Results:

The results below illustrate the average hydropower operations when using the following forecast information:

Baseline solution (gray line): the model is informed by a naïve forecast, which is the inflow climatology (based on past observed flows).

It represents the most basic information one can have if only historic observations (no numerical forecast models) are available.

Perfect solution (black line): the model is informed by a perfect knowledge of future inflows (actual observed flows).

It represents the best situation one could have if the future inflows were known in advance.

E-HYPE solution (cyan line): the model is informed by the 90-day accumulated inflows from the E-HYPE forecasting system.

It represents an intermediate situation between the previous two: the forecast information is used to improve the reservoir operation providing information on future inflows over the next 90 days, given updated meteorological and hydrological conditions.

Average monthly values of reservoir inflows (top left), daily electricity prices (top right), reservoir storage levels (bottom left) and hydropower revenue (bottom right) for the three solutions investigated (perfect in black, baseline in gray, E-HYPE in cyan), representing different forecast information used as inflow to a hydropower reservoir model.

• We can see that the inflow peak occurs during the melting period, in the 2nd half of July (top left panel).
• All solutions make use of this information and drawdown the reservoir level before the peak inflow (bottom left panel). The reservoir can then store the peak inflow for future use of this water.
• After the refill of the reservoir, all solutions generally keep high levels in the reservoir during the summer season (June-August) (bottom left panel).
• The baseline solution is the most conservative one: it keeps a larger storage buffer in the reservoir to avoid spilling in case of unexpected inflow peaks.
• This strategy, combined with some days of high electricity price in July (top right panel), allows the revenue from the baseline solution to be higher than the other two solutions during the summer (bottom right panel).
• The fall period (September- November) is when the forecast information brought by the E-HYPE modelling system becomes valuable:
  • the limited foresight of the baseline suggests adopting a greedy strategy that maximizes the revenue on the short term, i.e. drawing down the storage in the period September-December, when the price increases after the holiday period of August.
  • Conversely, both the Perfect Solution and the one informed by E-HYPE forecast capitalize on the forecasts of future inflows to allocate most of the production in the period January-March, when the price of electricity is higher than in the fall. Such delayed drawdown, efficiently scheduled in the high-price months, generates a large gain in terms of revenue over the baseline (as shown in the table).