Wisconsin, Madison, August 29 2018
Stephen J. Wright, a principal investigator in the Argonne-led MACSER project and a professor at the University of Wisconsin – Madison, gave a plenary talk on August 19, 2018, at the Nonlinear Model Predictive Control (NMPR2018) conference.
Model predictive control (MPC) is a control process widely used in chemical plants and oil refineries, power systems, and robotics. At the NMPR2018 conference in July, Wright addressed the challenges MPC raises with regard to optimization, including the need to tackle difficult, structured optimization problems in limited wall-clock time while meeting controller requirements such as stability.
In his lecture titled “Optimization and MPC: Some Recent Developments,” Wright noted that advances in machine learning may change this situation. He reviewed several recent proposals advocating the use of machine learning in control strategies in which the models are learned rather than specified, and he speculated about how recent developments in nonconvex optimization and in MPC might benefit each other.
Wright is one of a team of researchers participating in the MACSER (Multifaceted Mathematics for Rare, High-Impact Events in Complex Energy and Environment Systems) project, which is led by Argonne and funded by the U.S. Department of Energy. The project addresses critical mathematical challenges in the formulation of decision problems for energy systems. One of the objectives of the project is to synthesize modeling and machine learning techniques in control and optimization.
For further information about the conference, see the website nmpc2018.org.