The multiple-mean-reversion jump-diffusion model for Nordic electricity spot prices

Abstract

Electricity spot market prices are notoriously difficult to model, let alone predict, because of their extreme volatility. Such volatility is reflected in so-called price spikes that may increase the spot price by an order of magnitude as a matter of hours. Spot market price series are also subject to many other types of phenomena, such as periodicities at different scales, and to mean reversion. We introduce a model for electricity spot market prices that includes both spikes and mean reversion. The model is based on a jump diffusion process that is superimposed on a mean reverting Ornstein-Uhlenbeck model. Mean reversion takes place at several different time and price scales, so as to reproduce the observed behavior of spot market prices correctly. The parameters of the model are calibrated with the Nord Pool spot market hourly price series using a maximum likelihood approach. The simulated price series thus obtained very closely follows the statistical characteristics of the real price series.

Robust estimation and forecasting of the long-term seasonal component of electricity spot prices

“…These results question the predictive power of forward prices. The second alternative approach is to utilize fundamental information like weather (Jabłońska et al, 2011), generation and demand (Cartea et al, 2009) or inventory levels (Douglas and Popova, 2008). There are, however, some problems related with this.…”

Identifying Spikes and Seasonal Components in Electricity Spot Price Data: A Guide to Robust Modeling

“…Hambly et al (2009) suggest a specification of a jump-diffusion model including both a mean-reverting diffusion and mean-reverting jump process. Jabłońska et al (2011) use a multiple mean reversion jump-diffusion model with the mean reversion taking place at several different time and price scales.…”

“…While it is clear that price spikes should be captured by an adequate stochastic model, like mean reverting jump-diffusion (Bierbrauer et al, 2007;Borovkova and Permana, 2006;Cartea and Figueroa, 2005;Clewlow and Strickland, 2000;Geman and Roncoroni, 2006;Jabłońska et al, 2011;Nomikos and Soldatos, 2010;Seifert and Uhrig-Homburg, 2007;Weron, 2008) or a regimeswitching model (Becker et al, 2007;De Jong, 2006;Higgs and Worthington, 2008;Hirsch, 2009;Huisman and Mahieu, 2003;Weron, 2010, 2012;Keles et al, 2012;Mari, 2008;Mount et al, 2006;Weron, 2009), the literature does not agree on whether these observations have to be included or excluded in the estimation of the seasonal pattern.…”

A hybrid electricity price forecasting model for the Nordic electricity spot market

“…Regression models are considered to be functions of past price observations and exogenous explanatory variables such as electricity demand and meteorological conditions. These include threshold autoregressive models [13,14] and the autoregressive conditional heteroscedastic (ARCH) model of Engle [15] and its extended version GARCH [16][17][18][19]. Five short-term forecasting techniques were analyzed and evaluated in [8], and autoregressive integrated moving averages (ARIMA) have been applied to load forecasting [9,10].…”

“…The probability of a price spike occurrence and the value of a price spike are produced by a Gaussian Mixture model and K-nearest neighboring model, respectively. These include threshold autoregressive models [13,14] and the autoregressive conditional heteroscedastic (ARCH) model of Engle [15] and its extended version GARCH [16][17][18][19]. The results show that hybridization of the normal range price and price spikes forecasts may provide comprehensive and valuable information for electricity market participants.…”