Regularized Benders Decomposition for High Performance Capacity Expansion Models

We consider electricity capacity expansion models, which optimize investment and retirement decisions by minimizing both investment and operation costs. In order to provide credible support for planning and policy decisions, these models need to include detailed operations and time-coupling constraints, and allow modeling of discrete planning decisions. Such requirements result in large-scale mixed integer optimization problems that are intractable with off-the-shelf solvers. Hence, practical solution approaches often rely on carefully designed abstraction techniques to find the best compromise between reduced temporal and spatial resolutions and model accuracy. Benders decomposition methods offer scalable approaches to leverage distributed computing resources and enable models with both high resolution and computational performance. Unfortunately, such algorithms are known to suffer from instabilities, resulting in oscillations between extreme planning decisions that slows convergence. In this study, we implement and evaluate several level-set regularization schemes to avoid the selection of extreme planning decisions. Using a large capacity expansion model of the Continental United States with over 70 million variables as a case study, we find that a regularization scheme that selects planning decisions in the interior of the feasible set shows superior performance compared to previously published methods, enabling high-resolution, mixed-integer planning problems with unprecedented computational performance.
Comment: This work has been submitted to the IEEE for possible publication