Including snow pack series or preprocessed inflow series in inflow estimation

Theory¶

There is information about the future inflow stored in the snow pack, and for some areas, a long-term inflow trend may influence the future inflow.

It is possible to include the influence from snow pack series, or preprocessed inflow series like long-term inflow series, in the model adaptation.

This results in a generated inflow that is dependent on both the inflow of the previous week, and the additional inflow series:

z_{inflow,t} = \varphi_{inflow,inflow} z_{inflow,t-1} + \varphi_{snow} z_{snow, t-1}+\epsilon_{inflow,t}

(1)

The additional inflow series also influences the cuts:

\alpha_t - \sum_{r \in R} \pi_{rc} v_r - \mu_{inflow,c}\left( \varphi_{inflow,inflow} z_{t-1,inflow} + \varphi_{inflow,snow}z_{snow,t-1} \right) - \mu_{snow,c}\left( \varphi_{snow,inflow} z_{inflow,t-1} + \varphi_{snow,snow}z_{snow,t-1} \right) \leq \beta_c

(2)

How to include additional inflow series in the model estimation¶

By creating a dummy module and adding the additional inflow series (or snow pack series), the inflow series will be used in the estimation of the parameter matrix for the inflow model.

If not using pyprodrisk: Add snow pack and/or long-term inflow trends to the inflow archive (e.g. using the arcltm program)
Create a dummy-reservoir with the snow storage or long-term trend as inflow series and with 0 energy equivalent. Allow high maximal discharge and let the reservoir volume be large enough (e.g. the size of the largest reservoir of the watercourse). This allows for testing with all inflow models (also the principal component model).
Run Prodrisk. The dummy reservoir does not add any produced power, but the additional inflow series will influence the original inflow series.