The Analysis of Deviations on Measured Volumes Between OSBRA Pipeline Tank Farms and Its Customers

The Sa˜o Paulo–Brasi´lia pipeline (OSBRA) is a collection of installations built to transport, store and deliver Diesel oil, gasoline and Liquefied Petroleum Gas (LPG), from the Refinaria do Planalto (REPLAN) to several distribution companies located in the Brazilian Midwestern region. The volumes of these fuels are usually measured before deliverance at two serial measurement points: at the OSBRA tank farm installations and at the customer tank farm installations. At the OSBRA tank farms, the volume of each delivered batch product is usually measured by turbine type custody Measurement Station, called EMED. On the other side, at the customer tank farm, the received volume is usually measured by tank gauge systems or, in a few cases, by the customer EMED. The procedure used by PETROBRAS in Brazil is to compute the differences between these measurements and to compare these differences each month with a pre-established data range. If the differences computed are lesser than that benchmark values, the measurement is considered good. Otherwise, inspections of the measurement installations must be done, looking for the causes which deviates the measurement differences out of the desired range. That pre-established data range was built over historical data and using statistical tools. In this paper it is proposed to use statistical tools to analyze the behavior of the already existing historical data differences, rather than see only the differences magnitudes. These tools can give more information about each pair of measurement system, like the existence of systematic errors, the variability of each individual system and even if an observed difference greater than the pre-established data has real significance. These tools can be used to analyze other measurement system, like to verify the accuracy of operational measurement devices and to check if the mass conservation among the entries and withdraws of a system is being achieved. Furthermore, a real data analyze between two tank-farms is presented.