Information Batteries - Energy -> Data

The information battery concept might also be modelled with a series of Raspberry Pis equipped with voltage monitoring and larger supercapacitors in series with appropriate balancing, as well as larger solar panels, e.g. 40-100W. 

One design feature is parallel voltage regulators, where one regulator charges a small capacitor, e.g. 10,000uF and the other charges a couple of supercapacitors in series e.g. 25F.   The supercapacitors feed back to the small capacitor with a diode, and the computing power load goes across the small capacitor. 

For devices where people expect them to turn on instantly like with batteries.  When the sun comes out, the power circuit fills the small bucket (capacitor) quickly to the target voltage to start and run the device, while the larger bucket (supercapacitors) fill more slowly so that when they reach the target voltage, they provide stability from energy transients (like clouds passing) reducing the number of deep sleep cycles required.

 Less time working for investment of CPUs vs more time for greater investment in power stability.

At this point I have the Eaton hybrid supercapacitors, 220F, some specific solar panels I've tested (4.1V open circuit, so they naturally tap out at 3.6V below the diode, to keep the hybrid supercaps handy, with low part count.  No regulator is required to step around the hybrid supercapacitors, just a diode. 

I have prior experience with Espressif microcontrollers feeding their voltage systems back into the ADC with a voltage divider, so they can be programmed to go into deep sleep and then recover from intermittent power.  
Here's a video.  

In one of the images I've shared there's a Heltec WiFi Kit 32, hooked up to a Bosch pressure sensor, BMP280, but I may switch to another IoT such as the Adafruit ESP32-S2, which has the environmental sensor built in, or the Adafruit ESP32-S2 with TFT screen like the Heltec.   Initially it will not be important to conserve power, the main focus will be the software development for processing a research job with a mesh cluster.