Measurement-Based Nonlinear SPICE-Compatible Photovoltaic Models for Simulating the Effects of Surges and Electromagnetic Interference within Installations

Abstract

An emerging area of interest within photovoltaic (PV) centred research is the simulation of the propagation of electromagnetic interference (EMI) and surges within PV installations. An overarching constraint in all simulation-based research is the accuracy of the models employed. In general, for PV-focussed simulations, nonlinear models are utilised for direct current (DC) analyses, whilst linearised models are employed for analyses involving surges or conducted electromagnetic interference. For large-signal electromagnetic interference and surges, the following two problems arise: (1) the aforementioned linearisation is only valid for the small-signal case, and (2) as they are constructed using only DC measurements, traditional large-signal PV models are generally only valid for DC conditions. Therefore, neither of these approaches can properly represent real-world PV behaviour under dynamic conditions. To overcome this limitation, this article proposes a more suitable model, compatible with Simulation Program with Integrated Circuit Emphasis (SPICE), and which results from the combination of two sub-models: one for large-signal DC cases, and one for small-signal alternating current (AC) cases. Consequently, the combined model enables improved modelling of the effects of large-signal transient perturbations to be performed, as well as cases involving small-signal AC and large-signal DC perturbations. The model parameters are extracted using data from two different classes of measurement setups: the first utilised a vector network analyser (VNA) to produce small-signal AC impedance results (covering a frequency range between 1Hz and 50MHz), and the second produces DC current-voltage curves. Both classes of measurement setup consider the device under test at multiple operating points. Key results include: (1) an improved SPICE-compatible PV model which caters for large-signal transient simulations, as well as for small-signal AC and large-signal DC cases, (2) improvements in the modelling of reverse bias behaviour when compared to the standard SPICE diode implementation, (3) the correct implementation of a voltage-dependent total capacitance (suitable for large-signal simulations), (4) modelling parameters for both a small (10 W) and a large (310 W) PV module, (5) measurement results which de-embedded the parasitic effects of the test setups employed, and (6) above 1 MHz, the physical layouts of the cells within the PV modules begin to influence the observed impedances.

Modelling PV Module Lightning-Induced Response: The Oversimplified, the Overestimated, and the Overlooked

“…This distributed circuit layout, combined with the increased levels of detail in the new 3D models using a hybrid simulation methodology, produced impedance results which agreed better with the measured results and now included the resonant peaks and troughs at higher frequencies. This confirmed the layout-based postulation from [12], and provided options for greatly improved geometric modelling of PV modules for 3D CEM simulations.…”

“…The third study, [13], investigated the layout-based postulation from [12] by investigating the differential mode impedance observed between the output terminals of two types of wire-based equivalent models. These are referred to as the "simple" and the "complex" literature-based models.…”

“…The fourth study, [14], investigated the levels of coupling between the three internal loops of the same large PV module from [12] and [13]. This coupling was studied using a series of scattering parameter (S-parameter) measurements.…”

“…Explanations of the rest of the bypass diode and PV cell circuit parameters, as well as their values, are detailed in [11] and [12], respectively. For brevity, they are not repeated here.…”

“…The following section discusses how these setups will be examined to ensure that comprehensive insights are gathered. [12] and featured a voltage-dependent capacitance. For this study, the capacitance was set constant, as it was not compatible with the CST nonlinear circuit solver.…”