DESIGN AND DEVELOPMENT OF NEW VARIANTS OF SALP SWARM ALGORITHM FOR ENERGY RELATED OPTIMIZATION PROBLEMS

Abstract

The focus of this Thesis is to design and develop new variants of a Swarm Intelligence-based algorithm namely, Salp Swarm Algorithm (SSA) for solving some energy-related optimization problems. Two novel variants of the classical SSA are introduced in this Thesis. The first one is a Laplacian crossover operator-based Laplacian Salp Swarm Algorithm (LX-SSA) and the second variant is Quadratic Approximation Salp Swarm Algorithm (QA-SSA), which incorporates a Quadratic Approximation operator. The classical SSA suffers from drawbacks such as premature convergence, slow convergence speed and poor population diversity. Therefore, LX-SSA and QA-SSA have been developed to overcome these limitations and improve the performance of the classical SSA. To evaluate the performance of the proposed algorithms, extensive assessments are conducted on two sets of benchmark problems: 23 standard benchmark functions and IEEE CEC 2017 benchmarks functions and compared with some well-known Swarm Intelligence-based algorithms. The results obtained for these benchmark functions are analyzed numerically, graphically and statistically. The analysis reveals that LX-SSA and QA-SSA outperform the classical SSA and other considered algorithms in terms of convergence and solution accuracy. To comprehend the theoretical aspects of the classical SSA, LX-SSA and QA-SSA, it is essential to examine the structural behaviour of these algorithms. The Signature test is employed to analyze the structural bias of these algorithms. The results indicate that the classical SSA exhibits a strong center bias. On the other hand, QA-SSA is unbiased in early iterations, but a small bias is observed after some iterations and LX-SSA displays a bias in the form of multiple clusters throughout the search region.