In each electric power systems (EPS) beside voltage, frequency is one of the most important values in system state. The consequence of the unbalance between active power generation and consumption is electrical sinusoidal frequency deviations (voltage, current, etc.) from their nominal value. In order to establish balance between generation and consumption systems the fastest effect will be shown by using aggregate unit primary control via automatic increase of generation and energy storage in rotating masses of single units. This thesis investigates unit primary control area and its effect on frequency change. Experimental investigations of EPS dynamic are not usually possible because they are very complex, very expensive, adequate equipment is necessary, and they depend on EPS operational condition. Development of mathematical methods together with powerful computer systems makes theoretical investigations of EPS operational dynamics possible. In this dissertation two units mathematical model of minimal order is developed and studied using the software package MATLAB. Simplification of the mathematics is justified since the model is only approximate and may not represent the exact behavior of any particular location in the system, but is rather an estimate of the weighted average frequency of the entire Croatian EPS after disturbance occurred. The most important and the most influencing dynamical equivalent parameters of the Croatian electric power system are considered in order to ensure better insight on their effects to frequency response. Identification of these parameters is very difficult because they depend on EPS topological, operational and other conditions. Adequate identification method is applied for determination of investigated parameters with comparison of simulated and measurement frequency responses caused by serious disturbances. In Croatian EPS frequency recording, which has been going on for less than a year, there are not many recorded frequency responses after serious disturbances in the system. Some number of measurements are collected and used for parameters identification. As the number of measurement was determined calculation of midterm dynamics in time domain for the whole EPS of Croatia was made. For this purpose official classical software package MANISTA was used with detailed models of all elements in the system (this package was verified many times in project studies of dynamics Croatian EPS elaboration). Responses calculated in this way are used for parameters identification, too.