Weather-Corrected_Performance_Ratio NREL(2)

Given the goal of removing the seasonal variability of the PR metric without changing the PR value that is stated in the contract, we assert that it is possible to define a site-dependent average cell temperature to which the PR can be corrected. We will call this a “weather-corrected” PR because it corrects for most of the weather-related effects. Although it would be useful to correct the PR for every aspect of the weather, we propose here to correct only for weather variations that affect the module temperature (ambient temperature, wind, and irradiance). We do not attempt to correct for snow coverage, soiling, or irradiance variations that affect the PV efficiency (with the assumption that a high-quality installation does not suffer greatly from shunt and series resistance effects). While system tests could be corrected for snow or soiling, it is unlikely that a contractor would choose to run the test while the system is covered with snow or

1 Introduction: The Performance Ratio

is heavily soiled. By using a semiconductor reference-cell sensor in the plane-of-the-array, the seasonal spectral biases for the irradiance measurements are also minimized.

The purpose of this report is to 1) present the importance of using weather-corrected PR instead of uncorrected PR as a binding performance metric, 2) propose a method for applying the weather correction so as to remove the seasonal bias associated with variations in temperature, and 3) define a sample test protocol that can be referenced for contractual content. The report starts by quantifying the variability in PR that results from variation in temperature and by showing how this variability causes risk to all parties of the test, but can be removed by defining an annualized average temperature. Next, the report provides more detail about how the weather correction is constructed so as to remove the seasonal variability without changing the annualized PR and gives examples from plant acceptance testing around the world, showing how using the weather-corrected PR reduces the variability in the reported PR value. Finally, the report concludes by giving a step-by-step test protocol.

Variability of PR with Weather

Figure 1. Performance ratio (PR) calculated from measured data over 15-minute periods from a 24-megawatt (MW) facility.

The annual PR is not a stable function of the project weather file. The project weather file is the annual weather file of record used to determine energy generation expectations and set performance guarantees. The source may be a typical meteorological year (TMY) file or a combination of any other sources. It is recommended that all parties to the project agree to the data stored in the project weather file.

A project’s PR will change if a different weather file is used in the annual simulation — even though the plant design is unchanged. Table 1 shows the effects on PR as a result of changing either ambient temperature or wind.

Table 1. Effects of Annual Ambient Temperature or Surface Wind on PR Weather File

Baseline weather file

3° C Higher Annual Temperature

3 m/s Higher Annual Wind Speed PR 84.9% 84.0% 86.6% Difference -0.9% 1.7%

The lack of constancy of the PR as the weather file is varied is readily apparent. It is recommended that practitioners repeat this exercise of changing weather files on their own projects. The important recommendation is to have all parties to a project agree on a weather file (based on either historical data or data measured specifically for the project) before establishing PR guarantees. of -0.38%/°C, providing an example of how the PR may be expected to vary with variable weather.

Theoretical Approach As described in the introduction, the PR varies with changes in meteorological conditions (and thus throughout the year). Yet, the PR is an important metric to the industry. The goal of this report is to mitigate risk caused by the inexact nature of the PR by defining a modified metric: the weather-corrected PR.

To quantify this variability and show how it can be reduced or removed, we calculate PR using two different methods: the method outlined in IEC 61724, and a new method that corrects PR for site-dependent meteorological conditions. Simulations are presented for a facility located in the southwest United States. Equation (1) shows how the PR is traditionally calculated. Equation (2) shows the modifications to become a weather-corrected PR. The difference between the two is that the weather-corrected PR contains a term to translate modeled power to the average operating cell temperature. The operating cell temperature accounts for the effects of both the ambient temperature and wind (as well as the heating from the sunshine). The use of a matched reference cell to measure irradiance avoids the need to also correct for spectral variations. There is no attempt made here to correct for other weather effects, such as snow losses, soiling losses, or the effects of variable irradiance on efficiency. While corrections for these additional weather effects could produce more consistent results, Equation (2) provides a simple way to account for the primary effects.

(1)

(2) ????= ∑???????????????????? ????????????= ∑???????????? ????????_??????_?????? ? ??????????_?? ?????????

Where:

The summations are over a defined period of time (days, weeks, months, years)

PR = performance ratio (unitless)

PRcorr = corrected performance ratio (unitless)

ENAC = measured AC electrical generation (kW)

PSTC = summation of installed modules’ power rating from flash test data (kW)

GPOA = measured plane of array (POA) irradiance (kW/m2)

i = a given point in time

GSTC = irradiance at standard test conditions (STC) (1,000 W/m2)

Tcell = cell temperature computed from measured meteorological data (°C)

Tcell_typ_avg = average cell temperature computed from one year of weather data using the project weather file (°C)

δ = temperature coefficient for power (%/°C, negative in sign) that corresponds to the installed modules.

The motivation for amending the PR metric into a weather-corrected number is evident in Figure 2. This shows the uncorrected and corrected PR calculated from a simulation. (A simulation is used because it represents an ideal system where all aspects that contribute to electrical generation are controlled. This is required to show that a performance metric is not a consistent value. The reader is encouraged to repeat this analysis.)

In this plot, the PR is calculated for each month in the year’s simulation. The blue markers are the PR values calculated using Equation (1), and the red markers show the corrected PR for the same time computed using Equation (2). Note that the uncorrected PR changes by 10% over the year. This bias will result in false high values during the winter months (causing risk for the PV customer because a poor-performing plant might falsely pass the test during this time) and false low values during the summer months (causing risk for the PV installer). It is this instability in the metric that is the motivation for a corrected PR. Without the weather correction, PR is not consistent throughout the year.

Some have attempted to address this error by producing a table that states PR for each month. However, this is still a biased metric if the month is unseasonably warm or cool — resulting in a possible falsely high or low result. All parties to an agreement will carry weather risk during testing periods that may result in a false pass or fail if uncorrected measurements are used.

Figure 2. Corrected and uncorrected PR from simulated results.

It is recommended to use the weather-corrected calculation and the mutually agreed-upon project weather file if the PR is to be used for a contractual metric. This avoids the risk that the weather would produce erroneously high or low readings. As can be observed, the corrected data are more consistent through the year — a better choice for demonstrating contract guarantees.