A new hybrid Modified Firefly Algorithm and Support Vector Regression model for accurate Short Term Load Forecasting Abdollah Kavousi-Fard a , Haidar Samet b,⇑ , Fatemeh Marzbani c a Department of Electrical Engineering, Sarvestan Branch, Islamic Azad University, Sarvestan, Iran b School of Electrical and Computer Engineering, Shiraz University, Shiraz, Iran c American University of Sharjah, Sharjah, United Arab Emirates article info Keywords: Support Vector Regression (SVR) Modified Firefly Algorithm (MFA) Short Term Load Forecasting (STLF) Adaptive Modification Method abstract Precise forecast of the electrical load plays a highly significant role in the electricity industry and market. It provides economic operations and effective future plans for the utilities and power system operators. Due to the intermittent and uncertain characteristic of the electrical load, many research studies have been directed to nonlinear prediction methods. In this paper, a hybrid prediction algorithm comprised of Support Vector Regression (SVR) and Modified Firefly Algorithm (MFA) is proposed to provide the short term electrical load forecast. The SVR models utilize the nonlinear mapping feature to deal with nonlinear regressions. However, such models suffer from a methodical algorithm for obtaining the appropriate model parameters. Therefore, in the proposed method the MFA is employed to obtain the SVR parameters accurately and effectively. In order to evaluate the efficiency of the proposed methodology, it is applied to the electrical load demand in Fars, Iran. The obtained results are compared with those obtained from the ARMA model, ANN, SVR-GA, SVR-HBMO, SVR-PSO and SVR-FA. The experimental results affirm that the proposed algorithm outperforms other techniques. Ó 2014 Elsevier Ltd. All rights reserved. 1. Introduction In order to supply the energy demand constrained on the worldwide limited energy sources, it is essential to provide an accurate electrical load prediction. It is noteworthy that underesti- mation or overestimation in the electrical load forecast can intro- duce various challenges to the system operators. Problems in power quality or reliability and insufficient provided reserves stem from the energy demand underestimations. On the other hand, load forecast overestimation results in unnecessary establishments and spinning reserves, inefficient energy distributions, and increas- ing the operation costs. Therefore, precise load forecast is acquired by utilities and system operators in order to provide efficient unit commitment and load dispatching decisions, contingency plan- ning, and optimal load flow. Research studies around the accurate load forecast date back to the late 1960s. Based on the prediction horizons, the load forecast- ing problems can be categorized into three classes: long-term, medium-term, and short-term. Among them, the main focus is on the short-term load prediction which accounts for daily or weekly load estimations (Kavousi-Fard & Akbari-Zadeh, 2014). Moreover, short-term predictors are essential tools for system operation of utilities and electricity markets. There exists a large variety of prediction techniques in short-term forecast subclass. These methods can be divided into two main categories of: classical statistical algorithms and Artificial Intelligence based (AI) methods (Abdel-Aal, 2004; Che-Chiang & Chia-Yon, 2003; Metaxiotis, Kagiannas, Askounis, & Psarras, 2003; Park, El-Sharkawi, Marks, Atlas, & Damborg, 1991; Yalcinoz & Eminoglu, 2005). The statistical techniques include Auto Regressive Moving Average (ARMA) Mbamalu & El-Hawary, 1993, multiple linear regressions, and Kalman filter techniques (Guan, Luh, Michel, & Chi, 2013). The statistical methods identify the load pattern, then based on the obtained pattern, the time series analysis approaches are utilized to provide the future value of the measurements. Since the early 1990s, the AI techniques are among the most detailed studied forecasting methods. One of the most well-known models in the category of AIs is neural network (NN). In the ANN approaches, in order to obtain the future values, a nonlinear rela- tion is assumed between the previous values and some external variables. Neural fuzzy networks (Li-Chih & Mei-Chiu, 2008; Ling, Leung, & Lam, 2003) and NN variants (Niebur, 1995) have been employed in the last years to provide short term electrical load prediction. The NN models have been extensively used in various applications and they are considered as promising forecasting http://dx.doi.org/10.1016/j.eswa.2014.03.053 0957-4174/Ó 2014 Elsevier Ltd. All rights reserved. ⇑ Corresponding author. Tel.: +98 9132131154. E-mail address: samet@shirazu.ac.ir (H. Samet). Expert Systems with Applications 41 (2014) 6047–6056 Contents lists available at ScienceDirect Expert Systems with Applications journal homepage: www.elsevier.com/locate/eswa