Contino, FrancescoPinsar, DenisDenisPinsarPlancq, HadrienHadrienPlancq2025-05-142025-05-142025-05-142021https://hdl.handle.net/2078.2/22658The European Union intends to move towards zero carbon emission by 2050. There is no doubt that the decisions taken by the politicians in the coming years will play an important role in our ability to reach this emission reduction target. Unfortunately, the human brain is not capable of considering all the policy actions and their impacts in terms of cost and emissions reduction on a country-size whole-energy system. The aim of this master's thesis is to place the politician in different scenarios and determine which are the most appropriate levers to activate for the purpose of helping the energy transition. The policy framework of the politician is to impose taxes or incentives on resources and technologies used in the Swiss energy system. To do so, we have implemented a deep Q-learning model that will provide an appropriate path to follow in terms of political actions. This paper is a first attempt to apply a deep Q-learning model on a whole-energy system in order to find an appropriate series of actions which allows to reach a reduction emissions target while minimizing the cost. The results obtained on a simplified problem meet the objectives set. However, the computational time of the model representing the energy system (EnergyScope TD) is the limiting factor for considering scenarios that are closer to reality. This work is therefore the foundation on which improvements could be made to enable the model to obtain results suitable for the real world.Whole-energy systemDeep Reinforcement LearningDeep Q-learningEnergyScope TDEnergy transitiontext::thesis::master thesisthesis:30618