02-02-2013, 11:08 AM
Coordination of a multiple link HVDC system using local communications based distributed model predictive control
1Coordination of a multiple l.pdf (Size: 191.36 KB / Downloads: 22)
Abstract:
As the complexity of power networks increases, the installation of devices such as
High Voltage Direct Current links (HVDC) and Flexible AC Transmission Systems (FACTS),
and the use of advanced control techniques, can be used to improve network stability. Model
Predictive Control (MPC) is an example of such an advanced control technique. However, it
is often impractical to implement this technique in a centralised manner, as often the problem
can be too computationally complex or several independent controllers may be responsible
for different subsystems. Distributed approaches use communication between a number of
controllers to approximate control of a centralised system. In this paper it is proposed to use
distributed MPC for controlling a multiple link HVDC system using local communications only.
INTRODUCTION
Power networks are large, complex, highly interconnected
systems. As increasing demands are imposed on power
networks more advanced control techniques are needed
in order to maintain network stability. HVDC lines allow
for the efficient transmission of large quantities of power
over long distances. Moreover, due to their high level of
controllability these devices can improve transient stability
and power system damping (Kundur, 1994). In Erikkson
(2008), a multiple HVDC link system based on part of
the Nordic power grid is presented. The techniques used
to control this system were primarily centralised, nonoptimisation
based control techniques.
Model Predictive Control (MPC) (Rossiter, 2003) (also
known as Receding Horizon Control) is an optimisation
based control technique, in which the controller uses statespace
and output predictions to calculate optimal control
moves for the system. One of the main advantages of this
control technique, over non-optimisation based techniques,
is the systematic and intuitive manner in which constraints
are incorporated into the control system and the fact that
delays are naturally catered for. It is a mature technology
at this stage, with stability and robustness analysis well
established for the linear, time-invariant, centralised case.
THE MULTIPLE HVDC LINK SYSTEM
The continuous-time dynamics of the multiple link HVDC
system under study here are described in this section.
2.1 Multiple HVDC link system description
Fig. 1 shows the system to which we will in this paper
apply the generally applicable distributed MPC scheme
discussed in the next section. This system is based on the
multiple HVDC link system between Denmark, Norway,
and Sweden (Erikkson, 2008). It consists of 4 buses with
their own generation and loads. Both AC and HVDC
lines connect the buses. The HVDC lines are of the Line
Commutated Converter type (HVDC-LCC) (Pai et al.,
1981). Generation capacities and loads are kept constant
in this paper.
Description and state-space prediction
In MPC a control agent uses a discrete-time system
model that predicts the system’s future trajectory over a
prediction horizon in order to calculate optimal discrete
inputs over this horizon. Only the input for the first
discrete time step is applied. At the next time step a
new action is determined. The prediction horizon moves
forward in a receding manner each time step.
SIMULATION RESULTS
The distributedMPC scheme is used to control the coupled
HVDC link system. The control inputs, the HVDC line
powers, are common to all 4 agents and the AC connected
buses share interconnecting variables too. A simplification
here is to directly calculate and apply the HVDC powers.
However in a real system, currents are injected and so these
would have to be calculated from these powers.
One agent is assigned per HVDC link as the HVDC link
control agent, sending the HVDC power it calculated at
the end of each control cycle through the link. Communication
is needed between the agents in order to coordinate
the HVDC powers sent to and received by each agent.
CONCLUSIONS AND FUTURE RESEARCH
Here the application of a distributed MPC to a multiple
link HVDC system is proposed. The resultant set-point
tracking performance is significantly better than that of
a decentralised MPC controller and close to that of a
centralised MPC controller.
However this performance comes with a significant communication
and computational overhead.Ways of reducing
the computational and communication overhead will be
needed before the distributed MPC in this paper could
be implemented in reality. Furthermore, stability and convergence
guarantees should be investigated for this distributed
MPC technique. Communication delays and data
transmission errors are other issues that would affect the
control performance. These problems form the basis for
future research in this area.