In this paper a comparative fault location approach is presented for a 400 km HVAC transmission line under different fault conditions. The proposed study evaluates three fault location techniques: direct signal based estimation, Discrete Wavelet Transform (DWT) processed estimation and Long Short Term Memory (LSTM) deep learning based estimation. Four major fault cases are considered, including single line to ground(LG), double line ground (LLG) line to line (LL), and three phase faults (LLL), applied at different locations along the transmission line. The direct signal method estimates the fault location directly from the measured waveform, whereas the DWT based method extracts high frequency transient components generated during fault inception. The LSTM based method further improves the estimation accuracy by learning nonlinear fault patterns from the measured and processed signal behavior. The simulation results show that the direct signal method provides acceptable fault location estimation but suffers from higher deviation due to unprocessed transients and signal disturbance. The DWT method significantly improves the location accuracy by enhancing the fault-generated travelling-wave information. However, the LSTM deep learning method achieves the best performance among all compared methods. The mean absolute error is reduced from 1.125 km for the direct signal method to 0.20850 km for the DWT method and further reduced to only 0.03227 km using the LSTM method. Similarly, the average accuracy improves from 98.69762% for the direct method to 99.87012% for DWT and reaches 99.97766% for the LSTM based method. The obtained results confirm that the LSTM model provides superior accuracy, lower error variation, and better robustness for different fault types and locations. Therefore, the proposed LSTM based fault location approach is suitable for accurate and reliable transmission-line protection studies and can be extended for real-time smart grid protection applications