optimization target of the adaptive MPC algorithm module has a double target of "humidity tracking accuracy + CO2 tracking accuracy", and an objective function is: J =∑ω d ·(d 目标 (k+i|k)) 2 +∑ω c ·(c 目标 (k+i|k)) 2 +∑∑ω j ·u j (k+i-1) 2 wherein ω d represents the humidity tracking weight, d 目标 represents the target humidity value required for crop growth in the greenhouse environment, d(k+i|k) represents the indoor humidity value at the k+i time predicted at the k time, ω c represents the CO2 concentration tracking weight, c 目标 represents the target CO2 concentration value required for crop photosynthesis, c(k+i|k) represents the indoor CO2 concentration value at the k+i time predicted at the k time, ω j is the energy consumption weight of the jth type of equipment, u j (k+i-1) is the operating amount of the jth type of equipment at the k+i-1 time.

8. The adaptive greenhouse environmental control system of claim 1, wherein, The execution layer comprises a plurality of device clusters, and the device clusters are divided into temperature regulation clusters, humidity regulation clusters, light regulation clusters and carbon dioxide regulation clusters according to functions.

9. The adaptive greenhouse environmental control system of claim 1, wherein, The optimization layer constructs a three-dimensional evaluation model based on an environment deviation degree, an energy consumption cost and crop sensitivity, and solves an optimal device combination through a particle swarm algorithm.

10. The adaptive greenhouse environment control system of claim 9, wherein, The three-dimensional evaluation model has an expression of Priority=α·ΔEnv+β·Cost+γ·Sens, wherein ΔEnv is the environment deviation degree, Cost is the energy consumption cost, Sens is the crop sensitivity, and α, β and γ are weight coefficients.

Images