| 摘要 |
<P>PROBLEM TO BE SOLVED: To establish a prediction model which corresponds to periodic climates and a property change of a photovoltaic power generation panel, capable of performing future power generation amount prediction accurately. <P>SOLUTION: Past power generation amount achievement y<SB POS="POST">i</SB>, climate data x<SB POS="POST">i</SB>and measurement date and time t<SB POS="POST">i</SB>thereof are used to establish a set of input/output (z<SB POS="POST">i</SB>, y<SB POS="POST">i+h</SB>) for storing in a storage section 15 (S23). New climate data x<SB POS="POST">q</SB>and measurement day and time t<SB POS="POST">q</SB>thereof are used to establish input data serving as query z<SB POS="POST">q</SB>(S26). For all input data z<SB POS="POST">i</SB>and query z<SB POS="POST">q</SB>, similarity J(z<SB POS="POST">q</SB>, z<SB POS="POST">i</SB>) is calculated with a lapsed time based on a measurement day and time vector without the element of similarity between climate data and year (S29), an N data sets (z<SB POS="POST">i</SB>, y<SB POS="POST">i+h</SB>) corresponding to the ascending order of the similarity J (z<SB POS="POST">q</SB>, z<SB POS="POST">i</SB>) is selected (S30) and a prediction model f is established using the N data sets (z<SB POS="POST">i</SB>, y<SB POS="POST">i+h</SB>) (S31). <P>COPYRIGHT: (C)2013,JPO&INPIT |
