Exploring the added value of hydrological ensemble forecast optimization for hydropower scheduling
MSc topics in collaboration with SINTEF and Deltares
1 Introduction
Hydropower producers make commitments for energy production on the day ahead market (24 hours ahead). The energy price and the available water – the current storage and expected inflow – are important decision parameters for hydropower scheduling. However, as the inflow forecast is uncertain, a power supply commitment based on expected inflow comes with a certain financial risk: If the expected water is not available, the commitments cannot be fulfilled, which comes with additional costs. At the same time, hydropower production and reservoir management must account for operational constraints from flood protection and environmental obligations.
Within this project we want to investigate how ensemble optimization techniques contribute to better decision making for hydropower scheduling under hydrological inflow forecast uncertainty. In addition to better informed decisions in general, lower financial risks and better fulfilment of environmental obligations, we expect that this study contributes to a more efficient use of the existing hydropower fleet with the available water on the long term in the light of changing boundaries (climate and socio-economic development).
The study is divided into two sub-tasks, each to be addressed within one MSc project.
2 MSc projects
2.1 Operational hydrological forecasting for hydropower scheduling
2.1.1 Research question
How should a hydrological model for operational hydropower scheduling look like in terms of
- Spatial and temporal discretization
- Data needs
- Computational performance
- Accuracy?
The purpose of the hydrological model provides reservoir inflow time series for operational hydropower scheduling. The model output will be used as input by a reservoir optimization model (Section 1.2).
2.1.2 Activities
- Development of a distributed hydrological model (rainfall-runoff) with wflow hydrological modelling software
- Define the modelling area, temporal and spatial resolution
- Data collection from global and local data (static data: land use, elevation etc.)
- Model setup in wflow softwrae
- Integrate model into an operational decision support system (Delft-FEWS)
- Calibration and validation of the hydrological model
- Calibration and validation with the help of selected scenarios
- Analyse the model forecasting skills with respect to specific processes (snow melt) and scenarios (flood, drought)
- Operationalize model, ensemble forecasts
- Connect model to input data sources for regular automated update (once or multiple times per day), integration into an operational data management framework (Delft-FEWS)
- Process ensemble data to model input
- Display ensemble forecasts for decision support
- Generation of a scenario data set for a hindcast experiment to be used for Optimal hydropower scheduling under inflow ensemble forecast (Section 2)
Figure 1 Hydrological model wflow for a sub-catchment in Nord-Gudbrandsdal (Norway). Picture source: SINTEF/Deltares
2.1.3 Output
- A wflow model for a sub-catchment of interest in Norway
- A scenario dataset
- An MSc thesis describing development steps and requirements for development of a hydrological model to be used for operational hydropower scheduling
2.1.4 Software
2.2 Optimal hydropower scheduling under inflow ensemble forecast
2.2.1 Research question
How can optimization under an inflow ensemble forecast provide an added value for decisions on hydropower scheduling on real-time or short-term time scale (24 hours, multiple days)?
This project aims to develop an optimization model of a hydropower reservoir. This optimization model uses an inflow ensemble forecast (MSc project described in Section 2.1) to generate an optimal hydropower schedule. Different optimization techniques for ensemble forecast are available.
2.2.2 Activities
- Development of an optimization model for a hydropower facility (reservoir, generation units, river reaches)
- Application of different ensemble optimization techniques on the optimization model:
- Individual optimization of all ensemble members
- Holistic optimization: one release scheme that satisfies all ensemble members as good as possible
- Tree-based optimization: on specific branching points similar ensemble members are grouped and optimized together
- Process the optimization result to information that can be used for decision support
- Development of performance indicators (energy production, load balance, revenue, fulfilment of operational objectives, robustness of the schedule).
- Integration of the optimization model in an operational data management framework (Delft-FEWS)
- Hindcast experiment and evaluation of different ensemble optimization techniques in terms of performance indicators and user-experience of the decision support
2.2.3 Output
- An optimization model
- MSc thesis describing how different ensemble optimization techniques can provide decision support for hydropower scheduling and how to interpret the results.
2.2.4 Software
- SHOP[3] (SINTEF)
- Delft-FEWS (Deltares)
3 Contact information
[1] https://www.deltares.nl/en/software/wflow-hydrology/
[2] https://www.deltares.nl/en/software/flood-forecasting-system-delft-fews-2/
[3] https://www.sintef.no/en/software/shop/