Real-time electrocution detection system for HVEF based on ANFIS and CPSO_DP co-evolution mechanism

Adaptive neuro-fuzzy inference systems (ANFIS) provide powerful nonlinear modeling capabilities but are highly sensitive to noise and prone to suboptimal convergence when trained with conventional optimization methods. To address these challenges, this study proposes an enhanced ANFIS optimized by a convergence-speed-driven dynamic cooperative particle swarm optimization (ANFIS-CPSO_DP). The novelty of the method lies in dynamically regulating swarm population size according to convergence rate, enabling efficient transitions between global exploration and local exploitation. Theoretical justification is provided by analyzing convergence stability under dynamic population control, while experimental validation is performed on noisy electrocution detection tasks. Comparative results against PSO, CPSO, Differential Evolution, and JAYA optimizers demonstrate that ANFIS-CPSO_DP​ achieves the lowest RMSE, smallest error variance, and highest R² values across multiple noise levels. Furthermore, the error distribution reveals diagnostic features exploitable for fault identification, reinforcing the robustness and practical utility of the approach. This work fills the gap in noise-resilient ANFIS training by coupling adaptive population regulation with cooperative swarm intelligence, offering a reliable solution for real-time fault detection in noisy environments.