AI-Driven Agent-Based Simulation and Optimization for Collective Self-Consumption and Grid Balancing

Capabilities and Expertise: Our group specializes in the development of advanced Agent-Based Simulation (ABS) models and AI-driven forecasting algorithms specifically designed for electricity markets. We possess expertise in modeling strategic bidding behaviors using Deep Reinforcement Learning algorithms, including DDPG, SAC, TD3, PPO, and RDPG, to optimize producer performance and market equilibrium.

Specific Contributions to the Topic:

• AI-Driven Forecasting: We have developed machine learning models (e.g., LightGBM, LSTM optimized with Artificial Rabbit Optimization) for price forecasting, achieving superior accuracy in capturing complex, non-linear market dynamics.

• Collective Self-Consumption Optimization: Our "Agent Bidding Model" utilizes original genetic algorithms and Artificial Bee Colony optimization to model risk-sensitive decision-making.

• Grid Balancing: We will propose original algorithms for Distribution Locational Marginal Price (DLMP) calculation, which are crucial for managing congestion and voltage constraints in distributed energy systems.

• Social & Game-Theoretic Analysis: We conduct human-based experiments and game-theoretic modeling to analyze social interactions, consumer trust, and the impact of diverse energy resource compositions on market performance.

Previous Experience: We successfully led the the TÜBİTAK 3501 Career Development project focusing on AI-assisted learning and optimization for electricity markets. We have another pending proposal for TÜBİTAK 1001 project: "Analysis of DLMP-Based Electricity Markets with Agent-Based Simulation".

Pilots/Prototypes: We have developed a modular, open-source Python-based simulation platform for analyzing various market mechanisms (Uniform Pricing vs. Pay-as-Bid).