Zusammenfassung
This paper describes the extension of classical forecasting methods for an application to hierarchical data structures. We show that various methods of hierarchical coupling can improve forecasting results using hierarchical relationships considerably. In a concrete application we are able to reduce the relative error of a retail forecasting model by 10%.

Zusammenfassung
The research project IO.Netz(*) aims to improve the cur- rent process of long term distribution system planning. Analysing and planning today’s distribution systems is still characterized by isolated software tools so that network planners have to deal with a list of shortcomings. Furthermore they have to deal with an increasing complex environment and have to include additional aspects (e.g. uncertainty for the investment decisions with lower budgets; development of renewable sources). The challenge to embed decentralized renewable energy sources into the distribution network implies a tight integration of the software tool chain for planning decision support. This paper proposes to raise the synergies between the replacement strategies in asset management and investments driven by the inclusion of renewable sources. Our central approach estimates the realization probability of new decentralized generation sites, simulates grid development by a system dynamics approach, calculates investments under uncertainty and applies multi-criterial optimization based on the simulation model. (*) This work has been supported in the R&D project “IO.Netz” funded by BMWi, grant no.: 03ET1071