Two Time-scale Optimal Control Of Greenhouse Cultivation

Greenhouse cultivation enables growers to manipulate crops' growing environment,thus promoting the crop production and quality.However,traditional greenhouse control technology is mainly based on growers' experience,which lacks scientific explanations and guarantee of optimal profit.Optimal control algorithm aims at maximizing profit,and computes optimal control inputs based on greenhouse-crop model.Consequently,physical meanings of the optimal results can be scientifically explained through the model.Nevertheless,different time scales in greenhouse climate dynamics and crop growth increase the computation load,making the optimal control algorithm difficult to implement in real time.The method of time-scale decomposition separates optimal control of greenhouse cultivation in a slow problem,which relates to the crop growth,and a fast problem,which relates to the greenhouse climate.In this way,the computation load is largely reduced,making the on-line implementation of optimal control possible.In this thesis,the follow researches are performed to solve problems in two time-scale optimal control of greenhouse cultivation.First,a proper greenhouse-crop model is chosen for two time-scale optimal control.A fast pseudospectrol method is used for solving the slow problem and the fast problem in open loop.Physical meanings of the results are scientifically explained through the model.A digital optimal control algorithm is developed for on-line implementation.Taking results of the fast problem as an initial guess,it is used in the closed-loop receding horizon optimal control implementation.Second,two time-scale optimal control of Chinese solar greenhouse is researched for the lack of applying optimal control algorithms in Chinese solar greenhouse.The chosen Venlo greenhouse model is extended with a north wall and a thermal blanket to represent the model of Chinese solar greenhouse.The method of two time-scale decomposition is used in calculating the profit obtained through applying the optimal control algorithm in Chinese solar greenhouse.Consequently,the optimal control trajectory of the thermal blanket is obtained.Third,adaptive optimal control is researched for the lack of considering uncertain or time-varying model parameters in two time-scale optimal control of greenhouse cultivation.Candidate model parameters are chosen based on their physical meanings and the open-loop sensitivity of the profit to these parameters.Through a fast identifiability algorithm,two greenhouse climate model parameters are chosen for on-line estimation.Closed-loop sensitivity of the profit to these two parameters are analyzed.Measurement errors are included to test the tracking ability of the adaptive parameters.Fourth,double closed-loop optimal control is researched for the lack of solving the slow problem in closed loop.The feedback of crop growth information is obtained through the crop growth model and the greenhouse climate measurements.The updated slow problem is automatically solved using results from the last slow problem as an initial guess.The possible gain of profit through the double closed-loop is analyzed facing prediction errors in crop price,control cost or long-term weather.A more remarkable gain of profit through the double closed-loop is shown after introducing LED lighting.The feasibility of on-line implementing the double closed-loop optimal control is tested through the computation time.