Sensorless predictive control of voltage source inverters for renewable energies integration under unbalanced and distorted grid conditions

Three-phase grid-connected voltage-source inverters (VSIs) are widely used for renewable energies integration. Cost reduction and suitable operation under nonideal grid conditions are their important technical challenges. Accordingly, this paper proposes an efficient adaptive neural filter-based virtual flux (ANF-VF) estimator for sensorless control of a grid-connected VSI under unbalanced and distorted grid conditions. To perform sensorless predictive direct power control (PDPC), the grid voltage sensors are substituted by the ANF-VF estimator. This estimator includes an emulated ideal integrator in series with two simple ANFs. Lyapunov's theory-based convergence analysis is conducted for its optimal tuning. This is resulted in an accurate extraction of VF fundamental components. For more effectiveness under unbalanced grid conditions, an extension of original instantaneous power theory is introduced in the proposed VF-based PDPC (VF-PDPC). Effectiveness of the VF-PDPC is verified through simulation and experimental tests. A direct and smooth startup without initialization is accomplished under unbalanced grid conditions. Superiority of the VF-PDPC compared to the conventional PDPC is demonstrated. The proposal presents sinusoidal grid currents with low total harmonic distortion under unbalanced and distorted grid conditions. Moreover, the ANF-VF estimator illustrates best performances compared to the second-order generalized integrator-based VF estimator that uses measured grid voltages under nonideal grid conditions.