Mahi Ayman, Mariam Elaskary, Nour Elgarhy, Zeina Tamer
Supervised by: Eng. Yomna Hassan, Eng. Maha Sayed
November 8, 2021
Institutions have been rapidly redirecting their investments away from fossil fuels, creating a favorable scenario for the generation of clean energy. The wind industry, especially, has seen an exponential increase in recent years. Early fault detection creates an alternative for operation and maintenance (O&M), helping forgo costs before they reach a catastrophic stage since preventive methods are used to stop malfunctions before their occurrence, improving turbine reliability and allowing for maximum utilization and better control of the wind farms. In this document, deep learning methods such as Autoencoders-LSTM (Long short-term memory) are used on time series data produced from SCADA to reach the optimal results.
Energy is now a crucial part of living on Earth. Almost everything depends on the use of energy. For a long time, the main source of energy has been fossil fuels, which, however, comes with major drawbacks. The use of fossil fuels plays a major role in climate change and affects air pollution negatively, consequently, impacting humans and living creatures . People tended to move to less harmful energy sources, and the wind was one of those.
The first sale of commercial wind turbines was finalized in 1927, to a group of US farmers for what was considered a fairly high price, but it was the 1970s oil crisis in the Middle East that pushed various governments to fund research into renewable energy as a means not only for energy security but also to battle air pollution and climate change.
Maintenance of wind turbines used to occur twice a year within 6 months of each other. A thorough inspection would occur upon the system, along with repairs and replacements according to the technician. This process was deemed time-consuming and costly and that led to the introduction of condition monitoring and later machine learning for fault detection.
The use of renewable energy is now becoming the best alternative to reduce reliance on fossil fuels. As renewable energy proceeds in its rapid growth around the world, the need for prediction and forecasting tools increases as well. Machine learning provided a novel solution for this need, and as a result, renewable energy sources will become more reliable and affordable and increase in expansion and potential.
A major energy source is wind. While wind can be a solution to the major climate change resulting from the use of non-renewable energy, converting it to energy is a long process that includes major steps. Since wind turbines are the source of wind energy, they require maintenance as they are placed in remote locations, which subject them to failures, with harsh environmental conditions.
Failures in wind turbines can be costly. A current solution that is now being implemented is to detect possible faults using condition monitoring. It, however, needs knowledgeable professionals to analyze the data and perform the work. A possible improvement to condition monitoring is using machine learning models to develop systems that would eliminate the hassle of finding professionals for doing the job.
Renewable energy sources in 2013 accounted for an estimate of 19.1% of energy consumption. The Energy Roadmap aims that the energy supply specifically of wind to reach a number between 31.6% and 48.7% in 2050. This increase, however, will have a major impact on the O&M of the costs of wind turbines. That being said, systems that could monitor and detect potential failures in wind turbines would benefit the world greatly by limiting the time of downtimes and thus minimizing loss of revenue.
1.3 Problem Statement
To avoid wind interference near cities, wind turbines are mostly located in secluded areas. Due to this allocation, the operation and maintenance costs tend to surge and the conditions of it (O&M) become increasingly volatile due to unpredictable weather conditions. By predicting the faults in wind turbines before they occur and maintaining them accordingly and forecasting weather variables to predict wind power generation, the costs of maintenance will be reduced significantly and the wind farms will be able to operate sufficiently.
In our research, one of the major problems agreed upon in this area of study is the lack of labeled datasets due to security problems related to data sharing. This made solutions depending on deep learning difficult since it depends on a great amount of data. On a local level, we were also faced with the problem of the lack of datasets containing data from Egyptian wind turbines which led to creating a simulation with the available data.