Robust functional principal component analysis for detecting anomalous behaviors in electricity markets

This paper introduces a novel methodological framework to detect potentially manipulative behaviors in deregulated electricity markets using robust statistical tools. The work focuses on identifying outlying bidding patterns in daily auction microdata by modeling the shape of supply curves, rather than relying solely on price data, to better capture strategic market behavior. The approach is grounded in robust Functional Principal Component Analysis (FPCA), which enables the detection of anomalies in supply curve shapes while being resilient to outliers. A key innovation lies in applying the skewness-adjusted boxplot of Hubert and Vandervieren (2008) to the residuals from robust FPCA, enhancing sensitivity to asymmetric and extreme behaviors. Crucially, it is shown that the anomalies detected via robust FPCA differ significantly from those identified by classical outlier detection methods applied directly to price series, as the suggested method captures deviations in the underlying strategic behavior of market participants that are reflected in the structure of the supply curves, not necessarily in prices. Applied to the Italian day-ahead market, the method detects supply-curve anomalies that differ substantially from those identified by classical price-based techniques. A comparison with the LTSts and rolling-window filtering approaches confirms the distinct contribution of the proposed method, which identifies a complementary set of suspicious events potentially linked to strategic bidding behavior. The findings provide new tools for regulators to support market integrity and ensure compliance with transparency regulations such as the Regulation on Wholesale Energy Market Integrity and Transparency (REMIT).