Daily the demand for power and energy is increasing due to the growth in world population and the advancement of industries. Consequently, modern power systems are usually operated very close to the steady state stability limit to meet this growing demand. When an online generator trips or a heavily loaded transmission line is interrupted the balance between demand and generation is disturbed and the frequency begins to decline. This threatens the secure operation of power systems, hence necessitate the need for intelligent schemes that can effectively handle such power system disturbances. In situations where there is a severe and rapid decline in frequency, under frequency load shedding (UFLS) techniques are implemented to curtail portion of loads to balance the demand and generation. In this thesis, we propose two under frequency load shedding methods. The first method is the wide area measurement systems (WAMS) based comprehensive UFLS scheme. Phasor measurement units are used to provide the real time status of the power system and based on these measurement restorative actions are easily implemented. The main shedding decision parameters are the disturbance magnitude which can be estimated using the disturbance observers or the swing equation and the minimum frequency which can be obtained through frequency prediction techniques such as polynomial curve fitting. The second methods encourages customer participation in demand response, here the real time consumed power is considered a reserve which can be shed to restore the frequency. Simulations are done on selected three reduced power systems models and on a highly detailed IEEE 39 bus system using matlab/Simulink. Key words: demand response, disturbance estimation, frequency prediction, smart grids, under frequency load shedding, wide area measurement systems