Search

Grid‐interfaced solar photovoltaic (GIPV) system requires a robust control algorithm for satisfactory operation under grid voltage disturbances. In this study, an application of Newton's learning rule‐based total least‐square estimation employing single‐layer neural network structure is adopted to interface the photovoltaic unit with the utility grid. In addition, the weight‐updating mechanism is uniquely integrated with threshold neuron for attaining speedy extraction of fundamental load current component. In particular, the proposed control algorithm has the capability to transfer maximum active power to the utility grid/AC load at unity power factor while operating as distribution static compensator to offer various ancillary services including current harmonics attenuation and reactive current suppression. This will lead to increased device utilisation factor of the overall GIPV system during night time (i.e. in the absence of solar irradiance). Moreover, the proposed control scheme is designed and developed without any complex transformations and derivative terms, which results in less computational intensiveness. The GIPV system with three‐phase double‐stage configuration is modelled in MATLAB/Simulink software using sim‐power system tool. Finally, the adaptability of the proposed control algorithm has been verified and confirmed through 500 W laboratory prototype using low‐cost Arm Cortex‐M4 microcontroller under various operating conditions. [ABSTRACT FROM AUTHOR]

This paper makes two contributions. Firstly, it is shown that there may exist oscillations in distribution networks with physically close photovoltaic (PV) units due to their control interactions. The main reason for oscillations is changes in the dynamics of PV units with varying solar irradiance and the decentralized tuning of control parameters for the nominal operating condition. Eigenvalue analysis, nonlinear interaction index, and time-domain simulations are used to assess the degree and impacts of dynamic interactions. Secondly, a decentralized robust control design is proposed to ensure a non-interacting and well-damped response under varying operating conditions for physically close PV units. The control design process uses an estimate of changes in the model due to variations in solar irradiations, changing electric loads, and the dynamics of the physically close PV units. A minimally conservative method is used to capture the estimate as a norm-bounded uncertainty. Time-domain simulations on a test distribution system verify the predictions of the interaction analysis and demonstrate the performance of the proposed robust controller under different system contingencies. [ABSTRACT FROM AUTHOR]