Optimization of Non-Linear Problems Using Salp Swarm Algorithm and Solving the Energy Efficiency Problem of Buildings with Salp Swarm Algorithm-based Multi-Layer Perceptron Algorithm

Abstract

The aim of this paper is to evaluate the optimization capabilities of the salp swarm algorithm (SSA), a metaheuristic algorithm capable of addressing contemporary global challenges. The paper focuses on assessing SSA as an optimizer and observing its impact as a predictor in an example energy problem to gauge its predictive power. Salp swarm algorithm (SSA) distinguishes itself with its optimization capabilities, providing effective solutions to optimization problems. The quality, competitiveness, and efficiency of the algorithm were initially assessed using the CEC 2019 and CEC 2020 function sets. The results demonstrated that SSA is a competitive, effective, and up-to-date algorithm. This competitive nature suggests that SSA can be effectively employed across a wide range of problems. Therefore, the paper aims to evaluate its success in providing solutions to an energy prediction problem. In addressing the challenge of effective energy utilization, the accurate prediction of heat loading (HL) and cool loading (CL) factors, critical in building design, contributes significantly to the solution. In solving this problem, machine learning algorithms, specifically the multi-layer perceptron (MLP) as an artificial neural network architecture, were chosen. SSA was approached in a supervised manner, and a comparison with alternative metaheuristic algorithms was conducted. The obtained results indicate that the SSA-based MLP architecture (SSA-MLP) exhibits effective predictive capabilities in energy problems. By combining the optimization power of SSA and the learning capabilities of MLP, a robust solution with a competitive advantage in energy efficiency is presented.