COMING SOON! PQDT Open is getting a new home!

ProQuest Open Access Dissertations & Theses will remain freely available as part of a new and enhanced search experience at www.proquest.com.

Questions? Please refer to this FAQ.

Dissertation/Thesis Abstract

Electricity Consumption Forecasting in Chicago Area Using Artificial Neural Network (ANN) with the Nonlinear Autoregressive Network with Exogenous Inputs (NARX) Approach
by Ozturk, Berk, M.S., Southern Illinois University at Edwardsville, 2020, 90; 28259069
Abstract (Summary)

In this century, the use of electricity has vital importance, and the amount of electricity consumption has been increasing gradually. Companies and governments have made many investments in power generation plants to meet the increasing electricity consumption. However, while the amount of these investments is sufficient, it may cause an excess of energy production in some cases. Therefore, the importance of the proper management and planning of energy investments has increased. Well-organized energy generation and distribution planning provide an opportunity for companies and governments to reduce cost and increase in revenue. More or less electricity generation than a region demand can cause additional costs to the governments or electricity distribution companies. To prevent energy planning mistakes, the amount of future electricity consumption needs to be estimated accurately. Since many different variables can affect the amount of electricity consumption, it is also difficult to make accurate step-ahead predictions on electricity consumption. Because the weather condition parameters and population are the most significant variables, step-ahead predictions are usually made based on these variables. In the literature, there are many different studies to make future electricity consumption forecasts. While some of these studies apply traditional prediction methods, some studies use machine learning methods. In this study, a step-ahead prediction was performed using Artificial Neural Network (ANN), a machine learning technique, to predict future electricity consumption for a determined region. ANN was used to determine and model the significance of input variables such as monthly weather parameters and population of Chicago between 2001–2019 on the total electricity consumption. R-Studio, a package software, was used to deal with seasonality, trend, and missing values. After editing of the original dataset, ANN and Auto-Regressive Integrated Moving Average (ARIMA) were set up using R-Studio, separately. After determining the importance of the input variables and eliminating the non-important input variables using ANN in R-Studio, the MATLAB package program was used to make the step-ahead predictions. ANN was established again with the non-linear autoregressive network with exogenous inputs (NARX) approach in MATLAB to make the step-ahead prediction based on edited input variables. Then, the one-month-ahead electricity consumption estimation of the city of Chicago was performed, and the performance values of the models, Mean Square Error (MSE), Mean Absolute Percentage Error (MAPE), and Root Mean Square Error (RMSE) have been calculated. The calculated MSE, MAPE, and RMSE values of ANNs were compared with the MSE, MAPE, and RMSE values of Multiple Linear Regression (MLR), ARIMA, and other traditional estimation methods such as Moving Average (MA), Exponential Smoothing (ES), and Holt's Method (HM). Then, the conclusion was made based on the outcomes.

Indexing (document details)
Advisor: Chen, Xin
Commitee: Ko, Hoo Sang, Lee, H. Felix
School: Southern Illinois University at Edwardsville
Department: Industrial Engineering
School Location: United States -- Illinois
Source: MAI 82/8(E), Masters Abstracts International
Source Type: DISSERTATION
Subjects: Industrial engineering, Artificial intelligence, Mathematics
Keywords: Artificial neural network (ANN), Bayesian regularization, Forecasting, Machinelearning, Nonlinear autoregressive Network with Exogenous Inputs (NARX)
Publication Number: 28259069
ISBN: 9798569957996
Copyright © 2021 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest