Subsea Oil Spill Risk Management based on Sensor Networks

This thesis consists of the evaluation of sensor-based risk management against oil spills using an underwater distributed sensor network. The work starts by highlighting the importance of having a performing leak detection system both from an environmental, safety and economic point of view. The case study is the Goliat FPSO in the Barents Sea which has to meet requirements dictated by Norwegian authorities to prevent oil spills. The modeled network is made of passive acoustic sensors monitoring the subsea manifolds. These sensors send their local 1-bit decision to a Fusion Center which takes a global decision on whether the leakage is occurring. This work evaluates how the choice of adapted Fusion Rules (Counting Rule and Weighted Fusion Rule) can affect the performances of the leak detection system in its current geometry. It will also be discussed how different thresholds, selected for a specific FR or sensor test, can change the system performance. The detection methods are based on statistical signal processing adapted to fit this application within the Oil&Gas field. The work also proposes some new leak localization methods developed so they can be coupled with the proposed leak detection methods, giving a coherent set of operations that the sensors and the FC must perform. Performances of detection techniques are assessed balancing the need for high values of True Positive Rate and Precision and low values of False Positive Rate using indexes based both on the ROC curve (like the Youden's Index) and on the PR curve (the F-scores). Whereas, performances of localization techniques will be assessed on their ability to localize the spill in the shortest time; if this is not possible, parameters like the difference between the estimated and the real leak position will be considered. Finally, some tests are carried out applying the different sets of proposed methods.