DEVELOPMENT OF A PPO-BASED MODEL FOR OPTIMIZING INDUSTRIAL ENERGY CONSUMPTION

Abstract

The article considers a reinforcement learning model for optimizing industrial energy consumption using the Proximal Policy Optimization (PPO) algorithm. The proposed approach is aimed at supporting energy management decisions under variable production demand, dynamic electricity prices, equipment condition constraints and the availability of local generation and battery storage. The energy management task is formulated as a Markov decision process, where the state space includes production load, energy price, solar generation, battery state, equipment health and demand indicators. The action space is represented by a continuous control vector for changing equipment load levels, while the reward function combines energy cost reduction, demand satisfaction and equipment health preservation. The model is tested in the IndustrialEnergyEnv simulation environment and compared with a baseline control scenario. The results show that the PPO-based model can reduce total energy consumption and economic cost; however, additional reward balancing is required to ensure full production demand satisfaction