In this paper, a nominal SISO (Single Input Single Output) model of PHWR (Pressurized Heavy Water Reactor) type nuclear power plant is developed based on normal moderator pump-up rate capturing the moderator level dynamics using system identification technique. As the plant model is not exact, therefore additive and multiplicative uncertainty modeling is required. A robust perturbed plant model is derived based on worst case model capturing slowest moderator pump-up rate dynamics and moderator control valve opening delay. Both nominal and worst case models of PHWR-type nuclear power plant have ARX (An Autoregressive Exogenous) structures and the parameters of both models are estimated using recursive LMS (Least Mean Square) optimization algorithm. Nominal and worst case discrete plant models are transformed into frequency domain for robust controller design purpose. The closed loop system is configured into two port model form and H? robust controller is synthesized. The H?controller is designed based on singular value loop shaping and desired magnitude of control input. The selection of desired disturbance attenuation factor and size of the largest anticipated multiplicative plant perturbation for loop shaping of H? robust controller form a constrained multi-objective optimization problem. The performance and robustness of the proposed controller is tested under transient condition of a nuclear power plant in Pakistan and found satisfactory.