RESEARCH

On going research towards improving the operations and maintenance cost of onshore wind energy, lowering the levelized cost of energy and reducing uncertainties and around power production through bespoke data-driven, real-time forecasting algorithms and feature-detecting analytics with the capability of handling data of variable qualities, power production variation and early downtime detection with high accuracy.

Wind power forecasting

In the last few years, wind power forecasting has established itself as an essential tool in the energy industry due to the increase of wind power penetration in the electric grid. The team's present focus is on developing novel machine learning algorithms to effectively forecast wind power generation for short, medium, and long terms. Forecasting of the wind power generation may be considered at different time scales, depending on the intended application. From milliseconds up to a few minutes, forecasts can be used for the turbine active control. Such type of forecasts is usually referred to as very short-term forecasts. For the following 48–72 hours, forecasts are needed for power system management or energy trading. Bids for energy to be supplied on a day are usually required during the morning of the previous day. These forecasts are called short-term forecasts. For longer time scales (up to 5–7 days ahead), forecasts may be considered for planning the maintenance of wind farms, or conventional power plants or transmission lines. Maintenance of offshore wind farms may be particularly costly, so optimal planning of maintenance operations is of particular importance - another aspect which is presently investigated by the team. 
 

Real-time downtime detection of wind turbines

Complex systems are susceptible to many types of anomalies, faults, and abnormal behavior caused by a variety
of off-nominal conditions that may ultimately result in major failures or catastrophic events. Early and accurate
detection of these anomalies using system inputs and outputs collected from sensors and smart devices has
become a challenging problem and an active area of research in many application domains. A real-time downtime detection framework is actively pursued by the researchers to determine whether a sample is an anomalous one taking into consideration the trade-off between misclassification errors and detection rates. The proposed detection model not only identifies the downtime but also classifies it as a scheduled maintenance break, a fault, or even as an outlier - in a real-time framework.