Forecasting methods in AMIRIS

Are any AI or statistical methods used for forecasting? Are the agents learning and adapting their behavior? How do you implement forecasting without statistical or AI methods?

AMIRIS already comprises a forecasting agent that provides information on future prices to other agents. Currently, it does not rely on AI methods but requires information on future bids from the traders. This allows using perfect information for a flexibility agent (e.g. storage) but supports a single flexibility agent only. To allow having multiple competing flexibility agents, however, we at DLR are already experimenting with AI forecasting methods and plan to implement them in AMIRIS in the near future. Please see also this presentation.