Prescribed Robustness in Optimal Power Flow

TL;DR

This paper introduces a new method to determine adaptive safety intervals for stochastic resources in power systems, integrating physics and cost optimization to enhance safety and efficiency in decarbonization efforts.

Contribution

It presents a novel prescriptive approach using differentiable optimization and stochastic gradient descent to compute interpretable safety intervals for robust power flow optimization.

Findings

Effective safety intervals derived for stochastic resources.

Improved robustness in power flow operations.

Case studies demonstrate practical applicability.

Abstract

For a timely decarbonization of our economy, power systems need to accommodate increasing numbers of clean but stochastic resources. This requires new operational methods that internalize this stochasticity to ensure safety and efficiency. This paper proposes a novel approach to compute adaptive safety intervals for each stochastic resource that internalize power flow physics and optimize the expected cost of system operations, making them ``prescriptive''. The resulting intervals are interpretable and can be used in a tractable robust optimal power flow problem as uncertainty sets. We use stochastic gradient descent with differentiable optimization layers to compute a mapping that obtains these intervals from a given vector of context parameters that captures the expected system state. We demonstrate and discuss the proposed approach on two case studies.