Neuro-Fuzzy Adaptive Model Predictive Control for Enhanced Voltage Stability in Transmission Systems

Abstract Contemporary high-voltage (HV) transmission networks are increasingly strained by rapid load growth and the stochastic integration of renewable energy resources, forcing grids to operate perilously close to their critical stability margins. Voltage collapse—the progressive, irreversible decline of bus voltages culminating in widespread blackouts—represents the most severe consequence of this operational stress. While Model Predictive Control (MPC) offers systematic, constraint-aware trajectory optimisation for voltage regulation, conventional implementations rely on static weighting matrices that become suboptimal during severe disturbances. Here we present a novel Adaptive Neuro-Fuzzy Inference System (ANFIS)-based MPC strategy that simultaneously co-adapts both the state penalty matrix Q and the control effort penalty matrix R in real time—a dual-matrix adaptation capability not simultaneously provided by prior ANFIS-MPC approaches. A Sugeno-type ANFIS trained on 12 000 simulation samples (covering N-1, N-2, and renewable fluctuation scenarios) uses real-time voltage error and its rate of change to drive the adaptation. Validated on a high-fidelity model of the Ethiopian 400 kV and 230 kV transmission network under critical N-2 contingency conditions, the ANFIS-MPC achieves a 68.7 % improvement in voltage restoration speed (1.5 s versus 4.8 s for conventional MPC) and elevates the voltage nadir from 0.81 p.u. to 0.91 p.u., keeping the system above the critical V-Q nose-point. These findings establish that AI-driven dual-matrix adaptation is an essential advancement for securing long-distance transmission corridors in developing power infrastructures.