Cost reduction has always
been an important focus for industrial plants to improve their bottom lines. A
well-proven method to achieve and continuously sustain these improvements has
been the application of advanced process control (APC) technologies—more specifically, multivariable
predictive control (MPC) technology.
In recent years, energy
efficiency improvements have seen increased prioritization for sustainable and ongoing
operations of these same industrial plants. Although the application of MPC
technology has traditionally focused on the main production units, as the penetration
of the technology has progressed, industry has been applying this technology
increasingly on power generation, boilers and other utility systems.
As operator of the
North West Shelf Project, Woodside Energy Ltd. (Woodside) has used MPC
technology to optimize its LNG production trains at the Karratha Gas Plant
(KGP) in Western Australia for more than two decades (FIG. 1). Woodside has
collaborated with the author’s company, an Australia-based APC and optimization
consulting company, to grow the portfolio of MPC applications and execute many
revamp projects of older applications at the site over this period.
After an initial
investigation into how this technology could be utilized in the power
generation area, the author’s company was engaged to design, implement and
commission a solution that seeks to optimize the fuel efficiency of the entire
fleet of gas turbine generators. This case study describes the project applying
MPC technology as the solution to this optimization challenge.
Process description. KGP
utilizes ten gas turbine generators (GTGs) to supply power to the five-train
LNG production complex. An overview of the
KGP electrical power system is shown in FIG. 2. Six of these are driven by Frame 5 gas
turbines, while the other four are driven by General Electric LM6000 gas
turbines. The GTGs are arranged to supply the generated power to three
distribution networks, with interconnectors to ensure power can be transferred
from any network with excess generation capacity to another with a generation
shortfall.
The maximum power
generation capacity of these GTGs is a function of the prevailing ambient
conditions, pre-dominantly ambient temperature and wind effects. The maximum
power delivered by a gas turbine reduces as the inlet air temperature
increases, and this variation can be quite significant in the northwest of
Australia, especially in summer.
One key objective for
any power generation system is to ensure that sufficient power generation
reserves are available in the event of a GTG trip to prevent cascaded trip
events. This is typically referred to as “spinning reserve,” and is determined
by the spare capacity of each GTG. This becomes a moving target for the overall
system optimization.
Another complication
faced by most power generation systems is managing—or rather, preventing—under-frequency events resulting from
excessively large load demands on the online GTGs: this is especially
exacerbated when another GTG has tripped. To prevent this, the instantaneous
change in load demand is limited to a value (referred to as “load acceptance”)
that the system can handle while maintaining the system frequency. The result
of this approach is that actual (or true) power generation reserve is typically
somewhat less than the absolute spinning reserve. This does not necessarily
impact the management or control of the power generation system. It is,
however, something that should be acknowledged in the design of any automated optimization
of the system.
Operational challenges. The
location of the processing plant means that significant changes in the ambient
conditions not only affect the overall power generation capacity, but also has
a significant impact on the power demand from the production process. This is
not always a linear or predictable relationship.
The load demand on each
LM6000 GTG is set by the panel operator and adjusted on an ongoing basis to
manage the varying power demand resulting from changing process conditions and
objectives. A power management system (PMS) is used to manage the remainder of
the power distribution system by manipulating the load demand on the
Frame 5 GTGs. This includes generating the balance of the power demand,
subject to the system frequency and voltage.
The LM6000 turbines are
more fuel efficient than the Frame 5 turbines and have a greater power-producing
capability. The general objective is to maximize the load on the LM6000 GTGs to
leverage these improved efficiencies and deliver a net reduction in fuel gas
consumption and carbon dioxide (CO2) emissions across the 10 GTGs
operated at KGP.
Unfortunately, optimal
operation is not as simple as setting the load demand on each LM6000 GTG at the
maximum allowed, or just maximizing the load demand on the LM6000 GTGs. The
optimal LM6000 load is a dynamic target that is constrained by a variety of
factors, such as:
Technical considerations. The
main challenge was to develop a system that can continuously adjust the loads
on the LM6000 GTGs to adapt to the continually changing playing field, while
honoring the various process and equipment constraints.
The objectives for the
system/application were to:
Some of the options
considered and discussed during the initial design phase of the project were:
Existing PMS. One option considered
was to add the functionality to the existing PMS, hosted in a high-speed PLC
system. Numerous risks were identified that would put the success of the
project at risk:
Bespoke solution. Another option briefly
considered was to develop a bespoke (or custom) application to interface with
the DCS and the existing PMS. This idea was discounted as numerous risks were
identified that would put the success of the project at risk:
APC application. It was decided that by
leveraging Woodside’s existing
investment, support and acceptance of MPC technology at the KGP site, coupled
with the author’s company’s experience with the specific MPC technology, the
proposed solution was to develop an MPC application to optimize the fuel
efficiency of the entire fleet of gas turbine generators. This choice of
solution was based on:
An important design
decision was to retain the existing PMS to manage the load demands on the
Frame 5 GTGs, and therefore the system frequency and voltage. The typical
execution frequency of an MPC application is not fast enough (typically 15 sec–60 sec) to manage the
high frequency demanded by power systems (typically several times a second).
To maximize the LM6000
load demand, an MPC solution was developed that could replicate what KGP
operators previously did manually, while continually monitoring and considering
the various constraints at a much higher frequency than is realistic for a
human. The MPC application also enables this efficiency opportunity to be
chased 24/7 in concert with improved operating envelope adherence, as the MPC
application does not need to attend to other issues or to take breaks.
Project background. The
development of a new power generation MPC application was begun in early 2021
while navigating the various COVID-19 challenges and restrictions. The team was
fortunate enough to be able to complete the plant testing and initial
commissioning preparations onsite in March 2021.
Although many of the
technical steps of the project could be completed remotely, this underlined the
intangible benefits of being onsite to discover subtle clues and valuable bits
of information that can have a significant impact on the success and outcome of
an APC project. Although process control related projects are highly technical,
the human interaction with the panel operator (the ultimate customer) should
never be underestimated. Being onsite provided valuable opportunities to
discuss the solution with the operators, gather more information around the
usability features, and start to get them excited about the new tool they were
about to receive.
The development of the MPC
application was completed following the plant test period, and the application
was commissioned onsite in May 2021. Following the initial commissioning
period, ad hoc technical support to the site-based engineer and panel operators
was provided remotely (due to COVID-19 restrictions).
Project results. The
outcome was a robust and well-accepted MPC application that:
Examples of the results
are shown in FIGS. 3
and 4. HP
ACKNOWLEDGEMENT
The authors would like
to thank the North West Shelf Project joint venture partners for their
continued support of APC projects. They would also like to thank Woodside’s central
control room panel operators for their valuable input, buy-in and support
during the development of the APC application.
GEORGE LA GRANGE is a Principal
Consultant at Greenfern Dynamics. He has 27 yr of experience with APC in various
industries, including oil refining, LNG production, oil and gas, chemical
production and minerals processing. He earned a BE degree in chemical
engineering from the University of Pretoria and is a chartered professional
member of Engineers Australia.
DAVID ALLAN is a Process Control Engineer at Woodside Energy Ltd. He has 10 yr of experience in the oil and gas industry and 3 yr of experience as a process control engineer. He earned a BE degree in chemical and process engineering from the University of Western Australia.