2段階オートエンコーダの 発電所データでの異常予兆検知性能評価
英字タイトル:
Performance Evaluation of the Anomaly Detection by Two-Stage Autoencoder using Power Plant Data
著者:
田口 安則 Yasunori TAGUCHI 内藤 晋 Susumu NAITO 加藤 佑一 Yuichi KATO 中田 康太 Kouta NAKATA 富永 真哉 Shinya TOMINAGA 高戸 直之 Naoyuki TAKADO 三宅 亮太 Ryota MIYAKE 寺門 優介 Yusuke TERAKADO 青木 俊夫 Toshio AOKI 高森 由己夫 Yukio TAKAMORI 大熊 栄一 Eiichi OOKUMA
発刊日:
公開日:
anomaly detectionautoencodermultivariate time series dataplant monitoringtwo-stage autoencoder
Performance Evaluation of the Anomaly Detection by Two-Stage Autoencoder using Power Plant Data
著者:
田口 安則 Yasunori TAGUCHI 内藤 晋 Susumu NAITO 加藤 佑一 Yuichi KATO 中田 康太 Kouta NAKATA 富永 真哉 Shinya TOMINAGA 高戸 直之 Naoyuki TAKADO 三宅 亮太 Ryota MIYAKE 寺門 優介 Yusuke TERAKADO 青木 俊夫 Toshio AOKI 高森 由己夫 Yukio TAKAMORI 大熊 栄一 Eiichi OOKUMA
発刊日:
公開日:
カテゴリ: 論文
キーワードタグ:anomaly detectionautoencodermultivariate time series dataplant monitoringtwo-stage autoencoder
概要
In power plant, the operator monitors measured values of many installed sensors for operation and maintenance. Since the number that can be visually confirmed is limited, we have proposed two-stage autoencoder for anomaly detection. The latest version was developed to suppress false positives due to spurious correlations in training data, and its effectiveness was shown using simulation data. In this paper, the anomaly detection performance is shown using the operational data of an actual power plant. As a result, the effect of suppressing false positive detections was confirmed. It was also confirmed that a repair work overlooked by comparison methods was correctly detected as an anomaly.
全文掲載のPDFファイルダウンロード
会員の方はログインしてください
ログイン 会員登録