Micro-agriculture box multi-sensor fusion adaptive control method and system

Abstract

This invention discloses a multi-sensor fusion adaptive control method and system for a micro-agricultural box. The method simultaneously collects three types of data: environmental, cultivation medium, and crop images. It utilizes an improved unscented Kalman filter algorithm for multi-source data fusion and multi-parameter coupling calculation, combined with a BP neural network-optimized adaptive PID controller to achieve closed-loop control. The system also includes an image recognition dynamic correction unit and a fault diagnosis and tolerance unit. This invention addresses the pain points of existing micro-planting equipment, such as low control precision, large parameter coupling interference, inability to adapt to the entire growth cycle, and poor reliability for unattended operation. This invention improves temperature and humidity control precision to ±0.5℃ / ±3%RH, reduces pest and disease incidence by more than 95%, and achieves pesticide-free, unmanned adaptive planting, suitable for home and commercial micro-agricultural scenarios.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology for smart agriculture, specifically relating to a multi-sensor fusion adaptive control method and system for a micro agricultural box. Background Technology

[0002] With urban residents' increasing demand for pesticide-free, freshly picked and eaten ingredients, family mini agricultural boxes and commercial planting cabinets for restaurants have become emerging niche markets in smart agriculture.

[0003] Existing control schemes for micro-planting equipment mainly fall into two categories: one is fixed-threshold open-loop control, which only achieves simple on / off control of a single parameter; the other is traditional PID closed-loop control, which was developed for large-scale farm greenhouses and then simplified for use in micro-equipment. These schemes suffer from the following core drawbacks:

[0004] First, the control precision is insufficient. The existing solution only collects a single environmental parameter and does not consider the strong coupling relationship between temperature, humidity, CO2, light, and nutrient solution parameters (such as a 2%-3% decrease in humidity corresponding to a 1°C increase in temperature). This results in large control deviations, poor microenvironment stability, and an susceptibility to pests and diseases.

[0005] Second, it lacks adaptability. The existing solution uses fixed parameters, which cannot adapt to the different needs of different crops (leafy vegetables / spices / mushrooms), different growth cycles (germination period / seedling period / growth period / harvest period), or the changes in external environment between northern and southern regions and between winter and summer.

[0006] Third, there is no fault tolerance mechanism. The existing solution lacks sensor fault diagnosis and emergency control capabilities. When home users are away on business trips or restaurants are not managed by dedicated personnel, even minor malfunctions of sensors or actuators can lead to large-scale crop necrosis.

[0007] Fourth, the algorithm is incompatible with the hardware. Existing greenhouse control algorithms have high computing power requirements and high hardware costs, making them unsuitable for low-cost MCUs in micro-devices.

[0008] A search revealed that no complete technical solution described in this invention has been disclosed in the prior art, thus this invention possesses novelty and inventiveness. Summary of the Invention

[0009] This invention aims to solve the problems of low control accuracy, lack of full-cycle adaptability and poor reliability in the prior art, and provides a multi-sensor fusion adaptive control method and system for micro agricultural boxes.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] A multi-sensor fusion adaptive control method for a micro-agricultural box includes the following steps: S1: Obtaining information on the crop variety selected by the user, and retrieving the optimal environmental parameter matrix corresponding to the initial growth cycle of the crop from a pre-stored crop growth model library as the initial control target value; S2: Synchronously collecting multi-source sensor data of the target planting space at a preset frequency, including environmental parameter data, cultivation medium parameter data, and crop growth status image data; sequentially performing outlier removal, filtering and noise reduction, and normalization processing on the collected multi-source sensor data to obtain a standardized sensor dataset; S3: Using an improved unscented Kalman filter algorithm to perform multi-source data fusion on the standardized sensor dataset, establishing a multi-parameter coupling relationship model, calculating the mutual influence coefficients between each control parameter, and correcting the initial control target value based on the mutual influence coefficients to obtain the optimal dynamic control target value at the current moment; 4. Using the deviation between the optimal dynamic control target value and the real-time standardized sensor data as input, an adaptive PID controller optimized by a BP neural network outputs the corresponding control signal to the actuator, driving the actuator to complete closed-loop regulation; S5. Based on crop growth status image data, an image recognition algorithm identifies the actual growth cycle and health status of the crop. When the deviation between the actual growth status and the preset growth model exceeds a preset threshold, the parameter matrix of the crop growth model is corrected, and the optimal dynamic control target value is updated; S6. The sensor data and the feedback signal of the actuator are monitored in real time to determine whether there is a equipment fault. If a fault is determined, the system automatically switches to the preset fault-tolerant control mode and sends a fault warning message to the user terminal; S7. Real-time planting data and control data are uploaded to the cloud platform. The cloud platform optimizes the crop growth model library based on big data analysis and updates the optimized model to the local device.

[0012] A multi-sensor fusion adaptive control system for a micro-agricultural box includes: a multi-sensor acquisition module for synchronously acquiring environmental parameter data, cultivation medium parameter data, and crop growth status image data of the target planting space; a core control module, communicatively connected to the multi-sensor acquisition module, for preprocessing sensor data, multi-sensor fusion, adaptive control decision-making, fault diagnosis, and crop growth model management; an actuator module, electrically connected to the core control module, for receiving control signals from the core control module and executing corresponding environmental and medium adjustment operations; a human-machine interaction module, communicatively connected to the core control module, for allowing users to select crop varieties, view real-time data, receive early warning information, and remotely control the equipment; and a cloud platform module, communicatively connected to the core control module via a wireless network, for storing planting data, optimizing the crop growth model library, and enabling batch management of multiple devices.

[0013] Compared with the prior art, the present invention has the following beneficial effects: (1) Significantly improved control accuracy: Through multi-sensor fusion and decoupling, the temperature and humidity control accuracy is improved to ±0.5℃ / ±3%RH, the incidence of pests and diseases is reduced by more than 95%, and zero-pesticide planting is achieved. (2) Full-scene adaptation: Built-in crop growth model library + image recognition dynamic correction, automatically adapts to different crops, full growth cycle, different regions and seasons, without the need for professional agricultural knowledge. (3) High reliability: Fault tolerance mechanism ensures crop safety when equipment is abnormal, and has strong anti-interference ability. (4) Low cost and high adaptability: The algorithm can run on low-cost STM32 series MCUs, and the hardware cost is reduced by more than 60% compared with traditional greenhouse control systems. Attached Figure Description

[0014] Figure 1 This is a block diagram of the overall architecture of the adaptive control system described in this invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of protection of this invention.

[0016] Example 1: Hardware Configuration. The core control module uses an STM32F407 MCU; the multi-sensor acquisition module includes an SHT30 air temperature and humidity sensor, an MH-Z19E CO2 sensor, a BH1750 light sensor, a nutrient solution EC / pH sensor, and an OV2640 miniature camera; the actuators include a full-spectrum LED supplemental light, a semiconductor cooling chip, an ultrasonic humidifier, a ventilation fan, a nutrient solution circulating water pump, and an ultraviolet disinfection lamp; the human-computer interaction is via a WeChat mini-program. 2. Specific Implementation Steps: S1: The user selects "Italian lettuce" to plant in the mini-program. The system retrieves the lettuce full growth cycle parameter matrix (including specific temperature, humidity, light, and nutrient solution parameters for the germination, seedling, growth, and harvesting stages) from the crop growth model library. S2: After system startup, the multi-sensor module collects environmental and nutrient solution data every 10 seconds and crop images every hour; the preprocessing unit filters and normalizes the data, removes outliers, and outputs a standardized dataset. S3: The multi-sensor fusion unit employs an improved unscented Kalman filter algorithm to fuse multi-source data and calculate the coupling effect coefficient of "a 1°C increase in temperature leads to a 2.5% decrease in humidity," automatically correcting the humidity control target and avoiding the problem of insufficient humidity after a temperature increase in traditional control. S4: The adaptive PID control unit uses a BP neural network to dynamically adjust PID parameters based on real-time deviations. When light is insufficient, it automatically adjusts the LED supplemental lighting duty cycle and simultaneously increases CO2 concentration to match photosynthetic needs, achieving multi-parameter collaborative optimization. S5: The system uses an image recognition algorithm (YOLO lightweight algorithm) to identify the number of lettuce leaves and plant height. When the lettuce enters the next growth stage, it automatically switches the corresponding control parameters without requiring manual adjustment by the user. S6: When the system detects abnormal humidity sensor data, it determines a sensor fault, automatically estimates the ambient humidity using fused data from the temperature and CO2 sensors, switches to fault-tolerant control mode, and simultaneously sends a fault warning to the user via a mini-program. S7: The system uploads planting data to the cloud platform daily. The cloud platform optimizes the growth model based on big data and periodically updates the device via OTA. 3. Implementation Results In this embodiment, the growth cycle of Italian lettuce is shortened from the traditional 45 days to 32 days, with a 100% yield rate, zero pesticides throughout the process, and no professional operation required by the user, achieving completely unmanned planting.

[0017] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A micro-agriculture box multi-sensor fusion adaptive control method, characterized in that, Includes the following steps: S1: Obtain the crop variety information selected by the user, and retrieve the optimal environmental parameter matrix of the corresponding initial growth cycle of the crop from the pre-stored crop growth model library as the initial control target value; S2: Synchronously collect multi-source sensor data of the target planting space at a preset frequency. The multi-source sensor data includes environmental parameter data, cultivation medium parameter data, and crop growth status image data. The collected multi-source sensor data is then processed sequentially by outlier removal, filtering and noise reduction, and normalization to obtain a standardized sensor dataset. S3: An improved unscented Kalman filter algorithm is used to perform multi-source data fusion on the standardized sensor dataset, establish a multi-parameter coupling relationship model, calculate the mutual influence coefficients between each control parameter, and correct the initial control target value based on the mutual influence coefficients to obtain the optimal dynamic control target value at the current moment. S4: Taking the deviation between the optimal dynamic control target value and the real-time standardized sensor data as input, the adaptive PID controller optimized by the BP neural network outputs the control signal of the corresponding actuator to drive the actuator to complete the closed-loop regulation; S5: Based on crop growth status image data, the actual growth cycle and health status of the crop are identified through image recognition algorithms. When the deviation between the actual growth status and the preset growth model exceeds the preset threshold, the parameter matrix of the crop growth model is corrected and the optimal dynamic control target value is updated. S6: Monitor sensor data and actuator feedback signals in real time to determine if there is a fault; if a fault is determined, automatically switch to the preset fault-tolerant control mode and send fault warning information to the user terminal; S7: Uploads real-time planting and control data to the cloud platform. The cloud platform optimizes the crop growth model library based on big data analysis and updates the optimized model to the local device. Type claim 1 in the field.

2. A micro-agriculture box multi-sensor fusion adaptive control system, characterized in that, The system for implementing the adaptive control method of claim 1 includes: The multi-sensor acquisition module is used to simultaneously acquire environmental parameter data, cultivation medium parameter data, and crop growth status image data of the target planting space. The core control module is communicatively connected to the multi-sensor acquisition module and is used for preprocessing sensor data, multi-sensor fusion, adaptive control decision-making, fault diagnosis, and crop growth model management. The actuator module is electrically connected to the core control module and is used to receive control signals from the core control module and perform corresponding environmental and media adjustment operations. The human-computer interaction module is communicatively connected to the core control module and is used to allow users to select crop varieties, view real-time data, receive early warning information, and remotely control equipment. The cloud platform module is connected to the core control module via a wireless network and is used to store planting data, optimize the crop growth model library, and realize batch management of multiple devices.

3. The method of claim 1, wherein, In step S3, the multi-parameter coupling relationship model is used to characterize the mutual influence relationship between temperature, humidity, CO2 concentration, light intensity, nutrient solution EC value and pH value; the improvement of the improved unscented Kalman filter algorithm is that, based on the standard unscented Kalman filter, an adaptive attenuation factor is introduced to dynamically adjust the prediction covariance matrix.

4. The method of claim 1, wherein, In step S4, the BP neural network takes the control deviation and the rate of change of deviation as input, and the proportional coefficient, integral coefficient, and derivative coefficient of the PID controller as output. It optimizes the PID parameters in real time through online learning to adapt to the dynamic response requirements of different growth stages.

5. The system of claim 2, wherein, The core control module has a built-in fault diagnosis and fault-tolerant control unit, which is used to monitor sensor data and feedback signals from actuators in real time. When a device fault is detected, it automatically switches to a preset fault-tolerant control mode and uses fused data from adjacent sensors to replace the data collected by the faulty sensor, ensuring the continuous operation of the system under abnormal conditions.

Images