An artificial neural network (ANN) is developed for hybrid state estimation of a three tank system. The ANN structure uses a nonlinear autoregressive network with exogenous inputs (NARX) model for state estimation. The ANN provides corrections to the state estimates similar to an extended Kalman filter. Experimental results show the ANN approach achieves better performance than other controllers in terms of integral squared error and computation time for regulatory control, servo control, and in the presence of uncertainties like initial condition mismatches, plant parameter mismatches, and valve faults.