Although renewable energies such as photovoltaics are penetrating rapidly, in order to harness their varying output efficiently and robustly it is necessary to control the power systems with not only hardware facilities but also software systems. To this end, electricity storage devices from large stationary batteries to electric vehicles will be one of the most important hardware systems, which will require also scheduled operation based on forecasting output of renewables. As the forecast of renewables' output is a complex problem because of its large and varying uncertainty, a probabilistic forecast is better than a deterministic forecast. The aim of this paper is to advance a novel approach of probabilistic forecast of solar irradiation with consideration of temporal correlation as well as varying uncertainty by using copula-based Markov process and variable dispersion beta regression. Employing these methods, we have verified in terms of the performance indexes of reliability and sharpness that the marginal distributions of solar irradiations are well expressed by the beta distribution with varying dispersion as well as their temporal correlation can be well modeled by using Gumbel copula. This temporally-correlated model enables updating a day-ahead forecast by observed irradiations resulting in improving confidence intervals. In addition, probabilistic scenarios of irradiations can be generated by this model, which enables a precise probabilistic forecast of daily cumulative irradiation.
J-STAGEがリニューアルされました!https://www.jstage.jst.go.jp/browse/-char/ja/