Quantile-SMPC for Grid-Interactive Buildings with Multivariate Temporal Fusion Transformers

Spencer Hutchinson, Abraham P. Vinod, François Germain, Stefano Di Cairano, Christopher R. Laughman, Ankush Chakrabarty

Keywords: building optimization, predictive control, time-series, generative networks, stochastic optimization

Abstract: Exogenous disturbances such as occupancy and weather strongly affect the performance of grid-interactive buildings, making accurate probabilistic forecasting essential for robust control. We propose a new MQF2-TFT architecture that captures both temporal and cross-variable dependencies for multi-horizon disturbance forecasting. To integrate forecasts into stochastic MPC, we develop a quantile-based sample method that enforces chance constraints despite non-Gaussianity. Experiments with real office building data demonstrates improved key performance indicators over baselines.

Submission Number: 34