Advanced three-stage photovoltaic system phasor model for grid integration dynamic studies

Abstract

As the penetration of photovoltaic (PV) generation increases at different system voltage levels, their impact on the power grid operation and control is more important to analyze. Hence, suitable PV models must be developed with which it is possible to carry out dynamic analysis for fair-sized electrical networks. This is a demanding task, computationally-wise, for detailed switching-based PV models using small numerical integration steps required by the associated power electronic devices. In this sense, an advanced three-stage PV system phasor model is proposed in this paper whose hallmark resides in the proper characterization of the PV arrays and DC/DC converter through a nonlinear potential regression technique in combination with the generatrix method. As a result, the PV operating surfaces, as function of irradiance and temperature, can be determined and used for the calculation of virtual resistances relating to the PV optimal operation and internal power losses. The virtual resistances are seamlessly combined with a developed AC/DC converter phasor model for efficient dynamic simulations of electrical networks.The new PV model has been thoroughly validated using Simscape Electrical of Simulink where an electrical network with one PV system is dynamically simulated. The outcomes of the developed model are compared to those obtained from switching-based PV models based on different maximum power tracking strategies, i.e., the incremental conductance (IC) and perturbation and observation (P&O) algorithms. The impact of the PV system on the power grid operation and the accuracy of results are assessed, demonstrating that the proposed model favorably agrees with the benchmark switching-based PV models as errors inferior to 2 % are obtained. Furthermore, it is confirmed that the new PV system model enables faster dynamic simulations than what it is possible to achieve with switching-based models since it is 28.82 times faster than the PV model based on the IC algorithm and 31.80 times faster than that using the P&O technique.