Guo Xiaopeng
School of Economics and Management, North China Electric Power University, Beijing, 102206, China
Wang Yi
School of Economics and Management, North China Electric Power University, Beijing, 102206, China
ABSTRACT
This study explores two regression forecasting models of Relevance Vector Machine (RVM) and Support Vector Machine (SVM) to improve short-term load forecasting accuracy in small sample data sets. In the regression model of SVM, the study adopts Particle Swarm Optimization (PSO) algorithm to automatically select the penalty factor C and the parameteróof kernel functions, which can overcome the drawback that some specified parameters are often determined by experience. The regression model of RVM based on sparse Bayesian theory has many advantages, such as good generalization ability, no penalty factors setting. In order to verify the forecasting performance of two methods, the study uses power load data from a certain area of Fujian province. The results obtained present that they all have good forecasting accuracy. By contrast, the regression model of RVM has better performance, more stability and efficiency.
PDF References Citation
How to cite this article
Guo Xiaopeng and Wang Yi, 2013. Compare Relevance Vector Machine with Improved Support Vector Machine in
Short-term Power Load Forecasting. Information Technology Journal, 12: 3209-3213.
DOI: 10.3923/itj.2013.3209.3213
URL: https://scialert.net/abstract/?doi=itj.2013.3209.3213
DOI: 10.3923/itj.2013.3209.3213
URL: https://scialert.net/abstract/?doi=itj.2013.3209.3213
REFERENCES
- Hong, W.C., 2011. Electric load forecasting by seasonal recurrent SVR (support vector regression) with chaotic artificial bee colony algorithm. Energy, 36: 5568-5578.
CrossRefDirect Link - Li, C.B. and K.C. Wang, 2007. A new grey forecasting model based on BP neural network and Markov chain. J. Central South Univ. Technol., 14: 713-718.
CrossRefDirect Link - Muller, K.R., A. Smola, G. Ratsch, B. Scholkopf, J. Kohlmorgen and V. Vapnik, 1997. Predicting time series with support vector machines. Proceedings of the 7th International Conference on Artificial Neural Networks, October 8-10, 1997, Lausanne, Switzerland, pp: 999-1004.
CrossRef - Song, X.H., P.E. Zu, J. Yi and D. Liu, 2012. An optimally combined forecast model for long-term power demand based on improved grey and SVM model. J. Central South Univ., 43: 1803-1807.
Direct Link - Salcedo-Sanz, S., E.G. Ortiz-Garcia, A.M. Perez-Bellido, A. Portilla-Figueras and L. Prieto, 2011. Short term wind speed prediction based on evolutionary support vector regression algorithms. Expert Syst. Appl., 38: 4052-4057.
CrossRef - Shi, Y.Z., Y. Peng and H.C. Zhou, 2012. Research on mid-and long-term runoff forecast model with relevance vector machine. J. Dalian Univ. Technol., 52: 79-84.
Direct Link - Tan, Z.F., J.L. Zhang, L.Q. Wu, Y. Ding and Y. Song, 2011. A model integrating econometric approach with system dynamics for long-term load forecasting. Power Syst. Technol., 35: 186-190.
Direct Link - Tipping, M.E., 2001. Sparse Bayesian learning and the relevance vector machine. J. Machine Learn. Res., 1: 211-244.
Direct Link - Wu, C.H., G.H. Tzeng and R.H. Lin, 2009. A novel hybrid genetic algorithm for kernel function and parameter optimization in support vector regression. Expert Syst. Appl., 36: 4725-4735.
CrossRef - Kennedy, J. and R. Eberhart, 1995. Particle swarm optimization. Proc. IEEE Int. Conf. Neural Networks, 4: 1942-1948.
CrossRefDirect Link