MODEL PREDICTIVE CONTROL USING NEURAL NETWORK

Abstract

An efficient technique of establishing control of a linear or nonlinear process using Model Predictive Control (MPC) scheme by using Artificial Neural Network (ANN) and Adaptive Neuro Fuzzy Inference System (ANFIS) structure is presented. Time and efforts required to model a process by using ANN and ANFIS are significantly reduced in comparison with the conventional methods. Three modelling techniques are described, implemented and compared on the basis of accuracy and speed. The three techniques are Backpropogation, Extreme Learning Machine and Extreme learning ANFIS (EANFIS). EANFIS comes out to be as winner as it has good accuracy and high speed. After modelling of the process, optimization is performed to generate optimal control input by using Newton Raphson (NR) or Quasi Newton method. NR method is considered to be the fastest optimization method for quadratic Performance Function. Detailed derivation of the MPC algorithm is carried out for both ANN and EANFIS architectures using NR optimization. Simulation results show convergence to a good solution.