重回帰分析を用いた系統エリア大の風力発電・太陽光発電の出力予測と予測誤差特性の分析

抄録

Renewable power forecasting is one of key components to maintain the reliability of a power system with large-scale renewable integration and has been studied by use of a variety of approaches, such as area-wide numerical weather simulation, satellite image based forecasting and statistical model. In this paper, an approach to predict the area-wide total wind and photovoltaic power of one hour to one day ahead prior to the power delivery is presented, which makes use of multi regression technique and numerical weather prediction data distributed by the Japan Meteorological Agency. Tohoku area is taken as an example to show the prediction accuracy. The results show that, forecasting errors of the proposed models are a little higher than those of area-wide numerical weather simulation of earlier studies and further investigation in this approach is needed to improve the accuracy. According to the analysis of forecast error characteristics, both of the forecast error distributions of wind and photovoltaic power look well-fitted with Laplace distribution rather than Normal distribution.