Power system is a large system which is very sensitive to damage and failure especially during transmission line and power generating unit outage that is why contingency analysis is performed to investigate the power system security in case of outage to ensure that the system is working in safe limits. In an interconnected power system unscheduled line outage results safe operating limit violation effect on the rest of the network elements. Farther more it leads the system to cascaded outage, then to total blackout. In Ethiopia a total of 49 blackouts have been recorded from January 28th 2013 to mid of May 2016. One of major initiating event for the blackouts were high voltage transmission line fault. Cascading events from initiating event to the total blackout took several minutes. This implies that the system needs fast contingency analysis tool which asses the severity level of the outage thereby, enable to take corrective action accordingly. It is difficult to perform fast contingency analysis due to low accuracy in case of approximate methods and high computational time requirement to simulate all contingencies in case of full AC power flow method. This thesis proposed fast contingency analysis based on line outage performance index (PI) prediction using Artificial Neural Network(ANN) in MATLAB. The first part of this thesis covers about 132kv, 230kv and 400kv line outage contingency ranking of Ethiopian power system based on PI & then detail contingency analysis for the first top ranked severe line outage from each voltage level stated above using NR load flow method. As a results among each 132kv, 230kv & 400kv line outage; Dire Dawa III to Dire Dawa II, Tekeze to Axum & Gilgel GibeII to Wolayata SodoII were found to be the severe top ranked outage from their respective voltage level. As remedial action upgrading single circuit into double circuit, removing appropriate shunt reactors together with OCP and OCP together with proper line switching off has been suggested accordingly. As a results their impacts have been alleviated. The second part of the thesis is predicting 400kv line outage PI using RBFNN. It has been first trained by taking bus power injection along with line outage number as input and the corresponding PI as target output. The datasets were determined by AC load flow method for different loading conditions. RBFNN model performance analysis result shows that MSE, MAPE & R values between targets and predicted PI are 5.7783x10 -6 , 3.9% & 0.99997 v respectively. Therefore, RBFNN model effectively predict line outage PI with reasonable accuracy.