Preprocessing of Multi-Time Instant PV Generation Data

For the evaluation of system overlimit risk indices in a PV-integrated power system, PV generation data at specific instants of time (in each day for several years) are required to be collected. Such data have inherent annual periodic variations, which are different at various places. These variations are skewed and/or multimodal, which contributes significantly toward the overall variance of data and is primarily attributable to the Sun's position. This letter proposes a regression model that assumes the observed data as a function of few influencing factors related to the Sun's position and trend in data. Finally, the estimated variations using the developed model are removed from the data to characterize the unpredictable components.