In modern wind power plants, wake effects are the cause for significant losses in the annual energy production and active wake control methods can be used to mitigate the impact of wake formations created by the wind rotors. Wake steering is a particular active wake control method that uses the wind turbine yaw angle to steer emerging wake formations in a desired direction. In certain wind conditions, this can lead to an increase in the overall power production of the wind power plant. After a concept for a wind farm operation control using artificial intelligence has been proposed in previous publications, this investigation quantifies the gains in power production that can be achieved in a small wind farm. The process of generating training data for an artificial neural network is explained and the result of the network training process confirms sufficient accuracy for the task of finding optimal yaw angle set points. The steady-state increase in power production using an advanced wind farm operation control is determined for a large number of wind conditions. A detailed simulation model is used for the aerodynamic interactions between the turbines as well as the local operation control. The results show that the increase in power production that can be achieved using wake steering depends mostly on the wind direction and is highest in partial shadowing situations. Using optimal yaw angles values, the power production can be increased in the range of single-digit percentage points. The advanced wind farm operation control is thus shown to be effective in optimizing the wake interactions in a small wind farm with the goal of increasing the overall power production.

Wake steering is a control method for wind power plants that aims to increase the total power production. The yaw angles of individual wind turbines are adapted in order to redirect emerging wake formations in the desired direction. In previous publications, an AI-based wind farm operation control has been presented that uses artificial neural networks to determine optimal yaw angle set points for the purpose of wake steering. Significant increases in the steady-state power production of wind farms could be achieved for certain wind conditions. In this study, the performance of the wind farm control is investigated for dynamic wind direction changes at multiple wind speeds. The results show that for certain favorable wind directions, increases in power production are possible even if the rate of change is high. For other wind directions the rate of change needs to be lower in order to achieve positive effects, as the redirection of wake formations can lead to unintended aerodynamic interactions between the wind turbines. For sufficiently high rates of change, wake steering results in a lower power production than is achieved with the standard maximum power point tracking approach. The utility of wake steering thus depends on the rate of change of the wind direction in relation to the wake propagation delay as well as the wind direction itself and the specific wind farm layout. For limited rates of change, the investigated wake control method increases power production in the range of single-digit percentage points.

Active wake control for wind farms offers the potential to increase the total power production and reduce mechanical stress on turbine components by manipulating the propagation of wake formations through the wind farm. However, the determination of suitable set points for the wake control methods is challenging. Varying wind conditions and a large number of turbines mean that the impact of a large number of possible set point combinations must be continuously evaluated to ensure optimal performance. Reinforcement learning for active wake control has received increasing scientific attention in recent years because, as a type of machine learning model, it learns from its interaction with the wind farm, e.g. in an atmospheric simulation model, and generalizes an optimal policy that maps the current wind conditions to wake control set points. No explicit model building of the complex wake interactions is necessary. In this paper, a concept is presented that uses a function approximation agent to determine optimal set points for wake redirection considering the constraints of modern commercial wind farms. The relevant properties of wind farms are assessed and the agent structure and learning process is crafted to match the available measurement data as well as the wind field dynamics that complicate the evaluation of beneficial control actions in large wind farms. In the future, the resulting concept will be implemented and tested using the Smart Windpark Laboratory and a dedicated wake model.

Advanced operation control methods for wind power plants aim to increase the total power production or reduce structural loads on wind turbine components. The wake redirection control is used to increase the power production by adjusting the yaw angles of all wind turbines depending on the current wind conditions and the wind farm layout. The challenge for deploying the wake redirection control is to determine yaw angle set points that yield the optimal power production under varying wind conditions. In the project SmartWind, an active wake control is being implemented that uses AI-based methods to predict the impact of a large number of set point combinations on the power production of the wind farm. Its implementation shall be presented in this paper, including the AI algorithms and the set point scenario generation based on geometric considerations of the wind farms’ spatial layout. Simulation results using real wind farm parameters are used for validation. In the simulations, the advanced operation control increases the total power production by 2.1 % compared to the standard maximum power point tracking method and the wind field is used to explain the decision-making process. Finally, further optimizations to the operation control algorithm are proposed and the next steps towards the implementation in a real wind farm are illustrated.

This paper presents a review based on Axial Induction Control (AIC), Wake Redirection Control (WRC) and artificial intelligence to support operation and optimal control of a wind farm. In addition to that, a simulation is carried out using Sim Windfarm in Matlab/Simulink, to generate a wind farm model with 8 wind turbines based on the NREL 5MW reference turbine. The data generated from the model consists of 24 measured values of wind speed and total power production by the wind farm. The main objective of the simulations is to generate data that can be used to predict the power output of the wind farm. The power prediction is done using nonlinear auto-regressive network with exogenous inputs (NARX) neural network, in which the measured wind speed is used as input to the NARX network, while power production is used as a target. The NARX configuration is based on a two-layer feed-forward network, with a sigmoid transfer function in the hidden layer and a linear transfer function in the output layer. The NARX network performs a one step ahead prediction for 1.25 seconds with small error. The prediction result indicates a good performance as confirmed by the error autocorrelation plot, regression graph and the response plot generated by the network. This prediction can be used to improve WRC, AIC, or turbines control in the range of seconds.

An increasing amount of electrical energy worldwide is produced by renewable energy sources such as wind turbines. Modern wind turbines consist of multiple mechanical and electrical subsystems which influence each other during operation. In order to improve the wind turbine design, it is desirable to simulate its operating behavior under varying conditions using accurate simulation models. This paper aims to develop a simulation model for a doubly-fed induction generator wind turbine which considers the relevant characteristics of the many subsystems involved. In this model, the wind rotor blades, mechanical drive train, generator system, and the surrounding wind field are represented and modeled in sufficient complexity. In a multi-rate simulation, the model is verified using a machine startup scenario as well as a wind gust scenario. The simulation results show the interaction between the wind rotor, drive train, and generator system. During wind speed changes, the generated power fluctuates before reaching a new steady-state condition. Oscillations occurring in the drive train put stress on the mechanical parts of the wind turbine and influence the power, transmitted to the point of common coupling.