As one goes about evaluating an optimizer on-the-bench please keep in mind
how these algorithms work. It's not straightforward. They're not like a
classic switching power supply.
The unit optimizes power transfer by adjusting both its input operating
point and its output operating point simultaneously. Because all the units
are in series (with multiple series strings in parallel should the roof be
large enough) the job of the optimizer's feedback loops is to "present" an
output voltage of panel (a forcing function) and then adjust its output
voltage (while accepting the string current in common with all panels
connected in series) such that power transfer is optimized. If anything
changes on the output side of the optimizer such as the inverter presenting
a different voltage upon the entire string (another forcing function) in
response to string-to-string changes or more commonly changes within a
string (such as panel #3 becoming shaded) then the optimizer must
re-optimize its input and output parameters. Because each optimizer is
doing this independently there is opportunity for the many feedback loops
to become to excessively coupled causing the whole system to go unstable
(i.e. chasing its tail). The algorithm developers take great pains to
avoid these instabilities. Rapidly moving clouds or swaying branches
create rapidly changing incident radiation. This is worst case.
RFI wise we're looking at a PWM inverter with very sharp switching edges
(to keep efficiency high). The character of the RFI is always changing as
incident radiation conditions change. Substantial filtering is necessary,
especially common-mode currents. Yes, ground returns must be choked off
too. It's too bad the output side of each optimizer does not consist of
twisted pairs each returning to a central point within the main inverter
box where serialization would take place. This method would consume too
much costly Cu cabling and would be labor intensive. Because of the high
voltages involved in these strings and the need for fire safety, these
cables are costly.
When one tests optimizers on the bench I suggest trying different loads. A
simple resistor that is varied in value such that the output of the
optimizer is taking through it's voltage compliance is probably
sufficient. In the industry active loads are used. The latter can present
current or voltage forcing so as to better simulate actual operation. You
can also slide some cardboard over the array to observe the optimizer
adjusting to this situation.
How do I know this? My company developed the very first optimizers. These
did not generate RFI because a ham was involved (yours truly). However,
the cost of making quality optimizers with outstanding efficiency and
reliability was too high for a successful market run and the product was
discontinued while cheaper stuff began to hit the market. Meanwhile, the
installers went from pro's (who underwent technical training) to local
contractors that just install and don't understand the system. The cost of
panels, electronics and installation had to come way down for mass
adoption. Let's face it, today's residential solar systems are consumer
products.
I know Solar Edge because it was an early competitor. Suffice it to say
that the RFI problems being reported do not surprise me one bit..
Dennis AE6C
_______________________________________________
RFI mailing list
RFI@contesting.com
http://lists.contesting.com/mailman/listinfo/rfi
|