A linear dynamic model describing the behavior of temperature variation and thermal stratification in aquaculture ponds is used in developing optimal feedback controllers for a thermal destratification control system. Linear quadratic regulators (LQR's) with output feedback control of a linear-invariant system are chosen for this study. Both impulse and step disturbances are taken into account and an optimal proportional plus integral (PI) feedback controller is synthesized. To illustrate this procedure, the design is applied to control the thermal stratification in a typical shallow aquaculture pond. A sensitivity analysis is also performed to study the sensitivity of the objective function with respect to the controller gains. Feedback gains associated with temperature at middepth are shown to be far more sensitive compared with that of temperatures at other water depths. This result provides useful information in order to install the control devices at an optimum mixing depth.