| Environmental regulation is one of the most important aspects in contemporary horticultural facilities.Especially in recent years,the frequent occurrence of haze weather has caused low temperature and light environment in facilities,which has made the facilities light environment regulation becoming a necessary guarantee for the production of facilities in winter.Scholars at domestic and foreign have conducted extensive research on the model of light environment regulation coupled with environmental factors.However,due to the influence of the internal and external structures of the facility and the demand for crop diversity cultivation,the small environments and crop varieties in each area of the facility are not the same.The existing light environment control system regulates the entire facility through a single regulation model,and caused the phenomenon of the amount of supplemental light and actual demand cannot be accurately matched.Therefore,a set of facility light environment control equipment that supports remote updating of models has been developed,considering the actual production needs of light environment regulation in facility horticulture.To achieve the light environment regulation with various methods for the different region within the facilities at the same time,while providing the ability to use the equipment sustainably.The main research contents of this article include:?1?Based on the overall design requirements and the interactive requirements of the facility's production,an embedded control device software was developed.Embedded software is divided into two parts:backstage function module and man-machine interface according to the implementation method.Both are linked together through storage tables in storage space.The background function module realizes functions such as network data communication,monitoring data analysis,and control decision instruction encapsulation.The human-computer interaction interface displays the running results of the background function module and passes the user input to the background module,thereby completing the direct control of the monitoring node and the light-up node.?2?A simplified multi-layer perceptron computing module was designed,combining with the features of embedded devices.The MATLAB neural network toolbox structure was analyzed by using MATLAB as a model training platform,and deriving the weight threshold of multi-layer perceptron.Through the method of calculating the transform weight threshold by the matrix,the normalization and anti-normalization steps in the multi-layer perceptron calculation process are removed to realize a more precise and concise calculation module.After testing on the embedded platform,the storage problem of double-precision floating-point data was found and corrected.Finally,the average absolute error between the model calculated results and the MATLAB standard output reached 1.447×10-18,and the maximum relative error reached 2.125×10-18.?3?Designed data transfer protocol and remote service program between server and embedded device.Multiple data frames are transmitted between the embedded device and the server by constructing a data transmission protocol.The service program with monitoring data storage,model download,and remote device restart functions was designed by using python as a development language.It is deployed in the Ali cloud server ECS to ensure its stable operation.This makes it possible to load and adapt to various crop models under the limited storage resources of embedded devices. |
PDF Full Text Request