Robust Rule-Based Sizing and Control of Batteries for Peak Shaving Applications

TL;DR

The paper addresses robust sizing and control of batteries for peak shaving under forecast uncertainty. It introduces stochastically tuned, three-parameter rule-based controllers (RBCs) together with a CVaR-based robust training objective to minimize daily peaks and tail risk, optimized via gradient-free methods. Experiments on real yearly meter profiles show that RBC sizing yields smaller installations and more realistic LCOE estimates, often outperforming deterministic MPC, particularly for high-tail peak events, with adversarially tuned RBC achieving the lowest LCOE in many cases. The work delivers a lightweight, interpretable framework for robust battery sizing and control that remains practical for industry deployment under nonstationary demand.

Abstract

As the cost of batteries lowers, sizing and control methods that are both fast and can achieve their promised performances when deployed are becoming more important. In this paper, we show how stochastically tuned rule based controllers (RBCs) can be effectively used to achieve both these goals, providing more realistic estimates in terms of achievable levelised cost of energy (LCOE), and better performances while in operation when compared to deterministic model predictive control (MPC). We test the proposed methodology on yearly profiles from real meters for peak shaving applications and provide strong evidence about these claims.

Figures (5)

• Figure 1: Uncontrolled (violet) and controlled profiles for 6 time series and four compared controllers. Dashed lines represent the SoC of the battery. Left axis: power, kWh; right axis SoC.
• Figure 2: Boxen plots of quantiles of normalized daily maxima across meters, for different controllers and sizing methods. A: prescient sizing. B: RBC sizing. Top: lower quantiles. Bottom: higher quantiles. Outliers are not shown.
• Figure 3: Battery size boxen plots across series with the two sizing strategies. Left: prescient sizing; right: RBC sizing. Log scale.
• Figure 4: Boxen plots across meters of LCOE on the test set, under prescient sizing. Left: unnormalized distribution; the distribution of the LCOE estimated on the sizing set is shown in violet. Right: LCOE distributions normalized with the distribution of the LCOE estimated on the sizing set.
• Figure 5: Boxen plots across meters of LCOE on the test set, under RBC sizing. Left: unnormalized distribution; the distribution of the LCOE estimated on the sizing set is shown in violet. Right: LCOE distributions normalized with the distribution of the LCOE estimated on the sizing set.