High Performance Fault-Tolerant Control for DC-DC Buck-Boost Converters: An Unsupervised Learning-based Approach

This paper introduces an advanced learning-based fault-tolerant control strategy for DC-DC buck-boost converters, demonstrating significant improvements over classic approaches under challenging operating conditions. The proposed Learning-based Fault-Tolerant Control PI (LFTC-PI) integrates Fuzzy C-Means (FCM) clustering with adaptive parameter adjustment to enable fault detection and mitigation dynamically. In contrast to fixed-parameter robust controllers, which often depend on conservative designs, the LFTC-PI adaptively adjusts its control parameters based on operational data, allowing for prompt adjustments in response to component degradation and external disturbances. Comprehensive simulation studies demonstrate substantial performance improvements: reduced settling time, lower overshoot, and diminished steady-state error compared to classic fault-tolerant control strategies. Moreover, the proposed controller ensures stable operation even under simultaneous 40% inductance and capacitance degradation, with concurrent voltage and load disturbances. An analysis of computational complexity demonstrates the algorithm's suitability for implementation on standard embedded systems. This approach represents a transition from reactive to proactive fault mitigation in power electronics, with potential applications in electric vehicle charging infrastructure and renewable energy systems, where reliability is critical.