A stochastic model for scheduling energy flexibility in buildings Stig Odegaard Ottesen a, * , Asgeir Tomasgard b a eSmart Systems and Norwegian University of Science and Technology, Institute for Industrial Economics and Technology Management, Norway b Norwegian University of Science and Technology, Institute for Industrial Economics and Technology Management, Norway article info Article history: Received 27 September 2014 Received in revised form 12 May 2015 Accepted 13 May 2015 Available online xxx Keywords: Demand side management Smart grids Scheduling Short-term flexibility Stochastic programming abstract Due to technological developments and political goals, the electricity system is undergoing significant changes, and a more active demand side is needed. In this paper, we propose a new model to support the scheduling process for energy flexibility in buildings. We have selected an integrated energy carrier approach based on the energy hub concept, which captures multiple energy carriers, converters and storages to increase the flexibility potential. Furthermore, we propose a general classification of load units according to their flexibility properties. Finally, we define price structures that include both time- varying prices and peak power fees. We demonstrate the properties of the model in a case study based on a Norwegian university college building. The study shows that the model is able to reduce costs by reducing peak loads and utilizing price differences between periods and energy carriers. We illustrate and discuss the properties of two different approaches to deal with uncertain parameters: Rolling ho- rizon deterministic planning and rolling horizon stochastic planning, the latter includes explicit modeling of the uncertain parameters. Although in our limited case, the stochastic model does not outperform the deterministic model, our findings indicate that several factors influence this conclusion. We recommend an in-depth analysis in each specific case. © 2015 Elsevier Ltd. All rights reserved. 1. Introduction According to the IEA [1], demand side activities should be the first choice in all energy policy decisions that aim to create more reliable and sustainable energy systems. Demand side activities integrated with smart grid technologies [2] represent a wide vari- ety of benefits for different stakeholders in the energy value chain and society as a whole. Examples are: cost reductions for con- sumers, increased ability to integrate intermittent renewable po- wer generation and electric vehicles, improved energy system reliability and less costly network reinforcements [3e6]. Many studies quantify the potential benefits from demand side activities [7e11] with respect to reductions in cost, peak demand and emissions. In this paper we will present a decision-support model that can be used to control the demand side flexibility in a building. A price elastic inverse demand curve is a simplified represen- tation of flexible demand [12,13]. This representation is not suffi- cient to describe demand response, as it lacks an explicit link to the underlying physical energy system and thereby also an inter- relation between time periods. It disregards the fact that chang- ing the load in one period may affect demand and the feasible decision space in later periods. Several authors have addressed demand response in short-term multi-period optimization models. Conejo et al. [14] introduce a real-time electricity demand response model for a household or a small business where a minimum daily energy-consumption level must be met, constrained by maximum and minimum hourly load levels. Gatsis and Giannakis [15] split the load of a residence into three components: one “must-run”, one adjustable where the total amount must be met over the sched- uling horizon and finally one that can be reduced, but at the dissatisfaction of the end-user. In Refs. [16] and [17] the concept of deferrable loads is introduced with limits for start- and end-time in addition to minimum and maximum load levels and total load. A combination of the above-mentioned concepts is presented by Hong et al. [18], who describe an approach to allocate load among individual appliances. Loads are categorized into non-shiftable, shiftable and controllable. In addition to reducing cost by shifting loads from high-price to low-price hours, their model seeks to reduce the number of peak demand hours. Finally, [19] and [20] take into consideration that some types of loads cannot be inter- rupted or changed when first started. In real life, different appli- ances will fit into variations of the above representations. * Corresponding author. Tel.: þ47 90973124. E-mail address: stig.ottesen@esmartsystems.com (S.O. Ottesen). Contents lists available at ScienceDirect Energy journal homepage: www.elsevier.com/locate/energy http://dx.doi.org/10.1016/j.energy.2015.05.049 0360-5442/© 2015 Elsevier Ltd. All rights reserved. Energy xxx (2015) 1e13 Please cite this article in press as: Ottesen SO, Tomasgard A, A stochastic model for scheduling energy flexibility in buildings, Energy (2015), http://dx.doi.org/10.1016/j.energy.2015.05.049