This project aims to develop robust AI algorithms using various AI trainers offered by MATLAB. Scaled Conjugate Gradient, is utilized as the best option to train and test our ANN on a testing data set. The ANN learns from the training dataset and utilizes learning patterns to detect and classify faults by identifying abnormal voltage and current deviations from balanced conditions. The training dataset comprises all possible permutations of injectable faults in a three-phase power system (eleven in total). A MATLAB Simulink model is designed to create our testing and training datasets. It features two three-phase power sources: one to supply the voltage and another with a - 30degree offset to simulate a load. The transmission lines are connected to 11 fault injectors with each one consecutively injecting a 0.003s long fault at an interval of 0.10s. The scope records the Voltage and Current data values for every line which is then exported into a spreadsheet format. This file is fed into our MATLAB code to develop and train the AI model. The discrete sample time was set at 0.003s to keep the recorded data points under 5,000 to maximize efficiency and minimize overlapping. Unexpected failures in electrical power transmission lines, such as short circuits, can lead to severe economic losses and reduce system reliability. Quick identification and categorization of these faults are essential for safety. Traditional methods rely on human feature extraction and prior knowledge. This project develops a three-phase simulation model with fault injectors to create datasets for training and testing AI algorithms, aiming to improve fault detection and classification. Artificial Neural networks (ANNs) will be employed to analyze and classify fault data with high accuracy