Day-ahead Price Forecasting in Ontario Electricity Market Using Variable-segmented Support Vector Machine-based Model

Abstract

In this article, the wholesale price of the Ontario electricity market has been forecasted by splitting the time series into 24 series, one for each hour of the day. Then, a one-step ahead forecast for each hour of the next day for a test period of three years has been made using the respective hour–time series and by employing a support vector machine. A detailed sensitivity analysis was performed for the selection of model parameters. Furthermore, the performance of a support vector machine model has been compared with a heuristic technique, simulation model, linear regression model, neural network model, neuro-fuzzy model, autoregressive integrated moving average model, dynamic regression model, and transfer function model. It has been shown that the proposed variable-segmented support vector machine model possessed better forecasting abilities than the other models and its performance was least affected by the volatility.

Keywords:

• independent electricity system operator
• learning theory
• neural network
• price forecasting
• support vector machine
• price volatility

Acknowledgment

The authors are thankful to Robert Doyle and Charanjeet Minhas from Customer Communications Department of Ontario IESO for providing the past SSR data from their data bank and for other useful suggestions.