MIPRO 2015, 25-29 May 2015, Opatija, Croatia Review on Unit Commitment under Uncertainty Approaches Kristina Jurkovic, Hrvoje PandZic and Igor Kuzle University of Zagreb Faculty of Electrical Engineering and Computing, Zagreb, Croatia {kristina.jurkovic, hrvoje.pandzic, igor.kuzle} @ fer.hr Abstract - Wind power has already become an important renewable energy resource in many regions of the world. Because of its variability and uncertainty, integration of wind power presents a challenge that, if not adequately addressed, can jeopardize the operational reliability of a power system. Generally, generation unit commitment decisions are made once a day, i.e., the commitment decisions are made 24 or more hours ahead of the actual operation. Taking into account the uncertainty of wind power prediction, these decisions need to provide sufcient fexibility at a minimum price. This paper describes the current practice and analyzes unit commitment formulations available in literature highlighting their advantages and shortcomings. I. INTRODUCTION The primary concer in operating an electrical power system is to meet the demand for electricity at all times and under different conditions depending on the season, the climate, and the weather. Moder power systems are supposed to accommodate large total capacity of distributed, volatile generation, as well as large-scale price responsive demand and electric vehicles which dramatically increases both supply and demand uncertainty [1 ]-[3]. Because of its variability and uncertainty, wind generation impacts power system operation and can potentially jeopardize its reliability. To deal with the larger uncertainty on the net load (the difference between electricity demand and the output of non-dispatchable generation), power system operators are increasing the reserve margins, thus increasing the regulation cost [4]. In order to minimize the operating cost of non dispatchable resources, it is essential to derive a computationally effective approach to optimally select the units and their output level to preserve the operational reliability of the system. Unit commitment (UC), one of the most critical decision processes, is an optimization problem that generates the outputs of all the generators in a way that minimizes the system-wide fel cost. Features included in most moder unit commitment models include generator minimum and maximum generation limits, ramping limits, minimum up and down time constraints, time-dependant start up costs and transmission capacity limits [5J-[8]. The work of the authors is a part of the FENTSG - Flexible Energy Nodes in Low Carbon Smart Grid fnded by Croatian Science Foundation under project grant No. 7766 During the normal operation, system operator dispatches the committed generation resources to satisf the actual demand and reliability requirements. In the event of a signifcant deviation between the actual and the expected system condition, system operator needs to take corrective actions, such as committing expensive fast-start generators, voltage regulation or load shedding, to maintain system security. The main causes of the unexpected events come from the uncertainties associated with the load forecast error, changes of system interchange schedule, and unexpected transmission and generation outages. [9] Deterministic UC formulation is a traditional solution in which the net load is modeled using a single forecast for each wind far output and the associated uncertainty is handled using ad-hoc rules, i.e., the generating units are committed to meet the deterministic forecast and the uncertainty is handled by imposing reserve requirements [1 0]-[1 3]. Such an approach is easy to implement in practice, but the ad-hoc rules do not necessarily adequately account for this uncertainty. Namely, committing extra generation resources for reserve is economically inefcient, while the power system may still suffer from capacity inadequacy in case of a signifcant deviation between real-time and expected net load. There is a lot of research on optimizing the reserve requirements based on deterministic criteria [1 4]-[1 7]. In [1 4] a new technique to determine the SR requirements at each period of the optimization horizon is proposed using a costibeneft analysis. Similarly, in [1 5] the cost of interruptions is considered when optlmlzmg the scheduling of spinning reserve. In [1 6J a probabilistic analysis of the reserve requirements is taken into account. The authors of [1 7J show that reserve requirements cannot be specifed a priori without sacrifcing the optimality. A more rigorous approach is incororating uncertainty in the unit commitment model itself, which is the focus of this review paper. Section II describes stochastic unit UC, Section III robust UC frmulation, while Section IV describes interval UC formulation. Section V describes some recent advancement in hybrid UC models that combine the aforementioned forulations. Conclusions are duly drawn in Section VI. II. STOCHASTIC UNIT COMMITMENT Stochastic UC is based on probabilistic scenarios. A fnite set of scenarios is generated and assigned weight in proportion to their likelihood. Stochastic UC IS 1093