Yeonsoo Kim

PhD Candidate
Department of Chemical & Biological Engineering

Seoul National University

Yeonsoo Kim received her BS degree in Chemical and Biological Engineering from Seoul National University (SNU), Korea, in 2013. She is currently pursuing the PhD degree in Chemical and Biological Engineering from SNU. She received the first prize in Process Simulation Olympiad held by Korean Institute of Chemical Engineering (KIChE) in 2012 and the best poster presentation award in 2014 from KIChE. She has been a IEEE student member since 2017. She has conducted a Korean government project of detection and localization of small leakage in the water distribution network and an industrial research project of modelling and control of diesel aftertreatment system. Furthermore, she has published the results of the government project in the journal "Computers and Chemical Engineering" and has registered patents in Korea. The work of industrial research project has recently been accepted in the journal "IEEE Transactions on Control system technology". Her research interests include modeling, control, and optimal design of the diesel aftertreatment systems, as well as nonlinear control theory in general.

Hybrid Nonlinear Model Predictive Control of Diesel Aftertreatment System

A lean NOx trap (LNT) followed by selective catalytic reduction (SCR), LNT-pSCR, is one of the promising aftertreatment systems of vehicles. LNT adsorbs NOx during normal engine operation (lean mode), and additional fuel is injected (called post injection) to regenerate the LNT periodically (rich mode). In the rich mode, NH3 is produced in LNT, which can be absorbed and used as a reductant in pSCR. We propose a hybrid nonlinear model predictive control (NMPC) that determines whether the rich mode is activated at each sampling time. The obstacles to applying NMPC are as follows: 1) the manipulated variable is a binary variable (rich mode or lean mode) that makes the optimization problem into integer nonlinear programming, which is difficult to solve. 2) Since the engine operation is manipulated, the effect of the post injection is described and the cases where the post injection is not available from the viewpoint of engine operation should be considered. 3) NMPC should be solved within the sampling time for the real-time implementation. First, a post-injection map depending on the duration of maintaining the rich mode and the duration after the end of the rich mode is established based on the experimental data. Second, by successive linearization method and by considering only the feasible cases, the NMPC can be solved within the sampling time and the infeasible cases are excluded beforehand. The proposed strategy reduces NOx and produces NH3 more in LNT with shorter total rich duration than those of the existing logic.