Forecasting Turkey Electricity Prices by Alpha-StablePeriodic AutoRegressive Time Series

Abstract

Finite variance Gaussian time series are insufficient in forecasting irregular and sudden changes in electricity production and consumption due to the complexity of today's power grids. Such Levy type fluctuations in the generation and/or consumption of electricity might inevitably result in the impulsive behavior in the electricity prices. For this reason, using the alpha-stable time series (alpha(α)-stable Levy process), which have infinite variance and heavy-tailed distributions will be a more appropriate model for price forecasting. In this study, we have tested the alpha-stablePeriodic Autoregressive (PAR) time series with real data obtained from TURKISH MARKET & FINANCIAL SETTLEMENT CENTER (EPIAŞ) electricity prices and have compared this model with previously made forecasting stochastic models such as Gaussian-driven Autoregressive (AR). Our findings not only demonstrated the effectiveness of the alpha-stablePAR time series method in estimating electricity prices but also highlighted its superiority over the Gaussian-driven AR model for Turkish electricity prices. © 2023 IEEE.