Abstract
Industry 4.0 adoption in the oil and gas sector has introduced numerous AI-driven decision-making tools. However, data-driven modeling for oil platform simulators using historical production data is still relatively unexplored. Floating platforms, such as FPSOs, play a critical role in oil production, particularly in Brazil. Electrical power systems design for these platforms typically employs conservative risk measures, and historical data to determine electrical equipment factors is limited. Therefore, this paper presents the FPSO Power Demand Analytics (FPDA) tool for estimating electrical equipment load on oil and gas platforms, aiding daily production improvements, and supporting the design of new FPSOs based on operational history insights. FPDA is an open-source Python tool compatible with multiple platforms and utilizes historical data analytics and machine learning for modeling. Users can generate electrical diagrams and define loads through a configurable power flow model. FPDA consists of three distinct modules: Knowledge Discovery in Databases (KDD) for preliminary data analysis, Machine Learning (ML) for model training and demand forecasting, and Power Flow (PF) for aggregating demand projections and estimating generator/transformer demand. The algorithms were assessed using data from three different FPSOs with varying sampling periods and temporal horizons. Seven ML models were trained per device to predict active power demand based on process variables, and their performance was evaluated using two test scenarios and various statistical measures. The ML algorithms offered precise projections with minimal computational time, while the power flow module delivered consistent results requiring only a few seconds for simulation, making it suitable for planning environments.