In research, the overall power production in a wind farm is typically increased by employing model-based wind farm control. A controller, in an open-loop setting, operates based on the velocity of the wind flow, predicted by a wind farm model. For a controller to achieve the desired level of power production, a wind farm model has to be accurate and computationally tractable.

Generally, high-fidelity wind farm models are accurate but are computationally complex, making real-time control infeasible. This issue can, however, be addressed by employing closed-loop approach. In this approach, low- or medium-fidelity wind farm models, which are computationally tractable, are used, and their accuracy Is improved by employing an estimator.

In the previous researches, a centralized estimation approach was employed. In this framework, a single estimator is employed for estimating all the states (second-to-second wind field at turbine hub height) in a wind farm. Simulation results show that the accuracy of an open-loop model can be improved. However, the problem is the state size of the wind farm models. This leads to the objective of the thesis, which is “Can the accuracy of a wind farm model be improved while maintaining computational tractability?”.

In this thesis, distributed estimation is proposed as a solution to this problem. The basic idea behind distributed estimation is to distribute a wind farm into a number of small spatial domains (subsystems), and define a wind farm model for each of these subsystems to independently predict the wind flow in their respective spatial domain. For estimation, each sub-system employs an estimator to independently estimate their respective states, in parallel. In this thesis, based on the extent to which the states are estimated (measurement-update), model distribution, and size of the subsystems, four types of distributed architectures are devised, using the medium-fidelity model WindFarmSimulator (WFSim). Simulations show negligible loss in performance, and at the same time, the time taken for each iteration decreases drastically, making it computationally tractable.

In conclusion, distributed architectures are capable of improving the accuracy of the open-loop wind farm models, to the same level of accuracy offered by the centralized architecture, while maintaining computational tractability. Additionally, application of these distributed architectures for controller design will be a scope for future research in this topic.

Currently there is strong interest in the deployment of renewable energy sources such as wind, solar and hydro energy. This is driven in part due to the negative consequences of burning fossil fuels, such as health issues and climate change. To optimize the energy extracted in a wind farm the turbines require some control algorithm. Today, wind turbines that are part of a wind farm do not take neighbouring turbines into account when determining their control settings. This results in greedy control, where each turbine tries to align itself with the dominant wind direction and optimize its energy production using generator torque control and pitch control. During operation, each turbine creates a volume of slow-moving turbulent air behind its rotor, which is called a wake.

The issue with greedy control is that a turbine does not take into account where its wake will end up with respect to downwind turbines. By misaligning a turbine with the wind direction its trust can displace the wake. This will cause an efficiency loss on the misaligned turbine, but it can be used to increase the efficiency of a downwind turbine, leading to an increase in net power production. We know that a controller that uses yaw optimization can optimize the power production in a small scale experimental setup [Campagnolo et al., 2016]. In this particular research a line of three turbines produced 15% extra power compared to greedy control.

Wake redirection can be thought of as an attempt to efficiently mix the slower moving air in the wake with the faster stream surrounding it. This allows more of the total energy in the free stream to be extracted by any given wind farm. Yaw redirection has the potential to mix in the high velocity air that could otherwise pass through the empty space between the turbines. Tilt redirection has the potential to more efficiently mix the high velocity air that would otherwise pass over the wind farm into the air that hits the rotors [Annoni et al., 2017]. This work focuses on the development and implementation of a wake model that can be used to predict the effects of rotor tilting on the wake.

We started with the FLOw Redirection and Induction in Steady state (FLORIS) model as described in [Gebraad et al., 2014]. This model can be used to estimate the power production of a wind farm. It models the wake intensity and position and combines wakes when they overlap. The power production and wake characteristics of each turbine are predicted using three sub-models. There is one model for the wake intensity and its velocity profile, one to estimate wake deflection and one to determine how wakes are added to each other. In current practice, the wake deflection is modeled only in a horizontal plane, it can be driven by rotor yaw. In this work, the deflection model will be extended in such a way that the effects of rotor tilting on the wake position can be modeled.

Tuning a reduced order model such as FLORIS is notoriously difficult. In the FLORIS model, as described in [Gebraad et al., 2014] there are twelve hand tuned parameters. An important part of this research was attempting to simplify the tuning problem by making a robust parameter tuning procedure. A sensitivity analysis of the tuning parameters on the predicted power signals by FLORIS was performed. This analysis was used to identify situations where a subset of the model parameters is responsible for the variance in the predicted power production. Such situations with specific sensitivity were identified, but I lacked time and resources to fully leverage these and accurately tune the tilt extended model.

We stuck with the original nominal parameters of the different parts constituting FLORIS to conduct a case study. The case study compared predicted power increases by FLORIS with high fidelity simulations in two different Large Eddy Simulation (LES) packages. The case study was conducted by trying to optimize the power production of a small wind farm containing six turbines. The wind turbines were positioned in a two by three grid and the wind direction was aligned along the three turbines. The FLORIS model was used to optimize the yaw angles, tilt angles and both of them simultaneously for this layout. This led to four sets of control settings, the baseline case with greedy control and three optimized sets. We found that the FLORIS model strongly overestimated the power gain caused by turbine tilt. The main reason for this over-estimation seems to be that FLORIS currently has no implementation of the ground. In effect, the wakes in FLORIS can simply disappear into the ground. The case where only the yaw angles are optimized matched the high fidelity simulations to a higher degree confirming the work done in [Bastankhah and Port´e-Agel, 2016].

In conclusion, tilt control seems to have the potential to extract more energy from a certain atmospheric region. This is postulated because the air that gets forced on to rotors using tilt control would otherwise have passed unused over a wind farm. However, the model proposed in this thesis is insufficient to analyze the possible energy gains because it overestimates the effect of turbine tilting. This biggest problem with the FLORIS model is that it lacks a method for modeling the interaction between the ground and a wake. For future research the main priority should be implementing a solution to that problem.