Heterogeneous Unit Clustering for Efficient Operational Flexibility Modeling for Strategic Models

Abstract

The increasing penetration of wind generation has led to significant improvements in unit commitment models. However, long-term capacity planning methods have not been similarly modified to address the challenges of a system with a large fraction of generation from variable sources. Designing future capacity mixes with adequate flexibility requires an embedded approximation of the unit commitment problem to capture operating constraints. Here we propose a method, based on clustering units, for a simplified unit commitment model with dramatic improvements in solution time that enable its use as a submodel within a capacity expansion framework. Heterogeneous clustering speeds computation by aggregating similar but non-identical units thereby replacing large numbers of binary commitment variables with fewer integers that still capture individual unit decisions and constraints. We demonstrate the trade-off between accuracy and run-time for different levels of aggregation. A numeric example using an ERCOT-based 205-unit system illustrates that careful aggregation introduces errors of 0.05-0.9% across several metrics while providing several orders of magnitude faster solution times (400x) compared to traditional binary formulations and further aggregation increases errors slightly (~2x) with further speedup (2000x). We also compare other simplifications that can provide an additional order of magnitude speed-up for some problems.