Unit Commitment Based on Risk Assessment to Systems with variable Power Sources P. M. Fonte Electrical Engineering Department Lisbon Superior Engineering Institute, ISEL Lisbon, Portugal pfonte@deea.isel.pt Abstact- This paper presents the development of a complete methodology for power systems scheduling with highly variable sources based on a risk assessment model. The methodology is tested in a real case study, namely an island with high penetration of renewable energy production. The uncertainty of renewable power production forecasts and load demand are defned by the probability distribution function, which can be a good alternative to the scenarios approach. The production mix chosen for each hour results from the costs associated to the operation risks, such as load shed and renewable production curtailment. The results to a seven days case study allow concluding about the difculty to achieve a complete robust solution. Index Terms-Power generation scheduling, risk assessment, uncertainty. I. INTRODUCTION Increasing introduction of electric energy production with Renewable Energy Sources (RES), and mostly those with high variability, has created several challenges to the energy networks operators, especially in the scheduling. This problem is boosted in low power networks, particularly in islands without any connection to continental networks. Large variations on renewable production can introduce stability problems in the network, which can originate generation or load shed and, at limit, black-outs a strong possibility[ 1]. When available, the RES production allows thenal production decrease, especially during the peak load periods. Optimizing the number and the power of the on-line thenal units lowers cost and emissions. On the other hand, an extreme reduction of the thermal committed capacity can lead to a situation where the spinning reserves are not suficient to handle with great variations of load, renewable production or generation outages. Therefore, due to the uncertainty in load and renewable production forecast, it is sometimes hard to find a completely robust/economic scheduling solution. With this into consideration and for security, scheduling is generally done by a conservative way, with low risk, although sometimes far away from an optimal operation. As such there is the necessity to introduce Claudio Monteiro l Fernando Maciel Barbosa ! , 2 l FEUP - Engineering Faculty, University of Porto 2 INESC TEC Porto, Portugal cdm@fe.up.pt, fmb@fe.up.pt uncertainty of load/RES in scheduling for achieving a better management of the thermal unit' s commitment. The stochastic programming is an approach widely used to deal with the generation scheduling under uncertainty applying recourse problems, chance-constrained or robust optimization, with uncertainty described by scenarios [2] [17]. The scenario-based approach demands a great number of realizations in order to capture the temporal interdependence of the probabilistic behavior of the uncertainty. One of the main problems of this approach is that it is time consuming to solve all scenarios, being necessary to appeal to some kind of scenario reduction. To overcome this problem, this work develops a short-term scheduling approach to be used in insular power grids based on risk assessment, addressing the increase of variability and uncertainty created by RES. II. DESCRIPTION OF THE METHODOLOGY The proposed generation scheduling is designed to minimize the sum of the estimated costs based on risk cost analysis. These costs result from the estimated normal operation cost plus the estimated cost of operating outside normal conditions. It is understood as "abnormal" conditions if there is the necessity of load shed due to the lack of available thermal production or RES curtailment caused by the lack of load. The risk of load shed or RES curtailment and thermal production below the technical minimums are used to defne the objective function, as well as the probability of the thermal generators operating inside the appropriated ranges. Contrary to widely used scenarios-based approach, in this work it is proposed the probabilistic estimation of costs based on estimation risk, directly using the probability density function of the random variables. Knowing the probability function of net load (LN), obtained by load minus the renewable production (L-RES) [1],[5],[6],[9],[15], for each hour h of the scheduling period, the ability of each thermal GENeration mix SET (GENSET) to meet the net load is verified. Notice that in the risk assessment approach there are no infeasible solutions, only more or less costly solutions. These decisions have to be made to accept a risk as long as it 978-1-5090-3474-l/16/$3l.00 ©20l6 IEEE 3924