Short-term load forecasting using a hybrid model with a refined exponentially weighted fuzzy time series and an improved harmony search

• Published in: International journal of electrical power & energy systems Vol. 62; pp. 118 - 129
• Main Authors: Sadaei, Hossein Javedani, Enayatifar, Rasul, Abdullah, Abdul Hanan, Gani, Abdullah
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 01.11.2014; Elsevier
• Subjects: Algorithms; Applied sciences; Auto correlation function; Electric power generation; Electrical engineering. Electrical power engineering; Electrical power engineering; Exact sciences and technology; Exponentially weighted fuzzy time series; Fuzzy; Fuzzy logic; Fuzzy set theory; Harmony search; Mathematical models; Operation. Load control. Reliability; Power networks and lines; Proposals; Searching; Short term load forecasting; France; Europe; Auto correlation function; Exponentially weighted fuzzy time series; Harmony search; Short term load forecasting; Short term; Autocorrelation function; Error analysis; Error estimation; Demand forecasting; Time series; Forecasting management; Learning; Case study; Fuzzy algorithm; Load management; Hybrid model; Comparative study
• ISSN: 0142-0615; 1879-3517
• DOI: 10.1016/j.ijepes.2014.04.026

•A three Phases hybrid method was introduced for Short Term Load Forecasting.•The first Phase discusses the primitives for initialization and inputting.•The second Phase is linked with the Improved Harmony Search procedure.•The third Phase is concerned with a typical fuzzy time series algorithm.•The results have shown that proposed method was suppressed its counterparts. This article discusses the proposal of an enhanced hybrid algorithm. The algorithm focuses on a sophisticated exponentially weighted fuzzy algorithm that is aligned with an enhanced harmony search. Short-term load forecasting can be performed appropriately with this specific method. The initial phase of this research discusses the recognition of the fuzzy logical relationship order with the aim of autocorrelation analysis. The second phase aims at obtaining the optimal intervals and coefficients for adoption using training data set. The last phase seeks to apply the obtained information and attempts to predict a 48-step-ahead on Short term load forecasting (STLF) problems. It is essential to validate this process. To achieve this goal, eight case studies of actual load data from France and Britain (from 2005) were employed. These data were applied to both the developed algorithm and certain improved STLF predicting models. The subsequent errors from these models were compared. The results of the error analysis exhibit the advantages of the developed algorithm with respect to its prediction preciseness.