A storage of ICIS historical data at the NRC «Kurchatov institute». Potential and perspectives

28th Symposium of AER on VVER Reactor Physics and Reactor Safety (2018, Olomouc, Czechia)
[4] Core surveillance and monitoring

Authors

M.V. Khalizov, O.I. Sinegub, I.A. Eliseev, D.V. Vorobyeva (NRC Kurchatov institute)

Abstract

The in-core instrumentation system (ICIS) of the VVER was designed for a safe operation of a whole power unit and a nuclear fuel in particular.

As for now, the last generation of in-core instrumentation systems designed by the NRC “Kurchatov Institute” have been successfully deployed and has been operating at power units of Russian nuclear power plants (Balakovo NPP, Kalinin NPP, Rostov NPP, Novovoronezh NPP and Leningrad NPP) and foreign nuclear power plants (Kozloduy NPP, Kudankulam NPP, Tianwan NPP and etc.).

During its operation, ICIS stores a large amount of data into historical data archives. ICIS historical data archives are the prime source of the key scientific information for tasks such as:

- NPP operation analysis;
- Design and verification of a new reactor related software.

Today the amount of the available ICIS historical data surpasses more than one hundred of fuel cycles, totally exceeding approximately 25 TB of a highly compressed data.

The NRC “Kurchatov Institute” have developed special framework to provide access to the compressed ICIS historical data without a headache of deploying full local copy of the ICIS software. This framework includes:

- Tools for the direct work with data;
- Administrative tools for the maintenance of data storage;
- User’s API;
- Built-in visualization and analysis tools.

By the means of this framework scientists and engineers of the NRC “Kurchatov Institute” can easily access and work with any available on the server historical data via LAN.

The NRC “Kurchatov Institute” have proposed a bunch of machine learning projects focused on a development and implementation of empirical predictive models for ICIS. Some moderate progress was achieved with help of such large amount of data and modern open source data analysis/machine learning tools within Python eco system (mostly for rapid R&D).