翌日PV発電出力における10秒単位変動の信頼区間推定

抄録

Photovoltaic (PV) power generation systems have gained much attention against a background of energy and environmental problems. However, the output of PV system depends on climate conditions. Therefore, a mass introduction of PV results in the power quality deterioration in the distribution system. To overcome the difficulty, it is effective to introduce Low Voltage Regulator (LVR) to the distribution system. However, the application of LVR brings a new problem, i.e., a control delay for activating the tap after detecting the fluctuation. With this background, to provide more useful predictive information in advance, this paper deals with a novel prediction interval estimation method of 10 second PV fluctuation with two-dimensional kernel density estimation. We also carry out the prediction by using several different methods for comparison, e.g., the conventional method with one-dimensional kernel density estimation and the application of various smoothing methods to this conventional method. Finally, some numerical results, which demonstrate the effectiveness of the proposed method on the basis of several indexes, are also presented.