Abstract Views :187 |
PDF Views:0
Authors
Affiliations
1 EED, National Institute of Technology (NIT), Kurukshetra, IN
2 Tennessee Technological University (TTU), Cookeville, US
Source
Power Research, Vol 5, No 1 (2009), Pagination: 37-45
Abstract
In this paper, Support Vector Machine (SVM), a new machine learning technique based model to forecast price profile in a single settlement real-time electricity market has been presented. The proposed model has been trained and tested on data from the Victoria Electricity Market (VEM) to forecast the Regional Reference Price (RRP). The selection of input variables has been performed using correlation analysis, and in order to take advantage of the homogeneity of the time series, forty-eight separate SVMs have been used to predict the next-day price profile, with each SVM forecasting price for each trading interval. Forecasting performance of the proposed model has been compared with (i) an heuristic technique, (ii) a naïve technique, (iii) Multiple Linear Regression (MLR) model, and (iv) Neural Network (NN) model. Forecasting results show that the SVM model possesses better forecasting abilities than the other models and can be used by the participants to respond properly as it predicts the price before the closing of the window for submission of initial bids.
Keywords
Correlation Analysis, Learning Theory, Multiple Linear Regression, Neural Network, Price Forecasting, Support Vector Machine.
