Analysis of hourly metered data collected from 19 residential grid-tied PV systems in California helps to answer questions about actual system power output.
In recent years technology advances, environmental considerations, and economic factors have contributed to an increased interest in grid-tied photovoltaic (PV) systems for homes. Well-funded support and promotional programs in Japan, Germany, and California have created active markets in these locations for this renewable distributed-generation technology.
For homeowners, builders, or designers considering installation or specification of PV technology, access to reliable performance information is critical. Performance information for individual components under nominal rating conditions is readily available. System performance under actual operating conditions is another matter altogether—and this performance information is sparse.
To fill in this information gap, in January 1999 the California Energy Commission (CEC) and Regional Economic Research, Incorporated (RER) jointly developed a project to monitor the in-field performance of PV systems that had been funded in part by the CEC’s Emerging Renewables Buydown program.
RER,the company I work for, monitored the key performance parameters of 19 PV systems for which incentives had been distributed through the CEC’s Buydown program.We collected data on energy production, power output, and net impact on utility system loads from mid- February 2000 through December 2001. Because of monitoring system moves and other factors, varying quantities of data are available for the different sites.
To ensure that we would monitor a diverse range of PV systems,we selected systems based on the following criteria: geographic diversity, system size and configurations, equipment, retail suppliers, and installation vendors.The system sizes ranged from 1 kW to 12 kW (see Table 1).The sites are located all over California, from San Diego County in the south to Willits in the north. Several sites are located in coastal areas, while sites located well inland include Grass Valley and Mariposa in the Sierra Nevada foothills.
What Does That Power Rating Mean?
The power output of PV systems varies depending on irradiance level and module temperature, which makes specification of system AC capacity a complicated endeavor. Manufacturers of PV cells and modules typically rate their products at standard test conditions (STC), comprising 1,000 watts per square meter (W/m2) irradiance and a cell temperature equal to 25ºC. The resulting DC power output ratings are often incorporated into model numbers. Because these ratings constitute the most readily available size information,we used them when listing system sizes in Table 1.However, when these systems are actually operating in the field and receiving irradiance levels of 1,000 W/m2, cell temperatures often exceed 25ºC.The higher cell temperatures can cause observed power outputs to fall short of nominal nameplate ratings.
There are alternative approaches to developing system capacity estimates that are based on weather rather than on cell temperatures, and these estimates can be more representative of actual in-field conditions. One commonly used alternative rating system was developed by the Photovoltaics for Utility Scale Applications (PVUSA) national public-private partnership. The weather that constitutes PVUSA test conditions (PTC) consists of 1,000 W/m2 plane-ofarray irradiance, 20 °C ambient temperature, and wind speed equal to 1 meter per second (see “PV Vocabulary,” p. 29). Since cell temperatures influence power output, it is important to calculate what the cell temperatures would be for each system under these weather conditions. Estimations of what cell temperatures would be under PTC conditions, which will vary from system to system depending on a variety of factors,may be made using experimental or theoretical methods.
To estimate actual system AC capacity at PTC conditions,we collected hourly metered data for each system on plane-of-array solar radiation and module temperature. Specifically, the parameters we monitored included solar radiation (on plane of array);PV module temperature; whole-building electricity consumption (AC kWh) measured near point of interconnection with the utility (main breaker panel); inverter energy output (AC kWh); and PV array output (DC kWh).The monitoring system platform consisted of two data loggers, each with an internal modem. One logger was used to measure solar radiation and module temperature, and the other was used to measure AC and DC power quantities.
Power output at PTC conditions cannot be estimated directly with these data alone, because PTC conditions refer to specific ambient weather conditions. Instead, separate regression models were estimated for PV output versus module temperature and module temperature versus weather conditions.
Many Ways to Size
We used these regression analyses to calculate estimates of system AC power capacities for PTC conditions of 1,000 W/m2 and estimated module temperatures (see Table 2). We found on average that measured AC system capacity at PTC conditions was 38% less than nominal DC system size at STC conditions. The smallest discrepancy was 30%. This difference is attributable to factors such as wiring, module mismatch, and DC-to-AC conversion losses, as well as reduced output at PTC weather conditions compared to STC testing conditions. For each kW of nominal DC module capacity, typical AC system power output for 1,000 W/m2 plane-ofarray irradiance (1-sun conditions) and 68ºF ambient temperature was 620 watts. For 1-sun conditions and 100ºF ambient temperature, the estimate of typical AC system output falls to 575 watts.
There are many ways to rate system output, and it is important that customers and system integrators understand the basis for the sizing information supplied by manufacturers. Clearly, simply referring to a “1 kW PV system” is insufficient; at a minimum, the plane-ofarray solar radiation and ambient or module temperature associated with such a value should be presented alongside the size value.
Energy Production Varies
In addition to power capacities,we looked at several other important measures of energy production for the monitored PV systems (see Table 3). Daily average electricity production and plane-of-array irradiance are calculated directly from the hourly metered data. Average electricity production ranged from 1.2 to 45.4 kWh per day.
The considerable variability observed in these results is largely attributable to effects of system sizes, which vary by a factor of more than ten.However, tracking, orientation, slope, shading, module mismatch, module soiling, battery size, inverter settings, and the effectiveness of an inverter’s maximum power point tracking capability all contribute to the 40-fold range in electricity production that we observed. Mechanical damage and different levels of accuracy in nominal module sizes also affect electricity production.Tolerances of 5%–10% for actual module size at the time of purchase are common in the industry, and longterm performance guarantees exhibit considerable variability.And even for a specific product, the guaranteed module size can change over time.For example, actual size may be guaranteed to be 90% or more of the nominal size rating for the first 10 years, and 80% or more of the nominal size rating for 20 years.
To net out the effects of system size, tracking, orientation, and shading—and get a clearer measure of how well the systems convert solar energy coming in into electric energy going out—we normalized the energy production results for incident irradiance and system size. Our ability to net out shading is somewhat limited because it is possible for the irradiance sensor to be shaded while some of the modules are not shaded, or vice versa. Not surprisingly, the two systems exposed to the most solar radiation include automatic tracking systems. The orientation of the system exposed to the next most solar radiation is adjusted manually to improve system performance.
We found normalized energy production values that ranged from 0.23 (Wh/W)/(kWh/m2) to 0.67 (Wh/W)/(kWh/m2) (see “PV Vocabulary,” p. 29).Along with other factors, these normalized results reflect the effects of battery storage on PV system performance. Chemical reactions in storage batteries produce internal losses that occur continuously even if power from the grid is available and the batteries are not discharged to satisfy household electric loads. A portion of PV system output is used to make up for these standby losses, maintaining batteries in a fully charged state. The energy requirement for this type of battery charging is a function of battery type and storage system size, not PV array size.All else being equal, battery charging will have a larger negative influence on overall system performance for systems that include smaller PV arrays.
This effect appears to be strongly influencing the normalized AC energy production results for the Mariposa site. This system consists of a small PV array and a battery storage system that backs up all house loads. While the sizes of the battery storage systems are unknown because the CEC did not collect these data, systems configured in this manner likely would include larger battery storage systems than those designed to power only a few critical loads when grid power is unavailable. System DC output at this site is seen to compare favorably with performance observed for other systems. System AC net output, however, is significantly less than average, due to battery-charging requirements. When considering only those hours when the PV system is generating AC power, the average DC/AC conversion efficiency for the Mariposa system is 81%. This result is similar to the results calculated for other systems, and it exceeds the overall average efficiency for this site by a factor of approximately two.
A Year of Sun Power
A performance measure of particular interest to owners of residential systems is total electricity production per year. For most of the systems, less than one year of data was available. Therefore, to estimate annual energy production,we combined the normalized AC electricity production results (column C) with National Renewable Energy Laboratory (NREL) estimates (column D) of annual average plane-of-array irradiance for solar collectors. For each site, information concerning actual system orientation was used to select the annual average plane-of-array irradiance value most representative of actual site conditions.
The annual energy production ranged from 622 to 1,740 kWh per year per kW of nominal DC system size. The two systems whose annualized output exceeds 1,500 kWh/yr/kW are both tracking systems. The average for nontracking systems is 1,122 kWh/yr/kW.
The annual energy output estimates that we arrived at are based on an important assumption related to the treatment of shading effects of trees and other obstructions. These estimates are based on interval-metered plane-of-array irradiance measurements. In some cases these values include shading effects of trees and other obstructions in addition to shading effects of clouds. Annual average plane-of-array irradiance estimates presented in column D of Table 3 include shading effects of clouds only, not shading effects of trees or fixed obstructions. When normalized AC energy output results are combined with these unobstructed annual average irradiance data, resulting annual energy output estimates represent the output for a system free of shading obstructions. While this basis is ideal for developing information for consumer education purposes, in cases where obstruction shading is influential, actual output may be less than is indicated in Table 3.
The Value of Net Metering
Net metering arrangements allow participants to send surplus electricity to the grid during hours when renewable energy system output exceeds the rate of household electricity consumption. When this happens, participants become net generators of electricity rather than net consumers and effectively use the grid as a battery. For each site, the extent to which this happens is a function not only of system size but also of lifestyle, appliance types and fuels, number of people in the household, and the weather.
From the perspective of participants, a meaningful measure of net energy production performance may be the net extent to which electricity produced by renewable means displaces power generated by other means that would otherwise have been purchased to satisfy household electric loads. During the monitoring period, energy use varied substantially from participant to participant, averaging 24 kWh per day. Consumption values for particular sites ranged from 9 to 49 kWh per day (see Table 4). PV output as a percentage of total household electric energy use ranged from 3% to 139%.
On average, these PV systems deliver electricity to the grid during 47% of the hours when they are producing electricity. For the average monitored system, 41% of the electricity that is produced is sent to the grid. For these monitored systems, ability to net meter and use the grid as a battery is very important. (For PV’s potential usefulness during peak demand hours, see “Matching the Peak,” below.)
Not 100% Reliable
Like most energy conversion equipment, small PV systems are generally not 100% reliable. Hardware and software problems may jeopardize system performance, and the performance of several monitored systems changed through time. The tracking system at the Winters site included three separate, independently operating tracking systems. While details of tracking problems are unavailable, the data suggest that at least two of the trackers experienced equipment problems that reduced system performance substantially. The system installer was notified of the problem, and system performance improved a short time later. The manually adjusted system at the Willits site included unframed PV modules, several of which experienced glass breakage due to unknown causes. At some point during the monitoring period, the owner of the system replaced one of the affected modules with a smaller module from a different manufacturer. Finally, the 1.8 kW, two-inverter system in Cupertino experienced at least two inverter failures that required inverter replacement.
If distributed generation is to play an increasingly large role in the future, its overall reliability must be closely monitored. Detailed data collected for this project contributed to problem troubleshooting. It is likely that some fraction of systems not included in this monitoring project will experience similar problems at some time during their long life. To ensure satisfactory performance throughout the life of a system, some level of ongoing monitoring is necessary. Because system output is a function of weather conditions, both electric generation and weather should be accounted for in any ongoing monitoring plan. Data requirements of such a plan could vary from very minimal to very detailed. For this project, the average cost of the monitoring system hardware was approximately $2,700 per site. The design of a more widely targeted ongoing performance monitoring plan would have to carefully weigh the tradeoffs between the cost of collecting performance data and the value of those data.
Finally, it is worth emphasizing that the PV systems in this monitoring study included small systems that used the technology that was commercially available several years ago.The performance of larger systems could be substantially different, as could that of newer technologies that are now available from domestic and international suppliers. Newer systems may offer improved performance for homeowners who choose to produce their own power.
Enter your comments in the box below:
(Please note that all comments are subject to review prior to posting.)