Forecasting of volatility and risk premia in electricity markets

Battle Count: 73

Highly relevant for electricity market participants and quantitative traders in energy derivatives. The paper provides practical forecasting tools for the full within-day covariance structure of electricity prices, directly applicable to spread trading (peak vs baseload), portfolio risk management, and forward market pricing. The RCV-based risk premium predictions nearly double the R² of classical methods. However, the focus on electricity markets (rather than equities or FX) and the specialized infinite-dimensional framework somewhat limit broader applicability. The renewable energy covariates and quarterly HAR component are particularly timely given energy transition dynamics.

The theoretical framework involves infinite-dimensional SPDEs and Hilbert space operators, but the practical implementation reduces to matrix operations: kernel regression for de-trending, semigroup estimation via matrix ratios, and standard linear regressions for the HAR components. The DRD decomposition and separate modeling of volatilities and correlations add moderate complexity. The main challenges are: (1) proper de-trending with Epanechnikov kernel, (2) semigroup estimation, (3) ensuring positive definiteness of forecasted matrices, and (4) the model confidence set computation with block bootstrap. Overall, a skilled quantitative analyst could implement the core forecasting models in 2-4 weeks.

Data is publicly available on the ENTSO-e transparency platform. The methodology is fully specified with explicit equations for the RCV estimator, semigroup estimation, kernel de-trending, and all model specifications. However, the infinite-dimensional SPDE framework and some implementation details (e.g., exact kernel bandwidth selection, bootstrap procedures) require careful replication. The paper references prior work (Kloster and Benth, 2026) for foundational results.