Prediction of the operating control rod position of the HTTR with supervised machine learning

Ho, H. Q.   ; Nagasumi, Satoru; Shimazaki, Yosuke ; Hamamoto, Shimpei  ; Iigaki, Kazuhiko ; Goto, Minoru  ; Simanullang, I. L.* ; Fujimoto, Nozomu*; Ishitsuka, Etsuo   

During operation of the HTTR, hundreds of technical signals and operating conditions must be observed and evaluated to ensure safe operation of the reactor, for example reactor power, control rod position, coolant flow rate inlet/outlet, coolant temperature inlet/outlet, etc. The accumulated experiment data of the HTTR is not only important for the HTTR operation, but also for the basic development of the HTGR in general. Artificial intelligence (AI) and particularly machine learning (ML) are increasingly being used in various fields of research in modern science. They give the ability to make predictions as well as allow the extraction of key information about physical process from large datasets. Hence, there is a lot of potentials to apply AI and ML to predict the operating and safety parameters of the HTTR, and finally, a reactor simulator system for the HTTR could be expected by using the AI and ML algorithm. In this study, the control rod position of the HTTR is predicted based on ML without using the conventional neutronic codes. With the large accumulated data from operation history of the HTTR, the supervised ML with a linear regression algorithm was used. The linear regression algorithm finds a functional relationship between the input dataset (reactor power, burnup, etc.) and a reference dataset (control rod position), constructing a function that predicts control rod position from the other operation conditions. As result, the ML gives a good prediction of the HTTR control rod position with less than 5 difference compared to that in the experiment. This study is the initial step towards machine learning for research and analysis at the HTTR facility. With increasingly complicated experiments that create a large amount of data, ML is also expected to improve the design and safety analysis of the HTTR in the future.