Optimised Geographical Allocation of Wind Energy Capacity using a Mean-Variance Portfolio Algorithm for Clustered and Un-clustered Profiles

This paper explores the use of mean-variance portfolio optimisation to find the optimal geographical spread of wind energy capacity as a function of the variability of the cumulative wind power profile. The methodology is applied for the eight Renewable Energy Development Zones identified for the South African renewable energy program.The effects of using the Clustering LARge Applications algorithm as a data reduction pre-processor on the performance of the mean-variance portfolio algorithm is investigated. In this case the optimisation is performed using an optimal number of representative wind power clusters for the case study areas. The introduction of clustering reduces the computational intensity and processing time associated with the optimisation algorithm significantly, while still yielding good results for geographical allocation of wind energy capacity. A proposed method of combining mean-variance portfolio theory and Clustering LARge Applications algorithm to more accurately determine the optimal number of clusters needed is also presented.