A satellite itself which is on an orbit high enough to not interact with the atmosphere is essentially maintainance-free. Without atmospheric drag it won't need any course corrections to maintain its orbit. It just stays there indefinitely.
Now about it's operation: This really depends on what the satellite is doing. Do the instruments need regular instructions from earth? How difficult is it to come up with these instructions? Does it send signals back to earth? How much effort is needed for analyzing its transmissions? How strong is its transmission power and how much data does it transmit? Is it OK when it can only be reached from time or does it need constant contact to a ground station? When the latter, do its orbital characteristics allow to use existing ground stations or will it be necessary to build and maintain new ones? Are we talking about one satellite or a satellite system consisting of multiple ones (in that case you could likely reduce cost-per-satellite significantly through synergy effects)?
Depending on these factors, the cost can vary between zero and several million per year.
One random figure I could find: The European Galileo and EGNOS navigation systems together are estimated to cost 800 million Euro in maintainance per year after completion. These systems consist of 48 satellites. Does that mean one satellite of the system would cost 16.7 million per year? Unlikely, because you would need almost the same number of operators and ground equipment. Would it be possible to do it cheaper when it wouldn't be an EU government project hampered by countless political factors? Likely.