Deep Recurrent Neural Networks for Building-Level Load Forecasting

Load forecasting plays a crucial role in the day-to-day operations of electric utilities, especially in modern power systems, where a significant share of power generation is attributable to renewable sources. Over the years, several algorithms have been developed to tackle this problem, on time scales ranging from a few hours to several months. Most recent solutions have employed machine learning techniques such as deep learning to increase the granularity of the prediction, down to the single-building level. Here, we employ a framework based on long short-term memory networks to estimate the average power consumption of a single building equipped with solar panels. We show which measurements are more important for an accurate forecast and test several prediction horizons in order to find the best trade-off between training speed and prediction accuracy. Our results reinforce the notion that long short-term memory networks can be successfully used for short-to medium-term load forecasting.