The long-term trends on the electricity markets: Comparison of empirical mode and wavelet decompositions*

- An improved CEEMDAN-based approach to electricity price trend-cyclical component filtering is proposed
- A combined criterion for determining the modes to be included into the trend component is introduced
- CEEMDAN outperforms the EMD
- CEEMDAN is perform better or at least not worse then wavelet-smoothing
- The CEEMDAN approach does not require a prior choice of the parameters and demonstrates a certain degree of versatility

This paper proposes an improved approach to electricity prices trend-cyclical component filtering, which is based on the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN). A combined criterion for determining the modes to be included into the trend component is introduced. The performance of the proposed approach is compared with the ordinary empirical mode decomposition (EMD), as well as with the method of wavelet-decomposition well-known in the energy economics literature. We test it on four day-ahead electricity markets: the Europe-Ural and the Siberia price zones of the Russian ATS exchange, the PJM exchange of the USA and the APX exchange of the United Kingdom. Our results show that the proposed approach based on CEEMDAN and the combined criterion outperforms the standard EMD on all the four electricity markets, and on two of the studied markets (PJM, APX) it outperforms the wavelet-smoothing, while on the other two (ATS Europe-Ural and Siberia) it performs at least not worse than the wavelet-smoothing. At the same time, the proposed approach does not require a prior choice of the smoothing parameter, as in the case of the wavelet-decomposition, and demonstrates a certain degree of versatility on the studied markets.