Modeling Regime Switching in Day-ahead Market Prices Using Markov Model

The accurate price forecasting of electricity market is crucial for profit maximizing producers and consumers in liberalized power markets. In all market places (day-ahead, intra-day and real-time) accurate price prediction is needed to generate optimal bids and maximize the profit. This paper first presents three methods for forecasting day-ahead market prices, namely Generalized Autoregressive Conditional Heterosedastic (GARCH), Holt-Winter (HW) and Mean Reversion and Jump Diffusion (MRJD). These methods are based on three broad methodologies of time series analysis, exponential-smoothing and stochastic processes. The dynamics of hourly prices in day-ahead market are varying from day to day. Each forecasting tool is suitable to capture one type of price dynamics. To capture this phenomenon, we combine GARCH, HW and MRJD methods using proposed Markov switch. The proposed Markov model is tested using Nordic day-ahead prices.