CONWIND: Research on smart operation control technologies for offshore wind farms

A wind farm's profitability is hugely influenced by the wakes generated by the turbines, which are moving turbulent wind fields with decreased average wind speed. The increased size in offshore wind turbines leads to larger wakes, and thus the steering of these wakes (by steering the turbines) can have a big impact on the energy production and the wear and tear on the turbines. The profitability of the wind farm is also linked to accurate production estimates or production goals, which are easier met through a smart and coordinated control of the turbines. In order to improve current control algorithms of the turbines, an efficient and reliable prediction of the incoming wind field is needed. We are particularly interested in short term forecasting, ranging from 5 minutes to 1 hour. Improvements in the wind prediction will be investigated through the integration of measurements in the wind modelling as well as by the use of machine learning. This work is well underway, using data-assimilation of lidarwind measurements into a regional weather forecasting model. A machine learning model has also been developed for short term predictions, which has been trained and validated using wind measurements from FINO1. The next step is to obtain computationally efficient models that will evaluate the movement of the wakes in the wind farm and the impact on the turbines. In combination with the improved short-term forecasts we will then be able to obtain a wind farm model that will form the basis for the controller. We will use statistics to estimate the uncertain parameters that will be relevant for the wind farm behavior, and both physical models and machine learning techniques will be evaluated. The control objectives are to reduce the severity of loading and to distribute accumulated fatigue evenly over the turbines, while maintaining or increasing the power production and complying with grid constraints. Based on the wind farm model and the incoming data, the controller will send the appropriate command to all the turbines in the farm. Finally, the project will validate its findings through demonstrations in offshore wind farms or laboratories in China or Norway, or through numerical validations were appropriate. A first iteration of scenarios to test the algorithms on has been defined and will be further refined in the project.

-

Offshore wind is still lagging behind onshore wind when it comes to investments, though the potential is huge. In recent years, costs have been driven down, outpacing expectations, due principally to the rapid increase in turbine size. The increased size also leads to increased wakes behind the turbines, which are turbulent wind fields with decreased average wind speed. The wakes are not static, and lead to more fatigue loads on the turbine blades as well as a loss in energy production in the turbines hit by the wakes. The levelized cost of energy (LCOE), which is the break-even cost to produce energy, can decrease either by diminishing costs or by increasing the energy production. With an intelligent control system it is possible to work on both these options: one may optimize the energy production when the electricity price is high, and one may steer the wakes of the wind turbines in such a way that their impact on the next turbine is less damaging and thus reducing maintenance costs. The primary objective of the project is therefore to reduce the operating costs and increase the energy production of offshore wind farms through development of advanced control algorithms. Amongst the challenges is to obtain fast and accurate predictions of the wind field in the wind farm, wakes included, to provide the input for the controller. The controller itself must be capable of balancing various demands and take into account the various model uncertainties. To obtain a control system that increases the profitability of offshore wind farms by more than 2% will make a great impact on the profitability of offshore wind farms.