Learning to REDUCE: A Reduced Electricity Consumption Prediction Ensemble

Utilities use Demand Response (DR) to balance supply and demand in the electric grid by involving customers in efforts to reduce electricity consumption during peak periods. To implement and adapt DR under dynamically changing conditions of the grid, reliable prediction of reduced consumption is critical. However, despite the wealth of research on electricity consumption prediction and DR being long in practice, the problem of reduced consumption prediction remains largely un-addressed. In this paper, we identify unique computational challenges associated with the prediction of reduced consumption and contrast this to that of normal consumption and DR baseline prediction. We propose a novel ensemble model that leverages different sequences of daily electricity consumption on DR event days as well as contextual attributes for reduced consumption prediction. We demonstrate the success of our model on a large, real-world, high resolution dataset from a university microgrid comprising of over 950 DR events across a diverse set of 32 buildings. Our model achieves an average error of 13.5%, an 8.8% improvement over the baseline. Our work is particularly relevant for buildings where electricity consumption is not tied to strict schedules. Our results and insights should prove useful to the researchers and practitioners working in the sustainable energy domain. Introduction One of the critical challenges confronting modern societies is the need to attain energy sustainability. Buildings account for about 40% of the energy consumption worldwide (UNDP 2010) and novel energy optimization measures adopted in buildings can significantly contribute to energy sustainability. With the advent of Smart Grids, buildings are now being fitted with smart meters that record electricity usage every 15 minutes or less (Aman et al. 2015). Mining large amounts of electricity consumption data collected by smart meters provides insights into peak demand periods for buildings. Electric utilities can use these insights to ask building occupants and facility managers to reduce consumption during anticipated peak demand periods, a practice popularly know as Demand Response (DR). DR is defined as: “change in electric usage by end-use customers from their normal consumption patterns in response to changes Copyright c © 2016, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. Figure 1: Normal consumption, Reduced consumption, and DR baseline vis-a-vis a DR event. in the price of electricity over time, or to incentive payments designed to induce lower electricity use at times of high wholesale market prices or when system reliability is jeopardized” (FERC 2010). We address reduced consumption prediction during DR, which can help utilities in: • estimating the extent of potential reduction during DR before the DR event occurs (Chelmis et al. 2015); • performing dynamic DR at a few hours’ advance notice whenever necessitated by the dynamically changing conditions of the grid, such as due to the integration of intermittent renewable generation sources (Aman et al. 2015); • intelligently targeting customers for participation in DR based on a prediction of their reduced consumption and modifying such selection in real-time as needed (Ziekow et al. 2013); and • estimating the amount of incentives to be given to the customers (Wijaya, Vasirani, and Aberer 2014). Techniques that work well for normal consumption prediction, such as time series models, are ineffective for reduced consumption prediction due to 1) abrupt changes in the consumption profile at the beginning and end of the DR event (Figure 1); and 2) insufficient recent observations within the DR window for a time series model to be trained reliably. Instead, historical data from past DR events can be used as predictors for reduced consumption. Another challenge is that reduced consumption is affected by several factors such as the time of day and day of week, and DR factors such as curtailment strategy, human behavior, as well as environmental factors, such as temperature. Table 1: Normal consumption, Reduced consumption, and DR baseline: Key characteristics and challenges Prediction Task Goal Prior work Timing Historical Data Compute Requirements Profile changes Normal Consumption Planning, DR Several Outside the DR event Readily available Off-line or real-time Gradual Counterfactual DR Baseline Curtailment calculation Several During the DR event Readily available Off-line Gradual Reduced Consumption Planning, DR, dynamic DR None During the DR event Sparse or nonexistent Real-time for dynamic DR Abrupt at the DR event boundaries Our contributions in this paper are: • We identify key characteristics and challenges of reduced consumption prediction problem. • We use diverse predictors in a novel ensemble that use different sequences of daily electricity consumption on DR event days, as well as contextual attributes, for reduced consumption prediction. The low computational complexity of our model makes it ideal for real-time applications such as dynamic demand response (Aman et al. 2015). • We evaluate our model on a large real-world dataset from a university microgrid. Our model achieves an average error of 13.5%, an 8.8% improvement of over the baseline.