On Mitigating the Uncertainty in Renewable Generation in Distribution Microgrids

Arnab Dey, Vivek Khatana, Ankur Mani, Murti V. SalapakaarXiv 2020 · arXiv

Introduction

The global transition toward Renewable Energy Sources (RES) like wind and solar is essential for sustainability but introduces significant operational challenges. Unlike traditional fossil-fuel-based grids where large spinning turbines provide inherent stability, microgrids—localized power systems—must contend with the high intermittency and stochastic nature of renewables. This volatility creates a risk of "underproduction," where supply fails to meet demand, potentially compromising critical infrastructure such as hospitals or data centers.

This paper proposes a robust mathematical framework to manage this uncertainty. By treating renewable assets and battery storage as a financial portfolio, the authors develop optimization strategies and a "Black-Scholes-inspired" policy to provide "almost sure" power guarantees for critical loads.

System Architecture

The study considers a distribution microgrid comprising multiple renewable generation units (Photovoltaic (PV) and Wind Turbines) and a battery bank. These components are integrated through power electronic converters to a common bus that serves a local load.

Schematic of the microgrid system including PV, wind, and battery storage units.

Mathematical Framework and Methods

The problem of balancing supply and demand is addressed through three distinct mathematical lenses:

1. Short-term Power Demand Optimization

For microgrids without batteries, the focus is on the optimal "mix" of renewable sources. Assuming sources are uncorrelated (e.g., geographically dispersed wind farms), the authors apply Stochastic Optimization and Karush-Kuhn-Tucker (KKT) conditions. They adapt the "water-filling algorithm"—a technique from communications theory—to find the asset weights that minimize total variance while meeting a minimum power threshold.

2. Sustained Energy Demand Management

When batteries are integrated, the objective shifts to minimizing the expected squared mismatch between supply and demand over a time horizon. Renewable output is modeled using Geometric Brownian Motion (GBM), a stochastic process commonly used in finance to model stock prices. An incremental algorithm is proposed to dynamically determine the necessary battery reserve to buffer against generation dips.

3. Future-time Power Guarantees

The most innovative contribution is a provisioning policy for a "Renewable Generation Farm" (ReGF) to guarantee a specific power level at a future time $T_f$. Using Ito’s Lemma and solving a Partial Differential Equation (PDE), the authors derive a policy for adjusting active generation and battery blocks. This approach mirrors the Black-Scholes model used for pricing financial options, treating the risk of power outage as a financial risk to be hedged.

Key Results and Discussion

Numerical simulations validated the theoretical framework. The "Excess Production" (EP) and "Critical Production" (CP) solutions provide clear guidance for prioritizing assets based on their mean output and variance.

The dynamic tracking capabilities of the proposed policy are demonstrated in scenarios where renewable generation fluctuates. As shown in the figures below, the system successfully adjusts the required battery blocks to ensure the power portfolio meets the demand deficit, regardless of whether generation is initially above or below the target.

Dynamic tracking of power demand and required battery blocks for scenarios where generation is below (a) and above (b) the critical demand.

A key insight from the simulations is the sensitivity of the battery reserve to the time horizon. Counter-intuitively, for some realizations, a longer horizon might require fewer battery blocks at a specific instant because the system has more time to "correct" for the drift in renewable generation.

Conclusion and Future Work

The financial portfolio analogy provides a powerful toolset for microgrid reliability. By framing power management as a hedging problem, the authors demonstrate that "almost sure" guarantees are possible even with volatile sources.

Future research could extend this work by relaxing the assumptions of constant drift and volatility in the GBM model to better capture seasonal weather patterns. Additionally, incorporating more realistic battery physics—such as state-of-charge limits and degradation—would further bridge the gap between theoretical modeling and practical implementation.

Summary updated

BibTeX
@misc{dey2020mitigating,
  author = {Arnab Dey and Vivek Khatana and Ankur Mani and Murti V. Salapaka},
  title = {{On Mitigating the Uncertainty in Renewable Generation in Distribution Microgrids}},
  year = {2020},
  eprint = {2007.08641},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2007.08641},
}
Figure