Smart Bidding for Smart Homes: Multi-Market Electricity Trading

Abstract

\indent The 2021 Texas blackouts exposed the vulnerability of ERCOT’s deregulated electricity grid to extreme weather and price volatility, and these risks intensify with climate change, electrification, and rising electricity demand. At the same time, residential distributed energy resources (DERs)—rooftop solar, home batteries, and electric vehicles—are becoming more widespread and capable of providing valuable grid flexibility. Virtual Power Plants (VPPs) aggregate these resources to bid into wholesale markets, but typically rely on opaque, automated platforms that offer users little visibility or control. As a result, most households lack tools to understand or optimize their DERs for financial benefit. This thesis presents a transparent, data-driven framework that enables household DERs to actively participate in ERCOT’s two-settlement electricity market, where day-ahead bids must anticipate uncertainty in real-time prices and solar generation. We combine probabilistic forecasting with two-stage stochastic optimization to model household market decisions. Quantile LSTM (QLSTM) models, trained on historical price and solar data, generate scenario-based forecasts that feed into a two-stage optimization model capturing DER dynamics, including battery degradation, EV availability, and flexible load scheduling. To align forecasts with downstream outcomes, we fine-tune the QLSTM using a decision-focused learning (DFL) loss that minimizes regret in the two-stage problem. Simulations across four Texas cities and three levels of residential load profiles show that the baseline predict-then-optimize strategy consistently recovers most of the profit achievable under perfect foresight, while DFL improves robustness under volatile conditions. This work demonstrates how machine learning and optimization approaches can empower households to participate meaningfully in electricity markets by offering a user-centric alternative to DER optimization and supporting a more distributed, resilient, and responsive energy grid.