農産物貯蔵工場の知能的制御

概要

論文の詳細を見る
A new optimal control technique for ventilation in a fruit storage factory is proposed. The method is composed of two steps. In the first step, static model for calculating the evapotranspiration of orange (Iyokan) fruit is first constructed by using neural networks, and then the absolute humidity in the storage factory is calculated. In the second step, the dynamic model of absolute humidity as affected by ventilation is constructed by using neural networks, and then optimal value (6-step on-off ventilation times which minimize the deviation between the current values and the setpoint of the absolute humidity) is sought by using genetic algorithms. A three-layer neural network was effective for the identification of the non-linearity between temperature and evapotranspiration of fruit. The genetic algorithm allowed the optimal value to be quickly sought from the model simulation when the crossover and mutation rates had high values. Thus, the hybrid control technique involving neural networks and genetic algorithms is effective for the optimal control of environmental factors in the fruit storage factory.
日本生物環境工学会の論文