XAI based Energy Platform Development

Development of deep learning processing platform for detecting and responding to abnormalities of power facilities using Terabyte class power data

    Real-time analysis of power monitoring data streams is required to detect electrical equipment failure and response to that failure; failure to respond immediately could lead to major accidents such as blackout and fire.

Figure 1. XAI-based Energy Platform Architecture for Detecting Abnormalities of Power Facilities


1. Distributed Real-time CEP Processing Acceleration

  • Real-time detection of power patterns that have high probability of failure from data stream continuously collected by the real-time monitoring instrument
  • reducing the amount of information transferred between servers in multi-server deep learning
    Real-time detection of power patterns that have high probability of failure from data stream continuously collected by the real-time monitoring instrument & distributed task execution technology for deep-learning-based Complex Event of time series data stream.

Figure 2. Increase data generated and accelerate processing for complex event processing


2. Deep learning based CEP

    The amount of data which should be handled in real-time is exploding, and as a result, power Industry requires real-time analytics of huge amount of stream data. In this context, we consider the streaming data as events and explores the possibilities to combine them for detecting abnormalities. We study real-time detection of power patterns that have high probability of failure from data stream continuously collected by the real-time monitoring instrument. Furthermore, we utilize distributed task execution technology for deep-learning-based Complex Event of time series data steam to accelerate processing. Deep learning processing acceleration technologies which we consider also include using real-time dynamic task allocation in a heterogeneous acceleration environment which contains FPGA and GPU.

Figure 3. Deep Learning based CEP accelerated platform

    (1) Big data Processing

    Events should be handled with better performance by taking patterns into account multiple sources of data

    (2) Explainable AI

    • Using deep learning models such as LSTM with provides better performance and reliability than traditional Moving Average-based techniques in time-series forecasting
    • Deep Learning enables data attributes to be reflected by updating the model over time
  • heterogeneous many-core hardware system for UHPC


373-1, Guseong-dong, Yuseong-gu Daejeon, 305-701, Korea
Phone: +82-42-350-3495 FAX: +82-42-350-7260
Network and Computing Lab http://ncl.kaist.ac.kr
( Webmaster: iop851@kaist.ac.kr ─ Last update: 2018.09.27 )


Last update: 2017.09.11
Webmaster: iop851@kaist.ac.kr