Dissertation/Thesis Abstract

An Intelligent Maintenance System for Wind Turbines Using Optimization Models and Algorithms based on Artificial Neural Network and Exponentially Weighted Moving Average
by Arabgol, Samira, M.S., Southern Illinois University at Edwardsville, 2015, 77; 10188020
Abstract (Summary)

Wind power has been the world's fastest-growing source of renewable energy due to its reliable nature, sustainability, and efficiency. On-shore and off-shore wind turbines are widely used to convert wind energy into electricity. Since wind turbines are usually performing under severe environmental conditions, they must be monitored and maintained on a regular basis; however, it is difficult to access them as they are usually located at remote sites. Therefore, maintenance cost is one of the major issues that slow down the wind energy expansion. This research aims at minimizing maintenance costs by identifying optimal maintenance cycle in the long term and predicting failures in the short term.

Maintenance costs are incurred by preventive maintenance (PM), corrective maintenance (CM), and production loss due to turbine shutdowns. First, this study proposes a non-linear optimization model to minimize the total maintenance costs for long-term scheduling. However, the optimal maintenance schedule only reduces the probability of failures and does not eliminate them. In order to predict the failures in advance, this study investigates a predictive model for short-term predictions based on a heuristic model combining Artificial Neural Network (ANN) and Exponentially Weighted Moving Average (EWMA). The prediction model can detect small abnormalities on a wind turbine component before they cause major failures. These abnormalities could be due to a crack on turbine blade, a twist on internal wiring system, misplacing yaw sensor, etc. The models are implemented in an intelligent maintenance support system developed in Rstudio with two modules: 1) the optimization module that gets inputs from users and finds the optimal maintenance schedule; and 2) the prediction module that can be connected to the real-time database of a wind turbine, monitor its status, and alarm possible failures. A Monte-Carlo simulation study shows that the system is capable of detecting a large proportion of small failures, which could cause severe damages to wind turbines unless detected in advance, with a low number of false alarm.

Indexing (document details)
Advisor: Ko, Hoo sang
Commitee: Chen, Xin, Eneyo, Emmanuel
School: Southern Illinois University at Edwardsville
Department: Mechanical and Industrial Engineering
School Location: United States -- Illinois
Source: MAI 56/04M(E), Masters Abstracts International
Subjects: Engineering, Industrial engineering, Energy
Publication Number: 10188020
ISBN: 9781369760453
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy