Early fault prediction of a wind turbine using a novel ANN training algorithm based on ant colony optimization

Abstract

The technological developments in wind energy field have reduced the investment and the operation costs. For this reason, wind farms have become more popular around the world. Increasing the share of wind energy in the market has led to the need for secure, inexpensive, and effective monitoring and control approaches. In the present work, various monitoring and control tools, which are cheap and easy to implement in wind farms using existing system data are proposed. The primary purpose of this study is to offer a new methodology, i.e. an artificial neural network (ANN) design with a novel training algorithm called Antrain ANN, in order to explore the early fault detection in a wind turbine. Our case problem is the fault detection for a wind turbine. For this issue, we used real data consisting of 873 samples with 12 inputs and one output. The models used in the work try to forecast fault occurrence before 10 minutes it happens. The proposed Antrain ANN algorithm is compared with Quick Propagation, Conjugate Gradient Descent, Quasi-Newton, Limited Memory Quasi-Newton, Online Backpropagation, and Batch Back Propagation algorithms, respectively. The results have shown that the proposed novel approach has better results in the correct classification rates than other algorithms except the Quasi-Newton and Limited Memory Quasi-Newton ones.

Artificial Intelligence Based Prognostic Maintenance of Renewable Energy Systems: A Review of Techniques, Challenges, and Future Research Directions

“…In [113], an FFNN-based framework is proposed for the detection and classification of wind turbine sensor faults. There are some other interesting works relying on FFNNs for fault detection renewable energy systems [114]- [116].…”

Wind Turbine Multi-Fault Detection based on SCADA Data via an AutoEncoder

“…These costs can easily overcome approximately the 20% of the total energy generation cost for wind power plants (WPPs) [4]. When a WT has a lot downtime, in efficiency it means that productivity decreases while the operation and maintenance (O&M) costs increase [5], [6].…”

Accelerometer Mass Loading Study Based on a Damage Identification Method using Fundamental Laws in Closed Systems

“…Many different damage parameters were proposed in the early days, and it is seen that a significant part of these damage parameters is the functions of the modal parameters [6]. Later, it is seen that such damage markers used in advanced methods such as artificial neural networks [7,8] and deep learning [9] are tried to be associated with damage in real structures. For a damage indicator to be used in real structures, it is essential to show that it can be used effectively in theoretical systems.…”