Modeling dynamics of photovoltaic inverter with LCL-type grid filter

Photovoltaic generator is a unique power source with both constant-current and constant-voltage-like characteristics depending on the operating point. The operating point affects system’s dynamic response, because the generator has a varying dynamic resistance, which causes design constraints, e.g., to the control system design. Therefore, it is necessary to include the source-effect to the full-order dynamic model to predict inverter behavior accurately in its application area. A full-order small-signal model of the three-phase VSI-based photovoltaic inverter with an LCL-type output filter was derived in this thesis, which does not neglect cross-coupling effects or parasitic elements such as ohmic losses in inductors, capacitors and switches. Therefore, the model is also suitable for analysis in cases where strong cross-coupling between the d and q-channels is expected, e.g., when output filter components have large values at high power levels. The model was derived in the synchronous reference frame (i.e., in dq-domain), where the steady-state operating point required for linearization can be solved. Additionally, the effect of non-ideal source and load impedances were included in the model as they have a significant effect on inverter dynamics. The model is shown to give accurate predictions on the control-related transfer functions, which are essential in deterministic control design. Moreover, the closed-loop model allows the full-order output and input impedances to be accurately predicted which is important when analyzing the impedance-based stability of grid-connected PV inverter. The small-signal model has been verified by extracting frequency responses from a scaled-down prototype. The model uses a cascaded control scheme to regulate DC-link voltage and output currents as well as phase-locked-loop as the grid synchronization method. Furthermore, more sophisticated control systems, e.g., feed-forward, space-vector modulation with DC-link voltage sensing etc. can be included in the model. The measured transfer functions were found out to correlate very closely with the predictions