A Hybrid One-step-ahead Time Series Model Based on GA-SVR and EMD for Forecasting Electricity Loads

10.6180/jase.2017.20.4.08

Mu-Chiun Shiu；Liang-Ying Wei；Jing-Wei Liu；Deng-Yang Huang；Chien-Chih Tu；Kuo-Hsiung Liao

Electricity Load ； Support Vector Regression ； Empirical Mode Decomposition ； Time Series ； One-step-ahead Method ； Genetic Algorithm

淡江理工學刊

20卷4期（2017 / 12 / 01）

467 - 476

英文

Economic growth increases the demand for electricity, and forecasting electricity loads is critical for providing cheaper electricity. Conventional time series methods have been applied to forecast electricity loads. However, traditional statistical methods, such as regression models, are unable to address nonlinear relationships, such as those of electricity loads. Moreover, most timeseries models which use electricity load data with many factors, such as climate conditions and region environments, involutedly would reduce the forecasting performance. To overcome these problems and improve the forecasting ability of time series models, this paper proposes a hybrid one-step-ahead time series model that is based on support vector regression (SVR), empirical mode decomposition (EMD), and a genetic algorithm (GA) to predict electricity loads. The experimental results were generated from 2 electricity load datasets from various countries, and the proposed model was compared with several models. Our findings indicate that the proposed model outperforms the other approaches in terms of mean absolute percentage error (MAPE).