Operations modeling in the Iceland hydro dominated power system

The Icelandic power system is based primarily on hydroelectric and geothermal generation. The system has large resources compared to the size of Iceland's economy. The main utilization practice for electrical energy hitherto has been bulk Energy Intensive Industry (EII) with relatively flat and constant load. A possible future interconnection to neighboring countries via High Voltage Direct Current (HVDC) submarine cable could affect dramatically the operations planning framework, for instance with possibilities of linking to wind power, local and/or foreign and short term markets. This should affect factors such as operation strategies for hydro system and the planning and operations tools needed to realize this rent. Planning tools and modeling for hydro dominated systems may in some respects inherently be more complex than those of thermally dominated systems, due to the time-interdependence associated with reservoir storage and various constraints [4]. In this paper we give an overview of historical and current modeling practices used both as a tool to form system expansion policies and operations strategies in the long, medium and short term. The basic time frame requirements for such tools are discussed in this paper and how the may evolve in the future in terms of methodology and other important issues. Conclusions will be drawn regarding these planning tools and the future planning and operations framework in such a system, especialy the time frame issue. The long and medium term hydro-scheduling problem has been investigated since in the early part of the 20th century [14], applying various optimization and iterative simulation methodologies. The practices applied in the Icelandic system have traditionally been based on Scandinavian approaches, such as those originally suggested by [9], [17] and [2]. These methods in hydro dominated systems have been based on an iterative calculation of water value tables based on iterative dynamic programming like algorithms. Future methodology may involve LP or SDDP [16], but with a short term considerations (DTM, see below) to certain extent as outlined in this paper.