Missing Wind Speed Data: Clustering Techniques for Completion and Computational Intelligence Models for Forecasting

摘要

This paper addresses two objectives for the case of missing wind speeddata: (i) the implementation of clustering techniques for completion ofmissing wind speed data and (ii) the development of offshore windspeed forecasting models. Various clustering algorithms are comparedin terms of better portioning of the wind speed data. With the aid of arobust clustering tool, a more in depth analysis can be held on windspeed data drawing useful conclusions about the data structure. Inaddition, three novel techniques are developed for the completion of themissing data. Furthermore, two commonly used forecasting models areused, namely a Feed-Forward Neural Network (FFNN) and theAdaptive Neuro-Fuzzy Inference System (ANFIS). The purpose is totrain and test these models under the limitation imposed by theincomplete data set. The present paper serves as a necessary step in theproblem of handling incomplete wind speed data towards theassessment of offshore wind energy potential.