[2K01-05] Advancement of the Maintenance and Preservation Technology of the Plant 2

[2K02] Development of Abnormal Sign Detection System using AI for Nuclear Power Plant

(4)Improvement of Detection Algorithm with Two-stage Autoencoder

*Susumu Naito¹, Yasunori Taguchi¹, Yuichi Kato¹, Kouta Nakata¹, Isaku Nagura², Shinya Tominaga², Ryota Miyake², Toshio Aoki² (1. TOSHIBA, 2. TOSHIBA ESS)

Keywords:plant performance monitoring, plant health monitoring, early detection of anomaly signs, machine learning, deep learning, autoencoder

In a large-scale plant such as a nuclear power plant, thousands of process values are measured for the purpose of monitoring the plant performance and the health of various systems. It is difficult for plant operators to constantly monitor all of the process values. We present a new data-driven method to monitor many process values and to enable early detection of anomaly signs including unknown events with few false detections. We created a network configuration corresponding to the features of process values of a nuclear power plant. It is composed of two autoencoders. Each autoencoder learns signals of different characteristics to predict the normal state with high accuracy. We evaluated detection performances of this two-stage autoencoder with simulated process values of a nuclear power plant, a 1,100 MW Boiling Water Reactor having 3,100 analog process values. The two-stage autoencoder clearly showed a good anomaly detection performance with zero false detections in the steady state and the transient state, under noisy conditions.