Zhang Pingkang1, Wang Meng2, Zhao Dengfu3, Zhang Jiangshe4
(1. School of Economy and Finance, Xi'an Jiaotong University, Xi'an 710049, China; 2. Dispatching Center, Northwest ChinaGrid Company Limited, Xi'an 710048, China; 3. School of Electrical Engineering, Xi'an Jiaotong University, Xi'an 710049, China; 4. School of Sciences, Xi'an Jiaotong University, Xi'an 710049, China)
Abstract: A new algorithm
with high forecasting accuracy and global optimal property for peak load forecasting is
proposed based on the support vector machine (SVM) method, where the cross-validation is
introduced into hyper-parameter estimation in SVM to outperform the common cut and try
method. In addition to the load variables, the temperature information, weekday and
vacation information are taken into account in the input samples to improve the
forecasting accuracy. The practical examples show that the accuracy of the SVM is 0.4%¡«0.8%
higher than artificial neural network under the same load and weather conditions.
Keywords: peak load forecasting; support vector machine; kernels; cross-validation