System ATC estimation using Distributed Computing

Accurate and instant information about available transfer capability (ATC) is important in competitive electric power markets for providing non discriminatory open access transmission facility. ATC information conveys how much power can be transmitted through the power network over and above already committed uses. This paper presents, a Levenberg-Marquardt algorithm based neural network (LMANN) approach for fast and accurate estimation of system ATC with Distributed Computing. Effective input variables for neural network have been selected using entropy concept. For accurate estimation of ATC various contingency clusters were formed using Euclidean distance based clustering technique. For each contingency cluster separate LMANNs have been developed. All the proposed LMANNs have been trained and tested under distributed computing environment and a significant speed up in the training is obtained. The proposed approach has been examined on IEEE - 30 bus test system and found significantly efficient.