Improved winding design of a double fed induction generator (DFIG) wind turbine using surrogate optimisation algorithm

Abstract

The use of renewable sources of energy is becoming increasingly important role in electricity generation. Wind energy is the fastest growing renewable energy source and is making a significant contribution to meeting the energy demands while still reducing CO2 emissions. In designing generators for installation in wind turbines, characteristics of high efficiency and low cost are among the first to consider. As the number of installed wind turbines increases across the world, questions of turbine component failure and repair are also receiving much attention. This PhD starts with the investigation of modern wind turbine generator design with a focus on electrical generator and its operation. The finite element analysis of an off-the-shelf 55 kW doubly fed induction generator is used as a case study in order to investigate its design and improve the machine performance. The main work of this PhD is on a novel approach by surrogate-based analysis and the optimisation of winding design and rewinding design based upon the doubly fed induction generator for energy efficiency improvement. Surrogate models of the machine are constructed using Latin Hypercube sampling and the Kriging modelling. Having validated the surrogate models, the particle swarm optimisation algorithm is developed and applied to find the optimal solution. Assuming a winding failure occurs mid-life of the wind turbine, three optimisation plans have been studied for the repair and re-design of the stator and rotor winding separately and in combination. To validate the optimisation results, an improved testing standard is developed to test doubly fed induction generator. The original machine is then rewound following the optimised plan and tested to determine the difference in performance. By comparing the two machines, improved performance is achieved both in optimisation simulation and experiments. Finally an annual wind speed profile at a specific location (Albemarle site) is analysed to estimate wind power. The Weibull distribution of the wind speed data is combined with the turbine topologies for estimating the annual wind energy production. The annual power generation from the two machines is compared to validate the proposed technology.