ENERGY

Catching the SUN

By rotating solar panels to follow the sun across the sky, solar trackers can generate more
power. These solar power plants can be damaged by aeroelastic instability at modest wind
speeds. CPP Wind Engineering used simulation to determine the nature of the instability
and to identify operating procedures and design changes that can prevent them.

By Christian Rohr, CFD Manager and Peter Bourke, Operations Manager,

CPP Wind Engineering, St. Peters, Australia

Single-axis solar trackers that automatically rotate to follow the sun from east to west can generate 10 percent
to 30 percent more power than stationary or “fixed tilt” solar panels. Certain wind conditions can cause a
torsional instability that damages solar trackers. CPP was commissioned by several tracking companies,
including NEXTracker, to investigate these kinds of failures and develop a solution. CPP used ANSYS
Fluent computational fluid dynamics (CFD) software in conjunction with wind tunnel testing to re-create
the conditions under which the trackers became unstable. CPP identified the cause of the problem and
demonstrated how it could be solved through adjustments in operating conditions and design changes.

SOLAR TRACKERS
Single-axis trackers consist of photovoltaic panels mounted to a long shaft, called a torque tube, that
rotates the panels. The torque tube provides stiffness to resist wind forces, and some trackers also
have torsional dampers that look like automotive shock absorbers to reduce vibration. Most single-
axis trackers are mounted so that the axis of rotation is horizontal to the ground. The torque tube
is supported by vertical piers or posts at intervals along its span that are mounted to the ground.

A vortex forms along the upper side of the panel At 0.55 seconds, a vortex separates from the
during the simulation at 0.30 seconds. panel, and upward moment drops to zero.

42 I ANSYS ADVANTAGE ISSUE 1 | 2019

 When winds are high, trackers are often
rotated into a stow position. Traditionally,
this stow position places the panels
parallel to the ground to reduce horizontal FRAMING
MEMBERS
wind forces. In a series of incidents, PANELS

trackers from multiple suppliers stowed

in this manner have experienced large TRACKING

deflections in their first mode of vibration,

a helical twisting mode with opposite ends DAMPER TORQUE TUBE
+ MOTOR
of the solar panel rotating in opposite
directions. Reports from the field indicate GROUND
POST

that the panels were oscillating more than

20 degrees in the positive and negative
Typical solar tracker configuration
directions, causing damage.
In the flat stow position (parallel to the
ground), the trackers bear a resemblance to an airplane wing, so CPP engineers postulated that
their instability might be caused by flutter, which occurs when aerodynamic response amplifies
the vibration of blades. Calculations with handbook equations indicated that classical flutter was
not likely because the trackers cannot heave up and down while twisting because vertical motion is
constrained at regular intervals by the support posts. Furthermore, handbook equations could not
supply adequate results because they rely on assumptions that were not true for field conditions.
For example, most of the handbook equations pertaining to flutter are based on oscillations that are
much smaller than what was reported in the field with the solar trackers.
Neither CFD nor wind tunnel testing alone can provide a complete understanding of these
phenomena. Aeroelastic wind tunnel testing does not reveal the flow mechanism that causes the
instability, and the geometries and conditions that can be tested in this way are limited. CFD is
well-suited to simulating many different design points and virtually any geometry or condition
but requires validation comparison to physical testing when applied to a new problem. CFD also
provides a deeper dive into the physics than can be obtained with wind tunnel testing, providing full-
field pressure and velocity patterns that reveal the flow structures responsible for the tracker’s motion.

At 0.77 seconds, the leading edge of the tracker’s At 0.90 seconds, a vortex has separated and the
panel has rotated downward and a vortex has leading edge is about to rotate upward.
formed on the lower edge.

© 2019 ANSYS, INC. ANSYS ADVANTAGE I 43

Catching the Sun (continued)

CFD MODELING OF A SECTION

CPP engineers wanted to evaluate many different
design points in order to tell their clients which were
stable and which were unstable. To obtain short
solution times, they used a 2D CFD model. While this
model was not able to reproduce the 3D twisting seen
in real solar trackers, CPP engineers calibrated the
stiffness and damping of the 2D model to match the
behavior of real 3D trackers as observed in the wind
tunnel and in the field.
In the CFD model, the panel was mounted just
barely high enough above the ground to spin without
contacting the ground. The domain was meshed
Model used for wind tunnel testing
with approximately half a million quadrilateral cells
with refinement around the panel and wake regions.
Engineers applied the realizable k-ε turbulence NATURAL VENTILATION STRATEGIES
model to observe the threshold at which instability
The complementary benefits of wind
occurs. They ran a 10-second transient simulation
tunnel testing and CFD have also been
with minimal computing resources to determine the
used by CPP engineers on several
stability of the tracker.
occasions to improve comfort levels
Engineers substantially reduced the
in interior locations. Although wind
computational effort required to perform the
tunnel testing is the most appropriate
simulation by embedding the equations of motion
tool for accurately measuring wind
for this model into a user-defined function (UDF),
speeds and turbulence in urban
and using the UDF to calculate the deflection of the
environments, it is primarily limited
structure at each time step in the simulation. This
by Reynolds number effects to
information feeds back to the CFD solver, which
outdoor spaces. CPP engineers used a
adjusts a rotating mesh to alter the position of the
combination of wind tunnel pressure
tracker to account for the deflection. This approach
measurements and CFD studies
provided reasonable results quickly.
of internal spaces to determine
Engineers created a table of design points in
comfort levels in interior locations not
ANSYS Workbench by varying the windspeed, wind
accessible to wind tunnel testing. The
direction, elevation angle, structural stiffness and
computational models also enabled
damping of the tracker. They then entered values
evaluation of thermal loads generated
in the table of the design parameters that they
by occupants, equipment and solar
were interested in exploring. Next, engineers used
heating that cannot be measured in
Workbench to automatically generate models and
the tunnel.
execute CFD simulations to evaluate each of the
design points and store the results.

WIND TUNNEL TESTING VALIDATES SIMULATION

At the same time, engineers built a physical section model that matched the geometry of the 2D CFD
model and used a variable stiffness torsional spring to match the stiffness and damping of the CFD
model. The physical model was placed horizontally in
CPP’s atmospheric boundary layer wind tunnel. The
spring was attached to the axis of rotation, and angular
displacement was measured with a laser sensor. The
wind tunnel model showed good agreement with the CFD
results, providing assurance that the CFD model could be
used to accurately evaluate alternative designs.
Next, the engineers built a 3D CFD model to run
steady-state simulations of several key design points,
allowing them to check whether the behavior of these
points matched 2D CFD. They also built a fully three-dimensional aeroelastic wind tunnel model to
further validate both the 2D and the 3D CFD models. All of these simulations and tests correlated well
with each other.

44 I ANSYS ADVANTAGE ISSUE 1 | 2019

 WIND-DRIVEN RAIN
CPP engineers also synergistically
combined CFD with wind tunnel testing to
determine the extent of wet patches under
awnings and close to building entries. They
used both discrete particle models and
Eulerian multiphase strategies in ANSYS
Fluent and matched the predicted flow
field against concurrent pedestrian-level
wind tunnel testing to validate or scale the
CFD results.
CFD results for wind-driven rain

IDENTIFYING THE CAUSE OF THE INSTABILITY

The simulation showed that, in flat stow, a vortex forms above the leading edge of the tracker, creating
a substantial moment across the central chord. This causes the tracker to twist and, as it does, the size
of the flow separation increases and the zone of significant uplift crosses the midpoint of the chord. As
the vortex is shed by the tracker, the moment suddenly drops to zero. The tracker then bounces back
past the flat position so that its leading edge is twisted downward into the wind. A vortex then forms
on the underside of the leading edge, and the process described above is repeated. The simulations
showed that this instability is based on torsional divergence, and is not easily treated with damping.
Simulation showed that it is possible to excite twisting in the first mode shape with only a fraction
of the full span participating in the vortex shedding, and that significant amplitudes can be reached
in just a few cycles. Simulations were run at different conditions, and the time series was examined to
determine whether the tracker was stable or unstable.
Simulating the tracker over a wide range
Lock-in Threshold, CFD vs. Wind Tunnel
70
of conditions showed that the system has
a critical velocity above which instability
60 will always occur. The results revealed that,
50
CFD in many cases, instability could be avoided
Wind Tunnel
Wind Speed (m/s)

by stowing the trackers at an angle rather

40 than parallel to the ground. However, the
30
results also showed that tilting the tracker
introduces the potential for vortex lock-in
20 instability (where alternating vortices are
shed from the leeward face of the tracker).
10
Greater static loads and greater dynamic
0 excitation due to buffeting also result
0 0.05 0.1 0.15 0.2 0.25 0.3
Participation Fraction from higher tilt stow. These issues can be
addressed by increasing the stiffness and
< Critical wind speed determined by simulation vs. wind damping of the tracker.
tunnel testing
CPP engineers provided their clients
with maps of tracker angles, wind speeds and directions that indicate which types of trackers work
under which weather and operating conditions. They also provided guidance on what results could be
expected by upgrading the stiffness and damping of trackers. This project provides a good example of
how CFD and wind tunnel testing can be combined to solve problems that would have been difficult
or impossible to solve with either testing or simulation alone. Wind tunnel testing was also used to
validate both 2D and 3D CFD models. CFD, on the other hand, made it possible to identify the exact
cause and solve the problem. Engineers were able to evaluate many different design points in a
reasonable amount of time while providing voluminous diagnostic information on each point.
CPP Wind is supported by ANSYS Channel
Partner LEAP Australia. How Accurate Fluid Simulation Helps You
Develop Best-in-Class Products
ansys.com/accurate-fluid

© 2019 ANSYS, INC. ANSYS ADVANTAGE I 45