Long-term electrical energy consumption: Formulating and forecasting via optimized gene expression programming / Seyed Hamidreza Aghay Kaboli

Seyed Hamidreza , Aghay Kaboli (2018) Long-term electrical energy consumption: Formulating and forecasting via optimized gene expression programming / Seyed Hamidreza Aghay Kaboli. PhD thesis, University of Malaya.

	PDF (The Candidate's Agreement) Restricted to Repository staff only Download (1576Kb)
Preview	PDF (Thesis PhD) Download (4Mb) | Preview

Abstract

This study mathematically formulates the effects of two different historical data types, (i) electrical energy consumption in preceding years and (ii) socio-economic indicators (SEI) on electrical energy consumption (EEC) of ASEAN-5 countries, namely, Malaysia, Indonesia, Singapore, Thailand, and Philippines. Firstly, a multi-objective feature selection approach is developed in this study to extract the most influential subsets of input variables from each historical data type (EEC and SEI) with maximum relevancy and minimum redundancy for long-term EEC modeling. In the developed feature selection approach, multi-objective binary-valued backtracking search algorithm (MOBBSA) is used as an efficient evolutionary search algorithm to search within different combinations of input variables and selects the non-dominated feature subsets, which minimize simultaneously both the estimation error and the number of features. Then, in order to cope with the limitations of the existing artificial intelligence (AI) based methods, optimized gene expression programming (GEP) is applied to precisely formulate the relationships between historical data and EEC of ASEAN-5 countries. The optimized GEP as a recent extension of GEP approach is superior to other AI-based methods in giving an optimized explicit equation, which clearly shows the relationship between input historical data and EEC in different countries without prior knowledge about the nature of the relationships between independent and dependent variables. This merit is provided by balancing the exploitation of solution structure and exploration of its appropriate weighting factors through use of a robust and efficient optimization algorithm in learning process of GEP approach. To assess the applicability and accuracy of the proposed method for long-term electrical energy consumption, its estimates are compared with those obtained from artificial neural network (ANN), support vector regression (SVR), adaptive neuro-fuzzy inference system (ANFIS), rule-based data mining algorithm, GEP, linear, quadratic and exponential models optimized by particle swarm optimization (PSO), cuckoo search algorithm (CSA), artificial cooperative search (ACS) algorithm and backtracking search algorithm (BSA). The simulation results are validated by actual data sets observed from 1971 until 2013. The results confirm the higher accuracy and reliability of the proposed method as compared with other artificial intelligence based models. On the basis of the favorable results obtained, it can be concluded that recent enhancements in AI-based approaches, as in this study, could result higher accuracy with the least complexity for long-term EEC forecasting. Finally, future estimations of EEC in ASEAN-5 countries are projected up to 2030 by applying the rolling-based forecasting procedure on mathematical models derived from optimized GEP. Furthermore, EEC in ASEAN-5 countries is forecasted by autoregressive integrated moving average (ARIMA) model and first-order single-variable grey model (GM (1, 1)) and their forecasts are compared with those obtained by the proposed method.

Actions (For repository staff only : Login required)