A particle swarm optimization-based ensemble metaheuristic for long-term transmission network expansion planning

To significantly enhance the exploration and exploitation capabilities of particle swarm optimization, it is beneficial to hybridize complementary algorithms, integrate effective methods at specific stages, and trigger a ‘bail-in’ mechanism when encountering a local optimum. In this study, a Particle Swarm Optimization-Based Hybrid Framework (PSO-BHF) is proposed. The framework includes a probability calculation method that considers particle improvement after using Covariance Matrix Adaptation Evolution Strategy (CMA-ES) to promote the convergence rate. Additionally, an opposition-based learning method is integrated to extend search regions and maintain particle diversity if the search becomes stagnant, while a Success History Intelligent Optimizer (SHIO) is employed to prevent premature convergence. Finally, the Sequential Quadratic Programming (SQP) method is used to enhance exploitation around the best particle in the later stages of the evolutionary process. The novel framework was evaluated on CEC2017 benchmark functions and compared with 15 state-of-the-art PSO-based variants and 11 non-PSO-based algorithms. Ablation tests illustrate the effectiveness of individual mechanisms and their combinations. The framework is also applied to a real-world long-term Transmission Network Expansion Planning (TNEP) problem, which was released by GECCO 2023 and IEEE CEC 2023 for their joint competition on evolutionary computation in the energy domain. Experimental results demonstrate the impressive performance of the proposed framework compared to PSO-based variants and its effectiveness in the TNEP problem. However, it still needs time to catch up with the top tier of metaheuristics, specifically Differential Evolution (DE)-based algorithms.