Multi-ANN Predictive Control of Citrus Peel Supercritical Extraction Temperature
Considering nonlinearity, time-variation and inertia during temperature control of large supercritical extraction units, especially under the disturbance of system flow and pressure, a multi-artificial neural network (ANN) predictive control policy was proposed. It contains a radial basis function (RBF) ANN, aiming to approach nonlinear extraction temperature object and predicting output variable based on this model. There is also a back propagation (BP) ANN controller, seeking the optimal controlling signal by feedback correction and rolling optimization on purpose to overcome the time-variation and inertia. The experimental results indicate that this control strategy has excellent dynamic response performance, small steady state error and strong robustness.
Zhong Cao, Yinghe He, Lixian Sun and Xueqiang Cao
H. Y. Zhou et al., "Multi-ANN Predictive Control of Citrus Peel Supercritical Extraction Temperature", Advanced Materials Research, Vols. 236-238, pp. 1472-1479, 2011