Charging the Battery, or Not So Much

It should have been a simple task. Get a solar PV panel, charge controller, and a lead-acid battery to support the electronics on my solar hot water system, but it wasn't .

The solar PV system on solar thermal hot water heater has never worked properly. The system struggles to charge the batteries which typically end up being dead by the time I get home from work in the evening. I know the batteries work properly, because if I charge them on an external charger, everything works fine. The batteries have more than enough energy to run the differential controller all night.  

Some Background
  Not wanting to plug my new solar powered solar hot water system into mains power, I purchased two 20 watt solar PV panels and mounted them next to the solar hot water panels.  But for the solar PV panels to do their job, they need both a charge controller and a battery.
 
   For the charge controller, I found this 10 AMP CMP12/24 unit on ebay.  Sure it is an inexpensive Chinese made unit, but none of the brand name people had much in the way of small charge controllers.  It seems to do the job. 

   The charge controller is intended to be paired with one or two 12 volt lead acid batteries.   In my experience, these batteries only last about 4 years, so I wanted to try something different.

    Eneloop NiMH batteries have been serving me well in other applications, so I thought I would give them a try in this application.  The question was, how many batteries would I need to simulate the 24 volts of two lead acid batteries.  Since these batteries are typically listed as 1.2 volt, simple math show 20 batteries × 1.2 volts = 24 volts.  Perfect, right?  perhaps not as we shall see.

Since the beginning of operation of solar hot water system, there has been a problem with startup in the morning.  There didn't seem to be enough energy in the batteries to have the controller up and running all night long.  Therefore in the morning, the exhausted battery pack couldn't open the zone valves.  Typically the differential controller would try to open zone valve, but as soon as current started to flow from the batteries, the battery pack voltage would dip below 21 volts and the charge control would kill the power to the load.  This setup a cycle that could repeat about every 20 seconds for quite a long time, where the valves would start to open and then the power would get cut..

Collecting the data.
     As is often the case in problem solving, the first thing that needed to happen was to collect data so the problem could be understood.  I have previously used the Hobo U12-4 data logger to record temperature data, and the same system is capable of recording voltages.  Unfortunately, the Hobo data logger can only record voltages between 0 and 2.5 V.  To record higher voltages required a special cable to be made with a 2.5 mm stereo plug and a voltage divider circuit between the center pin and the shield.  The voltage divider used a 1Kohm and a 22 Kohm resistor.  This resulted in a gain of 23× and made the range expand to between 0 and 57.5 volts.   This is a larger range then needed, but those resistors were readily available.

   One side note about the Hobo U12-4 is that it forces all the grounds to be the same for all the four data channels it is acquiring data from.  This was a bit of a problem for my application.  I intended to record 1) the solar PV voltage, 2) the battery voltage, and 3) the load voltage.  Unfortunately that did not work with this system as forcing the grounds to be the same for those three items made the charge controller malfunction.  So, for my data collecting, I settled on only collecting the load voltage.


On the worst day
After recording data for several days, a pattern emerged.  On the worst days the load voltage looked like the following graph.

In the early morning hours, the voltage was switched off because it was below the 21 volt minimum.  In fact, the voltage was probably below 10 volts.  The pack was well and truly drained.  Starting about 7:00 in the morning, light began to fall on the panels.  Unfortunately, the battery voltage was so low, the charge controller tries to put the system into 12 volt mode which makes it impossible for the valves to open because they require 24 volts.
   The charge controller repeatedly turns the voltage on and off as the battery voltage fluctuates as the differential controller starts to turn on and draw loads.  This continues until about 11:00 when enough solar flux lands on the solar panels and somehow the charge controller decides that this is a 24 volt battery pack after all.
    Throughout the day, 27.5 volts is applied to the battery pack to charge it, perhaps ending at 5:00 pm.  From that time on, the differential controller draws about 1.5 watts from the pack which result in the battery voltage dropping slowly.  Around approximately 8:00 at night, it looks like one of the cells suddenly drops voltage significantly.  Later, at approximately 9:00 at night, another cell voltage collapses and the voltage drops below 21 Volts.
    From that point on, there is repeated cycling.  When the voltage drops below 21 volts, the charge controller cuts the power to the load.  With the load removed, the battery voltage recovers a little and once above 23 volts, the charge controller restores the voltage to the load.  This quickly drops the battery voltage back below 21 volts and the load is cutoff from the batteries, and the cycle continues.

On the most days
Most days didn't see the charge controller automatically switch to 12 volt mode as can be seen in the graph below.

 But in every other regard, the behavior is similar with a difficult cyclical startup in the morning and a cyclical end to the day.

 
So what is the problem
  One of the more troubling aspects of the data is that only 27.5 volts is applied to the pack during the day.  This amounts to (27.5V/20cells) = 1.38 volts per cell.  This should be enough to charge the batteries, but I have seen my smart charger apply 1.4V or more to these cells to properly charge them.  So perhaps a little more voltage per cell would be useful.

   However, there isn't any charging voltage adjustment on the low priced charge controller that I'm using.  An alternative is to use fewer batteries.
    Given how the battery pack is constructed, it would be useful to have a "dummy" battery that is just a shorted cell.  Fortunately I found some on ebay and purchased them as shown in the picture below.

So what is the solution
I decided to try 19 eneloop batteries and one dummy battery.  That would give (27.5V/19cell=) 1.45 volts/cell, which is significantly more than the 1.38 volts/cell previously available.  The results were immediate and dramatic.

After a day to let the batteries charge, for the first time, the battery pack was able to run the differential controller all night long and into the next day as can be seen in the graph.  The battery voltage never dipped down below 24 volts which is well above the 21 volt cutoff for the charge controller.  The difficult startup seen most mornings, with the cyclical behavior, was eliminated and it looks like the valves opened cleanly on the first time (although that behavior was not recorded).

Into the future
The only remaining question is can the batteries be maintained over, say, a week of cloudy days and then startup smoothly when the sun returns.  One possible solution to that problem would be a PV panel cut-off.  This would be a circuit that would cut power to the load unless the sun is shining brightly on the solar PV panels and producing at least some voltage (e.g. 24 volt).  This would prevent the drain on the battery in the evening and could make it possible for the battery pack to stay charged almost indefinitely.

Update:  Sure enough, after a couple of back to back cloudy days, the batteries were exhausted and power to the differential controller was cut by the charge controller.