Multiple Output Radial Basis Function Neural Network with Reduced Input Features for On-line Estimation of Available Transfer Capability
Abstract
In the deregulated power system, the Independent System Operator (ISO) has to update the value of Available Transfer Capability (ATC) on Open Access Same Time Information System (OASIS) for the secure bilateral/multilateral transaction planning. The off-line methods for calculating ATC requires large computation time and or not suitable for online estimation, hence the on-line updating of ATC requires an accurate method with lesser computation time. In this paper, Radial Basis Function Neural Network (RBFNN) has been proposed for on-line ATC estimation for both bilateral and multilateral transactions under normal and contingency condition. Multiple and Multi Neural Network is developed and their performance is analyzed. The training data for Neural Network is generated using Repeated Power Flow Algorithm (RPF). One of the challenges in the development of Neural Network in power system is the selection of suitable input variables because power system contains thousands of variables. For this a straight forward and quick procedure called Sequential Feature Selection (SFS) is used to extract the most influenced variables as features from a large set of variables. Simulation work is performed on standard IEEE 24 bus Reliability Test System (RTS) and the feasibility of implementation of the proposed Neural Network for on-line ATC evaluation is discussed. The Neural Network results are compared with RPF. Test result shows the effectiveness of the neural network approach for on –line estimation of ATC.