Electricity demand prediction using a hybrid approach / Syamnd Mirza Abdullah

Syamnd Mirza, Abdullah (2017) Electricity demand prediction using a hybrid approach / Syamnd Mirza Abdullah. PhD thesis, University of Malaya.

[img] PDF (The Candidate's Agreement)
Restricted to Repository staff only

Download (1327Kb) | Request a copy
    [img] PDF (Thesis PhD)
    Restricted to Repository staff only until 31 December 2019.

    Download (2361Kb) | Request a copy

      Abstract

      Electricity demand prediction is an important field of study that supports the government in developing a good economic and control plan for the future of electricity power generation. Various techniques and tools have been utilized throughout the history of such predictions, and different parameters have been analyzed. The main aims of studies in this field were to predict electricity demand and to minimize errors by analyzing various effects, such as that of the relation between the patterns of the data set and the utilized tools. In particular, this study focuses on reducing the degree of multicollinearity among independent variables to increase accuracy rate. In addition, the study aims to employ a combination system that accepts both linear and nonlinear patterns of the input data set to minimize the residual errors in prediction rate. To realize this objective, this thesis proposes a system that uses a hybrid approach that combines principal component analysis as a tool for lowering degree of multicollinearity, multiple linear regression (MLR) and a time series artificial neural network (ANN) to minimize errors. The novel electricity demand prediction model proposed in this thesis is called the principal component regression with back-propagation artificial neural networks model (PCR-BPNN). The data set fed into this model is the quarterly electricity usage in Malaysia from 1995 to 2013 provided by the Department of Statistics Malaysia. According to the performance indicators such as mean squared error, root mean squared error, and mean absolute percentage error, the PCR-BPNN model generates a more accurate predictions than previous methods such as principal component (PC)—MLR, PCNN, and PC-Support Vector regression models. The results indicate the expected electricity demand in Malaysia for 2020 will be 13702.91 Ktoe.

      Item Type: Thesis (PhD)
      Additional Information: Thesis (PhD) – Faculty of Economics & Administration, University of Malaya, 2017.
      Uncontrolled Keywords: Electricity demand; Hybrid approach; Department of Statistics Malaysia; PCR-BPNN model; PC-Support Vector regression model
      Subjects: H Social Sciences > HC Economic History and Conditions
      Divisions: Faculty of Economics & Administration
      Depositing User: Mr Mohd Safri Tahir
      Date Deposited: 11 Apr 2019 08:40
      Last Modified: 11 Apr 2019 08:40
      URI: http://studentsrepo.um.edu.my/id/eprint/9936

      Actions (For repository staff only : Login required)

      View Item