A Study on Optimizing Investment Value of Power Transformers Using Reinforcement Learning Algorithms

We are studying a risk-based asset management system for objective decision-making. In order to build an AMS, we must go through the APM and AIPM stages. In order to accurately verify the investment plan established based on risk, we must perform the AIPM stage using an appropriate optimization algorithm. However, the existing linear optimization technique does not consider many constraints and nonlinearity of PoF that must be considered in the process of establishing an investment plan for power facilities. Therefore, we must reflect the AIPM stage appropriately by optimizing using an optimization technique that considers many constraints and nonlinearity of PoF. Therefore, to solve this problem, in this paper, we propose a method to perform the AIPM stage by applying the DDQN (Double Deep Q-Network) method, a reinforcement learning algorithm that can accommodate various constraints and nonlinearity.