Decision-Focused Forecasting: A Differentiable Multistage Optimisation Architecture
This work addresses multistage decision problems for domains like energy and finance, where decisions have intertemporal effects, but it is incremental as it extends existing decision-focused learning to multistage contexts.
The paper tackles the problem of multistage decision-making under uncertainty by proposing a differentiable multistage optimization architecture for decision-focused forecasting, and it reports that the model outperforms existing approaches in energy storage arbitrage and portfolio optimization applications.
Most decision-focused learning work has focused on single stage problems whereas many real-world decision problems are more appropriately modelled using multistage optimisation. In multistage problems contextual information is revealed over time, decisions have to be taken sequentially, and decisions now have an intertemporal effect on future decisions. Decision-focused forecasting is a recurrent differentiable optimisation architecture that expresses a fully differentiable multistage optimisation approach. This architecture enables us to account for the intertemporal decision effects of forecasts. We show what gradient adjustments are made to account for the state-path caused by forecasting. We apply the model to multistage problems in energy storage arbitrage and portfolio optimisation and report that our model outperforms existing approaches.