Paper PowerGen 2013, ID T6S4P3
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Real-time and Intraday Optimization of
Multiple Power Generation Units
Dr. Rüdiger Franke
Hansjürgen Schönung
Marcel Blaumann Dr. Alexander Frick
Stephan Kautsch
ABB AG, Germany
Paper PowerGen 2013
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Introduction 2 / 15
1 INTRODUCTION ...................................................................................... 3
2 BRINGING MATHEMATICAL OPTIMIZATION TO REAL-TIME CONTROL ....................................................................................................... 3
3 INTERNAL OPTIMIZATION OF LARGE MULTI-UNIT PLANTS ... 6
4 COMBINED HEAT AND POWER PRODUCTION ............................... 9
5 REAL-TIME OPTIMIZATION OF POOLS OF BIOGAS PLANTS... 11
6 RENEWABLE SUPPLY MANAGEMENT ............................................ 12
7 INTRADAY OPTIMIZATION OF MUNICIPAL POWER ................. 13
8 CONCLUSIONS ....................................................................................... 15
Paper PowerGen 2013
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Introduction 3 / 15
1 Introduction
Facing increasing penetration of renewable energy, new opportunities arise for the trading of
power. Particularly the increased participation in grid services and intraday trading becomes
crucial for the economical result. This raises the requirement for increased flexibility on the
power generation side:
• Frequent updates during the day complement traditional day-ahead plans for
conventional power generation.
• Combined heat and power generation is changed from heat driven to electricity driven,
exploiting storage capacities on the heat side.
• The controllability of renewable generation units is increased.
This paper introduces a new optimization method that is placed in the real-time control of
power plants. This enables the pooling of individual power generation units and their
management like one large plant. The real-time optimization receives overall set points and
distributes them to each individual power generation unit, considering actual efficiencies,
process constraints and temporary limitations. The introduced hierarchy reduces overall
complexity and increases the flexibility. Moreover, it solves the power generation task at the
best point for the considered units.
The paper introduces the optimization method and lists five different application examples.
2 Bringing mathematical optimization to real-time control
Traditional unit commitment optimization focuses on day-ahead plans for power production
and trading. Well-established optimization tools find on top of plant information management
systems and interface with trading systems.
Paper PowerGen 2013
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Bringing mathematical optimization to real-time control 4 / 15
Unit 2DCS 2
Unit 1DCS 1
Unit …Other DCS/PLC
Engineering & Maintenance
Traditional Optimization
FleetManagement
PIMS
Scanner
Trading, Accounting
Figure 1: Classical system layout for plant optimization, e.g. unit commitment
Figure 1 shows the evolved classical system layout. Multiple production units are connected
through an automation network. A Scanner reads real-time data from the automation network
and provides it for the Plant Information Management Systems (PIMS). The PIMS provides a
process historian database and analysis functionalities. This primarily improved the
engineering & maintenance. Optimization algorithms traditionally run on top of the PIMS
system. Typically, human operators use such traditional optimization tools interactively.
Four new trends arose during the previous years:
1. The number of power production units significantly increases with the use of
renewable energy
2. The power production needs to be re-planned frequently during a day, in order to
account for fluctuations
Paper PowerGen 2013
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Bringing mathematical optimization to real-time control 5 / 15
3. The required optimization cycle times reduces from daily planning cycles down to
seconds, e.g. for pooling of secondary frequency control
4. The role of human operators changes from being part of the loop to supervision
Moreover, the mathematical optimization technology has maturated during the last years.
Modern optimization solvers treat large linear optimization programs, including also integers,
reliably in fractions of seconds. Even quadratic and nonlinear optimization algorithms
perform well in model predictive control applications in real-time.
The new trends require a shift from interactive use of optimization tools to automated
optimization in the real-time control system.
Unit 2DCS 2
Unit 1DCS 1
Unit …Other DCS/PLC
Scanner
Online OptimizationPIMS
Engineering & Maintenance
FleetManagement
Trading, Accounting
Figure 2: New system layout with online optimization
Paper PowerGen 2013
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Internal Optimization of large multi-unit plants 6 / 15
Figure 2 shows the new system layout. The optimization now directly accesses the automation
network through the Scanner. It only communicates data exchanged with the fleet
management, like planning data, through the PIMS.
This system layout provides for significantly improved reliability. For instance, the
optimization can still run if the PIMS goes offline – just updates of planning data would not
work anymore in this case. The direct communication also allows for significantly reduced
optimization cycle times down to about a second.
All communication between the Scanner, the PIMS and the Optimization goes through
TCP/IP sockets. This provides a lot of flexibility for different implementations, depending on
specific needs. In the simplest case, one server PC can run all functions (Scanner, Optimizer,
PIMS). Two redundant server PCs provide for higher availability. Moreover, the Scanner and
the PIMS can run on separate server PCs, in order to increase the throughput of data. Finally
yet importantly, the Optimization can share the use of already existing functions, like an
already installed Scanner or PIMS or of already existing hardware, e.g. for running the
Scanner on servers already existing in the plant control or SCADA system.
Traditionally a private automation network provides the communication with the power
generation units. Alternatively, virtual private networks (VPN) using the Internet or GPRS
enable the communication to small, distributed power generation units.
The new system layout particularly suits for the real-time optimization of actual plant set
points and for the intraday optimization of plant schedules. The following sections give an
overview about important use cases and exemplary installations.
3 Internal Optimization of large multi-unit plants
Two large power plants with an installed capacity of 6x500MW and
2x500MW/1x890MW/1x600MW, respectively, run a real-time optimization for plant set
points and secondary frequency control. The idea is to pool and optimize multiple units per
plant locally. At the same time, the optimization system automates the communication from
Paper PowerGen 2013
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Internal Optimization of large multi-unit plants 7 / 15
the load dispatcher to the control systems of the power generation units, in order to enable the
reaction on increasingly frequent updates, see Figure 3.
Figure 3: Principle of the Internal Optimization for a 6x500MW power plant
The optimization model covers efficiency curves and own consumption of the plant units,
besides many constraints ensuring the appropriate provision of primary frequency control,
secondary frequency control and required load ramps. The Modelica technology provides for
graphical structuring of the model equations and generation of efficient executable code, see
Figure 4.
Paper PowerGen 2013
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Internal Optimization of large multi-unit plants 8 / 15
Figure 4: Graphical formulation of the optimization model using Modelica
The graphical model editor exports compiled executable models to the online optimization.
In this case, two online optimizations are running in parallel in each plant. One optimization
adjusts the set points for the provision of secondary frequency control automatically in real-
time. The second optimization covers the plant schedule and proposes new set points. The
operator can influence the optimization by adjusting min/max constraints per plant unit. Once
acknowledged, the set points are transferred to all units automatically.
Standard operator graphics hides the optimization details from the operators. For instance,
standard faceplates serve as input for max and min bounds for feasible operating points. The
online optimization automatically determines the overall best point within the ranges specified
by human operators.
Paper PowerGen 2013
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Combined Heat and Power production 9 / 15
Figure 5: Operator graphics visualizing the status of all plant units together with optimiza-
tion results and linking to faceplates for operator inputs
4 Combined Heat and Power production
Combined heat and power production is key to improving energy efficiency. The physical in-
teraction between different components on heat side and on electricity side results in complex
constraints for the plant operation.
Paper PowerGen 2013
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Combined Heat and Power production 10 / 15
Figure 6: Plant model and Operator screen for combined heat and power production
Figure 6 shows the application to real-time optimization of plant set points. The plant model,
shown on the left hand side, considers:
• Efficiency curve and own consumption for each relevant plant component
• Couplings through steam headers
Appropriate assumptions need to be made, in order to enable online optimization in real-time.
The optimization builds on a well-tuned base control system for each unit. This means in par-
ticular that optimal coordination of multiple plant aggregates builds on well-controlled header
pressures and temperatures.
Real-time optimization treats a mathematical optimization program based on the physical
plant model. The optimization constraints and the objective cover:
• Demands for steam and electricity
• Externally given set points for the recovery boilers
• Constraints per plant component (e.g. min and max load)
• Fuel costs for coal and biomass
The real-time optimization of set points integrates with the regular plant operation in a
straightforward way. Standard operator screens visualize optimized set points besides actual
values. The base control systems take over optimization results either automatically or after
confirmation by human operators.
Paper PowerGen 2013
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Real-time optimization of pools of biogas plants 11 / 15
5 Real-time optimization of pools of biogas plants
Comparable fast load ramps make renewable generation units well suited for the provision of
grid services. This is why the direct marketing of renewable power is becoming increasingly
attractive. Individual renewable power plants are too small though. The pooling of small
renewable power plants is required, in order to achieve an overall capacity that is sufficient
for participation in the electricity market.
Online optimization manages pools of biogas plants for the provision of secondary frequency
control. The real-time optimization implemented on a central redundant server receives set
points from load dispatchers and distributes them to pooled plants. The IEC 60870-5-104
protocol is used in Virtual Private Networks (VPN) for the communication over the Internet.
Figure 7 gives an overview of the system structure.
RedundantProcess Database Real-Time Optimization IEC 60870-5-104 ScannerOperation & Management
Operations Client Engineering ClientImport / Export for
operational management
Service Client
Up to 500 production units,typically 500 kW per unit
4 GridOperators
Figure 7: Pooling of biogas plants for secondary frequency control
Paper PowerGen 2013
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Renewable supply management 12 / 15
6 Renewable supply management
Many renewable production units, like wind and solar, do not participate in the direct trading.
Instead, they are supplying an increasing amount of unmanaged power to the grid. This can
harm the grid operation and stability. Renewable supply management controls the amount of
renewable power fed to particular distribution networks. Online optimization offers proven
technology to break down overall set points to individual power generation units, fulfilling
legal requirements (like EEG Einspeisemanagement) and plant constraints. As a result, the
renewable generation units receive limits for their supply to the grid (see Figure 8).
Figure 8: Overall system layout for renewable supply management
(EEG-Einspeisemanagement)
Paper PowerGen 2013
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Intraday Optimization of municipal power 13 / 15
7 Intraday Optimization of municipal power
Storage capacities enable the temporal decoupling of power production and consumption.
Typical storages are latent heat stores of combined heat and power production and pump
stores. Moreover, the emerging electric mobility offers a huge potential for future storage
capacities.
From a control point of view, storages require a shift from real-time optimization of actual set
points to predictive planning of load trajectories. The planning needs to consider a multitude
of constraints. For instance, heat storages primarily need to ensure the supply of heat and car
batteries primarily have to power the engine.
Intraday optimization enables the optimal use of power generation units and storage
capacities, considering constraints and maximizing the economic result. It reacts on new
conditions by re-planning the power production. This maintains the overall balance, in order
to avoid the purchase of expensive regulating power from the transmission net.
Figure 9 shows the optimization model of multiple producers, consumers and storages that
form one regional balancing zone. Figure 10 shows exemplary intraday optimization results.
Triggered by an update of weather forecast data, a new prediction for the wind power arrives.
The wind power (thick green line) decreases after 12:00 and increases after 14:00, compared
to the original day-ahead plan (thin green line). The intraday optimization can avoid a
misbalance by re-planning the combined electricity and heat production (red lines). The heat
storage buffers excess heat during overproduction of electricity and releases it during
underproduction of electricity (black line).
Paper PowerGen 2013
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Intraday Optimization of municipal power 14 / 15
Figure 9: Graphical formulation of an intraday optimization model
Figure 10: Exemplary results of intraday optimization
Paper PowerGen 2013
Real-time and Intraday Optimization of Multiple Power Generation Units
© PSP-I42-, 2013-05-28
Conclusions 15 / 15
8 Conclusions
Traditional optimization tools run on top of process databases. They are used interactively and
the optimization results are typically transferred manually to the control systems of the opti-
mized power production units (e.g. via email or ftp to the plant and manual input into the con-
trol system). These tools and procedures run into limitations when facing an increasing num-
ber of small renewable power generation units and/or increasingly frequent re-planning cy-
cles.
Online optimization connected directly to automation networks enables the flexible reaction
on frequently changing production tasks. The plant control systems either receive optimized
set points automatically. Alternatively, standard operator graphics may present the optimiza-
tion results to human operators for acknowledgement. In both cases, the online optimization
not only increases flexibility, but also maintains the overall power production at the economic
best point.
Real-time optimization treats current plant set points. The solution time is short (down to frac-
tions of a second) thanks to the availability of mature numerical optimization solvers and
powerful computing hardware. This is why the optimization can even be placed into the
communication and sub-division of set points, e.g. for secondary frequency control or for re-
newable supply management. Moreover, this allows an interactive use. For instance, a human
operator can adjust bounds or constraints using regular process graphics and immediately see
the effect on the optimization results.
Intraday optimization not only covers current set points, but also predicts into the future. This
is required in the case of available storages, like heat buffers, pump storages or batteries. This
decouples the production from the consumption of electricity up to the extent of available
storages.
ABB has implemented the real-time and intraday optimization in OPTIMAX® PowerFit, now
basing on the online platform ABB Dynamic Optimization. The new optimization method has
proven successful in a number of different application projects.