Power-Based Generation Expansion Planning for Flexibility Requirements

TL;DR

This paper introduces a power-based generation expansion planning model that more accurately captures system flexibility, leading to more efficient investments and lower operating costs compared to traditional energy-based models.

Contribution

It develops a novel power-based GEP-UC model incorporating real-time flexibility and system constraints, improving investment efficiency and computational speed.

Findings

Power-based model reduces investments by 6-12%.

Operating costs decrease by 2-8% with the new model.

Semi-relaxed model is at least 10 times faster with minimal accuracy loss.

Abstract

Flexibility requirements are becoming more relevant in power system planning due to the integration of variable Renewable Energy Sources (vRES). In order to consider these requirements Generation Expansion Planning (GEP) models have recently incorporated Unit Commitment (UC) constraints, using traditional energy-based formulations. However, recent studies have shown that energy-based UC formulations overestimate the actual flexibility of the system. Instead, power-based UC models overcome these problems by correctly modeling ramping constraints and operating reserves. This paper proposes a power-based GEP-UC model that improves the existing models. The proposed model optimizes investment decisions on vRES, Energy Storage Systems (ESS), and thermal technologies. In addition, it includes real-time flexibility requirements, and the flexibility provided by ESS, as well as other UC constraints, e.g., minimum up/down times, startup and shutdown power trajectories, network constraints. The results show that power-based model uses the installed investments more effectively than the energy-based models because it more accurately represents flexibility capabilities and system requirements. For instance, the power-based model obtains less investment (6-12%) and yet it uses more efficiently this investment because operating cost is also lower (2-8%) in a real-time validation. We also propose a semi-relaxed power-based GEP-UC model, which is at least 10 times faster than its full-integer version and without significantly losing accuracy in the results (less than 0.2% error).