| Abstract: |
This paper introduces a novel probabilistic forecasting technique called
Smoothing Quantile Regression Averaging (SQRA). It combines Quantile
Regression Averaging - a well performing load and price forecasting approach -
with kernel estimation to improve the reliability of the estimates. Three
variants of SQRA are evaluated across datasets from four power markets and
compared against well-established benchmarks. Empirical evidence indicates
superior predictive performance of the method in terms of the Kupiec test, the
pinball score, and the conditional predictive accuracy test. Moreover,
considering a day-ahead market trading strategy that utilizes probabilistic
price predictions and battery storage, the study shows that profits of up to 9
EUR per 1 MW traded can be achieved when forecasts are generated using SQRA. |