Multi-model environmental modelling of energy-exergy efficiency using GUI-based aided design tools integrated with dependency feature analysis

Abstract

This research introduces a comprehensive multi-model framework for assessing energy-exergy (EE) efficiency, utilizing graphical user interface (GUI) based design tools in conjunction with linear-feature optimization techniques. The performance of several machine learning (ML) models, including the Adaptive Neuro-Fuzzy Inference System (ANFIS), Nonlinear Auto Regressive with exogenous inputs (NARX), Backpropagation Feedforward Neural Network (BFNN), Extreme Learning Machine Network (ELMN), and Feedforward Neural Network (FFNN) was evaluated, utilizing key statistical metrics throughout both training and testing phases. The results reveal that the ANFIS-M2, NARX-M3, and FFNN-M1 models achieved exemplary training accuracy, attaining an R² value of 1.0, underscoring their capacity to capture intricate nonlinear relationships effectively. Notably, during the testing phase, the FFNN-M1 model sustained its perfect R², while both ANFIS-M2 and NARX-M3 demonstrated robust predictive capabilities with R² values of 0.95. The BFNN and ELMN models also displayed commendable performance, yielding R² values between 0.75 and 0.97. In contrast, the ANFIS-M3 and BFNN-M3 models exhibited comparatively lower accuracy, recording R² values below 0.7 during testing. These findings underscore the efficacy of integrating GUI-based tools with linear-feature optimization for predicting exFergy efficiency. The study highlights the promising potential of FFNN and ANFIS models in enhancing the optimization of energy systems, thereby facilitating the development of more efficient computational frameworks for energy modeling applications.