Seite - 67 - in Intelligent Environments 2019 - Workshop Proceedings of the 15th International Conference on Intelligent Environments

For small and isolated electric systems, their inability to be tapped into a continental transmission grid for emergency power, results in reliability and other economies of scale impacts. Thus, the power systems operate on the premise that the load is uncontrollable and that system voltage, frequency and stability are primarily maintained through the real-time control of the generation. Frequency is a necessary parameter for load control in interconnected systems, as it is considered as a measure of mismatch between demand and generation. However, in the great majority of islands and isolated areas power systems that are not extensively interconnected with many generators and loads, will not maintain frequency with the same degree of accuracy and therefore cause an automatic load shedding or other control actions such as temporary power outages. This scenario is more probable during heavy loads periods, when the accuracy of the system is hard to maintain [20,29]. Whatever the case, effective planning requires a thorough understanding of the prevailing electricity demand patterns. Thus, for modeling and forecasting purposes the existing literature has extensively analyzed the main features of demand. Actually, electricity consumption is subject to great cyclical and seasonal effects (daily and weekly cycles, holidays), special events, nonlinearity of meteorological variables and possible nonlinear time dependence, etc [2,18,19,28,30]. However, the specificity of an open space, where the mobility of people outside and within a country is usually unrestrained, impedes the inclusion of an accurate variable that captures the daily population pressure effect. Contrastingly, this indicator can be accurately measured in islands and isolated territories that control their population transit. For instance, in the case of islands, where the only means of access and exit are ports and airports, it is expected that for security reasons there is a thorough control on the daily flow of incoming and outgoing passengers. Majority of islands have the possibility to assess the daily population’s weight present in their territory, which is generally characterized by its high fluctuations and seasonal aspects. A natural a priori hypothesis is that the daily electricity demand depends on the population stock, and most likely this dependence is very relevant in isolated territories where high seasonal fluctuations could easily affect the efficiency of the electrical system and provoke serious and costly problems. A high level of seasonality is a distinctive feature of coastal tourist regions, a problem that often characterizes islands too. Therefore, the possibility of having a daily population indicator will definitely fulfils the need of an accurate forecasting model that can predict future electricity demand, and provide the utility company with a model that reduces the gap between supply and demand and its concomitant cost. This article reports on the design and implementation of a medium-run forecasting model for daily system loads and an evaluation of the forecast performance of the Balearic Islands (Spain). An archipelago located in the western Mediterranean Sea that includes four inhabited islands Majorca, Minorca, Ibiza, and Formentera. The last two form what is known as the Pitiüses, a special unity in terms of electricity system. The Balearic Islands system supplies electricity annually to 1 million residents and 13 million tourists concentrated during summer months. The Balearic Islands are taken as a case study because of their geographical characteristics and relative importance of tourism in the region. M.BakhatandJ.RosselloNadal / ImprovingDailyElectricityLoadsForecasting 67