Achieving net-zero carbon emissions necessitates the major transformation of electrical grids into smart grids. In this context, urban districts play a crucial role in the flexible balancing of electricity demand and supply, which involves solving decentralized optimization problems. Such optimization problems rely on forecasts of local demand and supply, and require the systematic orchestration of data streams using cloud services. At the same time, it is necessary to automate both the design and operation of forecasting models in such services to keep pace with the increasing need for such locally adapted forecasts. Therefore, this paper proposes an automation level taxonomy to communicate the scope of automation in time series forecasting. Furthermore, we demonstrate a forecasting service that is used in a downstream demand-side management application and realized in the real-world project Smart East in Karlsruhe, Germany. Finally, we analyze existing forecasting services in the literature, categorize them according to the proposed automation level taxonomy, and compare them with our implementation.