Pareto-based design of thermophotovoltaic micro-combustors via a novel framework combining IGWO-tuned ANN, multi-objective multi-verse optimization, and ARAS-based decision making.

The efficient design of micro planar combustors (MPCs) is vital for advancing thermophotovoltaic (TPV) energy systems. However, current approaches often lack integration between predictive modeling, optimization, and decision-making processes. This study addresses that gap by introducing a novel, fully integrated framework that combines machine learning, metaheuristics algorithm, and structured multi-criteria decision-making. In the proposed framework, a multilayer perceptron neural network (MLPNN) tuned using an Improved grey wolf optimizer (IGWO) and particle swarm optimization (PSO) was developed to predict key performance indicators of the TPV system, including pressure drop (ΔP), output power (OP), and system efficiency (SE). The IGWO-MLPNN outperformed PSO-MLPNN for ΔP and OP with MAPE as low as 1.126% and 0.138%, respectively, while PSO-based model was superior for SE prediction (MAPE = 0.29%). Optimization was performed using multi-objective multi-verse optimizer (MOMVO) and NSGA-II, generating diverse Pareto-based solutions. Key Pareto results showed that optimal designs feature inlet velocities of 6.0-6.5 m/s, equivalence ratio ≈ 1.0, long tubes, wide spacing, and expanded tube diameters, balancing high power, low pressure drop, and efficiency. ARAS ranked Pareto solutions by prioritizing ΔP, OP, and SE, generating ten scenarios tailored to diverse TPV application needs. The framework provides a robust and scalable toolset for MPC optimization, enabling tailored, high-performance TPV designs adaptable to diverse real-world energy applications.