Resampling of Data for Offshore Grid Design Based on Kernel Density Estimation and Genetic Algorithm

Abstract

Offshore wind power has been a major focus in the renewable energy development in recent years, due to better wind speeds and wind energy are available offshore. Since the development (construction and grid connection) of offshore windfarms is relatively more expensive in nature, careful planning and design are needed to maximise the benefits of the offshore wind projects. Optimisation with only one operational state is not sufficient in grid design as the state of power system is not stationary due to the fluctuations of the wind power and power consumption. Eventually this leads to the fluctuation of the base load power generations. To account for this variability, the optimisation has to be done with many operational states. Historical data of power consumption at each load centre and simulation data of wind power have to be used to describe the system states. Ideally, the complete set of data should be used to describe the power system states but this could also lead to unsolvable case as there are too many unknowns involved in the calculation. To keep the number of states as low as possible to reduce the computation time, selection of smaller number of samples that can represent the whole data set has to be carried out. This involves detail studies of the statistical distributions of the data. This study is therefore dedicated to develop a procedure for selecting a set of statistically sound samples to represent the entire data set for grid design purposes.