Wake Flow Model for Wind Farm Control

Abstract

A hybrid methodology for the real-time analysis and modelling of wind turbine wake effect including wind speed and turbulence is presented in this paper. Large wind farms can take advantage of economies of scale by installing wind turbines in dense arrays. This reduces installation and interconnection costs as well as operation and maintenance costs. However, these turbines generate wake effects and thus suffer from array losses. These losses can be in the range from 20 – 45 %, severity of which depends upon different conditions [1]. A wind farm has to produce the maximum possible energy with limited number of turbines and minimum spacing between them. Traditionally wind farms operate with a greedy approach, in which each turbine increases its own power using Maximum Power Point Tracking (MPPT), ignoring the wake effects produced. This may not always be the best strategy. Upstream turbines will use significant and substantial quantities of the kinetic energy in wind, reducing wind speed for downstream wind turbines with increased turbulence. Lower wind speeds means less energy production and higher turbulence means increased fatigue loads. A coordinated control can be used to increase the efficiency and reliability of the wind farm. If the upstream turbine is de-rated so that the decrease in produced power is less than the increase in shadowed turbines powers, then there will be an overall increase in farm output whilst decreasing mechanical loads. This will increase wind farm life and will reduce cost of energy, which will be helpful in making wind energy competitive with conventional energy sources. Figure 1, presents a preliminary comparative analysis of a simple array of two 6MW wind turbines, quantifying the benefits of a coordinated control strategy. In the first case, the array is operated with the traditional greedy approach. Though upstream turbine is almost producing 6MW - its maximum - the farm output is 10.1MW. However, in the second case – when the upstream turbine is de-rated by (-12.16%) to produce 5.27MW - the farm output is 10.39MW. This represents a net increase of 2.87% in total farm output. In a larger farm the result would be even better. An optimizer can be used for selection of the best possible reference points for each turbine. This optimizer, if used, requires a fast processing wake flow model which gives both speed deficit and turbulence intensity at a downstream turbine with sufficient accuracy. CFD based wake models could produce accurate results but are computationally expensive. Despite some development, these models are still time consuming and are not the first choice for farm controller / optimizer, especially for real time control. In this work, a computationally effective wake model based upon the Jensen model [2] is presented which is combined the model developed in [3] to provide a holistic wake effect model including turbulence intensity. This model enables real-time computation of both multiple and partial wakes, enabling the use of wind farm level controllers to optimize yield and minimize fatigue. The integration of real-time wake modelling into a coordinated wind farm controller will significantly increase farm yield whilst significantly reducing upstream turbine fatigue.