Enhanced Representative Days and System States Modeling for Energy Storage Investment Analysis

TL;DR

This paper compares and improves models for evaluating energy storage investments in power systems with high renewable energy, introducing new models that better balance accuracy and computational efficiency.

Contribution

It proposes two novel models, SS-RFM and RP-TM&CI, that enhance existing methods by better capturing short- and long-term storage with minimal computational increase.

Findings

RP-TM&CI model reduces investment error by nearly 10 times

Enhanced models better represent both short- and long-term storage

All models are benchmarked against an hourly unit commitment model

Abstract

This paper analyzes different models for evaluating investments in Energy Storage Systems (ESS) in power systems with high penetration of Renewable Energy Sources (RES). First of all, two methodologies proposed in the literature are extended to consider ESS investment: a unit commitment model that uses the System States (SS) method of representing time; and another one that uses a representative periods (RP) method. Besides, this paper proposes two new models that improve the previous ones without a significant increase of computation time. The enhanced models are the System States Reduced Frequency Matrix (SS-RFM) model which addresses short-term energy storage more approximately than the SS method to reduce the number of constraints in the problem, and the Representative Periods with Transition Matrix and Cluster Indices (RP-TM&CI) model which guarantees some continuity between representative periods, e.g. days, and introduces long-term storage into a model originally designed only for the short term. All these models are compared using an hourly unit commitment model as benchmark. While both system state models provide an excellent representation of long-term storage, their representation of short-term storage is frequently unrealistic. The RP-TM&CI model, on the other hand, succeeds in approximating both short- and long-term storage, which leads to almost 10 times lower error in storage investment results in comparison to the other models analyzed.