A Sparse Coding Approach to Household Electricity Demand Prediction in Smart Grids

With the gradual deployment of smart electricity grids in many cities around the world, new opportunities arise for reducing energy usage and improving the control and management of the distribution network. At the individual household level, modeling and forecasting of single meter electricity load recently become possible with the large-scale deployment of smart meters. Compared to load forecasting at the city/system aggregate level, load forecasting for individual households is a much harder problem as the load is much more stochastic and volatile. In this paper we study the use of sparse coding for modeling and forecasting these individual household electricity loads. The proposed methods are tested on a dataset of 1000 households in the city of Chattanooga, TN of USA, for the period January 2011 to June 2012. We obtain a 10% improvement in the accuracy of predicting next-day total load, and a 20% improvement in that of predicting next-week total load. We also demonstrate an application of the load forecasting algorithm for raising consumers' awareness of their electricity usage through an application prototype for monthly electricity budgeting.