Simulation and control of windfarms

Thesis

This thesis examines the design of supervisory controllers for windfarms of pitch-controlled wind turbines. The control objectives are the maximisation of the financial income from the generated electricity and the minimisation of the turbines' fatigue damage. The design exploits the wide variations in the ratio of financial income to fatigue damage which are found both spatially across windfarms and as a function of time. The supervisory control strategy makes use of the ability of pitch-controlled turbines to operate with variable power set points; a capability which is rarely exploited in practice.

A windfarm simulation which has been developed for the purposes of testing supervisory controllers is described. It is shown that the simulation is a suitable test-bed for this application. Results are presented which demonstrate how the fatigue damage of a turbine's gearbox and structural components vary as functions of the mean wind-speed, turbulence intensity and power set point, both for isolated turbines and for turbines experiencing wake effects.

A lifetime performance function is proposed and 'ideal' power set point curves are evaluated using a genetic search algorithm. It is shown that significant improvements in performance can be achieved if the operation of the turbines is altered to take account of variable electricity tariffs.

A windfarm control strategy that splits the turbines into interacting and non-interacting categories is found to give good results. Using data generated by the simulation, it is shown that simple cost functions can be developed for non-interacting turbines which, when used in a controller, give performance that is close to the 'ideal'. A similar cost function is applied to a group of three interacting turbines, and it is found that substantial reductions in all measures of total annual fatigue damage are achieved for a small reduction in total annual financial income. The on-line implementation of windfarm supervisory controllers is discussed and the behaviour of a simple hill-climbing algorithm is examined using a simulated group of three interacting turbines.

Authors: Spruce, C; Christopher John Spruce

• Publication date: 1993
• DOI: 10.5287/ora-mqdaajbmj
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: supervisory control; optimisation; genetic algorithm; wind power plant; wind energy; control engineering; wind turbine; fatigue; mechanical engineering
• Subjects: Control engineering; Wind energy; Fatigue; Mechanical engineering