Double Closed-loop Optimal Control Of Greenhouse Based On Online Feedback Of Crop State

Through the simulation analysis of 'soft sensor' based double closed-loop optimal control of greenhouse,problems of realizing crop state online feedback based double closed-loop optimal control of greenhouse were researched.(1)'Dark time' was introduced in optimal control of greenhouse LED lighting to avoid continuous light.(2)Online feedback of crop state was realized through computer vision to handle crop model parameter errors.(3)IOT system which realized remote continuous control input for greenhouse optimal control was built for experiment verification.Main conclusions are as follow.1.'soft sensor' based double closed-loop optimal control of greenhouse can handle long-term prediction errors on weather and prices.The profit is increased by 2.98%when prediction errors are within 50%.Growth period is reduced to 20 days and the profit is increased by 24.5%when LED lighting is introduced.2.Lettuce model is calibrated through experiment data from a plant factory.‘Dark time'of LED lighting is realized through a switching function in optimal control.Profit under different electric energy prices and lettuce prices are calculated.This lays a foundation for the future research on optimal control of LED lighting in greenhouse.3.CNN model with deep learning and SVR,RF,and LR model with machine learning for greenhouse lettuce growth parameter prediction are compared.The CNN model shows higher accuracy than others.Accuracy can be increased up to 90%for group prediction.4.IOT system for optimal control of greenhouse with an interface for LED lighting is researched.In summer the wet pad and inside shading should work together to decrease the internal temperature to be within its upper limit.This influence should be involved in greenhouse climate modelling for optimal control in summer.