| dc.description.abstract | This thesis introduces a new method to estimate the long-term performance of
direct-coupled and maximum power-tracked photovoltaic (PV) systems without battery
storage. A number of models exist for estimating maximum power-tracked system
performance, but the maximum power-tracking feature is included here as a benchmark
for comparison to direct-coupled system performance, the true object of this research.
While the output from either type of PV system is dependent on weather and PV array
characteristics, the output from direct-coupled PV systems is dependent on the applied
load as well. As a result, estimating the performance of direct-coupled systems is more
complex than for maximum power-tracked systems.
The method developed here is computationally simple. A reduced set of hourly
weather data is generated from widely available long-term monthly-average global solar
and ambient temperature data. Correlations are used to estimate hourly weather variations
within a day and daily variations within a month. A small number of "typical day" groups
are used to approximate the long-term distribution of daily weather in each month. The
"typical days" within each group are assumed to be identical which reduces the number
of computations yet, subject to the accuracy of the correlations, retains the accuracy of
long-term simulations.
To assess the validity of the new method, a program titled DCPVSIMP (for
Direct-Coupled PV model, simplified version) was written. Five potential I-V curve sub-
models for this program were evaluated; the one which was selected corresponded well
with a large sampling of experimental I-V curve data from outside sources, and the
information required to use it is commonly available from PV manufacturers. A detailed
version of this program, identical in all respects except that it uses hourly TMY data,
titled DCPVDET, was also written to provide a basis for evaluating the weather
generation component.
Both the DCPVSIMP and DCPVDET versions were compared to two established
models for maximum power-tracking systems, PVFORM and PV f-Chart. Monthly and
annual estimates were within :t 1 % relative to PVFORM and + 5 to 6% relative to PV f-
Chart. PV f-Chart includes a variable correction factor which decreases estimates of the
absorbed radiation at off-normal incidence angles. For performance estimates at moderate
northern latitudes the overall effect of this term is about a 5% decrease in annual output
The difference between the DCPVDET model and a "5 typical day" version of the
DCPVSIMP model was found to be less than 1 % for maximum power-tracking systems.
Monthly and annual performance for over 800 cases using the DCPVDET model,
based on three locations and a large variety of direct-coupled resistive, fixed voltage, and
DC motor loads, was compared to 3, 5, 10, and 20 "typical day" versions of the
DCPVSIMP model. The overall annual % root mean square (RMS) difference between
the two models ranged from 3.8% for the "3 typical day" version to 3.2% for the "20
typical day" version, with most of the reduction in the % RMS difference occurring
between the 3 and 5 "typical day" versions (3.4% RMS for the 5 day version). For all
versions the % mean bias difference (MBD) was less than 1%. The worst monthly results
ranged between 5 to 6% RMS among the four versions tested, with a % MBD of less than
1%.
The body of the report includes a derivation of the new direct-coupled
performance estimating method and a description of the DCPVSIMP and DCPVDET
models, statistical evaluations of the models and their components, and a set of graphs
illustrating typical applications of the DCPVSIMP model for direct-coupled system
design. | en_US |
