Klaus Trangbaek

Bumpless Transfer between Observer-based Gain Scheduled Controllers

This paper deals with bumpless transfer between a number of observer-based controllers in a gain scheduling architecture. Linear observer-based controllers are designed for a number of linear approximations of a non-linear system in a set of operating points, and gain scheduling control can subsequently be achieved by interpolating between each controller. The Youla-Jabr-Bongiorno-Kucera (YJBK) parameterization is used to achieve a smooth scheduling between the controllers. This approach produces a scheduled controller as a linear fractional transformation between a fixed controller and a scheduling parameter. The approach is tested on a simple, but highly non-linear model of a fossil fuel power plant.

Plug-and-Play Control—Modifying Control Systems Online

Often, when new sensor or actuator hardware becomes available for use in a control system, it is desirable to retain the existing control system and apply the new control capabilities in a gradual fashion rather than decommissioning the entire existing system and replacing it with an altogether new control system. However, this requires that the existing controller remains in action, and the new control law component is added to the existing system. This paper formally introduces the concept of Plug-and-Play control and proposes two different methods of introducing new control components in a smooth manner, providing stability guarantees during the transition phase as well as retaining the original control structure. The applicability of the methods is illustrated on two different practical example systems, a livestock stable climate control system and a laboratory-scale model of a district heating system.

Hierarchical model predictive control for resource distribution

This paper deals with hierarchical model predictive control (MPC) of distributed systems. A three-level hierarchical approach is proposed, consisting of a high level MPC controller, a second level of so-called aggregators, controlled by an online MPC-like algorithm, and a lower level of autonomous units. The approach is inspired by smart-grid electric power production and consumption systems, where the flexibility of a large number of power producing and/or consuming units can be exploited in a smart-grid solution. The objective is to accommodate load variations on the grid, arising from varying consumption and natural variations in power production, e.g. from wind turbines. The approach presented is based on quadratic optimisation and has low algorithmic complexity as well as good scalability. In particular, the proposed design methodology facilitates plug-and-play addition of subsystems without controller redesign. The method is verified by simulating a three-level smart-grid power control system for a small isolated power grid.

Plug and Play Process Control Applied to a District Heating System

Abstract The general ideas within plug and play process control (P 3 C) are to initialize and reconfigure control systems just by plug and play. In this paper these ideas are applied to a district heating pressure control problem. First of all this serves as a concrete example of P 3 C, secondly some of the first techniques developed in the project to solve the problems in P 3 C are presented. These are in the area of incremental modelling and control and they make it possible to “plug” in a new sensor and actuator and make it “play” automatically.

Bumpless transfer between advanced controllers with applications to power plant control

This paper deals with bumpless transfer between a number of advanced controllers, e.g. in a gain-scheduling architecture. Linear observer-based controllers are designed for a number of linear approximations of the system model in a set of operating points, and gain scheduling control can subsequently be achieved by interpolating between each controller. We use the Youla-Jabr-Bongiorno-Kucera parameterization to achieve a differentiable scheduling between the controllers. This approach produces a controller as a linear fractional transformation between a controller and a scheduling parameter. In this paper we propose a systematic approach to achieve bumpless transfer between different nominal controllers. The approach is tested on a simple, but highly nonlinear model of a coal-fired power plant.

Exact power constraints in smart grid control

This paper deals with hierarchical model predictive control (MPC) of smart grid systems. The objective is to accommodate load variations on the grid, arising from varying consumption and natural variations in the power production e.g. from wind turbines. This balancing between supply and demand is performed by distributing power to consumers in an optimal manner, subject to the requirement that each consumer receives the specific amount of energy the consumer is entitled to within a specific time horizon. However, in order to do so, the high-level controller requires knowledge of how much energy the consumers can receive within a given time horizon. In this paper, we present a method for computing these bounds as convex constraints that can be used directly in the optimisation. The method is illustrated on a simulation example that uses actual wind data as load variation, and fairly realistic consumer models. The example illustrates that the exact bounds computed by the proposed method leads to a better power distribution than a conventional, conservative approach in case of fast changes in the load.

Robust quasi-LPV control based on neural state-space models

We derive a synthesis result for robust linear parameter varying (LPV) output feedback controllers for nonlinear systems modeled by neural state-space models. This result is achieved by writing the neural state-space model on a linear fractional transformation (LFT) form in a nonconservative way, separating the system description into a linear part and a nonlinear part. Linear parameter-varying control synthesis methods are then applied to design a nonlinear control law for this system. Since the model is assumed to have been identified from input-output measurement data only, it must be expected that there is some uncertainty on the identified nonlinearities. The control law is therefore made robust to noise perturbations. After formulating the controller synthesis as a set of linear matrix inequalities (LMIs) with added constraints, some implementation issues are addressed and a simulation example is presented.

Plug-and-Play Process Control: Improving Control Performance through Sensor Addition and Pre-filtering ⋆

An important issue in the area of reconfigurable systems is how to respond correctly if new components are added. We consider the problem of improving control performance for a system where a new set of sensors becomes available. It is assumed that a complete re-design of the control system is undesirable for various reasons. The sensor dynamics are unknown and must be identified via experiments. The paper demonstrates how new sensor information can be fused with existing sensor information and fed to the existing control system, either based on knowledge of the existing plant or in an entirely data-driven fashion. The method is illustrated on a numerical example.

Stable controller reconfiguration through terminal connections

Abstract Often, when new sensor and/or actuator hardware becomes available for use in a control system, it is desirable to retain the existing controllers and apply the new control capabilities in a gradual, online fashion rather than decommissioning the entire existing system and replacing it with the new system. This paper presents a novel method of introducing new control components in a smooth manner, providing stability guarantees during the transition phase, and which retains the original control structure.

Hierarchical Control for Smart Grids

Abstract This paper deals with hierarchical model predictive control (MPC) of smart grid systems. The design consists of a high level MPC controller, a second level of so-called aggregators , which reduces the computational and communication-related load on the high-level control, and a lower level of autonomous consumers. The control system is tasked with balancing electric power production and consumption within the smart grid, and makes active use of the flexibility of a large number of power producing and/or power consuming units. The objective is to accommodate the load variation on the grid, arising on one hand from varying consumption, and on the other hand by natural variations in power production e.g. from wind turbines. The high-level MPC problem is solved using quadratic optimisation, while the aggregator level can either involve quadratic optimisation or simple sorting-based min-max solutions. In this paper we compare the performance and computational complexity of these two solutions and find that the performance of the two algorithms are very similar, whereas the sorting-based algorithm is much faster than the quadratic optimisation-based algorithm, thus allowing to handle vastly larger numbers of consumers.