Energy system models (ESMs) are mathematical programs that represent the energy systems of countries or regions. Researchers use ESMs to investigate long-term scenarios describing the evolution of all sectors of the energy system of a given country or group of countries over the 21st century. These sectors can include buildings, transportation, industry, power generation, agriculture and more. ESMs, and energy systems analysis research, are well embedded in energy strategy deliberations worldwide and allow for more informed planning of capacity expansion and other investments by calculating the time evolutions of the whole system under different assumptions, e.g. development of CO2 price, availability of renewable energy, future demand, energy and climate change mitigation policies, etc.
Many ESMs are implemented in GAMS1. A model with decades of history is the TIMES energy systems modeling framework,which is developed by the Energy Technology Systems Analysis Programme (ETSAP) of the International Energy Agency (IEA) . TIMES is used by at least 50 countries worldwide and more than 120 research institutes, companies and governmental agencies.
As a very active research area, energy system modeling is challenging for even the best state-of-the-art solver algorithms. The increasing complexity of energy systems, e.g. due to the increasing shares of renewable energy, and the trend towards building models with higher levels of granularity on the one hand and increasing time horizons on the other hand, frequently result in challenging large-scale problems. When dealing with such problems, the targeted use of solver options can often lead to dramatic performance improvements. While the solvers used in GAMS include sophisticated heuristics to find a suitable parameterization of the solution algorithm, model developers can improve the solver’s performance by combining their knowledge on the model’s structure with the comprehensive information provided by GAMS in the output log and .LST files, in order to tweak some of the solver’s options.
As a concrete example, the recent ETSAP Webinar “CPLEX Barrier Options for TIMES models” provides an extensive overview on how to analyze and tune the CPLEX Barrier algorithm for a particular TIMES model, thereby showing the great potential for performance improvements when dealing with challenging large-scale LPs. This webinar provides quite the right amount of background information to provide modelers with useful insights to the complex internals of state-of-the-art optimization algorithms and puts those insights in relation to various potentially useful solver options whose impact on the solution time is worthwhile to explore. The insights and hints are by no means restricted to TIMES models but provide a useful cookbook to tune the widely used CPLEX barrier algorithm for large-scale LPs.
Watch the video here:
Download the presentation slides here: https://iea-etsap.org/webinar/CPLEX%20options%20for%20running%20TIMES%20models.pdf