Two strategies for control of nitrogen removal in an alternating activated sludge plant are compared. One is based on simple model predictions determining the cycle length at the beginning of each cycle. The other is based on simple rules relating present ammonia and nitrate concentrations. Both strategies are close in efficiency measured as effluent total (inorganic) nitrogen and both perform better than using fixed phase lengths for a test scenario describing a typical dry weather diurnal variation. After modifying the criterion functions of the rule based strategy the two strategies are found to produce equivalent controls, which means that the optimal criteria of the model based strategy can be fulfilled without the need for model predictions. (C) 1998 Published by Elsevier Science Ltd. All rights reserved.