Investigating the impact of data quality on the energy yield forecast using data mining techniques

Integration of renewable energy resources (RES) in the power system is increasing. However, the variability of PV output poses several challenges in maintaining reliable grid operation. This variability is a function of several meteorological variables. An accurate PV forecasting can tackle this issue in maintaining and scheduling stable grid operation. In this paper, we conduct a study to analyze the impact of using optimum combination (feature selection and extraction) of meteorological features and low dimensional subspace (dimensionality reduction) on the forecasting accuracy. We also assess and compare the output of the forecasting model when it is fed with all the input features present in the dataset with the case when we use low subspace of the dataset as an input to the model.