This subsystem provides power to all other subsystems.
The primary supply is likely to be 4x 18650 rechargeble lithium-ion power cells, with two cells in series, in parallel with two cells in series. Typical cells each provide 3.7 volts for ~3,000 mAh. The cells will be mounted to PCBs using clips such as these.
Solar cells will provide power input for a recharging circuit. Some examples of larger panels available. That same manufacturer has other products in various sizes that could be useful. If possible, the satellite should have solar cells on each of its six faces to ensure exposure no matter orientation. On the Build-a-Cubesat project’s Discord there has been some discussion about which models from AnySolar are practical for space applications. Fortunately their advice aligns with models I can already use in my design.
A charging circuit is not yet designed, but needs to consider the supply of solar energy, temperature of the power cells, and current charge state. The Microchip MCP9808 looks like a good candidate for a temperature sensor, if the charging circuit is controlled by a microcontroller. Simple thermocouples could also work.
The overall power bus for the multiple PCBs on the satellite will be 3.3 volts. Not sure what to do if some subsystems require 5 volts.
Power distribution needs to be controlled to avoid continuous drain. Circuits are needed for each subsystem to enable power on demand and monitor current draw. Overcurrent situations must shut down the power to that subsystem automatically.
Power management requires a dedicated microcontroller. Radiation hardened processors are available, and is sensible for this application given the critical nature of this subsystem. One such processor available in the Microchip ATmegaS128. This processor has a non-rad equivalent for development purposes, and I have very good experience using the ATmega processor family in my robotics projects.
Over at the Build-a-Cubesat project’s Discord there has been discussion about using CAN to communicate between boards, rather than I2c or SPI. I have some past experience using the Microchip ATmega16M1 with a built-in CAN controller. Microchip offers a radiation hardened version of something in the same family: ATmegaS64M1 which looks very promising.
The power management microprocessor will need to communicate with other processors on the satellite. CAN looks to be the best choice.