NUCLEAR TECHNIQUES, Volume. 45, Issue 12, 120602(2022)

Analysis and prediction of nuclear power plant operation events based on ARIMA-LSTM model

Qinmai HOU, Wei ZHU*, Xiang ZOU, Shixian LIU, and Yannong WU
Author Affiliations
  • Nuclear and Radiation Safety Center, Beijing 102445, China
  • show less
    Figures & Tables(7)
    Number and trend of operation events from 1991 to 2018
    M-K test results of the number of operation events
    ACF (a) and PACF (b) maps of first order difference
    Predicted values of combined model
    • Table 1. ARIMA model parameter estimation

      View table
      View in Article

      Table 1. ARIMA model parameter estimation

      ARIMA modelAR1AR2AR3MA1MA2AICR2
      ARIMA (3, 1, 2)1.256 3-0.477 6-0.132 7-1.997 00.998 9243.940.542
      ARIMA (3, 1, 1)0.526 8-0.082 9-0.250 4-1.000 0243.760.475
      ARIMA (2, 1, 2)1.352 7-0.624 0-1.992 60.999 6242.200.504
      ARIMA (2, 1, 1)0.561 1-0.205 0-1.000 0242.530.417
    • Table 2. Predicted values of operation events from 2019 to 2021

      View table
      View in Article

      Table 2. Predicted values of operation events from 2019 to 2021

      年份

      Year

      预测值

      Predicted values

      95%置信区间

      95% CI

      201928.02(-3.843 962, 59.888 68)
      202021.46(-13.263 979, 56.189 37)
      202120.06(-14.526 518, 54.653 04)
    • Table 3. Model evaluation parameters

      View table
      View in Article

      Table 3. Model evaluation parameters

      模型ModelMSERMSEMAPE
      ARIMA234.9215.3236.5%
      ARIMA-LSTM223.9414.9633.6%
    Tools

    Get Citation

    Copy Citation Text

    Qinmai HOU, Wei ZHU, Xiang ZOU, Shixian LIU, Yannong WU. Analysis and prediction of nuclear power plant operation events based on ARIMA-LSTM model[J]. NUCLEAR TECHNIQUES, 2022, 45(12): 120602

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Research Articles

    Received: Apr. 20, 2022

    Accepted: --

    Published Online: Jan. 3, 2023

    The Author Email: ZHU Wei (zhuwei@chinansc.cn)

    DOI:10.11889/j.0253-3219.2022.hjs.45.120602

    Topics