Iterative quality weighted interpolation for LPV‐MPC control of industrial heat exchanger under varying fouling conditions

Abstract

Fouling is an inevitable phenomenon that tends to change the dynamics of the heat exchanger (HE). An accurate prediction model describing these varying HE dynamics is demanded to implement model predictive control (MPC), which has been addressed in this work. A novel iterative quality weighted interpolation (IQWI) technique is proposed to determine the prediction model from the linear parametric varying (LPV) model. It uses a set of linear state‐space (SS) models, which have been identified from the input–output data of the industrial HE under various fouling conditions. The quality of the identified models is assessed using a fuzzy inference system (FIS) based on fit percentage and data quality. IQWI technique uses incremental thresholds for the model quality to iteratively generate interpolation curves. Each interpolation curve is capable of estimating the model parameters, which are fused using the distance weighted method to determine the optimal model parameters. The proposed technique is capable of improving the model accuracy by 24.43%. The use of a fouling‐based prediction model equips the MPC with accurate knowledge of HE dynamics, which enables it to provide efficient control. Experimental analysis under five different fouling conditions and comparative analysis with the existing controller indicates the efficiency of the proposed method.