Flattening the Duck Curve: A Case for Distributed Decision Making

TL;DR

This paper introduces a distributed optimization framework to coordinate distributed energy resources, effectively reducing ramping requirements caused by the duck curve in high renewable penetration scenarios.

Contribution

It presents a novel convex relaxation-based power flow model and an accelerated distributed optimization algorithm for grid resource coordination.

Findings

Reduced bulk system ramping by up to 23%

Effective coordination of distributed resources in high PV scenarios

Validated on a modified IEEE-34 node network

Abstract

The large penetration of renewable resources has resulted in rapidly changing net loads, resulting in the characteristic "duck curve". The resulting ramping requirements of bulk system resources is an operational challenge. To address this, we propose a distributed optimization framework within which distributed resources located in the distribution grid are coordinated to provide support to the bulk system. We model the power flow of the multi-phase unbalanced distribution grid using a Current Injection (CI) approach, which leverages McCormick Envelope based convex relaxation to render a linear model. We then solve this CI-OPF with an accelerated Proximal Atomic Coordination (PAC) which employs Nesterov type acceleration, termed NST-PAC. We evaluate our distributed approach against a local approach, on a case study of San Francisco, California, using a modified IEEE-34 node network and under a high penetration of solar PV, flexible loads, and battery units. Our distributed approach reduced the ramping requirements of bulk system generators by up to 23%.