Wind turbine power modelling and optimization using artificial neural network with wind field experimental data

The wake effect is a major and complex problem in the wind power industry. Wake steering, such as controlling yaw angles of wind turbines, is a proven approach to mitigate the wake influence and increase the power generation of a wind farm. This paper proposes a power prediction model and optimizes yaw angles to minimize the entire wake impact on wind turbines. The power model adopts the artificial neural network (ANN)with the consideration of the wake effect, so it is called ANN-wake-power model. The model can estimate the total power generation of wind turbines for given wind speeds, wind directions, and yaw angles. A case study has been conducted to introduce the modelling process. The experimental data of five wind turbines from an operating wind farm have been used to train and evaluate the model. The ANN-wake-power model has proven to be effective in estimating the power generation. It performs a good balance between computational cost and accuracy. Subsequently, the model is applied to optimize the yaw angles by using Genetic Algorithm. With the optimized yaw angle strategy, the total power ratio of wind turbines can reach 0.96 in all directions involved. For a row of wind turbines, the optimal yaw control strategy for each wind turbine is different. Finally, it is worth noting that, to achieve a good performance of the ANN-wake-power model, sufficient input data should be adopted in the training process.