Market-Based Coordination of Price-Responsive Demand Using Dantzig-Wolfe Decomposition Method

TL;DR

This paper introduces a novel distributed algorithm based on Dantzig-Wolfe decomposition for coordinating price-responsive demand in power systems, improving efficiency and cost savings.

Contribution

It is the first to apply Dantzig-Wolfe decomposition to coordinate self-benefiting demand response devices in a distributed manner.

Findings

Algorithm reduces generation costs and user payments.

Significant reduction in peak-to-average load ratio.

Fast convergence guaranteed for convex subproblems.

Abstract

With the increased share of Distributed Generation (DG) and Demand Responsive (DR) loads in the power systems, new approaches based on the game theory framework have been proposed to tackle the problem of coordination of Price Responsive Devices (PRD). The PRDs are modeled as self-benefiting players who try to optimize their consumption based on the price. In this paper, for the first time, a new algorithm based on the Dantzig-Wolfe (DW) Decomposition method to solve the coordination problem of self-benefiting PRDs in a distributed fashion has been proposed. By utilizing the distributed nature of Dantzig-Wolfe, the PRD's self-benefiting algorithms are modeled as sub-problems of the DW, and the coordinator (or the grid operator) who collects energy consumption of PRDs (their energy bid), solves the master problem of the DW and calculate the price signal accordingly. The proposed algorithm is fast since the subproblem in DW (which could be millions of PRDs) can be solved simultaneously. Furthermore, based on the DW theory, if the PRDs subproblems are convex, reaching the optimal point (Equal to Nash Equilibrium) in limited iterations is guaranteed. A simulation with 8 participant households has been conducted to evaluate the model. Each house is equipped with two types of loads: an Electric Vehicle (EV) as a sample of interruptible loads and an Electric Water Heater (EWH) as a sample of Thermostatically Control Loads (TCA). The results show that when the algorithm reaches the optimal point, the generation cost and the user payment (based on the marginal cost of generation) decrease. Furthermore, the aggregate load's Peak to Average (PAR) reduces significantly.