Electricity price forecasting has become a vital area of research after the global restructuring, deregulationand the introduction of competitive market in the power industries. Prior to deregulation, electricity priceswere normally regulated, and predetermined tariffs were provided to the buyers. The effort to design aneffectively operating competitive market that give players the accurate incentives were supposed to enhanceproduction efficiency and put a limit to market power. Thus, in deregulated electricity market, theplayers have more freedom (Papalexopoulos & Hesterberg 1990).In the early 1990s, deregulation of the Nordic market commenced in Norway (Fidalgo & Matos 2007),and now Estonia, Sweden, Denmark, Norway, Finland forms a single market in which electricity is sold ina common power exchange controlled by Nord Pool Spot AS (Nord Pool Spot, 2011). Its major marketplace is the day-ahead auction market Elspot, which determines the spot price.
“According to Nord PoolSpot (2011b), the power exchange serves the society by ensuring that the wholesale electricity trade hasa transparent pricing, the spot price is used as reference price in the electricity derivative market and thequotes for long-term contracts shows the expectation for future prices. Furthermore, it ensures the marketprovides a system for maintaining a balance between physical supply and demand (K¨ann¨o et al. 2013)Electricity is a distinct commodity in that it is not economically storable, thus, there is need to ensurethe quantity of power supplied into the grid is equivalent to consumption, so that the power systemis stabilized (Weron 2014). Also, the demand for electricity is dependent on intensity of business, theactivity of the day (working days and weekends, holidays, on and off-peak hours) and weather.
These factorsmake electricity prices to exhibit seasonality, major volatility and unanticipated significant price spikes.Trader, producers, and retailers, and traders rely on the forecasted values of electricity price to makedecision on the strategy “buy” and “sell” bids to broadcast for each hour of the day in the power tradingmarket. At the end of the bidding, the bid prices are fed into the prices determination system (Singhal& Swarup 2011). For selling offers, prices are selected in a descending order while for the buying offers,prices are selected in an ascending order until there is an equilibrium between the volume and price. Thus,accurate forecasting of electricity prices is vital for each entity to be involved in the bidding process. Industriesin the power exchange also need an accurate forecast of electricity price to be able to optimizethe use of their portfolio through bidding and hedging against price volatility and make maximum profit.According to studies, price forecasting can be classified into four categories in terms of time horizon,purpose of use, and resolution (Weron et al.
2006):1) Long-term price forecasting with lead time of several years it can be used for investment planningand profitability analysis. Such as deciding a new site of power plant.2) Mid-term price forecasting they are mostly used for balance sheet forecast and derivative pricing.It focuses on producing probabilities distribution of future estimated prices over certain period. Itspans form 1 week to a year (Kyriakides & Polycarpou 2007).3) Short-term price forecasting this type of price forecast is of more interest to players of auction-typespot markets whose bids quotation are expressed in terms of quantity and price. It is for forecast ofprices in less than 24-hours (Hahn et al.
2009).In the last few decades, researchers have proposed different approaches for electricity price forecasting suchas the linear regression models (Papalexopoulos & Hesterberg 1990), Artificial intelligence methods whichcaptures non-linear and complex effects (Huang et al. 2005), Autoregressive moving average exogenous variables(Carpinteiro et al.
2004), fuzzy methods Kim et al. (1995), Kalman filtering technology which hasbeen applied short-term price forecasting (Brown 1983), GARCH models have been used in original priceseries to simulate price spikes and hybrid models (Peng et al. 2005). The problem with these statistical approachis that they require more time to handle a big number of variable and they might not represent the1non-linear characteristics of complex prices.
ANN there are still some short comings in ANN, because thereis need to manually determine parameters and structure and also, it convergence is very slow in training.The GARCH model failed to incorporate spikes like the height which are always seen in the original prices.One of the major challenges of price forecast is how to select vital and appropriate variables withoutnecessarily reducing the capability of the predictor. Thus, in this research we aim at predicting electricityprice using then Random forest model in combination with ensemble system. We shall incorporate intoour model adequate variable that influences electricity prices and knowing fully well that weather has akey effect on electricity price we consider using weather from weather station rather than an aggregateddata which has been used in previous studies. Random forest is advantageous in feature selection: itrandomly creates various small decision trees to explain the data (James et al.
2013). Its random featureselection boosts the diversity of the system and enhances classification the performance (Cheng et al.2012).
Variables that appears often in these trees might have some measurable effect on what we aim atforecasting.2