Self-sufficient energy greenhouse and intelligent control method
By adjusting the tilt angle of photovoltaic panels and environmental parameters through an intelligent control system, the contradiction between cooling and lighting in tropical greenhouses has been resolved, achieving energy self-sufficiency and efficient power generation, and improving crop yield and the precision of environmental control.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SANYA NATIONAL INSTITUTE OF SOUTHERN BREEDING CHINESE ACADEMY OF AGRICULTURAL SCIENCES
- Filing Date
- 2026-05-26
- Publication Date
- 2026-06-23
AI Technical Summary
Tropical greenhouses face the challenge of balancing cooling and lighting. Traditional shading measures affect photosynthetic radiation, while fixed photovoltaic panels create a conflict between shading and power generation. Existing environmental control systems lack multi-factor synergistic optimization and adaptive capabilities, leading to issues with crop yield and energy consumption.
An intelligent control system is used to adjust the tilt angle of the photovoltaic panels. Combined with a model predictive controller and a reinforcement learning controller, the solar thermal environment is dynamically optimized. Energy self-sufficiency is achieved through a photovoltaic power generation and energy storage system, and the internal environmental parameters of the greenhouse are adjusted to meet the needs of lighting, cooling and energy.
It achieves efficient power generation in tropical greenhouses while dynamically balancing photovoltaic shading and daylighting, improving environmental control quality and crop yield, reducing energy consumption and ensuring system stability.
Smart Images
Figure CN122261306A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of intelligent control technology for facility agriculture, and in particular to an energy-self-sufficient fully automated greenhouse and its intelligent control method. Background Technology
[0002] In tropical regions like Hainan, high temperatures, strong radiation, frequent typhoons, and prolonged overcast weather pose severe challenges to crop production: strong sunlight in summer can scorch leaves, torrential rains and floods can cause root rot, and enclosed environments can easily lead to diseases caused by high temperature and humidity; while weak light in winter can cause crops to grow excessively and lose quality. Although modern greenhouses can provide a controllable environment, they generally suffer from poor heat insulation, high energy consumption, and high costs, making large-scale promotion difficult.
[0003] The core contradiction in tropical greenhouses lies in the difficulty of balancing cooling and light exposure: traditional shading and cooling measures simultaneously weaken photosynthetically active radiation. For example, most greenhouses in Hainan use external shading curtains with a 75% shading rate, adjusting plant light exposure by opening and closing the curtains during seasons with strong sunlight. However, Hainan's tropical or subtropical marine monsoon climate is highly variable, with daily sunlight naturally varying from morning (weak) to noon (strongest) to afternoon and evening (weak), plus cloudy skies and seasonal variations in sunlight. This complex variation in sunlight means that a single fixed 75% or other shading curtain cannot guarantee the stable light required for rapid plant growth, leading to reduced yields due to cooling. Furthermore, existing photovoltaic greenhouses mostly use fixed photovoltaic panels, causing permanent shading, and the optimal tilt angle for photovoltaic installation often conflicts with the greenhouse's light requirements, creating a conflict between power generation and planting space.
[0004] Therefore, there is an urgent need in this field to develop a greenhouse system that can dynamically balance photovoltaic shading and crop lighting, achieve energy efficiency and self-sufficiency, and possess intelligent collaborative control and adaptive adjustment capabilities for multiple environmental factors, so as to promote the energy conservation, intelligence and sustainable development of facility agriculture in tropical regions. Summary of the Invention
[0005] This application provides an energy-self-sufficient fully automated greenhouse and an intelligent control method. The greenhouse intelligently controls the tilt angle of photovoltaic panels and various environmental parameters inside the greenhouse through an intelligent greenhouse control system, effectively solving the long-standing contradiction between "cooling and lighting" in tropical greenhouses. While achieving efficient power generation, it dynamically optimizes the indoor light and heat environment, improving the quality of environmental control and crop yield.
[0006] Firstly, a fully automated, energy-self-sufficient greenhouse is provided, comprising a greenhouse body, a vacuum-insulated tempered ultra-white glass enclosure structure, a photovoltaic power generation and energy storage system, and a greenhouse intelligent control system.
[0007] The photovoltaic power generation and energy storage system includes: photovoltaic panels, drive brackets, and energy storage stations. The photovoltaic panels are installed on the top of the greenhouse body, and the drive brackets are connected to the photovoltaic panels.
[0008] The greenhouse intelligent control system includes a data acquisition module, a control module, and an execution module; the data acquisition module includes: a carbon dioxide sensor, a temperature and humidity sensor, and a light intensity sensor for acquiring indoor environmental parameters, and a weather station for acquiring outdoor environmental parameters;
[0009] The control module includes a model prediction controller and a reinforcement learning controller; the model prediction controller generates short-term optimal control instructions based on operating parameters, and the reinforcement learning controller correlates the reward function associated with the optimization objective of the model prediction controller to generate a long-term optimization strategy.
[0010] The execution module controls the drive bracket to adjust the tilt angle of the photovoltaic panel based on the control command, and regulates the ventilation equipment, wet curtain water pump, humidifier, CO2 injection device, air conditioning temperature control system, supplemental lighting and fertilization system in the greenhouse to regulate the internal environment of the greenhouse and meet the operating requirements of the greenhouse, including: lighting requirements, cooling requirements and energy requirements.
[0011] In conjunction with the first aspect, in some implementations of the first aspect, the model predictive controller constructs a multi-state variable dynamic equation based on the operating parameters, and solves for the short-term optimal control command with the objectives of stabilizing crop photosynthetically active radiation at a preset level and maximizing the power generation of the photovoltaic panel. The short-term optimal control command includes the target tilt angle of the photovoltaic panel.
[0012] It should be understood that the operating parameters include multiple types of state variables and multiple types of disturbance variables. The multiple types of state variables include plant dry weight, indoor CO2 concentration, temperature, absolute humidity, battery state of charge and temperature, and light intensity. The multiple types of disturbance variables include solar radiation, outdoor CO2 concentration, temperature, humidity, atmospheric pressure, wind speed, and wind direction.
[0013] It should be understood that by constructing multi-state variable dynamic equations through model predictive controllers, and solving for short-term optimal control commands containing the target tilt angle of photovoltaic panels with the dual objectives of maximizing crop photosynthetic effective radiation and photovoltaic power generation, this approach not only resolves the core contradiction between cooling and lighting, and between power generation and planting in tropical greenhouses, avoiding the industry pain point of "cooling equals reduced production," but also achieves energy self-sufficiency and cost reduction in greenhouses, while ensuring the accuracy of regulation and adaptability to climate scenarios.
[0014] In conjunction with the first aspect, in some implementations of the first aspect, the reinforcement learning controller employs the Deep Deterministic Policy Gradient Algorithm (DDPG), whose reward function is configured as follows:
[0015] ,
[0016] in, The model predicts negative values for the controller's performance metrics, encouraging control strategies to reduce energy consumption and avoid constraint violations. The negative value of the energy penalty for the model prediction controller is used to penalize power shortage, light loss, and excessive battery cycling; the reinforcement learning controller is used to generate the long-term optimization strategy.
[0017] It should be understood that this application designs the reward function of reinforcement learning (DDPG) as the negative of the sum of the stage cost and energy management penalty term of Model Predictive Control (MPC). This design directly transforms the stage cost function and energy management penalty of MPC into an immediate reward signal for reinforcement learning. Since the stage cost function and energy management penalty of MPC correspond to specific physical or economic objectives, the update direction of the policy parameters of reinforcement learning is guided in real time during the process of maximizing the cumulative reward through gradient descent, so as to simultaneously reduce energy consumption, increase output, meet safety constraints, and optimize energy use. This is equivalent to injecting the multi-objective optimization problem and domain knowledge encoded by MPC as a structured and dense supervision signal into the learning process of reinforcement learning, thereby avoiding ineffective or dangerous exploration and enabling the policy to quickly converge to a balance point that satisfies both short-term engineering constraints and achieves long-term adaptive optimization.
[0018] In conjunction with the first aspect, in some implementations of the first aspect, the control module further includes a parameter adaptive learning module, which is used to dynamically adjust the learnable parameters in the model prediction controller and the reinforcement learning controller through offline pre-training, online updating and transfer learning.
[0019] In conjunction with the first aspect, in some implementations of the first aspect, the control module controls the tilt angle of the photovoltaic panel in different modes to meet the operating requirements of the greenhouse, specifically including: light-priority mode, balance optimization mode, and time-segmented adaptive mode.
[0020] In conjunction with the first aspect, in some implementations of the first aspect, the power distribution logic of the energy storage station is as follows: the power generated by the photovoltaic panel is divided into three parts through a voltage regulator: the first part is used to charge the battery pack, the second part is directly supplied to the DC load, and the third part is supplied to the AC load after being converted by an inverter; when the photovoltaic power generation is insufficient, the battery pack discharges and supplies power to the load through the voltage regulator.
[0021] It should be understood that this application distributes photovoltaic power through a voltage regulator, which can reduce inverter losses and improve energy utilization efficiency when directly supplying DC loads. With the battery pack replenishment mechanism in case of photovoltaic power shortage, peak shaving and valley filling of electricity can be achieved, ensuring stable power supply to the greenhouse at all times, helping to achieve energy self-sufficiency, and reducing operating energy consumption and dependence on the external power grid.
[0022] Secondly, a method for intelligent control of a fully automated, energy-self-sufficient greenhouse is provided. This method is applied to a greenhouse as described in any implementation of the first aspect, and the method includes:
[0023] S1: After the system is powered on, it executes a self-test program to verify the communication status of the acquisition module, execution module and each device, and loads the offline pre-training parameters of the parameter adaptive learning module; it also collects multiple types of state variables and multiple types of disturbance variables in real time through the acquisition module and displays them visually.
[0024] S2: The model predictive controller in the control module performs rolling optimization based on the state variables, disturbance variables and system dynamics model to solve a finite-time optimization problem that includes control costs, output incentives and constraint penalties, and obtains short-term optimal control commands. At the same time, the reinforcement learning controller performs long-term policy optimization based on the reward function associated with the optimization objective of the model predictive controller.
[0025] S3: The execution module responds to the instructions of the control module to precisely regulate the greenhouse environment and equipment;
[0026] S4: The parameter adaptive learning module uses an online update mechanism to correct the learnable parameters of the model prediction controller and the reinforcement learning controller in real time based on the gradient feedback of the reward function; when the greenhouse crop is changed or the environment changes significantly, the trained parameters are reused and fine-tuned quickly through transfer learning.
[0027] In conjunction with the second aspect, in some implementations of the second aspect, step S3 specifically includes:
[0028] The system controls the drive bracket to adjust the tilt angle of the photovoltaic panels; regulates the ventilation equipment, wet curtain pump, humidifier, CO2 injection device, air conditioning temperature control system, supplemental lighting and fertilization system to control the temperature, humidity and CO2 concentration in the greenhouse; and dynamically allocates the photovoltaic power generation power and the charging and discharging power of the energy storage station. Attached Figure Description
[0029] Figure 1 This is a schematic diagram of a greenhouse body provided in an embodiment of this application.
[0030] Figure 2 This is a schematic diagram of a greenhouse intelligent control system architecture provided in an embodiment of this application.
[0031] Figure 3This is a flowchart illustrating the operation of a photovoltaic power generation and energy storage system, as provided in an embodiment of this application. Detailed Implementation
[0032] The terminology used in the following embodiments is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to also include expressions such as “one or more,” unless the context clearly indicates otherwise. It should also be understood that in the following embodiments of this application, “at least one” and “one or more” refer to one, two, or more than two. The term “and / or” is used to describe the relationship between related objects, indicating that three relationships can exist; for example, A and / or B can indicate: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character “ / ” generally indicates that the preceding and following related objects are in an “or” relationship.
[0033] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.
[0034] The harsh climate of tropical regions and the limitations of existing greenhouses severely restrict agricultural production: high temperatures and strong sunlight, torrential rains and typhoons, and weak sunlight in winter directly affect crop growth and quality. Traditional greenhouses face the contradiction of "cooling down means reducing yield," while fixed photovoltaic greenhouses suffer from permanent shading and the problem of "light and electricity competing for land." Existing environmental control systems mostly employ simplistic logic, lacking multi-factor collaborative optimization and adaptive capabilities, making it difficult to coordinate yield, energy consumption, and system stability. Therefore, there is an urgent need to develop a new type of greenhouse system that can dynamically balance shading and lighting, achieve energy self-sufficiency, and possess intelligent collaborative control and adaptive adjustment.
[0035] The technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings.
[0036] This application provides an energy-self-sufficient, fully automated greenhouse, which includes a greenhouse body, a vacuum-insulated tempered ultra-clear glass enclosure structure, a photovoltaic power generation and energy storage system, and a greenhouse intelligent control system.
[0037] The photovoltaic power generation and energy storage system includes: photovoltaic panels, drive brackets, and energy storage stations. The photovoltaic panels are installed on the top of the greenhouse body, and the drive brackets are connected to the photovoltaic panels.
[0038] The intelligent greenhouse control system includes a data acquisition module, a control module, and an execution module; the data acquisition module is used to acquire operating parameters; the control module generates control commands based on the operating parameters;
[0039] The execution module controls the drive bracket to adjust the tilt angle of the photovoltaic panel based on the control command, and adjusts the internal environment of the greenhouse to meet the greenhouse operation requirements, including: light requirements, cooling requirements, and energy requirements.
[0040] Figure 1 This is a schematic diagram of a greenhouse body provided in an embodiment of this application.
[0041] In one possible implementation, the greenhouse body is made of vacuum ultra-clear tempered glass, and the top is insulated with ultra-clear + vacuum layer + diffuse reflection tempered glass, with a multi-ridge single-slope roof structure; the interior of the greenhouse is divided into several independent compartments by partition walls, and each compartment is equipped with multi-layer three-dimensional retractable cultivation racks; the bottom of the greenhouse has a hollow structure and is equipped with rainproof electrically controlled louvers, air filters and insect-proof nets in sequence.
[0042] Furthermore, the slope of the multi-ridged single-slope roof is 15° to 20°, and an insulation layer with an aluminum foil reflective film is laid under the multi-ridged single-slope roof; the partition wall is composed of double-layer tempered glass sandwiched with vacuum insulation board; and sound-absorbing cotton is provided at the air outlet of the ventilation equipment.
[0043] Continue to refer to Figure 1 The greenhouse's main structure uses a steel frame as its skeleton, combined with double-layered vacuum-insulated glass to form a fully enclosed glass curtain wall, creating a transparent enclosed space. Its roof features a multi-ridged, single-slope design, with photovoltaic panels embedded in the roof area. A weather station 1 is also installed on the support structure above the roof to collect real-time environmental parameters from outside the greenhouse. The interior of the greenhouse is divided into several independent functional spaces, including a south greenhouse 3 and a north greenhouse 4. These spaces are separated by double-layered vacuum-insulated tempered glass walls, ensuring the independence of each area's environment. Each independent space is equipped with a separate double-door access passage, with M1020-sized metal-framed tempered glass doors at both ends. An area of approximately 1.5m is reserved between the two doors as an operating room 5, which houses displays showing the internal environmental data and algorithm operation status of each greenhouse.
[0044] In some examples, the control module includes a model predictive controller and a reinforcement learning controller. The model predictive controller generates short-term optimal control instructions based on the operating parameters, and the reinforcement learning controller correlates the reward function associated with the optimization objective of the model predictive controller to generate a long-term optimization strategy.
[0045] In some examples, the model predictive controller constructs a multi-state variable dynamic equation based on the operating parameters, with the objectives of stabilizing crop photosynthetically active radiation at a preset level and maximizing the power generation of the photovoltaic panel, and solves for the short-term optimal control command, which includes the target tilt angle of the photovoltaic panel.
[0046] In some examples, the reinforcement learning controller employs the Deep Deterministic Policy Gradient Algorithm (DDPG), whose reward function is configured as follows:
[0047] ,
[0048] in, The model predicts negative values for the controller's performance metrics, encouraging control strategies to reduce energy consumption and avoid constraint violations. The negative value of the energy penalty for the model prediction controller is used to penalize power shortage, light loss, and excessive battery cycling; the reinforcement learning controller is used to generate the long-term optimization strategy.
[0049] In some examples, the control module further includes a parameter adaptive learning module for dynamically adjusting the learnable parameters in the model prediction controller and the reinforcement learning controller through offline pre-training, online updates, and transfer learning.
[0050] In some examples, the execution module regulates the ventilation equipment, wet curtain pumps, humidifiers, CO2 injection devices, air conditioning temperature control systems, supplemental lighting, and fertilization systems within the greenhouse based on the control commands, in order to adjust the internal environment of the greenhouse.
[0051] Figure 2 This is a schematic diagram of a greenhouse intelligent control system architecture provided in an embodiment of this application.
[0052] In one possible implementation, refer to Figure 3 The greenhouse intelligent control system includes a data acquisition module, a control module, and an execution module. The data acquisition module consists of indoor and outdoor parts. The indoor part includes a carbon dioxide sensor, a temperature and humidity sensor, and a light intensity sensor installed in the greenhouse body; the outdoor part is a weather station 1, which acquires relevant data from the weather station.
[0053] Continue to refer to Figure 2The execution module coordinates and regulates the greenhouse environment and equipment based on the instructions of the control module. The specific equipment regulated includes: UV sterilization lamps, wet curtain water pumps, indoor (outdoor) circulating fans, humidifiers, LED supplemental lighting (i.e., lighting modules), air conditioners, etc.
[0054] Figure 3 This is a flowchart illustrating the operation of a photovoltaic power generation and energy storage system, as provided in an embodiment of this application.
[0055] In some examples, the controller controls the tilt angle of the photovoltaic panels in different modes to meet the operating requirements of the greenhouse, specifically including: light-priority mode, balance optimization mode, and time-segmented adaptive mode.
[0056] In one possible implementation, a light-priority mode is used, where the indoor PAR is less than the crop's lower photosynthetic limit threshold.
[0057] The control module (which uses a PLC as the core execution unit of the control module) outputs instructions to drive the photovoltaic panel to adjust to the maximum light transmission angle, so that sunlight can pass through the ultra-white low-E diffuse reflection glass into the greenhouse to the maximum extent. At the same time, the diffuse reflection glass makes the light on the plant canopy uniform and unobstructed, so that the crops can receive natural light from all directions. If the indoor PAR is still lower than the threshold after the angle is adjusted, the control module turns on the liftable plant supplement light for artificial lighting. At this time, the photovoltaic power generation gives way to the lighting needs and only uses the excess diffuse light to generate electricity.
[0058] In the balanced optimization mode, where the indoor PAR (Paracity) is within the optimal threshold range for crop photosynthesis, the control module adjusts the tilt angle of the photovoltaic panels based on a photovoltaic power model. This maintains stable indoor PAR while matching the photovoltaic panel's angle of sunlight to the solar radiation angle, enabling efficient power generation from excess solar energy. Furthermore, the control module fine-tunes the position of the cultivation rack based on the sunlight angle to ensure uniform light exposure for crop photosynthesis. Simultaneously, the posture detection components and light intensity detection components of the photovoltaic panel drive support collect data in real time. If changes in sunlight intensity cause PAR to deviate from the threshold, the control module immediately fine-tunes the tilt angle of the photovoltaic panels.
[0059] In the time-segmented adaptive mode, the control module combines the natural variation of sunlight exposure with pre-programmed and real-time detection to adjust the tilt angle of the photovoltaic panels according to the time of day: when the sunlight intensity is weak in the morning and evening, it is adjusted to a large light-transmitting tilt angle to ensure light intake; when the sunlight intensity is strong at noon, it is adjusted to a balance / power generation priority tilt angle to balance shading and power generation, without the need for manual intervention.
[0060] In some examples, the power distribution logic of the energy storage station is as follows: the power generated by the photovoltaic panels is divided into three parts through the voltage regulator: the first part is used to charge the battery pack, the second part is directly supplied to the DC load, and the third part is supplied to the AC load after being converted by the inverter; when the photovoltaic power generation is insufficient, the battery pack discharges and supplies power to the load through the voltage regulator.
[0061] In one possible implementation, refer to Figure 3 The photovoltaic power generation and energy storage system adopts a hybrid topology structure of "photovoltaic direct drive + energy storage buffer + AC / DC dual grid power supply". Its hardware architecture mainly consists of photovoltaic array units, central control and voltage regulation units, energy storage battery units, inverter output units and end load networks.
[0062] First is the photovoltaic array unit, where photovoltaic panels are laid on the top of the greenhouse as the main energy input.
[0063] In this embodiment, the photovoltaic power generation panel consists of 216 solar panels with a total installation area of 146.8 square meters and a conversion efficiency of 23%. The designed daily power generation is approximately 106 kWh, while the greenhouse's daily power consumption is approximately 84.2 kWh, achieving a power surplus of about 20%, thus meeting the system's energy-saving operation goal of "self-generation and self-consumption, with surplus power stored." Its output power follows the photovoltaic power model.
[0064] ,
[0065] Where κ is the photovoltaic conversion coefficient (unit: kW / (W / m²)), reflecting the proportion of solar energy converted into electrical energy, and ηPV is the photovoltaic panel efficiency, taking into account the effects of factors such as temperature and aging.
[0066] Secondly, there is the central control and voltage regulation unit, which is electrically connected to the output of the photovoltaic array unit and is used to convert the unstable photovoltaic voltage into stable DC power. The voltage regulator is rated at 200A and supports adaptive adjustment of multiple voltage levels, including 12V, 24V, and 48V. In terms of functional configuration, the voltage regulator is equipped with a real-time tracking module to track the maximum output power point of the photovoltaic array, serving as an energy hub connecting the generation end, energy storage end, and load end.
[0067] Next is the energy storage battery pack unit, which is bidirectionally connected to the central control and voltage regulation unit via a DC bus for storing and releasing electrical energy. In terms of hardware configuration, the energy storage battery pack unit consists of 16 lithium iron phosphate battery cells, each with a specification of 12V / 250Ah. The total rated capacity after series and parallel connection is 48kWh, and the charge / discharge efficiencies are η... ch and η disIts operating logic is based on the battery SOC kinetic equation:
[0068] ,
[0069] Where, μ 10 >0 indicates charging power, μ 10 <0 indicates discharge power, C batt For battery capacity, η ch η dis These refer to charge and discharge efficiency. When there is sufficient sunlight, the central control and voltage regulation unit controls the excess electrical energy to be stored in the energy storage battery pack unit; when there is insufficient sunlight at night or on cloudy or rainy days, the energy storage battery pack unit releases electrical energy to the load through the controller, realizing the functions of "peak shaving and valley filling" and emergency supply guarantee.
[0070] Next is the inverter output unit. The input terminal of the inverter output unit is connected to the DC output terminal of the voltage regulator or the energy storage battery pack unit, and its output terminal is connected to the AC load distribution box to convert DC power to AC power. In this embodiment, the inverter output unit uses a pure sine wave inverter with a rated power of 23kW to ensure that it can withstand the inrush current when all AC loads start and run simultaneously.
[0071] Finally, there is the load network, which is divided into DC power supply branches and AC power supply branches based on the power supply characteristics of the equipment. The AC power supply branch is connected to the output terminal of the inverter output unit and supplies power to the AC loads. The AC loads specifically include outdoor fans for overall heat dissipation and ventilation in the greenhouse, indoor fans for indoor air circulation and temperature equalization, sprayers for humidification and spraying operations, and germicidal lamps for disease control. The DC power supply branch is directly connected to the DC output terminal of the central control and voltage stabilization unit and supplies power to the DC loads to avoid inverter losses. The DC loads specifically include LED supplemental lighting to provide the spectrum required for crop photosynthesis, evaporative cooling system water pumps to drive water circulation in the evaporative cooling system, photovoltaic panel control motors to drive angle adjustment of the roof photovoltaic panels, and tissue culture rack telescopic motors to control the expansion and contraction of the three-dimensional culture racks.
[0072] Through the aforementioned hardware architecture, this system achieves end-to-end management from light energy acquisition, conversion, storage to distribution, ensuring the independent and stable operation of the greenhouse under off-grid or weak grid conditions.
[0073] Figure 3 This is a schematic diagram of a greenhouse intelligent control system architecture provided in an embodiment of this application.
[0074] In one possible implementation, refer to Figure 3The greenhouse intelligent control system includes a data acquisition module, a control module, and an execution module. The data acquisition module consists of indoor and outdoor parts. The indoor part includes a carbon dioxide sensor, a temperature and humidity sensor, and a light intensity sensor installed in the greenhouse body; the outdoor part connects to weather station 1 and acquires relevant data from weather station 1.
[0075] Continue to refer to Figure 3 The execution module coordinates and regulates the greenhouse environment and equipment based on the instructions of the control module. The specific equipment regulated includes: UV sterilization lamps, wet curtain water pumps, indoor (outdoor) circulating fans, humidifiers, LED supplemental lighting (i.e., lighting modules), air conditioners, etc.
[0076] This application embodiment also provides an intelligent control method for a fully automated, energy-self-sufficient greenhouse, which is applied to the greenhouse described in any of the foregoing examples. The method includes:
[0077] S1: After the system is powered on, it executes a self-test program to verify the communication status of the acquisition module, execution module and each device, and loads the offline pre-training parameters of the parameter adaptive learning module; it also collects multiple types of state variables and multiple types of disturbance variables in real time through the acquisition module and displays them visually.
[0078] S2: The model predictive controller in the control module performs rolling optimization based on the state variables, disturbance variables and system dynamics model to solve a finite-time optimization problem that includes control costs, output incentives and constraint penalties, and obtains short-term optimal control commands. At the same time, the reinforcement learning controller performs long-term policy optimization based on the reward function associated with the optimization objective of the model predictive controller.
[0079] S3: The execution module responds to the instructions of the control module to precisely regulate the greenhouse environment and equipment;
[0080] S4: The parameter adaptive learning module uses an online update mechanism to correct the learnable parameters of the model prediction controller and the reinforcement learning controller in real time based on the gradient feedback of the reward function; when the greenhouse crop is changed or the environment changes significantly, the trained parameters are reused and fine-tuned quickly through transfer learning.
[0081] In some examples, step S3 specifically includes:
[0082] The system controls the drive bracket to adjust the tilt angle of the photovoltaic panels; regulates the ventilation equipment, wet curtain pump, humidifier, CO2 injection device, air conditioning temperature control system, supplemental lighting and fertilization system to control the temperature, humidity and CO2 concentration in the greenhouse; and dynamically allocates the photovoltaic power generation power and the charging and discharging power of the energy storage station.
[0083] The above are merely preferred embodiments of this application. The scope of protection of this application is not limited to the above embodiments. Any equivalent modifications or changes made by those skilled in the art based on the content disclosed in this application should be included within the scope of protection recorded in the claims.
Claims
1. A fully automated, energy-self-sufficient greenhouse, comprising a greenhouse body, a photovoltaic power generation and energy storage system, and a greenhouse intelligent control system, characterized in that, The photovoltaic power generation and energy storage system includes: photovoltaic panels, drive brackets, and energy storage stations. The photovoltaic panels are installed on the top of the greenhouse body, and the drive brackets are connected to the photovoltaic panels. The intelligent greenhouse control system includes a data acquisition module, a control module, and an execution module; The data acquisition module includes: a carbon dioxide sensor, a temperature and humidity sensor, and a light intensity sensor for acquiring indoor environmental parameters, as well as a weather station for acquiring outdoor environmental parameters. The control module includes a model prediction controller and a reinforcement learning controller; the model prediction controller generates short-term optimal control instructions based on operating parameters, and the reinforcement learning controller correlates the reward function associated with the optimization objective of the model prediction controller to generate a long-term optimization strategy. The execution module controls the drive bracket to adjust the tilt angle of the photovoltaic panel based on the control command, and regulates the ventilation equipment, wet curtain water pump, humidifier, CO2 injection device, air conditioning temperature control system, supplemental lighting and fertilization system in the greenhouse to regulate the internal environment of the greenhouse and meet the operating requirements of the greenhouse, including: lighting requirements, cooling requirements and energy requirements.
2. The greenhouse according to claim 1, characterized in that, The model predictive controller constructs a multi-state variable dynamic equation based on the operating parameters, and solves the short-term optimal control command with the objectives of stabilizing the crop's photosynthetically active radiation at a preset level and maximizing the power generation of the photovoltaic panel. The short-term optimal control command includes the target tilt angle of the photovoltaic panel.
3. The greenhouse according to claim 2, characterized in that, The reinforcement learning controller employs the Deep Deterministic Policy Gradient Algorithm (DDPG), and its reward function is configured as follows: ， in, The model predicts negative values for the controller's performance metrics, encouraging control strategies to reduce energy consumption and avoid constraint violations. The negative value of the energy penalty for the model prediction controller is used to penalize power shortage, light loss, and excessive battery cycling; the reinforcement learning controller is used to generate the long-term optimization strategy.
4. The greenhouse according to claim 3, characterized in that, The control module also includes a parameter adaptive learning module, which is used to dynamically adjust the learnable parameters in the model prediction controller and the reinforcement learning controller through offline pre-training, online updating and transfer learning.
5. The greenhouse according to claim 1, characterized in that, The control module controls the tilt angle of the photovoltaic panels in different modes to meet the operating requirements of the greenhouse, specifically including: light-priority mode, balance optimization mode, and time-based adaptive mode.
6. The greenhouse according to claim 1, characterized in that, The power distribution logic of the energy storage station is as follows: the power generated by the photovoltaic panels is divided into three parts through the voltage regulator: the first part is used to charge the battery pack, the second part is directly supplied to the DC load, and the third part is supplied to the AC load after being converted by the inverter; when the photovoltaic power generation is insufficient, the battery pack discharges and supplies power to the load through the voltage regulator.
7. A fully automated intelligent control method for an energy-self-sufficient greenhouse, the method being applied to a greenhouse as described in any one of claims 1 to 6, characterized in that, The method includes: S1: After the system is powered on, it executes a self-test program to verify the communication status of the acquisition module, execution module and each device, and loads the offline pre-training parameters of the parameter adaptive learning module; it also collects multiple types of state variables and multiple types of disturbance variables in real time through the acquisition module and displays them visually. S2: The model predictive controller in the control module performs rolling optimization based on the state variables, disturbance variables and system dynamics model to solve a finite-time optimization problem that includes control costs, output incentives and constraint penalties, and obtains short-term optimal control commands. At the same time, the reinforcement learning controller performs long-term policy optimization based on the reward function associated with the optimization objective of the model predictive controller. S3: The execution module responds to the instructions of the control module to precisely regulate the greenhouse environment and equipment; S4: The parameter adaptive learning module uses an online update mechanism to correct the learnable parameters of the model prediction controller and the reinforcement learning controller in real time based on the gradient feedback of the reward function; when the greenhouse crop is changed or the environment changes significantly, the trained parameters are reused and fine-tuned quickly through transfer learning.
8. The method according to claim 7, characterized in that, Step S3 specifically includes: The system controls the drive bracket to adjust the tilt angle of the photovoltaic panels; regulates the ventilation equipment, wet curtain pump, humidifier, CO2 injection device, air conditioning temperature control system, supplemental lighting and fertilization system to control the temperature, humidity and CO2 concentration in the greenhouse; and dynamically allocates the photovoltaic power generation power and the charging and discharging power of the energy storage station.