الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 186 |  |  |
| | | from | 186 | | |
Abstract
The abrupt shutdown of wind turbines is a major concern for operators and manufacturers alike, so early fault detection is an important factor that reduces energy loss. As most wind turbines have a Supervisory Control and Data Acquisition system (SCADA), exploiting this data and using it to reach a reliable maintenance management program is a vital requirement for operators and manufacturers. This thesis employs SCADA data for early detection of malfunctions in a new way. Unlike conventional fault detection and useful life determination methods, which rely on measuring the physical state of the components, this thesis presents a new method for early fault detection and useful life determination by measuring the performance of the components. It provides a method for early detection of faults in the early stages of their occurrence and also predicts faults for a period of approximately three months
This method is based on the analysis of the turbine performance and the calculation of the power losses, which turns into a rise in temperature. The work was carried out in 3 stages. In the first stage, data are collected and stored to be used in the next stages with a focus on gearbox oil temperature, production capacity, and wind speed, as these are the main factors used in the study.
The second stage is to use SCADA data in comparing the performance of adjacent turbines to identify turbines with different performance that is by comparing the wind speeds of adjacent turbines, K-means clustering algorithm is used to group the turbines into groups according to the wind speed, then comparing the performance of identical wind speed turbines, by choosing a set of parameters that explain the performance and its efficiency, the mode is calculated for each parameter and the parameters of all the adjacent turbines are compared with this mode, then decision tree technique is used to determine the different turbines in performance.
This program designed to compare turbine performance all the time, and identifying the beginnings of a breakdown by showing a difference in performance, the actual faults are discovered using the cluster and decision tree, which allows the maintenance department the ability to prepare and work on corrective maintenance early, and plane the proper inspection to determine the causes of the malfunction, the output of the program is used to determine how the failure and the path of collapse occurred, and the effect of defect in a certain component on the performance of the rest of the various components..
The third stage neural network provides an accurate prediction of approximately 94% of the turbine performance, this allows the maintenance department to take preventive measures to prevent damage from occurring basically, and build a reference for the turbine performance. the turbine performance is always compared to this reference. This reference takes into consideration the ageing of the turbines; this comparison helps maintenance management in early fault detection, which helps in making the right decisions, reduces turbine failure and secondary collapse of various components, which saves spare parts, reduces the turbine downtime, and increases the economic viability of wind turbines.
In other words, the technique used allows dealing with the damage from the beginning of its occurrence and also anticipating the damage to try to avoid it
The focus of this thesis is on the gearbox, which is the most expensive and most change cost of component, despite focusing on the gearbox, this method is still usable for monitor the performance of different systems in the turbine.