No abstract

No abstract

Abstract. A hybrid neural network-based method is presented to predict day-ahead electricity spike prices in a deregulated electricity market. First, prediction of day-ahead electricity prices is carried out by a neural network along with pre-processing data mining techniques. Second, a classifier is used to separate the forecasted prices into normal and spike prices. Third, a second neural network is trained over spike hours with selected features and is used to forecast day-ahead spike prices. Forecasted spike and normal prices are combined to produce the complete day-ahead hourly electricity price forecasting. Numerical experiments demonstrate that the proposed method can significantly improve the forecasting accuracy.

Various machine learning-based methods and techniques are developed for forecasting day-ahead electricity prices and spikes in deregulated electricity markets. The wholesale electricity market in the Province of Ontario, Canada, which is one of the most volatile electricity markets in the world, is utilized as the case market to test and apply the methods developed. Factors affecting electricity prices and spikes are identified by using literature review, correlation tests, and data mining techniques. Forecasted prices can be utilized by market participants in deregulated electricity markets, including generators, consumers, and market operators. A novel methodology is developed to forecast day-ahead electricity prices and spikes. Prices are predicted by a neural network called the base model first and the forecasted prices are classified into the normal and spike prices using a threshold calculated from the previous year’s prices. The base model is trained using information from similar days and similar price days for a selected number of training days. The spike prices are re-forecasted by another neural network. Three spike forecasting neural networks are created to test the impact of input features. The overall forecasting is obtained by combining the results from the base model and a spike forecaster. Extensive numerical experiments are carried out using data from the Ontario electricity market, showing significant improvements in the forecasting accuracy in terms of various error measures. The performance of the methodology developed is further enhanced by improving the base model and one of the spike forecasters. The base model is improved by using multi-set canonical correlation analysis (MCCA), a popular technique used in data fusion, to select the optimal numbers of training days, similar days, and similar price days and by numerical experiments to determine the optimal number of neurons in the hidden layer. The spike forecaster is enhanced by having additional inputs including the predicted supply cushion, mined from information publicly available from the Ontario electricity market’s day-ahead System Status Report. The enhanced models are employed to conduct numerical experiments using data from the Ontario electricity market, which demonstrate significant improvements for forecasting accuracy.

Various machine learning-based methods and techniques are developed for forecasting day-ahead electricity prices and spikes in deregulated electricity markets. The wholesale electricity market in the Province of Ontario, Canada, which is one of the most volatile electricity markets in the world, is utilized as the case market to test and apply the methods developed. Factors affecting electricity prices and spikes are identified by using literature review, correlation tests, and data mining techniques. Forecasted prices can be utilized by market participants in deregulated electricity markets, including generators, consumers, and market operators. A novel methodology is developed to forecast day-ahead electricity prices and spikes. Prices are predicted by a neural network called the base model first and the forecasted prices are classified into the normal and spike prices using a threshold calculated from the previous year’s prices. The base model is trained using information from similar days and similar price days for a selected number of training days. The spike prices are re-forecasted by another neural network. Three spike forecasting neural networks are created to test the impact of input features. The overall forecasting is obtained by combining the results from the base model and a spike forecaster. Extensive numerical experiments are carried out using data from the Ontario electricity market, showing significant improvements in the forecasting accuracy in terms of various error measures. The performance of the methodology developed is further enhanced by improving the base model and one of the spike forecasters. The base model is improved by using multi-set canonical correlation analysis (MCCA), a popular technique used in data fusion, to select the optimal numbers of training days, similar days, and similar price days and by numerical experiments to determine the optimal number of neurons in the hidden layer. The spike forecaster is enhanced by having additional inputs including the predicted supply cushion, mined from information publicly available from the Ontario electricity market’s day-ahead System Status Report. The enhanced models are employed to conduct numerical experiments using data from the Ontario electricity market, which demonstrate significant improvements for forecasting accuracy.