Comparación de técnicas de relleno de datos faltantes de la variable velocidad de viento de los años 2014 al 2021

Abstract

Based on the meteorological study of the wind speed, which is one of the important variables that allowed the generation of energy from wind sources, being the latter that was tested to be able to know its consistency, to this was added that the studies in meteorology are very changeable due to different variables that occurred with the passing of time. The main objective of this work was to compare the techniques for filling missing data of the variable Wind Speed from 2014 to 2021, in the province of Chimborazo. The data matrix of this study was obtained from the Alternative Energies and Environment Group (GEAA), with a total of 703248 data, where through a purification of information a sample of 10964 missing data was taken from the 11 stations. The descriptive statistical analysis showed a minimum wind speed (1,000 m/s) at the Cumandá station and a maximum speed (20,703 m/s) at the Tixán station; the Rosner’s test for outliers revealed (677) outlier data. From the function (V IM.impute.pmm()) missing data were identified, where the most representative data were found at the Atillo, Multitud and Tunshi stations in the study years. With the Random Forest, Mean (MICE), Hot Deck and Iterative PCA Imputation methods, it was possible to impute and make the necessary adjustments in each station, graphically it was observed that Alao had a good adjustment between the real data and the imputed data, by means of the metrics EMP, EMC and DAM the Quimiag station presented low values 0.0021, 0.0002 and 0.0021 respectively for the mean (MICE) and Hot Deck techniques being those that best adjusted to the real data