Abstract:
|
One of the major challenges for energy companies is to adapt their process plants to
the continuous improvements of available technologies, so as to make their old plants
as competitive and cost-efficient as the new ones. Along these lines, process stream
data was recently collected for analysing opportunities for improved process
integration of the Hydrocracker Unit of a major oil refinery located in Lysekil on the
West Coast of Sweden. However, inconsistencies in the process data measurements,
e.g. energy balances that do not add up, made the study cumbersome. For analysing
heat exchanger networks it is essential to establish sets of process data with balanced
heat balances for the existing heat exchangers. The aim of this thesis project was to
develop a computer-based solution for systematic analysis, identification and
correction of the “raw” data obtained from process data measurements in order to
acquire such a consistent set of data.
With this purpose, a tool for Data Reconciliation and Gross Error Detection for
process stream data was developed using Visual Basic in Microsoft Excel. The tool is
based on the Modified Iterative Measurement Test. A second tool, which is easier for
handle large data sets and especially designed for networks with non-linear constraints
was also developed. This second tool is only able to solve Data Reconciliation
problems, so it is targeted for sets of data where there are exclusively random errors.
Both developed tools were used to analyse the data set collected from the refinery’s
Hydrocracker Unit with the purpose of generating a consistent set of data with
balanced heat exchangers. The solution proposed is an energy balanced network,
where from the 32 temperature measurements, all the reconciled values, except two,
are within the specified bounds indicated. The two reconciled temperatures outside
the bounds are the ones in which the presence of a gross error has been confirmed.
Since this is a preliminary study, the solution proposed must be taken as a
recommendation. |