Hybrid support vector regression in electric load during national holiday season

Hybrid Support Vector Regression in Electric Load During National Holiday Season

 
Abstract: This paper studies non-parametric time-series approach to electric load in national holiday seasons based on historical hourly data in state electric company of Indonesia consisting of historical data of the Northern Sumatera also South and Central Sumatra electricity load. Given a baseline for forecasting performance, we apply our hybrid models and computation platform with combining parameter of the kernel. To facilitate comparison to results of our analysis, we highlighted the results around MAPE-based and R ² -based techniques. In order to get more accurate results, we need to improve, investigate, also develop the appropriate statistical tools. Electric load forecasting is a fundamental aspect of infrastructure development decisions and can reduce the energy usage of the nation.

Modeling public holidays in load forecasting: a German case study

Abstract:  We address the issue of public or bank holidays in electricity load modeling and forecasting. Special characteristics of public holidays such as their classification into fixed-date and weekday holidays are discussed in detail. We present state-of-the-art techniques to deal with public holidays such as removing them from the data set, treating them as Sunday dummy or introducing separate holiday dummies. We analyze pros and cons of these approaches and provide a large load forecasting study for Germany that compares the techniques using standard performance and significance measures. Finally, we give general recommendations for the treatment of public holidays in energy forecasting to suggest how the drawbacks particular to most of the state-of-the-art methods can be mitigated. This is especially useful, as the incorporation of holiday effects can improve the forecasting accuracy during public holidays periods by more than 80%, but even for non-holidays periods, the forecasting error can be reduced by approximately 10%.

CLICK HERE to order complete paper

北京交通大学

Measurement and Prediction of Regional Traffic Volume in Holidays

Zhenzhu Wang – Yishuai Chen – Yuchun Guo – Yongxiang Zhao – Weikang Tang

Beijing Jiaotong University

Chao Zeng – Jingwei Chen

Join-Cheer Software Co., Ltd., Beijing, China

Abstract: Accurate regional traffic volume projection is important for department of transportation to plan investments, and also helps forecast oil or electric energy demand and CO ₂ emissions. Based on a 4.5 years’ daily traffic volume measurement data of the highway network of Guizhou province of China, this paper conducts a comprehensive measurement analysis of the network’s traffic volume growth pattern and proposes a new time series model, which improves the projection accuracy of non-holiday and holiday traffic considerably. We first find that the holiday traffic volume is considerably higher than that on the neighboring non-holidays (e.g., 1.88 times), which could bring tremendous pressure on the road network. We then find that the traffic of network increases exponentially, in particular, the increase rates in holidays are higher than those in non-holidays. Thus, we propose an Exponential-Growth (EG) holiday component model, which models the holiday component with exponential growth. Experimental results show that our model considerably improves the holiday traffic’s prediction accuracy compared with the existing models. For instance, for the first day of National Day holiday, which is usually the heaviest day in a whole year (from Jan. 1 to Dec. 31), the model decreases the prediction relative error from 18.7% to 7%.

CLICK HERE to order complete paper.