Abstract: Environment radiation monitoring system (ERMS) around nuclear power plant can provide real-time and continuous monitoring data, which is the most important peripheral supervision facility of nuclear power plant and provides data basis for radiation environment assessment. In order to master the characteristic elements that affect the quality of radiation data and timely detect environmental radiation anomalies, data feature exploring and prediction research of γ radiation dose rate data were carried out. A preprocessing method of HPIC dose-rate data based on singular spectrum analysis was proposed to learn the increase trend and inflection point details from its historical data. According to the multidimensional characteristics of data, a SSA feature fusion parallel optimization model prediction framework was designed, and simulation experiments were carried out. Data of 11 automatic radiation monitoring stations around Ningde Nuclear Power Plant in Fujian Province and Vertical Total Electron Content (VTEC) were used for experimental verification. The experimental results show that the feature fusion network model achieves good prediction performance and accuracy for γ radiation dose rate prediction.

Key words: time series, radiation environment, HPIC, γ radiation dose rate, singular spectrum analysis, feature fusion network