Bi-Level stochastic approximation for joint optimization of hydroelectric dispatch and spot-market operations

Abstract

We propose a bi-level formulation for the joint optimization of spot-market operations and optimal dispatch of hydroelectric power generation. The outer problem maximizes the net-profit from interacting with neighbours in the wider energy spot-market, where profit is the difference between the (non-convex) revenue generated from export and the cost for generating and transmitting the exported power. The latter is the optimal solution of the inner optimization problem, a stochastic linear program that determines the minimum-cost dispatch solution that meets existing local demand and planned export, subject to uncertainty in local demand and generation from run-of-The-river dams. The outer problem is solved using stochastic approximation, where the cost function gradient is obtained using parametric programming techniques on the inner problem. Experimental results establish the efficacy of this bi-level approach. We also provide some preliminary analysis of the convergence of this method.