Nowadays environmental problems, high energy costs, and reduction in the fuel resources have become as the reality of life, that makes the study of all energy-consuming processes; in the way of improving their total efficiency; inevitable. In every industrial process which is based on fluid transfer from one pressure level to a higher level, pumps are used as a main part, and for this reason, pumps are considered as a primary consumers of energy. In this thesis, separate parts of a pump-induction motor drive, and also the loss model for each part are studied and for each part a model is proposed, which although with suitable precision, it doesn’t need any complicated and extra calculations. Hence, in order to model the pumps and the hydraulic system, which is a nonlinear and complicated system, the neural network is applied. Different efficiency optimization approaches in the induction motor drive are studied, and it is shown that an appropriate approach for efficiency optimization in the pump-induction motor drive is the minimum loss search approach. Of the criticisms on this approach are its slowness and the oscillation round the optimum point. The search tables approach also exists as a controlling approach which is based on practical experiment. Of the criticisms on the search tables approach is that, in practice, the number of test points are limited; and also if the parameters of motor and load would change over the time, these tables must be modified and prepared again. In the present thesis, a fuzzy search control approach is proposed in the pump-induction motor drive. The proposed controller, by integrating the search table approach with the minimum loss search approach, and by considering the necessary strategies in design, could reduce the problems existed in both approaches.