Rüzgar santrallerindeki operasyonel iyileştirmelerin tahminlenmesi

The operational optimizations, referring to the upgrades on wind turbines, can be very expensive; on the other hand, it is very complicated to assess the level of improvement they provide. Because of the inability to make reliable estimates on improvement levels, the plant owners are often reluctant to invest in upgrades. Like the OEM power curves, the improvement percentages for the upgrades, represent merely a reference and might differ for better or worse in the actual environmental conditions of the plant. The evaluations can not be done with a simple comparison of the pre-upgrade and post-upgrade performance, due to the complexity of the variables affecting power production and high levels of uncertainty of the environmental variables. In this research, we aim to study a machine learning approach implemented on wind farm level to evaluate the impact of operational improvements. Our approach consists of modeling the power output of the farm using a group of turbines referred to as the control turbines. The control group will not be upgraded to form the baseline for the pre-upgrade conditions. This baseline is later used to make a reliable comparison with the conditions after improvements are implemented.