Evaluation of Methods for Long-term Hydropower Planning in Electricity Market Models

This work compares three approaches for long-term hydropower scheduling in electricity market models: the established EMPS and BID3 models, and a proposed look-ahead dispatch method. Using out-of-sample simulations over 100 weather years, BID3 delivers the lowest average electricity costs and avoids load shedding, whilst calibration improves results for all models.

This paper, presented by Svenska Kraftnät, evaluates and benchmarks three methods for long-term hydropower planning: EMPS, BID3, and a new look-ahead dispatch approach. The analysis uses a two-area test system modeled on Swedish conditions, with extensive scenario testing across 100 synthetic weather years. Key findings include:

  • Performance: BID3 outperforms the other models in terms of average electricity costs and reliability, as it avoids load shedding even before calibration. EMPS and the look-ahead dispatch both experience some load shedding, which increases system costs
  • Calibration impact: All models benefit significantly from calibration. For BID3, reducing risk aversion by adjusting water value bidding steps lowers costs. For EMPS and the look-ahead dispatch, increasing risk aversion reduces load shedding and overall costs
  • Model characteristics: BID3’s approach to water value calculation, which considers interdependencies between reservoir regions, is particularly effective in the two-area system tested. The look-ahead dispatch, while slightly less optimal in cost, is simpler to implement and more adaptable to systems with additional flexibility resources such as batteries or hydrogen storage
  • Practical implications: The study underscores the necessity of ongoing model calibration to reflect operational realities and changing system dynamics. It also highlights the potential of the look-ahead dispatch method for future electricity markets with high shares of renewables and storage, due to its flexibility and ease of integration

Overall, the research provides actionable insights for system operators and market modelers, emphasizing that while BID3 currently offers the best performance in the tested scenario, all methods require careful tuning.