An efficient implementation of GeneralizedPredictive Control using multi-layer feed forward neural network as the plant's nonlinear model is presented. Two algorithm i.e. Newton Raphson and Levenberg Marquardt algorithm are implemented and their results are compared. The details about this implementation are given. The utility of each algorithm is outlined in the conclusion. In using Levenberg Marquardt algorithm, the number of iteration needed for convergence is significantly reduced from other techniques. This paper presents a detail derivation of the neural generalized predictive control algorithm with Newton Raphson and Levenberg Marquardt as the minimization algorithm. A simulation result of Newton Raphson and Levenberg Marquardt algorithm are compared. Levenberg Marquardt algorithm shows a convergence of a good solution. The performance comparison of these two algorithms also given in terms of ISE and IAE.

No abstract

+PVTQFWEVKQP ,Q WKH PDQXIDFWXULQJ LQGXVWU\ WKH UHTXLUHPHQW IRU KLJKVSHHG IDVWUHVSRQVH DQG KLJK SUHFLVLRQ SHUIRUPDQFHV LV FULWLFDO 0RGHO SUHGLFWLYH FRQWURO 03& ZKLFK ZDV GHYHORSHG LQ WKH ODWH ·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

An efficient implementation of Model Predictive Control (MPC) using a multilayer feed forward network as the plants linear model is presented. This paper presents a comparison between the Generalized Predictive Control and Neural Generalized Predictive Control with Newton-Raphson as minimization algorithm. Three different linear models are taken and their performances are tested. Simulation result shows the effect of neural network on Generalized Predictive Control for linear systems. The performance comparison of these system configurations has been given in terms of ISE and IAE.

No abstract