Like any device or system, all electricity devices requires periodical maintenance. Maintenance in electricity systems is a source of large costs; in the EU the maintenance costs amount to between 4% and 8% of the total sales turnover. In vertically integrated systems the strategic maintenance of power plants' and network's components is performed in an integrated fashion by the monopolist, whereas in those market based, these problems are responsibility of the GenCos and of the Transmission System Operator (TSO) respectively.
The maintenance activities are indeed complex even to classify. For instance if we define Preventive Maintenance in an abstract way as a general process carried out at predetermined intervals or according to prescribed criteria and intended to reduce the probability of failure or the degradation of the functioning of an item, we can distinguish:
Basically the medium term perspective coincides with one to three months ahead, and with this horizon in mind the maintenance refers to Condition based maintenance and most importantly to Opportunistic maintenance. In details the main goal of the maintenance processes in electrical systems are identical to the long term but things are somehow easier to approach:
The maintenance process here is simpler because we do respect the long term scheduled issues and we just adjust mainly for opportunistic mainentances, these are usually based on statistic evidence. Some more optimization oriented approach has been proposed also for the short term in coordination with the long term, see e.g.].
Related to Prognisis/Condition Assessment is the optimizatin problem of finding the best replace/repair/maintain decisions for each individual pieces of equipment. This can be embedded in a wide planning context: possible change in demand/generation patterns affects decisions on whether to replace/upgrage equipment or maintain the existing equipement. There may be a lot of uncertainly about the state of old equipment (e.g it may not be known how often a transformer has been overloaded) so the problem is stochastic. If this is to be decomposed into decisions about individual bits of equipment then what is needed is an estimate of the cost to the whole electricity system of removing the equipment for maintenance or replacement and the cost of unexpected breakdown. In a decentralied system these cost for planned and unplanned unavailabily need to be built in to contracts.
 M.K.C. Marwali and S.M. Shahidehpour. Coordination between long-term and short-term generation scheduling with network constraints. IEEE Transactions on Power Systems, 15(3):11611167, 2000.
Dr David Flynn, Heriot-Watt
Prof. Ken McKinnon, University of Edinburgh