We propose a method that combines data-driven and physics-informed approaches to best estimate net load based on historical data.