Estimation and forecast accuracy of regional photovoltaic power generation with upscaling method using the large monitoring data in Kyushu, Japan

摘要

In order to optimize photovoltaic (PV) output curtailment control, forecasting a regional PV power generation are an important issue. Its estimation is also important as a basic step prior to forecasts. Upscaling algorithm is general approach for evaluating and forecasting a regional PV power generation because the number of monitored plants is usually limited. However, the method leads to large error when the characteristics of monitored plants differ from those of unknown plants in a region. In this paper, we analysed the errors on estimation and forecast of regional PV power generation with upscaling method by using monitoring data obtained from 2219 small PV plants in Kyushu, Japan. As the results, random sampling method has sufficient accuracy for day-ahead and short-term forecasts in case of the large number of reference plants, and unlike forecasts the minimum estimation error does not remain flat and continued to decrease as the number of power plants increased.