GNN-PPO-Based Multi-Stage Security-Constrained Transmission Expansion Planning for Renewable Energy Integration

The increasing integration of renewable energy resources, rapid load growth, and stringent reliability requirements have significantly increased the complexity of long-term transmission expansion planning. This paper proposes a Graph Neural Network-Assisted Proximal Policy Optimization (GNN-PPO) framework for multi-stage security-constrained transmission expansion planning (SC-TEP) for renewable energy integration. The proposed approach combines graph neural networks to capture network topology and operational characteristics with proximal policy optimization to determine sequential transmission investment decisions over a long-term planning horizon. The planning framework incorporates AC power flow analysis, voltage security constraints, thermal loading limits, renewable energy integration scenarios, and N−1 contingency assessment within a unified optimization environment. The methodology is validated on the South Sulawesi, West Sulawesi, and Southeast Sulawesi (SSWSSS) interconnected power system over the 2025–2034 planning horizon using projected load growth and planned generation expansion scenarios. Simulation results identify an optimal expansion strategy consisting of five transmission reinforcement projects with a total investment cost of 113 MUSD. The selected projects maintain bus voltages within the acceptable range of 0.95–1.05 pu throughout the planning horizon, with a minimum voltage of 0.9611 pu in 2034 and no voltage violations under normal operating conditions. The proposed framework also facilitates renewable energy integration while maintaining acceptable voltage performance. Overall, the results demonstrate that the integration of graph-based feature extraction and reinforcement learning provides an effective and scalable tool for security-constrained transmission expansion planning in large-scale renewable-oriented power systems.