時系列観測データに基づく太陽光発電システムの出力予測

抄録

Systems that generate electricity from renewable energy sources have become a focus of interest, and photovoltaic(PV) systems are being rapidly introduced. A particular problem of PV renewable energy systems is that their output power fluctuates greatly according to weather conditions. Connecting large numbers of unstable power generators such as PV systems to the power grid could compromise the quality of the power supply. To prevent degraded power quality, it is importallt to develop methods of predicting how constantly fluctuating power sources will change. In this paper, the MT system (T method-1) is used to predict electric power one hour in advance from time-series data obtained from meteorological observations. When the method was tested against actual observational data recorded at one-hour intervals, it was confirmed that the selection of the periods covered by the unit space and the signal data, the sizes of the hourly changes in feature variables, and transformations of these variables all contributed to improving the predictions.