Fault Analysis, Detection and Estimation for a Microgrid via H2/H∞LPV Observers

This works addresses the problem of Fault Detection and Diagnosis and provides a solution based on the use of multiple Linear Parameter Varying (LPV) extended-state observers coupled with simple search algorithm. This methodology is applied to a Grid-Connected Hybrid Power Plant, with different renewable sources, such as photovoltaic panels, wind power generation and the use of biomass. This plant might present different possible faults, that can lead it not to comply to its operational constraints, such as communication problems, valve malfunctions, vapor leakages and others. All these possible faults are carefully categorized based on empirical information from real plants in Brazil. The Fault Detection and Diagnosis (FDD) system designed aims to estimate and categorize these faults and, to do so, the proposed LPV observers are derived from LMI computation of the mixed H2/H∞ norm minimization, in such a way to reduce the effect of noise and external disturbances upon the fault estimation. Through high-fidelity simulations, the benefits of the presented method is discussed.