Nonlinear Model Predictive Control of MSWC plants

Project P.M.J. Van den Hof (Paul)
Keywords:Process technology, Model predictive control, Identification and estimation
Sponsored by:TNO

The incineration of municipal solid waste (MSW), i.e. household waste, is used for the reduction of the amount of waste and for the production of energy. It is typically performed at a plant of the form that is depicted below.

Such a MSW combustion plant is subject to both economic and environmental operational and, thereby, control objectives. Economic objectives are e.g. maximization of the waste throughput, maximization of the energy output and maximization of the lifetime of the components of the MSW combustion plant. Environmental objectives are e.g. upper bounds imposed on potentially contaminating components of the flue gas. Part of these objectives are supporting each other, for example maximization of the waste throughput implies maximization of the energy output, and part of the objectives are conflicting, for example the objective of maximization of the waste throughput and energy output conflicts with the demand of maximizing the life time of the components of the MSW combustion plant.

Among many MSW combustion plant managers there is a need to improve their process operation performance. This is due to the ever becoming more stringent environmental regulations and ever growing higher energy demands. An essential tool that these managers have available for the fulfilment of the ever becoming higher and tighter economic and environmental objectives is (apart from operators) a combustion control system (which, by the way, does not contain the flue gas cleaning equipment). Such a combustion control system is typically a network of proportional and, sometimes, integrating (PI) controllers which at best fulfils the mentioned objectives in a suboptimal manner. An alternative which is thought to be able to deliver a much better control performance than such conventional combustion controllers is model predictive control (MPC). The reason(s) for this expectation is that MPC is thought to be able to deal much better with the following typical characteristics of the MSW combustion control problem: (i) multiple, conflicting objectives, (ii) the multivariable interacting nature of the process and (iii) constraints.

The aim of the research project is to investigate the feasibility of MPC as a tool for improving the process operation performance of MSW combustion plants. Aspects of the research are, amongst others, (i) modelling of MSW combustion plants via linear system identification techniques and (ii) nonlinear MPC using a first-principles model that describes the main dynamical phemona taking place during the MSW combustion process and which is of low complexity (i.e. with respect to the number of (differential) equations of the model). If all works out well, (N)MPC will be tested on a real-life large-scale MSW combustion plant.

© Copyright Delft Center for Systems and Control, Delft University of Technology, 2017.