Subcontractor: Dr. Alex Cronin, University of Arizona

Subcontractor: Dr. Alex Cronin, University of Arizona

NREL Contract 99043, “Study Degradation Rates of Photovoltaic (PV) Modules Deployed in Arizona” NREL Technical Monitor: Dr. Sarah Kurtz

Description:Ninth Monthly Report

Authors: Steve Pulver and Alex Cronin

Date: February 8, 2009

This report consists of five parts and two separate documents:

“shade.pdf” and

“PV System performance comparisons.pdf”.

1) We present a new manuscript on the effect of partial shade. We propose to condense this 6-pg draft into a second PVSC submission from Cronin’s group. We can discuss collaborating on this. The manuscript is presented as a separate document, “shade.pdf”.

2) A review of PV system performance at the TEP yard is in a 38-page seminar presentation

“PV System performance comparisons.pdf”. Cronin gave this seminar on Feb 2, 2010 at the University of Arizona Physics Dept. We propose to develop this into a longer manuscript.

3) Steve Pulver has done new statistical analysis to find the probability of various differences between the absolute and relative degradation rates. Steve is writing a manuscript section on this. We propose to include this section in the full paper submission to PVSC. A preview of his result is below.

4) Potential causes for degradation are identified by examining the TEP hardware. The CIGS GSE modules have two spots that appear burnt, and some regions where the electrodes have delaminated. The PB Solarex MST50 thin film silicon modules exhibit electrochemical weathering. Photos of these are in ““PV System performance comparisons.pdf”.

5) For completeness, our PVSC abstract on degradation is reproduced below.

Figure 3.1 P(shift | data).

Figure 3.1 is a result from Steve Pulver’s new statistical analysis of final yields from 20 systems at TEP. It is consistent with the shift of 1.6% presented in our PVSC abstract draft. The result shown in Figure 3.1 is derived with joint probability function analysis. Steve is writing a manuscript section that will describe how he has computed the probability of shift given the data, P(shift | data).

Our PVSC abstract and 3-page paper on degradation are reproduced below. Before publication of a final IEEE proceedings, we anticipate incorporating Steve’s new section and the results shown in Figure 3.1.

Abstract:

A method to report PV system degradation rates without using irradiance isdemonstrated. It is shown that, given 20 PV systems in a data set, it ispossible to find relative and also absolute degradation rates from kWh(AC)measurements alone. This method was developed and tested by comparing thedaily output from 20 PV systems in the Tucson Electric Power solar test yardover a five year period. This method of analysis is significant becausedistributed PV systems are often deployed with electric energy meters butwithout sensors for irradiance. The approach described here can be appliedto any set of ten or more PV systems in a region in order to report absolutedegradation rates of each individual PV system.

The manuscript below has not changed since the last time we circulated it by email.

Measuring Degradation Rates without Irradiance Data

Steve Pulver1, Daniel Cormode1, Alex Cronin1, Dirk Jordan2, Sarah Kurtz2, Ryan Smith2

1: University of Arizona, Tucson, AZ 2: NREL, BoulderColorado

We demonstrate a novel method to measure PV system degradation rates without using sensors for irradiance. By comparing the output of 20 PV systems over five years, we deduced rates of change for each system’s performance ratio [1] with very good precision and competitive accuracy. This is significant because distributed PV systems are often deployed with electric energy meters but without sensors for irradiance. With many such systems in a region, our method can provide absolute degradation rates.

Most studies of degradation use irradiance data because annual insolation fluctuates by 15% (max to min over 20 years in Tucson), yet PV system degradation rates of interest are often less than 0.5% per year. Furthermore, when reporting such small rates of change, the instability or calibration drift of an irradiance sensor often limits the accuracy with which PV systems can be monitored. Methods presented here address these issues.

The fact that we can report relative degradation rates with high precision is no surprise, especially for PV systems located near the same address, facing the same direction, and operating simultaneously for several years. This amounts to using PV system yields as a proxy for irradiance. A bigger challenge is to find absolute degradation rates. We explored ways to do this and then used irradiance data to verify these methods.

One simple method assumes that the PV system exhibiting the smallest loss in annual yield is stable, i.e. not degrading at all. However, we find that more accuracy is obtained by allowing for erroneously high (sometimes positive) rates of change that result from noise in the data. We do this by shifting the entire set of results until the rates of change for the “best” two PV systems are one standard deviation above zero. This is based on the notion from statistics that one sixth of a (large and normally distributed) set of measurements are expected to be too high by at least one standard deviation. Using this method we found that relative degradation rates had to be shifted down by 1.6 % /yr to report absolute degradation rates without irradiance data. The additional uncertainty caused by shifting the data this way can be estimated from the error bars on the data, and is the subject of continuing research.

We studied degradation rates for 22 grid-tied PV systems based on data provided by Tucson Electric Power. All 22 PV systems are located in Tucson, AZ, and most have been monitored with a utility kWh meter since 2003. Ignoring data before or after system hardware changes still leaves data spanning at least three years for 19 systems. The PV module type and nameplate power rating are listed in Table I for each system. All systems use maximum power point tracking inverters. Our analysis focused on daily system yields calculated by dividing measured daily energy output [kWh] by the nameplate power [kW] of the modules [1]. System yield data is shown in Figure 1. We used global horizontal irradiance data from the Tucson AZMET station [2] to independently find absolute degradation rates for each system, and compared these to degradation rates found without using irradiance data, as shown in Figure 2.

Because of several potentially confounding problems such as different temperature coefficients of efficiency, different amounts of mutual shading, different start and stop dates of operation, fluctuations in irradiance from year to year, and the fact that horizontal global irradiance was measured 8 miles away from the PV systems, we assessed the uncertainty for each degradation rate by re-filtering the data in several ways. For example, we filtered the data for sunny days or partly cloudy days, for different times of day or by different ranges of irradiance, by different ranges of AC power, or by different start and stop dates for the analysis. We also used fitting procedures based on various annual periodic functions multiplied by linear decay rates as shown in Figure 1.

a) b)

Figure 1. (a) Daily PV output from system 9 normalized to its rated power (red circles) and a model that includes solar position, temperature fluctuations, and a decay rate of 0.5 %/yr. Plot (b) shows the same data normalized to the average of 18 PV systems. The remaining undulations are due to relative performance differences. Compared to the average of 18 PV systems, system 9 is improving by 1.6% per year.

Figure 2. Degradation rates for 20 PV systems reported with three methods. Rates determined without irradiance data (triangles) have a small uncertainty, and are consistent with methods that utilize irradiance data (diamonds, circles). In addition to horizontal global irradiance data, utilizing a solar position algorithm (SPA) to fit daily yields as shown in Figure 1a reduces uncertainty somewhat (diamonds as compared to circles).
Table 1. PV module types and degradation rates of each system. Precision in degradation is listed for methods without irradiance (WOI) and with irradiance (WI) data. The duration of data we analyzed from each system is listed under (yrs).
Material / Sys.
# / Make (Model) / KW / yrs / Change
%/yr WOI / Precision
%/yr
WOI WI
a-Si / 1 / Solarex (MST-43MV) / 3.00 / 3 / + 0.1 / 1.0 0.6
CIS/CIGS / 2 / Global Solar (GG-112,13309) / 1.44 / 5 / - 2.9 / 0.2 0.6
3 / Shell Solar (ST40) / 1.52 / 5 / - 2.7 / 0.2 0.5
HIT (Si) / 4 / Sanyo (HIP-G751BA2) / 1.34 / 5 / - 1.1 / 0.1 0.2
5 / Sanyo (HIP-J54 BA2) / 1.44 / 5 / - 0.4 / 0.2 0.5
MJ-Si / 6 / BP Solar (MST50MVHS) / 1.50 / 5 / - 4.5 / 0.3 0.3
8 / BP Solar (MST50 MVHS) / 1.50 / 3 / - 1.3 / 0.8 0.6
9 / UniSolar (US-64) / 1.54 / 3 / + 0.1 / 1.1 0.7
px-Si / 10 / BP Solar (BP 3150U) / 1.50 / 4 / - 0.4 / 0.2 0.8
11 / BP Solar (SX140S) / 1.40 / 3 / + 0.5 / 0.5 1.6
12 / Kyocera (KC150G-A) / 1.35 / 3 / + 0.7 / 0.7 1.6
13 / Schott (ASE-300-DGF/50) / 1.26 / 2 / + 0.3 / 2.5 4.5
14 / Schott (ASE-300-DGF/50) / 1.20 / 3 / + 0.3 / 0.2 0.8
15 / Schott (ASE-300-DGF/50) / 22.7 / 4 / - 1.3 / 0.7 0.6
16 / Schott (ASE-300-DGF/50) / 22.7 / 5 / - 2.1 / 0.4 0.6
x-Si / 18 / AstroPower (API-165-MCB) / 1.48 / 3 / - 1.7 / 1.3 2.5
unknown / 19 / Unknown; 10 mi from yard / 21.6 / 4 / - 0.4 / 0.7 0.4
20 / Unknown; 6.5 mi from yard / 108 / 3 / - 4.0 / 0.6 1.2
21 / Unknown; 6.5 mi from yard / 108 / 4 / - 3.1 / 0.5 0.5
22 / Unknown; 2.0 mi from yard / 1.20 / 5 / - 0.7 / 0.2 0.4

For systems with 5 years of data, degradation rates were determined without irradiance data (WOI) with a precision on average of 0.23 %/yr, and rates determined with irradiance data (WI) were found with a precision on average of 0.36 %/yr. For systems with 4 years of data, WOI precision was on average 0.53 %/yr and WI precision was on average 0.55 %/yr. For systems with only 3 years of data, degradation rates WOI were determined with a precision on average of 0.81 %/yr and rates WI were determined with average precision of 1.20 %/yr.

In conclusion, we used several analysis techniques to study how PV systems change over time. Comparing results from the different techniques verifies that one can determine absolute degradation rates without using irradiance data. The increased precision of the method that does not use irradiance data may have widespread application for data sets obtained from distributed PV systems without high quality irradiance data.

[1] B. Marionet.al., “Performance Parameters for Grid-Connected PV Systems,” Proceedings of the 31st IEEE Photovoltaic Specialists Conference, 2005.

[2]