Session: 01-02: AI for Energy Sustainability II

Paper Number: 142294

142294 - Energy Management in Renewable Energy-Based Distributed Generation Using Artificial Intelligence Optimization Technique 

Abstract: 

As the importance of renewable energy is emphasized, the installation of renewable energy is increasing. This is accompanied by a rise in usage of various energy sources, such as non-renewable energy sources, energy storage systems, and hybrid energy sources, to accommodate the burgeoning renewable energy capacity. Optimizing the energy management among these diverse energy sources complicates problem-solving and is time-consuming. To address this chronic problem, this study employed an artificial intelligence optimization technique within complex distributed generation. The object of optimization was a distributed generation system consisting of renewable energy, a gas turbine, a fuel cell, an adiabatic compressed air energy storage, and batteries. The optimization technique combines regression models of energy sources, constructed using machine learning, with a metaheuristic algorithm. To train the regression models, rigorous physics-based modeling of energy sources is required. Physics-based modeling not only reduces the risk of overfitting during the training process but also enables accurate simulation of operating characteristics. It is also crucial for verifying the operational feasibility of newly proposed energy systems. Constructing regression models for energy sources based on physics-based modeling and utilizing these models for optimization can preserve the operating characteristics of each energy source while improving computational efficiency. Genetic algorithm was selected as the metaheuristic algorithm, and the combination of the regression models was used as a fitness function of genetic algorithm. This optimization technique can be extended to include new energy sources, broaden the observation target, and add objective functions, allowing for extensive customization based on the optimization purpose. In the optimization process, meteorological data and demand profiles are provided by users, and the configuration of energy sources and objective functions are selected. The installation amount of renewable energy and the demand size can either be specified as input values or determined as optimal values through the optimization. The results of the artificial intelligence optimization technique identify the optimal dispatch for each energy source. To evaluate the performance of the optimization technique, it was assumed that renewable energy, the gas turbine, and the compressed air energy storage are operated in the distributed generation. The installed amount of renewable energy was optimized, and the demand size was specified with maximum electric demands of 20 MW and 50 MW. Operational cost per total output energy for distributed generation and self-sufficiency of distributed generation were considered as objective functions. At a maximum electric demand of 20 MW, the optimal load sharing rate of renewable energy was 45%, with operational cost and the self-sufficiency at 78 $/MWh and 85%, respectively. At a maximum electric demand of 50 MW, the optimal load sharing rate of renewable energy was 50%, with operational cost and self-sufficiency at 65 $/MWh, and 58%, respectively. In both scenarios, surplus electricity from renewable energy was curtailed instead of being used to charge the compressed air energy storage, as charging would require purchasing extra electricity from an external grid, leading to higher operational cost and lower self-sufficiency. This study demonstrated that optimal energy management for renewable energy-based distributed generation can be achieved through the artificial intelligence optimization technique. This technique offers significant customization based on the optimization goals and can be expanded as new energy sources are incorporated, reflecting its high adaptability and flexibility.

Presenting Author: Tong Seop Kim Inha University

Presenting Author Biography: Prof. Kim received a PhD degree in Mechanical Engineering from Seoul National University, Korea in 1995. He had research experiences at various institutions, such as Korea institute of Science and Technology, Institute of Steam and Gas Turbines at Aachen University of Technology, Germany, and Turbo and Power Machinery Research Center at Seoul National University, Korea. He joined the Dept. of Mechanical Engineering at Inha University in 2000. He also has visiting research experiences at Solar Turbines Inc. and UC Irvine, both in California, USA. His research interests include performance design and analysis of advanced energy systems including combined power generation cycles and aero-thermodynamic design and analysis of thermal power systems such as gas turbines and their components. He has authored 194 peer reviewed journal papers. He has served as an editorial memeber of several journals and now is serving as the editor of Journal of Mechanical Science and Technology in the subject area of thermal and power engineering. He also has served as the chair of the Thermal Engineering Division of the Korean Society of Mechanical Engineers (KSME) and the Pressident of the Korean Society for Fluid Machinery.

Authors:

Hyerim Kim Inha University
Tong Seop Kim Inha University