2017 IEEE PES Innovative Smart Grid Technologies Conference - Latin America (ISGT Latin America). Quito, Ecuador 978-1-5386-3312-0/17/$31.00 ©2017 IEEE Importance of Hourly Multi-Bus Unit Commitment Models in the context of High Adoption of Variable Renewable Energies: A Chilean Example Fernanda Ávila Javier Ayala Pablo Cerda Alejandro Navarro-Espinosa* Samuel Córdova Hugh Rudnick* Systep/Universidad de Chile* Systep Systep/Pontificia Universidad Católica* Santiago, Chile Santiago, Chile Santiago, Chile Abstract— The increasing levels of Variable Renewable Energies (VRE) require additional levels of flexibility in present power systems due to their intermittency and the resulting needs of reserve for increasing/reducing the generation with enough speed to maintain the system balance. In this context, the traditional long-term hydrothermal optimal dispatch models are not sufficient to adequately represent this situation, particularly in systems with hydro reservoirs that can store water from one period to another. These models do not use sequential representation of time periods (on the contrary, they work with load blocks) and they do not incorporate most of the non-linear constraints to model specific characteristics such as: minimum power, ramp up/down rates, stabilization times, minimum time in service, minimum down-time, etc. Therefore, more detailed models must be used to comprehensively represent the impact of VRE in the system operation. To assess the true cost of power system operation with high adoption of VRE, this work develops a detailed multi-bus hourly unit commitment model with all non- linear constraints, showing the importance of this type of model for assessing cycling and start-up in the new VRE paradigm. Index Terms — unit commitment, MILP, cycling, hourly unit commitment. I. INTRODUCTION Nowadays, climate change is one of the most important problems that our society is facing and to tackle it, specific mitigation and adaptation measures can be taken [1]. Among the mitigation actions, it is possible to find the energy generation through clean energy. Thus, several countries have incorporated renewable energy targets in their energy policies. For instance, the European Union is expected to generate, by 2020, 20% of its energy demand from renewables energies [2]. In the case of electricity, UK has set a target of 30% [3] and Chile set a goal of 20% by 2025 [4]. In this context, power systems around the globe are incorporating Variable Renewable Energies (VRE), particularly wind and sun, which will require additional levels of flexibility. Flexibility should be understood as the capability of a power system to support the variability and uncertainty produced by VRE [5]. In fact, the changing nature of wind and solar power generators requires additional reserves to provide power if the primary sources (e.g., wind or sun) are not available in a specific moment. Furthermore, the system must be able to change the power output of other technologies (ramp up or down) to maintain the power-demand balance when the variable energy resources are fluctuating. This intermittency will require much more flexibility from the thermal fleet in order to cope with the high variation from period to period for VRE (e.g., sun and wind). This flexibility will trigger the cycling of thermal units, which refers to the need for these units to be turned off and on or to be operated at part-load or technical minimum more often. To cope with the increased variability, thermal units must cycle in order to maintain system balance [6]. This cycling entails additional costs for thermal power plant operators, due to extra fuel costs and wear and tear of components [7]. To measure those flexibility requirements, the traditional long-term hydrothermal optimal dispatch models are not enough to adequately represent this situation, particularly in systems with hydro reservoirs that can store water from one period to other. Hence, to solve the problem in reasonable computational times some simplifications are included. For example, the months are represented though a by-block modelling, losing the chronological representation of hours along the year, and they do not properly model (or even they do not consider) most of the non-linear constraints required to properly model specific characteristics related closely to flexibility requirements. For instance, they do not include technical minimums, ramp up/down rates, stabilization times, minimum time in service, minimum down time, etc. The unit commitment (UC) formulation, which determines the units that should be online and offline throughout a period represented chronologically [8], is useful to assess flexibility requirements. Most UC problems cover short-term time horizons (typically one day or one week) [9],[10],[11] and some of them have also incorporated the stochasticity related to VRE [12]. For instance, [13] deals specifically with wind uncertainty whilst [14] copes high renewable penetration (sun and wind). However, those models do no incorporate the long term reservoir management (i.e., Hydrothermal systems) due to the large mathematical dimensionality and computing times [15]. To overcome that challenge, this work develops a detailed ___________________________________________________________ This work was supported by Systep Ingeniería y Diseños (www.systep.cl), and was developed by its Research & Development and Market Intelligence departments.