Inefficient control of the water level in a steam generator of a nuclear power plant is responsible for frequent unscheduled reactor trips. This problem is particularly critical at low power, when the steam generator exhibits the "swell and shrink" phenomenon and flow rate measurements are highly unreliable. The design of a proper controller capable of avoiding expensive shutdowns is eagerly sought for increasing the availability of the plant. In this paper, we present an extension of an adaptive, stable, fuzzy controller, whose design parameters are optimized via a genetic algorithm. Computer simulations confirm that the devised controller bears good performances in terms of small oscillations and fast settling time even in the presence of steam flow disturbances.