Feedback chaotic growth optimizer for parameter extraction of a novel direct current arc model

Direct current (DC) arc faults are a leading cause of fire incidents in photovoltaic (PV) systems. Accurate modeling of DC arc faults is essential for understanding the underlying mechanisms of DC arcs and for developing effective detection strategies. In this study, we propose a novel model for DC arcs, referred to as the exponent segmented noise model. This model effectively characterizes arc noise by establishing an exponential relationship between frequency values and spectral energy. To enable precise parameter extraction from the exponent segmented noise model, we introduce a new metaheuristic algorithm called the feedback chaotic growth optimizer (FCGRO). FCGRO improves upon the traditional growth optimizer (GRO) by integrating feedback operators and chaos mechanisms. Firstly, the convergence performance of FCGRO is rigorously evaluated through comparative experiments on three well-established benchmark engineering optimization problems. Subsequently, based on data collected from an established experimental platform, the proposed FCGRO and eight state-of-the-art algorithms are employed to extract parameters of the exponent segmented noise model for DC arc faults. The FCGRO achieves an overall average root mean square error (RMSE) of 0.0418 with a standard deviation of 0.00818, representing reductions of at least 10.43 % and 26.86 %, respectively, compared to the other eight methods. These results indicate that FCGRO delivers more accurate and stable parameter estimations than the competing algorithms. Regarding computational efficiency, FCGRO has an average processing time of 9.969 s, ranking it third among the nine evaluated methods, which confirms its competitiveness in terms of speed. Finally, compared with existing DC arc models, the proposed exponent segmented noise model reduces RMSE by an average of 53.26 %, demonstrating its superior modeling capability.