After spending considerable time trying various settings and operating strategies, I now have the inverter working, after a fashion. Princeton Power has been only marginally helpful during the process, given that this is an old unit, out of warranty. On thing they did tell me is that normally a 100kW inverter is connected to a battery bank of capacity around 200kW-hours. My bank has capacity of only around 15kW-hours, or 7.5% of “normal”. So I should expect some challenges.

Knowing that my batteries were somewhat discharged, I decided to bring them to a full state of charge. This would make my battery voltage as high as possible, and possibly avoid a low battery trip and allow the inverter to run. Success! The inverter locked into run mode! Makes lots of interesting noises.
Neither charging or discharging, the inverter claims a steady draw from the batteries of 4-5 amps. Princeton says this is normal. With my small battery bank, I will need to make sure I go to charge mode every hour or so to counteract this drain. I also need to make sure that I charge before any shutdown, to make sure my inverter will re-start when needed.
Now that the inverter is running, I need to interact with it. So I need to get the LAN interface working. This required many steps over several days, including locating the port on the inverter, running a cable to my laptop, purchasing a cross-over adapter, setting the IP address on my laptop LAN adapter, downloading and installing Java (needing to use IE, it doesn’t seem to work in Chrome??). Thank goodness for Tyler, my IT helper. Finally I can communicate. It should be easier that this.
The Princeton web interface works well, pretty much as described in the manual. I can quickly change parameters. More importantly, I can command a power flow to or from the battery bank. From the panel interface, I was only able to command power to flow from battery to grid. Not so handy when I have batteries that require frequent charging.
When I command a 1kW flow from batteries to grid, I observe about 14 amps DC. That’s more like 5kW. When I command a 1kW charge, a see 14 amps the other way. I also see in charge mode the battery box trip out after a minute or so because of a high voltage input.
Clearly I have a calibration issue on the inverter on the DC side. Amps reported by the inverter appear to contain a -0.4 amp offset compared to the amps reported by the battery box – easy to live with. But voltage is quite wrong as measured and reported by the inverter. Voltage as reported by the battery box is validated as correct by my hand meter in any operating mode. In inverter idle mode, inverter reports a DC voltage about 6 volts lower that actual. In run mode, regardless of amp flow, the inverter reports a DC voltage 40-50 volts below actual. Yikes! Princeton tells me that the box has gone out of calibration (no kidding) but I can’t persuade them to tell me how to fix it.
I decide to work around the problem by adjusting the battery settings in the inverter control to much lower values – values that I would expect the inverter to observe in its un-calibrated form at max drain, hold, and max charge conditions. Using these settings, I am now able to keep the inverter running for hours at a time. So far, I have been unable to have the inverter automatically switch to charge mode when battery voltage drops too low – I need to manually command a charge using the web interface. When charging or discharging, I command only 1kW (and get much more) so that I do not exceed the amp rating of the system.
Finally I have something that seems to work after a fashion, and might be used to conduct experiments, provided that one is respectful of system limits. At one time I thought that I was going to need a different inverter, but it looks like we can work with this one. A victory!
One thing I notice about the inverter – the manual says that it should continue to power the critical AC load from the batteries if the grid power drops out, as long as the batteries hold up. I should be able to get short time driving a 5kW load with my batteries. However, when I kill the AC power, my inverter does nothing of the sort – it simply dies. What’s that all about? You would think that at least the control logic in the inverter would stay alive. This leads me to trace wires in the inverter cabinet to figure out the control power scheme. Much to my disappointment, I discover that the control power is derived from the grid 480 3-phase, and has nothing to do with the DC side!!! How could it possibly stay alive? Obviously it can’t. Not good for the experiments my customers intend to run. So I need another work-around. Additional digging reveals that I need 110V and 220V single phase to keep the controls alive. I need to make those AC voltages from the DC battery voltage, or need to use the inverter produced 480 3-phase through some sort of switch. To keep costs down and the system simple, I elect to do neither. Rather, I’ll run a separate 220V cord with neutral from the wall to keep the controls alive. Not so good for a functional micro-grid system, but OK for the experiments I need to run, for now.
To have a useful system, I need to be able to log the key parameters. On the 480 3-phase side, our Keysight PA2203 Power Analyzer is just the thing, now that it is back from repairs. This is a wonderful device. I hooked it up, plugged it in, turned it on. It looked at the inputs and it immediately knew exactly what to do without my intervention. Without any button pushing, it reports volts, amps, and power waveforms by phase. One or two button pushes give analysis and totals. One or two additional button pushes and I manage this screen capture:

A couple more button pushes and I create a log in csv format that contains more numbers than I know what to do with. This is a tremendous instrument! Easy to use! I wish everything worked this easy.
Now I need to capture and record the data from the DC side – battery volts and amps. These are reported by the inverter, but we know that those numbers cannot be trusted. I need to be careful what I hook up to the batteries directly, as there is high common-mode voltage to earth ground created by the inverter. I opt to purchase analog output modules that can be plugged into the existing Carlo Gavazzi digital meters for volts and amps that are on the battery box. Expensive (about $300 for the pair) but should be easy solution.
After plugging in the modules, I need to program the meters to recognize them. This requires using a cryptic button pushing method that takes a little time to get used to. When I dig in, I discover that the meters have been password protected. Yikes! Trying number combinations did not work. Nobody from SBS (manufacturer of the box) remembers. Finally I get the back-door password from a friendly tech at Gavazzi.
Now I run a cable from these analog outputs and connect it to analog inputs of my Keysight 34972A data acquisition system. This box is somewhat older technology that the Keysight power analyzer, which means that the human interface is little more challenging. But still not difficult. I manage to correctly scale and offset the inputs, and figure out how to log the volts and amps at any desired frequency to a csv file.
The logs created when the inverter is running show some unusual behavior – lots of transients that have duration of 100-300 milliseconds. It appears that for this short period of time, the analog output generated by the Gavazzi meters is going to zero. Moreover, the analog output for amps STAYS at zero when the inverter is discharging the batteries. This is a puzzle. When the battery box is interfaced to the NHR charger, there are no such transients – the volts and amps log exactly as expected. Below are volts and amps traces created from csv data generated by the 34972A during inverter and NHR charger operation. These should be similar, but obviously they are not.


While operating the system one day trying to track down the source of the noise, the amp meter on the battery box temporarily flickered and went blank. Is this related to the noise? Those meters are powered by two of the batteries in the middle of the battery string, nominally at 24VDC. With my hand meter, one of the those batteries measured 6.1 volts and the other around 10 volts – that’s bad. And I notice that those two batteries are different than the others – they have been changed out before. I replace those batteries with new ones. I reason that this is a design problem with the battery box – those two batteries see a steady drain from the controls of about 0.5 amp that keeps the solenoids locked in. Yet there is no way to independently charge them. Not wanting to keep changing batteries, I move the control power over to a bench power supply. Everything is working again, but the noise problem remains.
In an attempt to circumvent the noise issue, I decide to try connecting the battery box volts and amps (volts across a shunt) directly to the datalogger. Right away, the logger tells me that the volts signal was out of range (max 300 volts on the logger) so I disconnect that line. I started logging the amps. After about a minute, there is a pop and smoke from the datalogger – that’s never good. I shut it down and look at the data – the noise is still there. A day or two later, I reconnect the logger to the analog outputs as before, to see what damage I caused to the logger. Amazingly, it all still works. Not sure what blew out, but apparently it was not important for this type of measurement. Lucky!
Lamenting about my noise problem with a friend led to a quick visit by Mark Bolent, a local electrical engineer, a friend of my friend. In about 20 minutes Mark identified what we think is the noise culprit – low power wiring in the battery box running right next to high power cables. We think that AC transients generated by the inverter on the DC buss are inductively coupled to the low power wires, causing interference with the meter power and signals. The NHR charger generates very clean DC, so there are minimal transients and therefore no interference caused when operating with that device.
I’m working now to re-route the wiring in the battery box. Stay tuned ……..