Date of Award

Spring 1-1-2014

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

First Advisor

Michael Brandemuehl

Second Advisor

Robert Erickson

Third Advisor

Gregor Henze

Fourth Advisor

Moncef Krarti

Fifth Advisor

Christopher Cameron

Abstract

This research develops a comprehensive methodology and model for accurate prediction of power losses caused by nonuniform electrical characteristics and operating conditions in grid-tied photovoltaic systems, as well as the potential for increased energy capture in systems which employ sub-array power optimizers (microconverters or microinverters). Investigation of these topics provides a framework for more accurate loss modeling and determination of power optimizers' value in a variety of scenarios, enabling future PV research and maximizing the value of PV systems in the built environment.

A custom multitracer, which records simultaneous module-level I-V curves, is designed and built to collect data on 27 PV installations in the Southwestern U.S. The resulting measured dataset, including over 500 modules of crystalline silicon and thin film technologies, indicates that commonly-used, single diode PV generator modeling methodologies often incorrectly predict PV performance at low and medium light levels. A new modeling methodology and parameters are proposed, demonstrating an improved way to incorporate low light data to increase prediction accuracy for crystalline silicon and thin film arrays.

A unique, detailed annual simulation environment for PV system modeling is developed, allowing user-input electrical characteristics and operating conditions at the PV cell level. It is designed specifically to model electrical mismatch and partial array shading, and use of power optimizers to mitigate related energy losses. The resulting simulations, combined with the module-level I-V curve dataset, are used to predict annual mismatch losses caused by module-to-module performance variation in each monitored array. The losses, representing energy that may be directly recovered using power optimizers, are moderately low for most of the tested arrays.

Annual simulations of realistic shading scenarios and PV array configurations show percent annual energy losses that are 2-3 times the annual percent incident light lost in partially shaded arrays. Sub-module or even module level simulations predict nearly the same shading losses as cell level simulations, demonstrating opportunities for model simplification. Arrays with power optimizers can recover 30-45% of the energy loss. In the example scenarios, power optimizers are most advantageous at the module or cell levels, adding little benefit at the string or bypass diode sub-module levels.

Share

COinS