Similarity search in electricity prices: An ultra-fast method for finding analogs

Abstract

Accurately predicting electricity prices allows us to minimize risks and establish more reliable decision support mechanisms. In particular, the theory of analogs has gained increasing prominence in this area. The analog approach is constructed from the similarity measurement, using fast search methods in time series. The present paper introduces a rapid method for finding analogs. Specifically, we intend to: (i) simplify the leading algorithms for similarity searching and (ii) present a case study with data from electricity prices in the Nordic market. To do so, Pearson's distance correlation coefficient was rewritten in simplified notation. This new metric was implemented in the main similarity search algorithms, namely: Brute Force, JustInTime, and Mass. Next, the results were compared to the Euclidean distance approach. Pearson's correlation, as an instrument for detecting similarity patterns in time series, has shown promising results. The present study provides innovation in that Pearson's distance correlation notation can reduce the computational time of similarity profiles by an average of 17.5%. It is worth noting that computational time was reduced in both short and long time series. For future research, we suggest testing the impact of other distance measurements, e.g., Cosine correlation distance and Manhattan distances.