Real-time optimisation of natural gas pipelines intransient conditions is a challenging problem. Manypipeline systems are, however, only mildly non-linear even in large transients such as compressorstation (CS) shutdown or startup. A dynamic,receding horizon optimisation problem is defined,where the free response prediction of the pipelineis obtained from a pipeline simulator and theoptimal values of the decision variables areobtained solving an approximate QuadraticProgramming (QP)problem where the cost function isthe energy consumption of the CSs. The problem isextended with discrete decision variables, theshutdown/start-up commands of CSs. A Mixed LogicalDynamical (MLD) system is defined, but the resultingMixed Integer QP problem is shown to be very high-dimensional. Instead, a sequence of QP problems isdefined resulting in an optimisation problem withconsiderably smaller dimension.The receding horizon optimisation is tested in asimulation environment and comparison with data froma true natural gas pipeline shows 5 to 8 % savingsin compressor energy consumption.
Hans AaltonHans Aalto: Doctor of technology in Automation Technology atHelsinki University of Technology. Presently works as VicePresident, Automation Engineering at Neste Jacobs Oy, Porvoo,Finland
Über den Autor
Hans Aalto: Doctor of technology in Automation Technology at
Helsinki University of Technology. Presently works as Vice
President, Automation Engineering at Neste Jacobs Oy, Porvoo,
Finland
Klappentext
Real-time optimisation of natural gas pipelines in
transient conditions is a challenging problem. Many
pipeline systems are, however, only mildly non-
linear even in large transients such as compressor
station (CS) shutdown or startup. A dynamic,
receding horizon optimisation problem is defined,
where the free response prediction of the pipeline
is obtained from a pipeline simulator and the
optimal values of the decision variables are
obtained solving an approximate Quadratic
Programming (QP)problem where the cost function is
the energy consumption of the CSs. The problem is
extended with discrete decision variables, the
shutdown/start-up commands of CSs. A Mixed Logical
Dynamical (MLD) system is defined, but the resulting
Mixed Integer QP problem is shown to be very high-
dimensional. Instead, a sequence of QP problems is
defined resulting in an optimisation problem with
considerably smaller dimension.
The receding horizon optimisation is tested in a
simulation environment and comparison with data from
a true natural gas pipeline shows 5 to 8 % savings
in compressor energy consumption.
