Optimal control of a wind turbine with de-icing system through ice forecasting and observations

Wind Power
Period of realization: December 2018 – March 2023

This project is made possible thanks to the financial participation of the Government of Quebec, via InnovÉÉ and of the Government of Canada, via the Natural Sciences and Engineering Research Council of Canada NSERC.

Project Description

The Saint-Philémon Wind Farm (STP) in Quebec and the Glen Dhu Wind Energy LP (GDWE) in Nova Scotia are two wind farms operated by Capstone Ltd. that are suffering significant energy losses due to ice accretion on their turbines.

The objective of the project is to control automatically, i.e. without human intervention, the de-icing system of the Enercon turbines installed at the two wind farms as a function of ice forecasts. When combined with meteorological forecasts provided by Environment and Climate Change Canada, the Nergica-developed GPEO icing model can be used to predict icing at the wind farm locations and to quantify ensuing energy losses.

Whenever an icing event is predicted, the turbines’ ice protection system is automatically activated at the optimum moment to limit the impacts of ice accumulation on wind farm production. This control has allowed operators to maximize the use of the wind turbine anti-icing system in a proactive manner.



  • Automatically advising the wind farm operator of predicted icing events.
  • Sending commands in Python to the wind turbine controller.
  • Optimizing energy gains as a function of the icing event.
  • Ensuring that the developed turbine control is low priority compared to the internal turbine control and the wind farm manager’s control.



  • Analyzing the results of Nergica’s GPEO icing model and emailing alerts whenever an icing event is predicted.
  • Developing a customized solution for the wind farm controller (OPC server) using existing Python libraries.
  • Developing control logics based on energy losses predicted by Nergica’s icing model.
  • Cancelling the control if wind turbine parameters differ from expected values and automatically inform the operator.


  • Icing alerts were implemented at the STP and GDWE wind farms. Highly appreciated by Capstone, these alerts were configured for 3 wind farms in Nova Scotia and 7 in Ontario.
  • Automatic start-up of the turbine blade de-icing system before performance losses are even detected by the turbine.
  • Energy gains comparable to proactive human intervention for the tested turbines.
  • Icing alerts sent automatically at 6-hour intervals.
  • Creation of an automated turbine control for the two wind farms.