Thu.2 10:45–12:00 | H 3013 | VIS
.

Numerical Methods for Bilevel Optimization and Applications (1/2)

Chair: Alain Zemkoho Organizer: Alain Zemkoho
10:45

Anthony Dunn

Foam prediction to improve biogas production using anaerobic digestion

Anaerobic digestion is used in renewable energy plants to produce biogas from recycled biological waste. Within the reaction digester, a phenomenon known as foaming can occur at irregular time intervals which results in a reduced yield of biogas and requires an extensive repair period. We show that machine learning based techniques (relying on single and two-level optimization) allow for the development of a predictive maintenance model that recognises periods of high foam risk. This enables the plant operators to apply suitable changes in the operations system to reduce the risk

11:10

Andrey Tin

Gauss-Newton-type method for bilevel optimization

This article examines the application of Newton-type methods for overdetermined systems to find solutions of bilevel programming problems. Since the Jacobian of the system is non-square, we discuss Gauss-Newton method and Newton method with Moore-Penrose pseudo inverse to solve the system. To our knowledge, this is the first time these methods are discussed in literature to solve bilevel programs. We prove that the methods can be well-defined for the introduced optimality conditions. In numerical part we compare the results of the methods with known solutions and performance of 124 examples.

11:35

Alain Zemkoho

joint work with Yekini Shehu, Phan Tu Vuong

An inertial extrapolation method for convex simple bilevel optimization

We consider a scalar objective minimization problem over the solution set of another optimization problem. This problem is known as simple bilevel optimization problem and has drawn a significant attention in the last few years. We propose and establish the convergence of a fixed-point iterative method with inertial extrapolation to solve the problem. Our numerical experiments show that the proposed method outperforms the currently known best algorithm to solve the class of problem considered.