Some thoughts in C,I,A:
Open ports (actually the programs listening and responding at them) may reveal information about the system or network architecture. They can leak banners, software versions, content, the fact a system is there at all (instead of dropping the packet) and what type of system it is (for example, nmap can fingerprint systems). Rook's answer got me thinking about this.
Without open port controls, software can open any candidate port and immediately communicate unhindered. This is often relied upon by games, chat programs and other useful software, but is undesirable for malware.
The network stack and the programs at open ports, even if the requests are invalid, still process incoming traffic. Even if electricity isn't an issue, technological solutions still have limited resources: degraded or denial of service results from finding a way to commit a port, network stack, computer, its hardware, network, or the people so they can't do much else.
Related to integrity and availability, an overwhelming amount of events and their logs can hide malicious activity (such as exploiting something you aren't looking at, to gain access) and lead to administrative fatigue and error. Potential misuse of certain services, by forcing the system to participate in DDoSing someone else is also possible.
Returning to Rook's answer, the smaller your attack surface the less control of your resources (and possibly other people's) you give to potential attackers.