畳み込みニューラルネットワークを用いた太陽電池システムの異常種類の判定

カテゴリ: 論文誌(論文単位)

グループ名: 【B】電力・エネルギー部門

発行日: 2021/06/01

タイトル(英語): Fault Classification of Photovoltaic Power Plants using Convolutional Neural Networks

著者名: 五井　雅登（東京理科大学），片山　昇（東京理科大学），盛田　克彦（太陽誘電（株）高崎グローバルセンター），大川　浩（太陽誘電（株）高崎グローバルセンター），小杉　明史（太陽誘電（株）高崎グローバルセンター），今井　庸二（太陽誘電（株）高崎グローバルセンター）

著者名(英語): Masato Goi (Tokyo University of Science), Noboru Katayama (Tokyo University of Science), Katsuhiko Morita (TAIYO YUDEN CO., LTD. Takasaki Global Center), Hiroshi Okawa (TAIYO YUDEN CO., LTD. Takasaki Global Center), Akifumi Kosugi (TAIYO YUDEN CO., LTD. T

キーワード: 太陽光発電システム，異常検知，異常分類，機械学習，畳み込みニューラルネットワーク　　photovoltaic power plants，anomaly detection，classification of faults，machine learning，convolutional neural network

要約(英語): Recently, faults of photovoltaic power plants are becoming a serious issue because it may prolong the payout time for power plant installation. Under this situation, not only detection of the photovoltaic module itself but also classification of the kind of faults has been attracting much attention. This study applies supervised the machine learning algorithm using neural networks to fault classification, which is able to lower maintenance costs. However, a large amount of input and output data are required to obtain enough estimation accuracy for machine learning models. In fact, in a single photovoltaic power plant, anomalies are not frequent and it is difficult to collect data to withstand practical use.In this study, a numerical simulation to generate a large amount of voltage data when the photovoltaic power plant has fault has been developed and the voltage data was generated to train machine learning models. The generated voltage data reproduce measured data from an actual photovoltaic power plant. For faults such as cell fault and shadow, the proposed method functions properly and classified the anomaly at high accuracies.

本誌: 電気学会論文誌B（電力・エネルギー部門誌） Vol.141 No.6 （2021）

本誌掲載ページ: 464-472 p

原稿種別: 論文／日本語

電子版へのリンク: https://www.jstage.jst.go.jp/article/ieejpes/141/6/141_464/_article/-char/ja/