Pipelines are a crucial component of the global transporting of petroleum products. Nonetheless, they are susceptible to leakage caused by aging and corrosion. It is therefore important to identify the leaks as soon as possible and have them repaired to avert serious ramifications. In this paper, a reliable robust leak detection system (LDS) that employs neural networks (NNs) in the analysis of pressure fluctuations and detection of leaks is proposed. The approach incorporates a novel data-set generation stage, preprocessing of the data, extraction of input features and design of the neural network to identify and localize leaks. Pressure drop ratios or changes (dP) are computed to check for deviations and establish a baseline for leak detection. Matlab and Simulink are employed to generate reliable operational scenarios with which real-life-like situations for creating the model are acquired. By optimizing the NN architecture and hyperparameters, nonlinear relationships in pipeline data were effectively modeled. Through careful preprocessing and logical decision-making processes, false alarms were significantly reduced, enhancing the reliability of the system in practical applications. The system demonstrated real-time monitoring with rapid leak identification. Adaptability to varying noise levels and operational conditions highlighted the robustness of the model. In testing, the proposed system demonstrated consistently low errors, with Mean Squared Error (MSE) reaching approximately 1.7e-05 for leak localization, ensuring robust real-time pipeline leak detection under diverse flow conditions.