Restoration of electrical power systems after severe disturbances is characterised by high time pressure and low experience of the operating personnel. The thesis describes a flexible approach for training and/or guidance of operators in system restoration, for the first time realised as a generic solution applicable to any given disturbance situation in any power system. Under consideration of the actual states of network and power units the system step by step suggests operational measures to be per-formed. The implementation also comprises components for knowledge acquisition, consistency check and explanation of reasoning; optionally an acoustic user dialogue can be activated. Coupling of the entire system with an operator training simulator gave the opportunity for verification with a large variety of scenarios on different power systems replicated under operational realism.