Hourly Electricity Price Forecasting for the Next Month Using Multilayer Neural Network

Load and price forecasting are key challenges for current electricity market participants. Load and price in electricity markets have complex peculiarities, such as nonlinearity, being nonstationary and irregular. Accurate short-term forecasting, such as hourly electricity price forecasting (EPF) for the next month gives pivotal information to power producers and consumers to enhance precise techniques to maximize their profit. This paper deals with short-term hourly EPF for the next month (January 2006), using the historical hourly data for the year 2005 as a training set. A new approach of multilayer neural networks is applied in composite topologies in order to improve forecasting accuracy. The intent is to study the behavior of diverse composite topologies to compare the best performance indices evaluated by the mean absolute percentage error and mean square error. The performance of different topologies is compared to identify the best connection architecture. The data used in the forecasting are hourly historical data of the temperature, electricity load, and natural gas price from the Australian electricity markets.