The rapid growth of photovoltaic (PV) systems has made them a central component of renewable energy integration. However, PV inverters, which are responsible for converting direct current (DC) into alternating current (AC) for grid compatibility, remain highly vulnerable to faults. Issues such as sensor offsets, partial shading, and grid disturbances can cause performance degradation, instability, and even disconnection from the grid. These challenges pose significant risks to system reliability. Conventional control strategies, including proportional–integral–derivative (PID) controllers, often fail to adapt dynamically to such disturbances. Similarly, standard Model Predictive Control (MPC) approaches with fixed parameters may be ineffective under fault conditions. This has created the need for advanced fault-tolerant control (FTC) methods.
Recent studies have explored residual-based fault detection, where discrepancies between expected and actual outputs are monitored to identify faults, as well as finite-state-machine (FSM)-based frameworks to govern system behavior under different operating conditions. However, integrating these approaches directly into MPC for real-time adaptation has been relatively unexplored, particularly for PV inverters.
This work introduces a novel FSM-guided adaptive MPC framework aimed at detecting faults and adjusting control strategies on the fly. The proposed scheme is designed not only to identify fault conditions but also to reconfigure its optimization priorities to maintain stability and performance despite variations in operating conditions and reference signals. The approach is validated using a detailed discrete-time simulation of a PV inverter, where fault offsets and varying reference signals are introduced to stress the system.
The FSM operates with two modes: normal and faulty. Fault detection relies on monitoring the residual, which measures the deviation between the system’s actual and nominal outputs. When the residual surpasses a threshold, the controller switches to the faulty mode, assigning greater importance to minimizing output errors while allowing for sharper control actions. Once the residual returns below a recovery threshold, the FSM transitions back to the normal mode with smoother control behavior.
Simulation results over 350 time steps show that the proposed adaptive MPC effectively maintains stability and tracking performance under fault conditions. When a disturbance is applied, the FSM correctly identifies the fault and transitions the controller to a more aggressive setting. This reduces tracking error compared to the normal operating mode and ensures the inverter output closely follows the reference despite the disturbance. Once the disturbance subsides, the controller smoothly transitions back to its original configuration. Additional analyses, such as state trajectories, phase plots, and control cost evaluations, confirm both the stability and robustness of the system.
Overall, the FSM-adaptive MPC framework improves fault tolerance in PV inverters by combining residual-based detection with dynamic tuning of control parameters. The method shows significant improvement in tracking accuracy during faults while maintaining overall system stability. Beyond its strong simulation performance, this framework addresses pressing reliability challenges in renewable energy integration. Future work will focus on expanding the FSM to multiple states, refining fault detection thresholds, and conducting hardware-in-the-loop experiments to demonstrate its practicality for real-world grid-connected PV systems.
