Cost-Optimal Operation of Energy Storage Units: Impact of Uncertainties and Robust Estimator

TL;DR

This paper introduces a new robust optimization method for cost-effective operation of energy storage units under uncertainty, demonstrating improved robustness and lower costs compared to traditional methods using real-world forecast data.

Contribution

The paper presents a novel robust optimization approach for energy storage operation that leverages statistical measures and a sliding window technique, outperforming existing methods in cost and robustness.

Findings

Lower acquisition costs with the proposed method compared to traditional robust optimization.

Enhanced robustness under forecast perturbations.

Applicable to general real-time energy storage optimization problems.

Abstract

The rapid expansion of wind and solar energy leads to an increasing volatility in the electricity generation. Previous studies have shown that storage devices provide an opportunity to balance fluctuations in the power grid. An economical operation of these devices is linked to solutions of probabilistic optimization problems, due to the fact that future generation is not deterministic in general. For this reason, reliable forecast methods as well as appropriate robust optimization algorithms take an increasingly important role in future power operation systems. Taking an uncertain availability of electricity into account, we present a method to calculate cost-optimal charging strategies for energy storage units. The proposed method solves subproblems which result from a sliding window approach applied on a linear program by utilizing statistical measures. The prerequisite of this method is a recently developed fast algorithm for storage-related optimization problems. To present the potential of the proposed method, a Power-To-Heat storage system is investigated as an example using today's available forecast data and a robust statistical measure. Second, the novel approach proposed here is compared with a common robust optimization method for stochastic scenario problems. In comparison, the proposed method provides lower acquisition costs, especially for today's available forecasts, and is more robust under perturbations in terms of deteriorating predictions, both based on empirical analyses. Furthermore, the introduced approach is applicable to general cost-optimal operation problems and also real-time optimization concerning uncertain acquisition costs.