A kind of edge-end cooperation-based multi-span greenhouse environment control system and method

Abstract

The application discloses a kind of based on edge-end cooperation's continuous ridge greenhouse environment control system and method, system includes: terminal perception execution layer contains multimodal sensor network and intelligent execution mechanism group;Edge intelligent control layer specifically includes: multi-source data fusion module, for fusing the measurement value of multimodal sensor network, output environment state optimal estimation value;Model predictive control module is used as input with environment state optimal estimation value and weather forecast, solves optimal control sequence in future time domain by rolling optimization, and issues control instruction to intelligent execution mechanism group;Fault diagnosis and fault-tolerant control module is used to obtain the current feedback value, position feedback value and / or angle feedback value of execution mechanism in real time, and compare with the instruction reference issued by model predictive control module, execute fault diagnosis and hierarchical fault tolerance.The application is based on the three-module closed-loop architecture of edge-end cooperation, realizes the intelligent fault tolerance of execution mechanism and the accurate control to continuous ridge greenhouse environment.

Description

Technical Field

[0001] This application relates to the intersection of precision control of agricultural facility environment and industrial Internet of Things technology. Specifically, it relates to an environmental control system and method for multi-span greenhouses based on edge-end collaboration. Background Technology

[0002] The intelligent environmental control system for multi-span greenhouses utilizes sensors, automatic control algorithms, and communication technology to automatically monitor and regulate multiple environmental factors inside the multi-span plastic film or glass greenhouses, such as temperature, humidity, light intensity, and CO2 concentration, ensuring that crops are always in a suitable growing environment. This intelligent control system can reduce energy consumption and labor costs while guaranteeing crop yield and quality, minimizing crop stress caused by improper control, reducing energy consumption and maintenance manpower, and improving the overall reliability and autonomy of the system.

[0003] Existing technologies mainly employ logic control based on a single factor threshold or traditional proportional-integral-derivative (PID) control. For example, when the temperature exceeds a set value, a skylight or fan is opened; when sunlight is too strong, a shade net is deployed. The drawbacks of these methods are: the control strategies are fragmented, easily leading to control conflicts, such as a sudden drop in humidity due to opening windows for cooling; there is a lack of real-time sensing of the physical state of the actuators (including motor current and film position), making it impossible to detect jamming or slippage faults; the system continues to blindly issue commands even when equipment fails, leading to equipment damage or environmental loss of control; and the control decisions lack foresight, relying solely on reactive adjustments based on current deviations, resulting in frequent equipment start-ups and shutdowns and energy waste. Summary of the Invention

[0004] To address the shortcomings of the existing technologies, this application discloses a multi-span greenhouse environmental control system and method based on edge-end collaboration. Based on a three-module closed-loop architecture with edge-end collaboration, it achieves intelligent fault tolerance of the actuators and precise control of the multi-span greenhouse environment through data fusion, real-time feedback collaboration of model predictive control (MPC), and fault diagnosis and tolerance modules. Specifically, the technical solution of this application is as follows: In a first aspect, this application discloses an environmental control system for multi-span greenhouses based on edge-end collaboration, comprising: The terminal sensing and execution layer includes a multimodal sensor network and an intelligent actuator group; the intelligent actuator group includes at least one motor-driven actuator integrated with a current detection module, a position sensor and / or an angle sensor; The edge intelligent control layer, deployed on the edge computing gateway, specifically includes: A multi-source data fusion module is used to fuse the measurement values ​​of the multimodal sensor network and output the optimal estimate of the environmental state. The model prediction control module is used to take the optimal estimated value of the environmental state and the weather forecast as input, solve the optimal control sequence in the future time domain through rolling optimization, and issue control commands to the intelligent actuator group. The fault diagnosis and fault-tolerant control module is used to acquire the current feedback value, position feedback value and / or angle feedback value of the actuator in real time, compare them with the instruction benchmark issued by the model prediction control module, perform fault diagnosis and graded fault tolerance, and feed back the diagnosis results to the model prediction control module in real time.

[0005] In some implementations, the aforementioned edge-to-edge collaborative multi-span greenhouse environmental control system further includes a cloud platform operation and maintenance layer: The cloud platform operation and maintenance layer is communicatively connected to the edge intelligent control layer, and is used to receive environmental data, control records and fault logs reported by the edge intelligent control layer, and to perform self-evolutionary optimization of the internal parameters of the model prediction control module and the judgment threshold of the fault diagnosis and fault tolerance control module through machine learning methods.

[0006] The output of the multi-source data fusion module is connected to the input of the model prediction control module. The control command output of the model prediction control module is connected to the reference input of both the intelligent actuator group and the fault diagnosis and fault tolerance control module. The diagnostic result feedback of the fault diagnosis and fault tolerance control module is connected to the correction input of the model prediction control module, forming a closed-loop autonomous system of "perception-decision-execution-monitoring-fault tolerance".

[0007] In some implementations, the multi-source data fusion module combines multi-source sensor measurements with a greenhouse physics model based on a discrete-time Kalman filter algorithm. The discrete-time Kalman filter algorithm includes: setting temperature and humidity as state variables, and the measured values ​​of each sensor node as observation variables; A state equation is established based on the greenhouse heat and mass balance, and an observation equation describing the mapping relationship between the observed variables and the state variables is constructed. During runtime, the environmental state at the current moment is predicted a priori using the state equation, and then the observed variables are corrected by Kalman gain weighting to output the optimal estimate that integrates the physical model and multi-sensor information.

[0008] In other embodiments, the model predictive control module incorporates a greenhouse dynamic model coupled with multiple environmental factors, and solves for the optimal control sequence within a finite time domain through rolling optimization. The greenhouse dynamic model includes constructing a temperature differential equation: ; ; Among them, T in T represents the indoor air temperature of the greenhouse. out Where is the outdoor air temperature of the greenhouse, S is the solar radiation intensity, and q is the outdoor air temperature of the greenhouse. her q represents the thermal power of the heating system. ven q is the ventilation rate. fog ρ is the water supply rate of the atomization system, ζ is the air density, λ is the specific heat capacity of air, μ is the latent heat of vaporization of water, A is the total heat loss coefficient, V is the greenhouse planar area, and V is the greenhouse volume. The effective transmittance of solar radiation by greenhouse covering materials.

[0009] The model prediction control module is also used to periodically collect weather forecast data as feedforward perturbations for the rolling optimization.

[0010] Optionally, the actuator may include one or more of the following: an electric film rolling device, an external shading net motor, an internal insulation curtain motor, and a circulating fan.

[0011] In some implementations, the fault diagnosis and fault tolerance control module is also used to execute the following fault diagnosis logic, including jamming fault determination and slippage fault determination; When the current feedback value is greater than the first current threshold and the duration of the state exceeds the first time threshold, and the change in the position feedback value and / or the angle feedback value is less than the first change threshold, it is determined to be a stuck fault. When the absolute value of the error between the commanded position and / or commanded angle and the position feedback value and / or the angle feedback value is greater than a first error threshold, and the current feedback value is less than a second current threshold, a slip fault is determined.

[0012] In other embodiments, the fault diagnosis and fault tolerance control module is also used to execute the following hierarchical fault tolerance strategy, including first-level fault tolerance and second-level fault tolerance; Level 1 fault tolerance is for minor faults. The automatic action sequence is as follows: first, run in reverse at the first angle, then wait for the first duration, and then run in forward again, with the number of loops not exceeding the preset number. Level 2 fault tolerance is applicable when Level 1 fault tolerance fails or is directly determined to be a serious fault. In this case, the faulty device is locked and reported to the cloud platform, and the model prediction control module is triggered to perform system-level dynamic reconstruction control.

[0013] In other embodiments, the model prediction control module is also used to: upon receiving a system-level dynamic reconfiguration request triggered by the fault diagnosis and fault tolerance control module, automatically correct the current control strategy and adjust the target setpoints or operating weights of other healthy actuators in the control loop to which the faulty actuator belongs, so as to compensate for the lack of environmental regulation capability due to the faulty actuator, thereby maintaining the stability of the overall greenhouse environment.

[0014] The second invention, this application also discloses a method for environmental control of multi-span greenhouses based on edge-end collaboration, the method being implemented based on the multi-span greenhouse environmental control system based on edge-end collaboration described in any of the above embodiments, and including the following steps: The terminal sensing and execution layer collects greenhouse environmental data and the current feedback value, position feedback value and / or angle feedback value of the actuator; The multi-source data fusion module of the edge intelligent control layer performs Kalman filtering fusion on the multi-sensor data and outputs the optimal estimate of the environmental state. The model predictive control module takes the optimal estimated value and weather forecast as input, solves the model predictive control optimization problem based on the greenhouse dynamic model, generates the optimal control sequence, and sends it to the intelligent actuator group. The fault diagnosis and fault tolerance control module monitors the current feedback value, position feedback value, and / or angle feedback value of the actuator in real time, compares them with the instruction benchmark issued by the model prediction control module, performs fault diagnosis and graded fault tolerance, and feeds back the diagnosis results to the model prediction control module in real time.

[0015] In some embodiments, the method for environmental control of multi-span greenhouses based on edge-end collaboration further includes: The model predictive control module dynamically corrects the optimal control sequence based on the feedback fault diagnosis results. The cloud platform operation and maintenance layer receives environmental data, control records, and fault logs reported by the edge intelligent control layer, and uses machine learning methods to perform self-evolutionary optimization on the internal parameters of the model prediction control module and the judgment threshold of the fault diagnosis and fault tolerance control module.

[0016] Compared with the prior art, this application has at least one of the following beneficial effects: 1. Completely resolves control conflicts and improves control accuracy: This application solves the temperature and humidity coupled control conflict problem at the algorithm level by using MPC multivariate collaborative optimization of the model predictive control module. Simulation comparison shows that under the same disturbance, this application improves the temperature control accuracy from ±2℃ of traditional threshold or PID control to ±0.5℃, and the humidity control accuracy from ±15%RH to ±5%RH, without any drastic fluctuations caused by neglecting one control over the other.

[0017] 2. Significantly improves system reliability: This application achieves real-time detection of early faults such as "jamming" and "slippage" in actuators through hardware improvements and precise fault diagnosis of the fault diagnosis and fault-tolerant control module, with a fault detection rate of ≥95%. The hierarchical fault-tolerant strategy, especially the system-level dynamic reconfiguration control, ensures that the overall system function is not lost when a single actuator fails, and the mean time between failures (MTBF) is expected to be improved by more than 50%.

[0018] 3. Significant Energy Saving and Consumption Reduction: The MPC forward-looking control of the model predictive control module in this application avoids ineffective and conflicting frequent start-ups and shutdowns of equipment in response to sudden disturbances, achieving "on-demand regulation". Compared with traditional control, the overall system energy consumption is expected to be reduced by 15%-20%.

[0019] 4. Achieving unmanned intelligent operation and maintenance: The cloud platform in this application utilizes a self-evolving model based on big data, enabling the system to continuously adapt to local climate and crop varieties and optimize control strategies. Simultaneously, the fault diagnosis and fault-tolerant control module provides precise fault location and tiered alarms, reducing the manpower costs for the operation and maintenance of large-scale greenhouse clusters by approximately 70%. Attached Figure Description

[0020] The preferred embodiments will now be described in a clear and easy-to-understand manner, in conjunction with the accompanying drawings, to further explain the above-mentioned characteristics, technical features, advantages, and implementation methods of this application.

[0021] Figure 1 This is a structural block diagram of an embodiment of an environmental control system for a multi-span greenhouse based on edge-end collaboration according to this application; Figure 2 This is a structural block diagram of another embodiment of an environmental control system for a multi-span greenhouse based on edge-end collaboration according to this application; Figure 3 This is a flowchart illustrating the steps of an embodiment of an environmental control method for multi-span greenhouses based on edge-end collaboration according to this application. Figure 4 This is a flowchart illustrating the steps of another embodiment of the environmental control method for multi-span greenhouses based on edge-end collaboration according to this application. Detailed Implementation

[0022] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application can also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0023] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or sets.

[0024] To keep the drawings concise, each figure only schematically shows the parts relevant to the invention, and these do not represent the actual structure of the product. Furthermore, to facilitate understanding, in some figures, only one of components with the same structure or function is schematically depicted, or only one is labeled. In this document, "one" not only means "only one," but can also mean "more than one."

[0025] It should also be further understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0026] Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0027] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the specific implementation methods of this application will be described below with reference to the accompanying drawings. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings and other implementation methods can be obtained based on these drawings without creative effort.

[0028] Multi-span plastic film greenhouses are currently the main form of facility agriculture. However, existing environmental control technologies reveal the following fundamental shortcomings when dealing with complex, coupled, and nonlinear environmental models: 1. The control strategy is too simplistic, leading to conflicting outcomes: Most existing systems are based on threshold-triggered logic for a single factor. For example, a simple action like opening a skylight to cool down the room may cause a sudden and drastic drop in indoor humidity, stressing moisture-loving crops; or closing ventilation to increase temperature may result in excessively low CO2 concentrations. This strategy of controlling tightly coupled parameters such as temperature, humidity, and CO2 in a fragmented manner is the core reason why current systems negatively impact crop quality and yield.

[0029] 2. "Blind" operation of actuators leads to poor system reliability: Actuators such as electric film rolling machines and shade nets often fail during operation due to mechanical deformation, foreign object jamming, and motor aging. Current technology lacks precise perception of the real-time physical state of the actuators (such as torque and position), causing the system to issue commands "blindly." This can lead to equipment overload and burnout at best, and structural damage at worst due to uncontrolled film rolling, failing to meet the stringent requirements of modern agriculture for "≥95% operational reliability" and "no jamming."

[0030] 3. Short-sighted control decisions and low energy efficiency: Traditional PID or simple logic control only makes reactive adjustments based on the current deviation and cannot predict future environmental changes (such as sudden cloud formations or temperature drops), resulting in frequent start-stop of actuators and large overshoot, which not only shortens the life of equipment but also causes significant energy waste.

[0031] In summary, while existing technologies have initially constructed intelligent control frameworks based on "cloud-edge-device" or "perception-decision-execution," their core focus remains on optimizing the "environment-crop" model to generate better control commands. This assumes that the actuators are idealized and can always execute commands accurately. However, in actual multi-span greenhouse production, physical failures of the actuators (especially electric film rollers), such as jamming and slippage, are the primary cause of control failures, equipment damage, and even production accidents. Existing technologies completely lack the ability to perceive the physical state (torque, position) of the actuators in real time, and have not established a fault-tolerant mechanism for "how the system autonomously maintains environmental stability when the actuators malfunction."

[0032] Therefore, there is an urgent need in this field for an integrated solution capable of forward-looking collaborative decision-making, with self-awareness and fault tolerance of actuator status, to fundamentally solve control conflicts and reliability problems. The technical problem to be solved in this application not only includes multi-factor collaborative control, but more importantly, it is about how to construct a highly reliable closed-loop control system that can sense the physical state of the actuator in real time, autonomously diagnose faults, and still ensure the basic stability of the greenhouse environment through system-level control strategy reconfiguration when the actuator partially fails.

[0033] Reference manual attached Figure 1 As shown, one embodiment of the multi-span greenhouse environmental control system based on edge-end collaboration of this application specifically includes: The terminal sensing and execution layer comprises a multimodal sensor network and an intelligent actuator group. The intelligent actuator group includes at least one motor-driven actuator integrating a current detection module, a position sensor, and / or an angle sensor.

[0034] The multimodal sensor network involves deploying multiple sensor nodes in different greenhouse zones, integrating sensors for temperature, humidity, light, and CO2. The system collects noisy measurements from all sensors at a fixed frequency (e.g., once every 10 seconds) and transmits this data along with timestamps to the edge computing gateway.

[0035] Optionally, the actuator includes one or more of the following: an electric film roller, an external shading net motor, an internal insulation curtain motor, and a circulating fan. This application uses an electric film roller as a representative actuator to describe the device and method in detail. It should be understood that the technical solution of this application is also applicable to other types of actuators, including but not limited to external shading net motors, internal insulation curtain motors, and circulating fans. These actuators are similar in working principle to or substantially similar to the electric film roller. Those skilled in the art, based on the content described in this application, can easily apply them to the above-mentioned or other conceivable actuators, and all should fall within the protection scope of this application.

[0036] The key improvement lies in integrating a high-precision current detection module (such as a Hall sensor) and an absolute angle encoder into the electric film winding machine, for real-time monitoring of the motor operating current and the absolute position of the film winding shaft. This forms the hardware foundation for implementing the fault diagnosis and fault-tolerant control module described below.

[0037] For example, the actuator includes an electric film winder, which integrates a current detection module for real-time monitoring of the motor's operating current and an absolute angle encoder for real-time monitoring of the absolute position of the film winding shaft. Optionally, the current detection module is a Hall sensor with a range of 0-5A and an accuracy of ±1%. The absolute angle encoder has a resolution of 1024 PPR, providing real-time current and position physical status data for the fault diagnosis and tolerance module.

[0038] The edge intelligent control layer, deployed on the edge computing gateway, specifically includes: The multi-source data fusion module is used to fuse the measurement values ​​of the multimodal sensor network and output the optimal estimate of the environmental state.

[0039] The model predictive control module is used to take the optimal estimated value of the environmental state and the weather forecast as input, solve for the optimal control sequence in the future time domain through rolling optimization, and issue control commands to the intelligent actuator group.

[0040] The fault diagnosis and fault-tolerant control module is used to acquire the current feedback value, position feedback value and / or angle feedback value of the actuator in real time, compare them with the instruction benchmark issued by the model prediction control module, perform fault diagnosis and graded fault tolerance, and feed back the diagnosis results to the model prediction control module in real time.

[0041] Specifically, the output of the multi-source data fusion module is connected to the input of the model prediction control module. The control command output of the model prediction control module is simultaneously connected to the reference input of the intelligent actuator group and the fault diagnosis and fault tolerance control module. The diagnostic result feedback of the fault diagnosis and fault tolerance control module is connected to the correction input of the model prediction control module, forming a closed-loop autonomous system of "perception-decision-execution-monitoring-fault tolerance".

[0042] Based on the above embodiments, this application discloses another embodiment of a multi-span greenhouse environmental control system based on edge-end collaboration, as detailed in the appendix to the specification. Figure 2 As shown, it also includes the cloud platform operations and maintenance layer: The cloud platform operation and maintenance layer is communicatively connected to the edge intelligent control layer, and is used to receive environmental data, control records and fault logs reported by the edge intelligent control layer, and to perform self-evolutionary optimization of the internal parameters of the model prediction control module and the judgment threshold of the fault diagnosis and fault tolerance control module through machine learning methods.

[0043] Specifically, in this embodiment, the system consists of three parts: a terminal perception and execution layer, an edge intelligent control layer, and a cloud platform operation and maintenance layer, communicating via 4G / 5G or LoRa networks. The terminal perception and execution layer includes a multimodal sensor network and a group of intelligent actuators. The edge intelligent control layer, deployed on the field edge computing gateway, is the core of this application, running three major algorithm modules. The cloud platform operation and maintenance layer provides remote visualization, data storage, and algorithm self-evolution capabilities.

[0044] In other implementations, the cloud platform's operations and maintenance layer uses digital twin technology to achieve remote 3D visualization monitoring. It collects massive amounts of "environment-control-fault" data, utilizes machine learning (such as LSTM and reinforcement learning) to self-optimize the MPC model parameters and fault diagnosis thresholds of the edge gateway, and then distributes the optimized model to the edge gateway to achieve continuous system evolution.

[0045] The core innovation of this application lies not in the "cloud-edge-device" three-layer architecture itself, but in the specific settings and unique closed-loop collaborative workflow of the three core algorithm modules deployed within the edge intelligent control layer: the output of the multi-source data fusion module serves as the input to the model predictive control module. The control commands generated by the model predictive control module drive the actuators, while its command values ​​serve as a benchmark reference for the fault diagnosis and fault tolerance control module. The diagnostic results of the fault diagnosis and fault tolerance control module are fed back to the model predictive control module in real time for online correction of the control strategy or triggering fault-tolerant actions.

[0046] This application provides another embodiment of an environmental control system for multi-span greenhouses based on edge-end collaboration. Based on any of the above embodiments, the multi-source data fusion module combines the multi-source sensor measurements with the greenhouse physical model based on the discrete-time Kalman filter algorithm.

[0047] The discrete-time Kalman filter algorithm includes: setting temperature and humidity as state variables, and the measured values ​​of each sensor node as observation variables.

[0048] A state equation is established based on the greenhouse heat and mass balance, and an observation equation describing the mapping relationship between the observed variables and the state variables is constructed.

[0049] During runtime, the environmental state at the current moment is predicted a priori using the state equation, and then the observed variables are corrected by Kalman gain weighting to output the optimal estimate that integrates the physical model and multi-sensor information.

[0050] Specifically, the multi-source data fusion module, based on an improved Kalman filter algorithm, receives noisy measurements from all sensors and, by combining them with a simplified physical model of the greenhouse, outputs a more spatially uniform and temporally smoother "virtual optimal estimate".

[0051] Algorithm sets state variable X k = [T k H k ] T ; Observed variable Z k = [T1, T2, …, H11, H2, …] T .

[0052] Define the state equation: X k = F·X k-1 + W k ; Observation equation: Z k = C·X k + V k .

[0053] Where T is the indoor temperature, H is the indoor humidity, T1, T2, …, H1, H2, … are the measurements from each sensor, F is the state transition matrix constructed based on the greenhouse heat-mass balance, and C is the observation matrix that maps the real state to the measurement space of each sensor. k ~N(0,Q) represents process noise, V k ~N(0,R) represents the observation noise.

[0054] In some alternative implementations, the multi-source data fusion module performs the following methods, including: A simplified physical model of the greenhouse is established. Specifically, a simplified physical model is constructed to describe the dynamic changes in temperature and humidity inside the greenhouse. This greenhouse physical model is used to express the influence of factors such as solar radiation, outdoor temperature, and ventilation rate on indoor temperature and humidity.

[0055] Define state variables and observation variables. Specifically, define the key parameters that need to be estimated in the greenhouse environment system as state variables, including indoor temperature and indoor humidity. Acquire data from a multimodal sensor network, optionally including temperature, humidity, light, and carbon dioxide sensor data. Define the acquired values ​​as observation variables; each node integrates a temperature sensor and a humidity sensor, and the measurements from all nodes together constitute the observation variables.

[0056] The state equation and observation equation are constructed. Specifically, the state equation reflects the transition relationship between the state variables at the previous moment and the state variables at the current moment, reflecting the natural changing trend of the greenhouse environment when there is no external observation input. The observation equation describes the mapping relationship between the measurement values ​​of each sensor and the actual environmental state.

[0057] The statistical characteristics of process noise and observation noise are defined. Through experimental calibration, the values ​​of the process noise covariance matrix in the state equation and the observation noise covariance matrix in the observation equation are determined, enabling the filter to perform optimal weighted fusion between predicted and measured values. Optionally, after experimental calibration, the process noise covariance matrix Q = 0.01, and the observation noise covariance matrix R = 0.1.

[0058] Perform Kalman filter iterative calculations. Within each sampling period, execute the state prediction step and the state update step sequentially. The state prediction step predicts the prior state estimate and its covariance at the current time step based on the optimal estimate and state equation from the previous time step. The state update step uses the actual measured values ​​of the observed variables at the current time step to calculate the Kalman gain and corrects the prior state estimate to obtain the posterior state estimate at the current time step, which is the required virtual optimal estimate.

[0059] The multi-source data fusion module in this embodiment can not only filter out noise, but also estimate the virtual value of the area not covered by the sensor through the physical model, providing a more reliable control basis for the model predictive control module than any single sensor.

[0060] This application provides another embodiment of an environmental control system for multi-span greenhouses based on edge-end collaboration. Based on any of the above embodiments, the model predictive control module incorporates a greenhouse dynamic model coupled with multiple environmental factors, and solves for the optimal control sequence in the future finite time domain through rolling optimization.

[0061] Specifically, the model predictive control module receives the "optimal estimate" from the multi-source data fusion module, user-defined values, and weather forecasts. Through rolling optimization, it solves for the optimal control sequence in the future, fundamentally avoiding control conflicts.

[0062] The greenhouse dynamic model includes constructing a temperature differential equation: ; ; Among them, T in T represents the indoor air temperature of the greenhouse. out Where is the outdoor air temperature of the greenhouse, S is the solar radiation intensity, and q is the outdoor air temperature of the greenhouse. her q represents the thermal power of the heating system. ven q is the ventilation rate. fog ρ is the water supply rate of the atomization system, ζ is the air density, λ is the specific heat capacity of air, μ is the latent heat of vaporization of water, A is the total heat loss coefficient, V is the greenhouse planar area, and V is the greenhouse volume. In practical implementation, the module establishes a temperature differential equation, the physical meaning of which is: the rate of change of greenhouse temperature = - heat dissipation from the building envelope - sensible heat loss from ventilation - atomization evaporation cooling + active heating + solar radiation + outdoor temperature compensation + sensible heat exchange from ventilation.

[0063] This greenhouse dynamic model reflects the interaction of multiple factors in greenhouse thermal dynamics and is the core foundation of the Model Predictive Control (MPC) module.

[0064] Specifically, in this greenhouse temperature dynamic model, the meanings of each symbol are as follows: T in The temperature is the indoor air temperature of the greenhouse, expressed in °C.

[0065] T out The outdoor air temperature of the greenhouse is expressed in °C.

[0066] S represents solar radiation intensity, measured in W / m². 2 .

[0067] qher This represents the thermal power of the heating system, measured in W.

[0068] q ven Ventilation rate, unit: m 3 / s.

[0069] q fog The water supply rate of the atomization system is expressed in g / s.

[0070] ρ is the air density, with units of kg / m³. 3 .

[0071] ζ is the specific heat capacity of air, expressed in J / (kg·K).

[0072] λ is the latent heat of vaporization of water, expressed in J / g.

[0073] μ is the total heat loss coefficient, with units of W / K.

[0074] A represents the greenhouse floor area, in meters (m²). 2 .

[0075] V represents the volume of the greenhouse, in meters (m³). 3 .

[0076] The effective transmittance of solar radiation by greenhouse covering materials.

[0077] In other embodiments, the model prediction control module is also used to periodically collect weather forecast data as a feedforward perturbation for the rolling optimization.

[0078] Specifically, in each control cycle (e.g., 5 minutes), an open-loop optimization problem with a finite time domain (e.g., a 30-minute prediction time domain) is solved. The objective function is to minimize the deviation of environmental parameters from the setpoint and the change in the control variable. The key is to introduce short-term weather forecasts (temperature, light intensity, wind speed) obtained from the cloud platform as feedforward disturbances into the model to achieve forward-looking control.

[0079] Output: Outputs a set of optimal control sequences, such as "Within the next 30 minutes, the sunroof opening will gradually increase from 20% to 40%, and the external sunshade will simultaneously expand to 50%." The system only executes the optimal command for the current moment.

[0080] This application provides another embodiment of a multi-span greenhouse environmental control system based on edge-end collaboration, which is based on any of the embodiments of the above system. The fault diagnosis and fault tolerance control module is also used to execute the following fault diagnosis logic, including jamming fault determination and slippage fault determination.

[0081] When the current feedback value is greater than the first current threshold and the duration of the state exceeds the first time threshold, and the change in the position feedback value and / or the angle feedback value is less than the first change threshold, it is determined to be a jamming fault.

[0082] When the absolute value of the error between the commanded position and / or commanded angle and the position feedback value and / or the angle feedback value is greater than a first error threshold, and the current feedback value is less than a second current threshold, a slip fault is determined.

[0083] Specifically, the fault diagnosis and fault-tolerant control module monitors the current and angle data of the electric film winding machine in real time and compares them with the command benchmark issued by the model prediction control module to achieve accurate fault diagnosis and graded fault tolerance.

[0084] In practice, the fault diagnosis and fault tolerance control module executes the following method flow: Establish a benchmark model. Develop a three-dimensional "current-angle-time" curve model for each actuator under normal operating conditions (e.g., normal current range 0.5-2A, angle change rate 0.5-2° / s).

[0085] Quantitative diagnostic logic. Pre-set threshold conditions for current, position, and angle.

[0086] Example: Stuck fault: If the current I > 3A (threshold) and the duration Δt > 5 seconds, while the angle change Δθ < 1°, it is determined to be stuck.

[0087] Slip fault: If the error between the command angle and the feedback angle is |θ cmd -θ feedback If the angle is greater than 10° and the current I is less than 0.3A, then it is determined to be slippage.

[0088] In some alternative implementations, the fault diagnosis and fault tolerance control module is also used to execute the following hierarchical fault tolerance strategy, including first-level fault tolerance and second-level fault tolerance.

[0089] Level 1 fault tolerance is for minor faults. The automatic action sequence is as follows: first, run in reverse at the first angle, then wait for the first duration, and then run in the forward direction again. The number of loops does not exceed the preset number.

[0090] Level 2 fault tolerance is applicable when Level 1 fault tolerance fails or is directly determined to be a serious fault. In this case, the faulty device is locked and reported to the cloud platform, and the model prediction control module is triggered to perform system-level dynamic reconstruction control.

[0091] Example: Level 1 fault tolerance (minor fault): Automatically executes a "back-forward" strategy. That is, reverses the direction by 5° → waits for 10 seconds → resumes forward operation to attempt to eliminate the bottleneck. This process is repeated no more than 3 times.

[0092] Level 2 fault tolerance (critical fault): If Level 1 fault tolerance fails, the device will be immediately locked and the issue reported to the cloud platform. Simultaneously, system-level dynamic reconfiguration control will be triggered.

[0093] Based on the above embodiments, this application provides another embodiment of an environmental control system for multi-span greenhouses based on edge-end collaboration. The model predictive control module is further configured to: automatically correct the current control strategy after receiving a system-level dynamic reconfiguration request triggered by the fault diagnosis and fault-tolerant control module, and adjust the target setpoints or operating weights of other healthy actuators in the control loop to which the faulty actuator belongs, so as to compensate for the lack of environmental regulation capability due to the faulty actuator, thereby maintaining the stability of the overall greenhouse environment.

[0094] For example, if the skylight gets stuck, the model predictive control module automatically corrects the strategy, increases the opening of the side windows and starts the circulating fan, using system-level means to compensate for the ventilation volume and ensure that the crop environment does not get out of control.

[0095] Based on the same concept, this application also discloses a method for environmental control of multi-span greenhouses based on edge-end collaboration. The implementation of this method is based on the multi-span greenhouse environmental control system based on edge-end collaboration described in any of the above embodiments. Specifically, an embodiment of the multi-span greenhouse environmental control method based on edge-end collaboration of this application is provided in the appendix to the specification. Figure 3 As shown, it specifically includes: S1, the greenhouse environment data and the current feedback value, position feedback value and / or angle feedback value of the actuator are collected through the terminal sensing and execution layer.

[0096] S2, the multi-source data fusion module of the edge intelligent control layer performs Kalman filtering fusion on the multi-sensor data and outputs the optimal estimate of the environmental state.

[0097] S3, the model predictive control module takes the optimal estimated value and weather forecast as input, solves the model predictive control optimization problem based on the greenhouse dynamic model, generates the optimal control sequence, and sends it to the intelligent actuator group.

[0098] S4, the fault diagnosis and fault tolerance control module monitors the current feedback value, position feedback value and / or angle feedback value of the actuator in real time, compares them with the instruction benchmark issued by the model prediction control module, performs fault diagnosis and graded fault tolerance, and feeds back the diagnosis results to the model prediction control module in real time.

[0099] Based on the above embodiments, this application discloses another embodiment of a multi-span greenhouse environmental control method based on edge-end collaboration, as detailed in the appendix to the specification. Figure 4 As shown, it also includes: S5, the model prediction control module dynamically corrects the optimal control sequence based on the feedback fault diagnosis results.

[0100] S6. The cloud platform operation and maintenance layer receives environmental data, control records and fault logs reported by the edge intelligent control layer, and performs self-evolutionary optimization on the internal parameters of the model prediction control module and the judgment threshold of the fault diagnosis and fault tolerance control module through machine learning.

[0101] This application discloses another embodiment of an environmental control method for multi-span greenhouses based on edge-end collaboration. Taking a GSW8440 type multi-span greenhouse (area 3000㎡) as an example, the specific implementation method of this application is described in detail to fully disclose the technical solution.

[0102] 1. Hardware deployment.

[0103] Sensor Network: Based on the greenhouse structure, 12 sensor nodes are evenly deployed to ensure that environmental parameters are spatially covered without any blind spots.

[0104] Intelligent actuators: Hall effect current sensors (range 0-5A, accuracy ±1%) and absolute rotary encoders (resolution 1024PPR) are installed on all electric film roll-up units for sunroofs and side windows.

[0105] Edge computing gateway: Install an industrial-grade edge computing gateway, such as the NVIDIA Jetson Orin NX, in the control room. It has a 6-core CPU and 8GB RAM to meet the real-time operation requirements of the three major algorithm modules.

[0106] Communication network: The edge gateway connects to the cloud platform via 4G / 5G, and the sensors communicate with the gateway via LoRa.

[0107] 2. Software configuration and running examples.

[0108] Software environment: The gateway runs on the Linux operating system. The core algorithm multi-source data fusion module and fault diagnosis and fault-tolerant control module are written in C++ to ensure real-time performance. The MPC optimization problem of the model prediction control module is solved using Python and the CVXPY optimization library.

[0109] Parameter settings: Set the MPC control cycle to 5 minutes and the prediction time domain to 30 minutes.

[0110] Operational Example: One midday, sunlight suddenly increased. The multi-source data fusion module combined data from various points and output the "optimal estimate" of the temperature reaching 28℃. The MPC controller of the model prediction control module, based on the prediction model and the 30-minute weather forecast (no rain, continuous sunlight), calculated the optimal coordinated sequence of "opening the skylight to 40% and simultaneously deploying the external shading to 60%," avoiding excessively low humidity caused by simply opening the window. After the command was issued, the fault diagnosis and fault tolerance control module monitored in real time that the motor current of one skylight was consistently >3A and the angle change within 5 seconds was <1°, immediately identifying it as a slight jamming. The fault diagnosis and fault tolerance control module automatically executed the first-level fault tolerance: "retreat 5° → wait 10 seconds → proceed again," successfully eliminating the fault. Fault diagnosis was completed within milliseconds; the first-level fault tolerance action was completed within approximately 15 seconds, with no impact on the greenhouse environment, and an automatic log message "Slight jamming of skylight No. 3 has been resolved" was reported to the cloud platform.

[0111] 3. Cloud platform self-evolution.

[0112] After collecting data on the "stuck" events and the long-term operation of the skylight, the cloud platform used machine learning analysis to discover that the skylight was prone to stuck when humidity exceeded 80% and the opening degree exceeded 70%. The platform automatically optimized the MPC control weights for the skylight in the model predictive control module and distributed the updated model to the edge gateway. Subsequently, when the system predicts high humidity conditions, it prioritizes increasing the side window opening to prevent the skylight from entering the high-risk zone, thus achieving continuous performance evolution of the system.

[0113] Experimental scenario: A simulated malfunction occurs when the electric film reel of a sunroof is slightly jammed due to the entry of a foreign mechanical object.

[0114] Compared to existing technologies: The decision engine continuously sends "open to 50%" commands to the faulty skylight based on rising ambient temperature. Because the system cannot detect abnormal motor current and unchanged angle, the commands are "blindly" repeated. The motor remains stalled, causing a sharp rise in current, and after approximately 30 seconds, it either triggers overload protection (if applicable) or burns out. Because the system receives no feedback, it continues to attempt to open the skylight, leading to its complete failure. The greenhouse temperature rises rapidly, and other actuators (such as side windows) cannot compensate because they are not linked, causing heat stress to the crops.

[0115] When a fault occurs, the fault diagnosis and fault-tolerant control module monitors in real time that the current is consistently >3A and the angle change is <1°. Within 5 seconds, it accurately determines that it is a minor jamming and immediately executes the first-level fault-tolerant strategy: "Reverse operation 5° → Wait 10 seconds → Resume forward operation." The jammed foreign object is released during the reversal, the fault is eliminated, and the skylight opens normally. The entire process takes approximately 20 seconds and has no significant impact on the greenhouse environment. The system also automatically reports the event to the log, and the cloud platform records the event.

[0116] Experimental conclusions: Comparative analysis demonstrates that this application, through combined current-position sensing and hierarchical fault-tolerant control, not only avoids physical damage to the actuator but also ensures the continuity and stability of greenhouse environment control under fault disturbances. Existing technologies, lacking this core mechanism, inevitably lead to control failure under the same fault scenarios.

[0117] Through the above specific implementation methods, those skilled in the art can clearly understand how this application solves the long-standing problem of precise control of multi-span greenhouses through two core innovations: "closed-loop collaboration of three major algorithm modules" and "fault-tolerant control based on current-position joint sensing".

[0118] The environmental control system and method for multi-span greenhouses based on edge-end collaboration in this application have the same technical concept, and the technical details of the embodiments of the two are applicable to each other. To reduce repetition, they will not be repeated here.

[0119] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of program modules is merely an example. In practical applications, the above functions can be assigned to different program modules as needed, that is, the internal structure of the device can be divided into different program units or modules to complete all or part of the functions described above. The program modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software program unit. Furthermore, the specific names of the program modules are only for easy differentiation and are not intended to limit the scope of protection of this application.

[0120] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

Claims

1. A multi-span greenhouse environmental control system based on edge-end collaboration, characterized in that, include: The terminal sensing and execution layer includes a multimodal sensor network and an intelligent actuator group; the intelligent actuator group includes at least one motor-driven actuator integrated with a current detection module, a position sensor and / or an angle sensor; The edge intelligent control layer, deployed on the edge computing gateway, specifically includes: A multi-source data fusion module is used to fuse the measurement values ​​of the multimodal sensor network and output the optimal estimate of the environmental state; The model prediction control module is used to take the optimal estimated value of the environmental state and the weather forecast as input, solve the optimal control sequence in the future time domain through rolling optimization, and issue control commands to the intelligent actuator group. The fault diagnosis and fault-tolerant control module is used to acquire the current feedback value, position feedback value and / or angle feedback value of the actuator in real time, compare them with the instruction benchmark issued by the model prediction control module, perform fault diagnosis and graded fault tolerance, and feed back the diagnosis results to the model prediction control module in real time.

2. The environmental control system for multi-span greenhouses based on edge-end collaboration as described in claim 1, characterized in that, This also includes the cloud platform operations and maintenance layer: The cloud platform operation and maintenance layer is communicatively connected to the edge intelligent control layer, and is used to receive environmental data, control records and fault logs reported by the edge intelligent control layer, and to perform self-evolutionary optimization of the internal parameters of the model prediction control module and the judgment threshold of the fault diagnosis and fault tolerance control module through machine learning methods.

3. The environmental control system for multi-span greenhouses based on edge-end collaboration as described in claim 1, characterized in that, The multi-source data fusion module is based on the discrete-time Kalman filter algorithm and combines the measurements from multiple sensor sources with the greenhouse physical model. The discrete-time Kalman filter algorithm includes: setting temperature and humidity as state variables, and the measured values ​​of each sensor node as observation variables; A state equation is established based on the greenhouse heat and mass balance, and an observation equation describing the mapping relationship between the observed variables and the state variables is constructed. During runtime, the environmental state at the current moment is predicted a priori using the state equation, and then the observed variables are corrected by Kalman gain weighting to output the optimal estimate that integrates the physical model and multi-sensor information.

4. The environmental control system for multi-span greenhouses based on edge-end collaboration as described in claim 1, characterized in that, The model predictive control module incorporates a greenhouse dynamic model coupled with multiple environmental factors, and solves for the optimal control sequence within a finite time domain through rolling optimization. The greenhouse dynamic model includes constructing a temperature differential equation: ； ； Among them, T in T represents the indoor air temperature of the greenhouse. out Where is the outdoor air temperature of the greenhouse, S is the solar radiation intensity, and q is the outdoor air temperature of the greenhouse. her q represents the thermal power of the heating system. ven q is the ventilation rate. fog ρ is the water supply rate of the atomization system, ζ is the air density, λ is the specific heat capacity of air, μ is the latent heat of vaporization of water, A is the total heat loss coefficient, V is the greenhouse planar area, and V is the greenhouse volume. The effective transmittance of solar radiation by greenhouse covering materials; The model prediction control module is also used to periodically collect weather forecast data as feedforward perturbations for the rolling optimization.

5. The environmental control system for multi-span greenhouses based on edge-end collaboration as described in claim 1, characterized in that, The actuator includes one or more of the following: an electric film rolling device, an external shading net motor, an internal insulation curtain motor, and a circulating fan.

6. The environmental control system for multi-span greenhouses based on edge-end collaboration as described in claim 1, characterized in that, The fault diagnosis and fault tolerance control module is also used to execute the following fault diagnosis logic, including jamming fault determination and slippage fault determination; When the current feedback value is greater than the first current threshold and the duration of the state exceeds the first time threshold, and the change in the position feedback value and / or the angle feedback value is less than the first change threshold, it is determined to be a stuck fault. When the absolute value of the error between the commanded position and / or commanded angle and the position feedback value and / or the angle feedback value is greater than a first error threshold, and the current feedback value is less than a second current threshold, a slip fault is determined.

7. The environmental control system for multi-span greenhouses based on edge-end collaboration as described in claim 1, characterized in that, The fault diagnosis and fault tolerance control module is also used to execute the following hierarchical fault tolerance strategy, including first-level fault tolerance and second-level fault tolerance; Level 1 fault tolerance is for minor faults. The automatic action sequence is as follows: first, run in reverse at the first angle, then wait for the first duration, and then run in the forward direction again, with the number of loops not exceeding the preset number. Level 2 fault tolerance is applicable when Level 1 fault tolerance fails or is directly determined to be a serious fault. In this case, the faulty device is locked and reported to the cloud platform, and the model prediction control module is triggered to perform system-level dynamic reconstruction control.

8. The multi-span greenhouse environmental control system based on edge-end collaboration as described in claim 7, characterized in that, The model prediction control module is also used to: upon receiving a system-level dynamic reconfiguration request triggered by the fault diagnosis and fault tolerance control module, automatically correct the current control strategy and adjust the target setpoints or operating weights of other healthy actuators in the control loop to which the faulty actuator belongs, so as to compensate for the lack of environmental regulation capability due to the faulty actuator, thereby maintaining the stability of the overall greenhouse environment.

9. A method for environmental control of multi-span greenhouses based on edge-end collaboration, characterized in that, The method is implemented based on the edge-to-end collaborative multi-span greenhouse environmental control system according to any one of claims 1-8, and includes the following steps: The terminal sensing and execution layer collects greenhouse environmental data and the current feedback value, position feedback value and / or angle feedback value of the actuator; The multi-source data fusion module of the edge intelligent control layer performs Kalman filtering fusion on multi-sensor data and outputs the optimal estimate of the environmental state. The model predictive control module takes the optimal estimated value and weather forecast as input, solves the model predictive control optimization problem based on the greenhouse dynamic model, generates the optimal control sequence, and sends it to the intelligent actuator group. The fault diagnosis and fault tolerance control module monitors the current feedback value, position feedback value, and / or angle feedback value of the actuator in real time, compares them with the instruction benchmark issued by the model prediction control module, performs fault diagnosis and graded fault tolerance, and feeds back the diagnosis results to the model prediction control module in real time.

10. The environmental control method for multi-span greenhouses based on edge-end collaboration as described in claim 9, characterized in that, Also includes: The model predictive control module dynamically corrects the optimal control sequence based on the feedback fault diagnosis results. The cloud platform operation and maintenance layer receives environmental data, control records, and fault logs reported by the edge intelligent control layer, and uses machine learning methods to perform self-evolutionary optimization on the internal parameters of the model prediction control module and the judgment threshold of the fault diagnosis and fault tolerance control module.

Images