Multi-Objective Optimization Design of An Integrated Energy System with Ice Storage Based on Deep Reinforcement Learning

ABSTRACT: This study proposes an integrated energy system that combines photovoltaic power, wind power, battery storage, and ice storage to meet the electricity and cooling demands of buildings. The model for the ice storage tank incorporates the nonlinear ice-melting characteristics. An improved Multi-Objective Proximal Policy Optimization algorithm is employed for multi-objective optimization. In a case study of an office building in Shanghai, the optimization results demonstrate that the proposed method reduces daily operating costs by 6.52% and improves the CO₂ emission reduction rate by 9.54%. The results demonstrate that the synergistic operation of electrical and ice storage effectively maintains supply-demand balance across different seasons. Sensitivity analysis further reveals that a 40% reduction in the unit cost of ice storage leads to a 5.7% decrease in battery capacity and a significant drop in grid dependency from 28.9% to 15.3%, highlighting the critical role of reducing ice storage costs in improving the system’s economic viability and renewable energy integration capability.