Application of artificial neural network (ANN) for frequency response in power system dynamics

Abstract

Frequency is regarded as a paramount index of the operation of power system because it can reflect the dynamic energy balance situation between load and generating power. Under steady state conditions the total power generated by power stations is equal to the system load and losses while frequency normally operated at a nominal value. However, the frequency can deviate from its nominal I value due to the transient events occurred in the power system dynamics. Therefore, an effective method for frequency estimation is an important task in the operation of power systems. This paper presents a new application of Artificial Neural Network (ANN) for frequency response in system dynamics. A Neural Network seems to be an ideal solution for a quick and accurate way to determine the frequency response compared with standard dynamic simulations. In order to perform the ANN, the power flow solution is obtained first for the system to be studied. During this steady state condition power system shows a balanced system. The purpose of load flow simulation is to get some operating parameters which influence the system frequency behavior. The transient simulation of a power system is then simulated by DigSilent Simulator to analyze the frequency response of the system when it is subjected to a disturbance. For a realistic transient analysis and simulation, all generators are modeled by controller's parameters such as prime mover (turbine), governors and voltage controller parameters. Simulations were carried out on the IEEE 9-Bus Test System considering load injection on the system. The data collected from transient simulation are then used as inputs to the ANN while the frequency response of the systems as the ANN output. The lavernberg- Marquardt optimization utilizing very fast propagation algorithm has been adopted for training feed-forward Neural-Network. To verify the effectiveness of the proposed application of ANN method, its performance is compared with the actual value from transient simulation. The ANN provides promising results in terms of error estimation, accuracy and computation time.