@inproceedings{scholars17267,
         journal = {2022 IEEE International Conference on Computing, ICOCO 2022},
       publisher = {Institute of Electrical and Electronics Engineers Inc.},
           title = {Hydrocarbon Flow Metering Prediction using MLP and LSTM Neural Networks},
           pages = {419--424},
            note = {cited By 0; Conference of 2022 IEEE International Conference on Computing, ICOCO 2022 ; Conference Date: 14 November 2022 Through 16 November 2022; Conference Code:186566},
            year = {2022},
             doi = {10.1109/ICOCO56118.2022.10031800},
             url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-85148453365&doi=10.1109\%2fICOCO56118.2022.10031800&partnerID=40&md5=f094f3eded94d49b66fd70b1c9c3d239},
        keywords = {Flow measurement; Flowmeters; Hydrocarbons; Learning systems; Time series, Flow measurement;; Hydrocarbon metering; Input variables; Machine learning, long short-term memory, multilayer perceptron, flow measurement; Machine-learning; Multilayers perceptrons; Neural-networks, Long short-term memory},
        abstract = {Accuracy and integrity of metering data are required for commercial purposes and as a commitment between suppliers and customers. One of the important variables to monitor hydrocarbon metering is volumetric flow measurements, where measurement errors may significantly impact the operator. This project aims to develop a neural network-based algorithm to predict flow measurement patterns using onshore metering data. After pre-processing and statistical analysis, the metering data is used to train models using Multilayer Perceptron (MLP) and Long Short-Term Memory (LSTM) networks. Both models were trained and tested with different combination of input variables and several hyperparametric settings. The best MLP model was trained using Pressure, Temperature and 15 Time-shifted Flow as input variables, yielding a Mean Absolute Percentage Error (MAPE) of 0.96. Furthermore, two versions of LSTM models-Time-Series LSTM and Single-layer LSTM - are also trained and tested, giving satisfactory performance with Flow variable as the input. Time-Series LSTM model has a better performance with an MAPE of 0.47. {\^A}{\copyright} 2022 IEEE.},
          author = {Hong, D. G. K. and Hisham, S. B. and Yahya, N.},
            isbn = {9781665489966}
}