Evaluation of uncertainties in linear energy system optimization models using HPC and neural networks

Abstract

Within the interdisciplinary BMWK-funded project UNSEEN, experts from High Performance Computing, mathematical optimization and energy systems analysis combine strengths to evaluate uncertainties in modeling and planning future energy systems with the aid of High Performance Computing (HPC) and neural networks. Energy System Models (ESM) are central instruments for realizing the energy transition. These models try to optimize complex energy systems in order to ensure security of supply while minimizing costs for power production and transmission. In order to derive reliable and robust policy advice for decision makers, hundreds or even thousands of ESM problems need to be solved in order to address uncertainties in a given model and dataset.Mixed-integer linear programs (MIPs), a direct extension of Linear programs (LPs), can be used to formulate and compute more concrete and realistic energy systems. Since the availability of fast LP solvers is a major prerequisite for optimizing MIPs, the development of an open-source scalable distributed-memory LP solver, called PIPS-IPM++, was started in a preceding project and can already outperform state-of-the-art solvers. A second prerequisite for efficient MIP solving is the availability of MIP heuristics. For this purpose, we develop a generic MIP framework including reinforcement learning methods. Moreover, we aim to implement an efficient automated HPC workflow for generating, solving, and postprocessing numerous ESM problems with a special structure in order to develop new tools for better predictions about the future of our energy system. This novel approach couples multiple existing and new software packages to achieve the project goals.