PV System Risk Assessment

 

Part of determining the longevity and return on investment of a solar photovoltaic (PV) system is understanding the probability that there will be repair costs related to wind damage. Some of the challenges related to this analysis are unique to solar, and some of them are still poorly understood. 

 

Impact of Project Location on Wind Loads

The starting point for such a probability analysis is the wind climate at the project’s location. A regional analysis of the likelihood of extreme winds has already been conducted for all of the USA, and is incorporated into the wind map in the American Society of Civil Engineers Minimum Design Loads for Buildings and Other Structures (ASCE 7). Similar maps exist in most countries, including Mexico, Canada, and Australia. These maps are created using wind data from airports, which are typically enveloped conservatively. However, it is often the case that little or no wind data is available from the sparsely populated areas where many utility-scale solar projects are installed. In some situations, a “special wind region” with locally high winds can be present, but not quantified in the building code, or wind loading standard. In these cases, the designer needs to conduct an assessment of the wind loading risk. This may include a discussion with the local building department, analysis of wind data from nearby sources, and investigation for evidence of high winds at the site – the presence of a nearby wind farm, perhaps, or anecdotal accounts of empty trucks being knocked off the highway on windy days. 

 

In the western USA, downslope winds and thunderstorms are of particular concern. Downslope winds, which can approach hurricane strength, develop in mountainous regions with a steep lee (down-wind) slope. Thunderstorm winds rise up very quickly, and can come from any direction. Wind speeds can jump from 20 mph to 60 mph in 2-3 minutes, particularly when there is a downburst. Tracking systems, which often move to a stow position (typically flat or near flat) in high winds to reduce wind loads on the system, should be designed to provide enough time to get into stow position. In addition, stow mechanisms should be designed to function in the event of a communications failure or utility outage. Conversely, downslope winds tend to be very directional and can last all day, or even for several days. It is important to know if the system is vulnerable to these wind directions. 

 

The question of vulnerability leads us to the second step in the analysis: determining the wind loads on the structure. There are two kinds of wind loads to consider – static loads and dynamic loads. 

 

Static Wind Loads on PV Structures

Static wind loads on the structure and on smaller components, such as modules and connecting hardware, must be determined as part of the engineering design process. Once static loads are determined, the load amplification, if any, from the structure’s dynamic response should be accounted for (see the following section).

 

Static wind loads are provided in the building code for many structures, but PV is not yet addressed in existing codes. Nearly all of the loads in the codes come from wind tunnel testing of scale models of the structures. ASCE 7-16 will introduce new provisions for wind loads on roof-mounted solar, which were derived from this type of wind tunnel testing. Most of these provisions are based on SEAOC PV2, “Wind Design For Low-Profile Solar Photovoltaic Arrays On Flat Roofs”. This document provides a lot of background information on how wind loads vary from one installation to the next, and is a must-read for those interested in understanding wind loads on low-profile, commercial roof-mounted systems. 

 

Despite the name, most static loads in the code are actually of very brief duration, the result of rapidly changing fluid dynamics. The design loads can last just long enough to fully load the structure. For roof-mounted solar, this might mean that the weight of a module and its ballast is exceeded for a tenth of a second or less. It is tempting to dismiss this as too brief to cause a significant failure. However, once the panel lifts, the aerodynamics can change for the worse, leading to more uplift force. For this reason, average forces from tests should never be used – peak wind loads must be used for design. Measuring peak wind loads in the wind tunnel requires specifically designed data acquisition systems which can collect data at very high sampling rates. Additional information of the requirements for such testing is available in industry literature. (Kopp, G.A., and Banks, D. “Use of Wind Tunnel Test Method for Obtaining Design Wind Loads on Roof-Mounted Solar Arrays”. Journal of Structural Engineering 139 (2) (2013): 284-287)

 

Dynamic Wind Loads on PV Structures

Dynamic loads increase the loads above those predicted with static coefficients because they include the effect of rhythmic motion of the structure on the loads. One of the best known examples of a structure that failed due to dynamic wind loads is the Tacoma Narrows Bridge. Every structure has modes of vibration, and each mode has a natural frequency. For example, many fixed-tilt systems will have a mode in which they sway back and forth, in a north-south direction. Typically, the natural frequency for this motion is between 2 Hz and 5 Hz. If the rack is buffeted by gusts of wind at its natural frequency, the rack will begin to sway back and forth at this frequency, which in turn, will cause inertial loading on the support posts. 

 

There is very little energy in the wind at these frequencies. However, fixed-tilt ground-mounted systems are generally built in large arrays. The rows of racks create wakes with smaller, higher frequency gusts relative to the approaching flow, due to vortex shedding. Interestingly, this is particularly pronounced in the second row from the north edge in south-facing fixed-tilt arrays.  

 

The effects of dynamic wind loading are sometimes expressed as a dynamic amplification factor (DAF), which is a multiplier on the static wind loads to account for the dynamic or inertial loading. As the name implies, the DAF is always greater than 1.0.  In a wind storm with a 90 mph gust, an array of 13 ft wide (4 m) rack tilted at 20° will be subjected to a significant amount of wind energy with excitation frequencies between 1 and 3 Hz in the array interior. If the natural frequency of the mounting system is 2 Hz, the DAF can be above 5 if the damping is low (under 1%). 

 

Clearly, ignoring this effect will significantly increase the likelihood of a failure. For this reason, a rule of thumb of keeping the natural frequencies above 4 Hz is sometimes used for such systems. Note that this frequency threshold will increase for smaller structures or when higher wind speeds are likely (such as in hurricane zones).

 

Dynamic wind loading is well understood for tall buildings, and as a result many structural engineering firms are familiar with methods of integrating wind tunnel test data with structural properties to predict dynamic loading. Solutions can involve increasing the damping and/or the natural frequency. 

 

Dynamic wind loads are most evident for winds approaching from the high side of the array (i.e. the north side in a south-facing fixed-tilt PV array), so that changes in array alignment to avoid the worst winds at a site can reduce this effect as well.  

 

Risk Assessment

The last step in this process is to incorporate the likelihood of a component (such as a pier) breaking under a certain wind load. Code calculations take a pass/fail approach to this issue, which has a built-in assumption, but all components will have a range of strengths. Statistics about the material strength can be combined with statistics about the wind loading (a combination of everything discussed above) to produce a fragility curve, which shows the probability of a given amount of damage over the life of the structure (see Figure 1). 

 

The y-axis on the fragility curve can be converted to a cost of repair. The cost of moving between the curves on the plot is the cost of increasing the strength of the members or structure. 

 

To create a fragility curve, it is necessary to understand how the component is affected by wind loads. Unfortunately, this is poorly understood for one of the key components: the PV modules themselves. Existing mechanical loading test protocols for PV modules do not reflect the actual wind loading environment experienced by the modules; simply put, the load tests are static, whereas the real wind loads vary rapidly in time, and are not uniform across the panel. As a result, modules may fail the tests when they are in fact suitable, or be installed and warrantied only to experience unexpected failures and degradation. With an industry trend toward reducing frame rigidity, and glass and cell thickness, it may be important to screen PV modules for long-term resistance to the type of wind loading it will be subjected to in the real world. Unfortunately, such a test protocol does not yet exist.

 

Summary

Geographic location, surrounding terrain, PV system geometry and the dynamic response of the mounting structure govern the magnitude of wind loads on PV systems. The dynamic response of many PV mounting systems is often overlooked and can result in a significant increase in wind loading relative to the static loads that are typically estimated by PV system designers. The risk of a structural failure in a PV system is best assessed by a statistical analysis (fragility curve) that combines statistical data for material strengths and historical wind data for the location. Statistical data that quantifies the impact of real-world wind loading on PV modules is currently unavailable but is needed to fully assess the risk of PV system damage due to wind loading.

 

 

 

 

Colleen O’Brien, P.E. is a Principal Engineer at DNV GL, where she manages the Commercial Solar Due Diligence group. 

 

 

 

 

 

Dr. David Banks is a leading expert on aerodynamics and wind tunnel testing and leads Solar Services and Special Projects at CPP.

 

 

DNV GL | www.dnvgl.com