Model Prediction XGBoost for Perspective of Closing Cernavoda Reactor 1 and the Impact on the Romanian Energy System

Abstract

Abstract Employing artificial intelligence (AI) and machine learning (ML) technologies to analyze the progression of energy markets is imperative, as these methodologies facilitate the modeling of diverse domain-specific data to forecast developments in an exceedingly volatile market. Utilizing machine learning algorithms to model data concerning the Romanian energy sector proves beneficial, particularly in circumstances where significant changes necessitate adaptation to internal consumption demands, while mitigating substantial fluctuations in electricity prices. Such fluctuations can profoundly impact crucial national economic indicators.. In this paper, we aim to simulate such a situation for the moment when Romania will be obliged to shut down and modernize Reactor 1 at Cernavodă due to reaching the standard operating time for the CANDU model. We will use an XGBoost for Regression algorithm in which we will input data about the national energy system (consumption, energy quantities according to the energy system's production structure by resource types) from the period 2020-2023, and then we will simulate the shutdown of the reactor to analyze how the national energy market is influenced. Our findings not only reveal the indispensable role of predictive analysis through ML/AI models in anticipating market changes, but also provide practical insights for decision-makers aiming to optimize the national energy system so that when Reactor 1 at Cernavodă is shut down, the domestic market is affected as little as possible, considering that this reactor has supplied the national system with approximately 700 MWe/h over the past 20 years.