Lars Peter Jensen
Flipped or blended learning
Flipped learning is a form of blended learning that replaces transmission based lectures with more participative, interactive and collaborative learning opportunities.
Activities are typically undertaken before, during and after class, freeing in class time to participate in activities and engage with concepts at a higher level.
Examples of how it can be done and why it is a good idea will be presented and pro’s et con’s identified for discussion.
Stochastic Safety Analysis of Stochastic Hybrid Systems
This work establishes an optimisation scheme for computing safety of stochastic hybrid systems. Specifically, the scheme computes the probability of reaching the set of unsafe states from a set of initial states. We address hybrid systems with both sporadic and state dependent switches. The method combines two approaches: the barrier certificate method and potential theory.
Assistive Exoskeletons for the Elderly
In this talk, I will present the Exo-Aider project, which is on the design of assistive exoskeletons for elderly and handicapped people. Challenges in designing exoskeletons range from designing a lightweight mechanical construction with small but powerful actuators to detecting the intention of the wearer of the exoskeleton. I will address some of these challenges and demonstrate a prototype of the exoskeleton.
Bounding Polynomials and Rational Functions in the Tensorial and Simplicial Bernstein Forms
In this talk we show bounds for the range of polynomials and rational functions which are obtained by the Bernstein expansions.
The main original concepts of this talk are as follows:
• Sharpness, monotonicity and inclusion isotonicity of the bounds
• Subdivision of simplex and convergence properties
• Minimization and certificates of positivity in the Bernstein basis
• Constant and affine lower bounding functions for polynomials and rational functions'
Karl Damkjær Hansen
Kugle, The Sociable Ballbot
The robot Kugle is a ballbot; a mobile robot balancing on a ball. The aim of Kugle is to develop a motion system that improves the interaction between mobile robots and humans.
Kugle is Danish for ball, which is the central mechanism in the robot’s locomotion. Balancing on a ball lets the robot move holonomically and rotate in place. One advantage of this type of locomotion is smoother motion. Robots relying on omnidirectional wheels for holonomic movements are hindered by the small diameter of the unactuated perimeter wheels, as small obstacles produce near impulsive motions to the robot, making it shake. The ball with a larger diameter makes for a smoother traversal of small obstacles. The base of the robot can also be smaller than robots that need several wheels. Especially if the robot is tall.
An effect of the ball mechanism is that the robot exhibits a dynamic balancing behaviour. This seems to be an advantage when interacting with humans. Humans tend to move slightly when interacting with each other both involuntarily to balance and also larger movements to accommodate for changes in the interaction context, e.g. a participant entering a group conversation.
Bernstein Bases in Control – What’s with all the fuzz
Although I’ve been at the department for about 2½ years by now, this will be my first FLL. I’ll use this opportunity to shed some light on the work I’ve done in the CodeMe project on the use of Bernstein bases to synthesize Lyapunov functions.
The conditions on a Lyapunov function to certify some stability property of a given system requires the determination of the sign of two functions. For non-linear systems this can be tricky task. When synthesizing using the Bernstein basis the problem boils down to the solvability of a certain linear problem. I’ll cover this derivation and show some examples using the software developed in the project.
A brief introduction to sequential Monte Carlo methods for state filtering and parameter estimation in hidden Markov models.
A still increasing availability of cheap computational resources has allowed sequential Monte Carlo methods to become ubiquitous for solving the state filtering problem in hidden Markov models. However, using these methods for estimating static parameters in the same class of models remains an unresolved issue. The purpose of this talk is to present basic principles upon which the sequential Monte Carlo methodology is built, stressing the reason why these methods can fail in the system identification problems. Considering a few simple examples, the first part of the talk compares classical approaches for state filtering with the basic, bootstrap, particle filter. The following part introduces sequential Monte Carlo methods. The talk concludes by showing a number of selected techniques for the parameter estimation.
Carsten Skovmose Kallesøe
The section has for a while worked on control of water distribution systems. This work has led to the research grant SAMPLE. The lecture will present some of the pre-results leading up till the SAMPLE project and the research goals for the SAMPLE project.
Robust Fault Detection for Uncertain Time-Delay Switched Fuzzy Systems
Many of the physical systems are multi-model inherently, and some dynamical systems are necessary to describe their behavior. A switched system can be effectively used to model these systems. Moreover, nowadays, fault detection problem is one of the most important problems in different sensitive and expensive industries. This presentation is about robust fault detection problem for a class of switched fuzzy systems.
Creating High-fidelity Mechanical Matlab Simulation Models With a Few Clicks
This will not be a traditional lecture, it will be a demonstration of how mechanical models designed in SolidWorks can easily be exported to MATLAB Simulink for visualisation, control system design, and validation. It will be demonstrated on the section’s favourite robot; the CrustCrawler.
Jan Dimon Bendtsen
Control of District Heating System with Flow-dependent Delays
All flow systems are subject to transport delays, which are governed by the flow rates in
the system. When the flow rates themselves are control inputs, the system becomes subject to input-
dependent state delays, which poses significant theoretical problems. In an earlier paper, we proposed
a guaranteed stable control design for a system of this type; in this paper, we provide experimental
evidence of the usefulness of the design
Abdul Rauf Khan
Classification of Noisy Data: An approach based on Genetic Algorithm and Voronoi tessellation
Classification is a major constituent of the data mining toolkit. Well-known methods for classification are either built on the principle of logic or on statistical reasoning. For imbalanced and noisy cases, classification may however fail to deliver on basic data mining goals, i.e. identifying statistical dependencies in data. In this talk I will introduce a different strategy for data mining based on partitioning of the feature space through Voronoi tessellation and Genetic Algorithm, where the latter is applied to solve a combinatorial optimization problem. In this short presentation, I will present some of the preliminary results for the comparison of the suggested methodology, with well-known classification methods such as (SVM, KNN, and ANN).
Jesper Viese Knudsen
Coordination and Control of Generator Sets
In this talk, I will introduce my industrial PhD project with the above title, starting with a brief presentation of the company I’m working for – DEIF. Then, a short description of the technical setup of diesel driven generator sets will be followed by the presentation of results obtained with a self-tuning LQR controller. Finally, I will introduce the optimization problem currently under investigation and present the direction in which I’m looking at the moment – which is genetic algorithms.
Almost Everywhere Stability
Anders Rantzer has shown that the existence of a Lyapunov-like density function for ODE's implies a weaker version of global stability, called almost everywhere stability. I will present a short summary of the relevant literature and discuss some new results and conjectures.
Karl Damkjær Hansen
What is ROS?
ROS is the hype in robotic programming. But how does it work and what makes it better than all the competition? I will go into the technicalities, looking at the plumbing connecting the individual processes running on the robot computer. I will take a broader view on all the available tools and software packages available, plus take a birds eye view on the entire eco system that makes ROS so great.
Anders la Cour-Harbo
Status on drone activities at AAU
I will give a brief overview of the research activities in the drone field at AAU, including the recent UAS-ability grant, the national drone strategy, and the DroneSafety funding application.
Overview of the OpSys Project
In this talk I will give an overview of the activities and goals within the EUDP project “OpSys” with the longer and more descriptive title “Under Floor Heating and Heat Pump Optimization”. I will also present some of our preliminary results, which have been submitted to the IFAC World Congress next year. Here we use neural networks to predict future indoor temperatures based on various amounts of information that might be available. These prediction capabilities are then used to find optimal future set-point trajectories for the floor heating system. This allows the heating system to account for the slow dynamics of the house and potentially suppress future expected disturbances better. I will also talk a bit about possible future directions of our research
Tom Nørgaard Jensen
Application of a Novel Leakage Detection Framework for Municipal Water Supply on AAU Water Supply Lab
This work was presented at the recent Systol International Conference on Control and Fault-tolerant Systems and is a joint work between Carsten Kallesøe and I.
The paper deals with isolation of leakages in a water supply system using a reduced network model. Detecting and isolating leakages fast is very important not only to save water but also to avoid destruction of roads and houses. We propose a heuristic leakage detection/isolation algorithm which uses the reduced network model to estimate nominal behaviour. The model is adaptive and thus adapts slowly to changes in the network. However, some leakages change the behaviour instantly, and the deviation between pressures estimated by the reduced network model and the actual measured pressure indicates the location of the leakage.
The proposed algorithm is tested on the AAU Water Supply Lab which is a laboratory setup that emulates a water distribution network. The test shows that the algorithm is in fact able to indicate in which area of the network a leakage has appeared.
Introduction to the WallMoBot Project
In this talk I will introduce the WallMoBot project. The WallMoBot is a collaborative robot for mounting wall panels such as glass, concrete, and drywall in buildings, and is intended to remove the heavy lifting from the workers. I will explain our role in the project and what challenges we are faced with when designing a control system for the robot. Questions such as semi-autonomy vs. full-autonomy will be presented along with cool animations!
Juan de Dios Flores Méndez
Compliant Robotic Motion
Modern robotic manipulators not only move according to prescribed kinematic references, but also in accordance with force and torque requirements when in contact with materials and other robots. This talk will give a general scope into Juan’s PhD project and show why it is relevant for future industrial applications.
Jan Dimon Bendtsen
Stepwise Commissioning of a Steam Boiler with Stability Guarantees
It is well known that multi-variable (MIMO) controllers can achieve significantly better performance than single-loop controllers when controlling multi-variable plants. Nonetheless, MIMO controllers are seldom used in industry.
One of the obstacles preventing widespread use of multi-variable controllers appears to be a certain distrust among operators with respect to their increased complexity. In this work, we present a stepwise commissioning strategy based on the so-called Youla-Kucera parametrization that allows for gradual transition from a set of single-loop controllers (e.g., PI) to a MIMO controller while guaranteeing nominal stability before, during and after the transition. The method has the added benefit of allowing to "roll back" the transition, i.e., recover the original single-loop controllers, in case anything goes wrong along the way. The method has been tested on a small industrial steam boiler and measurements show a clear performance improvement after transition.
Direct Georeferencing in Aerial Photogrammetry
Photogrammetry deals with the reconstruction of shapes and positions of objects which are visible on photographs. Especially aerial photogrammetry, which uses small unmanned aerial vehicles (UAV), is popular in recent years due to its low cost and time demands. This talk will introduce basic techniques of georeferencing in photogrammetry and also real flight data and data processing will be presented.
Fully Probabilistic Design
In this talk I will present a stochastic control design in the Kullback-Leibler sense. The technique is called the Fully Probabilistic Design (FPD) and consist in essence of finding a stochastic control law, given in terms of a probability density function (pdf), minimizing the Kullback-Liebler distance between pdfs describing a given closed loop system and a desired closed loop system.
PID Control with Robust Disturbance Feedback Control
This work was presented and published at The 2015 IEEE Multi-Conference on Systems and Control (MSC2015).
Disturbance Feedback Control (DFC) is a technique, for augmenting existing control systems with an extra feedback for attenuation of disturbances and model errors. In this work, we analyze the robustness and performance of a PID-based control system with DFC. We propose a simple grid-based search algorithm that can be used to find DFC to achieve robust stability and performance.
Tom Nørgaard Jensen
Adaptive Reference Control for Pressure Management in Water Networks
This work was presented at the ECC conference this summer. We consider water networks with a single pressure actuator and several consumers. Under mild assumptions on the consumption pattern and hydraulic resistances of pipes, we use properties of the network graph and Kirchhoff’s node and mesh laws to show that simple relations exist between the actuator pressure and critical point pressures inside the network. Subsequently, these relations are exploited in an adaptive reference control scheme for the actuator pressure that ensures pressure constraints at the critical points are met. Numerical experiments underpin the results.
Model Predictive Control of a Wave Energy Converter
The talk will give result from a paper presented at MSC2015 in Sydney last month with the above title, authors: Palle Andersen, Tom S. Pedersen, Kirsten M. Nielsen and Enrique Vidal.
In this paper reactive control and Model Predictive Control (MPC) for a Wave Energy Converter (WEC) are compared. The analysis is based on a WEC from Wave Star A/S designed as a point absorber. The model predictive controller uses wave models based on the dominating sea states combined with a model connecting undisturbed wave sequences to sequences of torque. Losses in the conversion from mechanical to electrical power are taken into account in two ways. Conventional reactive controllers are tuned for each sea state with the assumption that the converter has the same efficiency back and forth. MPC's are designed for each sea state using a model assuming a linear loss torque. The mean power results from two controllers are compared using both loss models.
Simulation results show that MPC can outperform a reactive controller if a good model of the conversion losses is available.
Safety Analysis of Stochastic Dynamical Systems
This presentation outlines a method for verifying the safety of a stochastic system. In particular, it will be shown how to compute the largest set of initial conditions such that a given stochastic system is safe with some given probability.
To compute the set of initial conditions we rely on the moment method that allows an infinite dimensional optimization problem on measures to be formulated as a polynomial optimization problem. Subsequently, the optimization problem is relaxed to obtain a finite dimensional polynomial optimization problem.
Lakshmi Munukutla - Guest Researcher at AAU (Fulbright Scholar)
Overview of Dye Sensitized Solar Cells.
This is a presentation to introduce Prof. Munukutla to the Automation and Control faculty, students and staff at Aalborg University, Aalborg, Denmark. The presentation covers brief background of her home institution, Arizona State University, USA and research outcomes of Dye Sensitized Solar Cells (DSSCs), which is her recent research activity. DSSCs are third generation solar cells gaining popularity due to its low cost combined with a relatively simple manufacturing process compared to crystalline solar cells. The DSSC technology is not matured to achieve production levels yet but it is promising. The presentation covers the fabrication process of the DSSCs and their characterization results.
Optimization of Heat Pump Operation with Genetic Algorithms
Some complex real-world problems can be difficult/impossible to solve with classical gradient-based optimization techniques due to smoothness and continuity requirements. Advances in computational power has made heuristic techniques such as genetics inspired search algorithms a viable option for such complex systems. In this talk I will present some of our recent work within project GreenFlex, where we have been investigating the potential use of genetic search algorithms for optimization and system identification of a district heating plant in Sønderborg. The plant consists, among others, of a complex system of absorption cycle heat pumps and setting good/optimal operating set-points for this system is a challenging task for the plant operator. A goal has there been to make an assistive tool for set-point optimization.
Abdul Rauf Khan
Statistical Data Mining for Efficient Quality Control in Manufacturing
Extensive use of machines, flexible/re-configurable manufacturing and transition towards the fully automated factories call for intelligent use of information recorded during the manufacturing process. Modern manufacturing processes produce Terabytes of information during different stages of the process e.g sensor measurements, machine readings etc, and the major contributor of these big data sets are different quality control processes. In this talk I will present methodology to extract valuable insight from manufacturing data. The proposed methodology is based on comparison of probabilities and extension of likelihood principles in statistics as a performance function for Genetic Algorithm.
Investigations of the Dynamics in a Sliding Mode Control System.
Discontinuous dynamics appears often in sliding mode control. Based on this, the sliding mode control of a mechanical system is studied, where the focus is on the two-dimensional discontinuous SDE.
I will present different approaches we have applied in order to obtain further information on ``the simple discontinuous SDE'' and the related discontinuous SDE.
Karl Damkjær Hansen
Socially Assistive Robots
What are socially assistive robots? We will try to answer this question by attempting to define the term robot. Once this is defined, we will look at specializations like mobile robots, social robots, assistive robots and finally socially assistive robots. We will then survey the past and current work in these areas performed in this section and the rest of the world.
Tom Nørgaard Jensen
A Distributed Algorithm for Energy Optimization in Hydraulic Networks
This is a joint work between Carsten Kallesøe from Grundfos, Rafael and myself, which was presented at this summers IFAC world congress. We develop a distributed control for a class of hydraulic networks with the purpose at maintaining a given reference pressure across selected valves distributed throughout the network while minimizing the electrical energy consumption of the actuators (pumps). We then use 'handwaving' arguments to argue that the developed control fulfills this objective.
Rasmus Lundgaard Christensen
The Purpose of Error Detection on Reefer Containers
This talk is meant as a presentation of the contents of my project. I will describe the driving factors behind the project, describe a common reefer container operation cycle, and finally present how we intend to detect the errors. Beware, the project plan is still under development, and this talk is meant as a project pitch! It will be very illustrious and not very technical.
Abdul Rauf Khan
Data-Driven Approach for Intelligent Control of Production Line: An Introduction to MADE9.5
It’s a common practice in production world that the quality of the product is examined by very sophisticated tests and technology on different stages of the production line. Where this practice insures the quality of the product this also involves the allocation of precious resources (time, labor, infrastructure, etc.). During the different quality control stages/tests significant amount of data is generated. These information-rich large data sets can be useful to understand/learn the behavior of the process. In MADE 9.5, we will use Data mining/Machine learning methods to learn the behavior of the process from the historical data and solve the problem of process optimization for the future production.
Jan Dimon Bendtsen
"Freezing fish the PDE way"
In this joint work (with C. Backi from NTNU), the stability properties of a nonlinear partial differential equation (PDE) with state-dependent parameters is investigated. Among other things, the PDE can be used to describe the temperature distribution in media that undergo phase change, such as freezing of foodstuff. We show that for certain forms of coefficient functions, the PDE converges to a stationary solution given by (fixed) boundary conditions that make physical sense, and illustrate the results with colorful numerical simulations.
An Industrial Model Based Disturbance Feedback Control Scheme
In this talk, I will present a Disturbance Feedback Control (DFC), which is original control algorithm by Fuji Electric.
The key of DFC is feedback function “L”. L can be tuned independently for disturbance response and L decreases the negative impact of disturbance and model errors.
DFC could be applied widely to industrial process systems because field engineers need a simple controller which can tune easily.I will show you some examples at FLL.
Voltage Stability! Why Should YOU Care?!?
In this talk, I will present power systems voltage stability. I will formally introduce it by a simpel example and explain how voltage instability can occur by presenting the 2004 Athens blackout. Afterwards, I will analyse what some of the components leading to the blackout was, and why this is interesting for us (or at least those of us studying smart grids). Finally, i will discuss the role of demand management, seen from a voltage stability point of view, and come with an interesting conclusion. Fear not! even though the word stability is mentioned, this talk will NOT be very technical :-)
CodeMe – a project on polynomials in control
In this talk, I will give an outline of the CodeMe project that has recently started. The aim of the project is to develop algorithms and software for analyzing and designing nonlinear (polynomial) control systems. There exists software for analyzing polynomial systems; however, these tools have scalability issues, and are mostly based on checking sufficient conditions. We will attempt to address these issues by not relying on sum of squares as our certificate of positivity.
A Sliding Manifold Specific to Purely Magnetic Attitude Control Problem
Magnetic torquers are spacecraft actuators producing torque via interaction with the local geomagnetic field. They are especially preferable in attitude control systems of small satellites due to their relatively low mass, volume and energy consumption. There have been many studies in literature that aim to obtain high performance and/or strong stability results by proposed controllers for a control system based on solely three orthogonally placed magnetic actuators. However, there is still need of such studies that consider the problem under environmental disturbance and model uncertainty, especially in its nonlinear representation.
This Friday I will present you a sliding manifold that consists of a conventional part, which is known to be specific to the problem of attitude control by using directly torque producing actuators (momentum exchange devices and thrusters), and an additional term that is proposed as specific to the purely magnetic attitude control problem. The proposed manifold makes the application of the equivalent control method to the problem possible. It also satisfies the reaching condition for the case only under environmental disturbances, which means that, in that case, the state trajectories governed by various reaching laws arrive at the sliding manifold from any initial position in state-space. After giving some simulation results, I hope to discuss with you the stability of the proposed sliding manifold.
Investigations of the Switching Mechanism
In this talk I will present investigations of the switching mechanism, which occurs on a switching surface. The switching mechanism is presented in a deterministic and in a stochastic model. Probabilistic results and a comparison between the two frameworks in a two-dimentional state space will be given.
Karl Damkjær Hansen
Genetic Algorithms - Search and Optimization Inspired by Biology
Many meta-heuristic search methods are inspired by natural phenomena, this is also true for the genetic algorithm. Natural populations of individuals show great adaptability to changing environments. One of the reasons for this is Darwin's notion of "survival of the fittest". The genetic algorithm employs this notion to exhibit robustness in the face of changing and different search spaces.
I will present the workings of a simple genetic algorithm and outline John Holland's schemata analysis to illustrate why this algorithm works. Finally, I will show how I have used the genetic algorithm for finding a near-optional trajectory for a helicopter.
Parameter Estimation for Stochastic Hybrid Systems with State-Dependent Switching Noise
I will discuss a method for estimating parameters of a class of stochastic hybrid systems with piecewise linear partitioned state space, and state-depending switching governed by an intensity function. In brief, this method is of maximum likelihook type and rest upon an idendification of the trajectories of the stochastic hybrid system with a point process. This is joint work with Jakob (math), Henrik and Rafal.
GreenFlex, Absorption, Modelica and other Hot Stuff
In this talk I will present the newly started GreenFlex project. Shortly, the objective in project GreenFlex is to develop and demonstrate a software tool named FlexIQ that can support economic optimal planning and control of integrated green energy plants in mobilizing flexibility for power and heat markets. I will also present the case study which is Sønderborg district heating and some of the initial investigations we have made. Finally I will talk a bit about the Modelica modeling language and absorption heat pumps. In all, a lot of "hot stuff".
Luminita Cristiana Totu
On Designing Demand Response for a Population of Thermostatic Loads
A single thermostatic unit (e.g. refrigerator) can be modeled as a hybrid system with temperature as the continuous state and two discrete modes corresponding to "on" and "off" power consumption. The switching between the discrete modes is normally based on the thermostat law. In order to get some control over the power consumption of the refrigerator, a "smart" thermostat law is introduced. This takes into account an external signal and introduces randomized switches. Populations composed of many thermostatic units can be modeled using probability distribution (densities) as the system state. The goal of the control problem is to design the external signal broadcasted to the "smart thermostats", such that the power consumption of the population can follow a given reference. The approach I have been investigating is model based, and this talk will present a number of modeling points. Some of these are (generalized) Fokker Planck operators, population heterogeneity, observability of the probability distribution from realistic measurables. Equipped with a model, designing the control can be done as an open loop task, or in feedback. An open loop (non-convex) optimization and interesting numerical results are presented. The architecture of a feedback solution is also presented, but this part is not yet complete.
Palle Andersen / Tom S. Pedersen
ForskEL Project "Digital Hydraulic Power Take Off for Wave Energy"
The section has received funding from ForskEL for a part of the project "Digital Hydraulic Power Take Off for Wave Energy".
The project is a cooperation between the company Wave Star and three AAU departments. In our part the goal is to optimize power take off of float type wave engine of which a prototype is situated in the sea near Hanstholm. New digital hydraulic equipment is investigated by other partners.
We will present the principle of the engine and a model of the float. What is known about optimizing the power take off of this type of engine for regular sinusoidal waves will be summarized. Finally problems with irregular waves, fatigue and power take off efficiency will be addressed.
Model Reduction of Linear Switched Systems by Restricting Discrete Dynamics
We present a procedure for reducing the number of continuous states of discrete-time linear switched systems, such that the reduced system has the same input-output behavior as the original system for a subset of switching sequences. The admissible set of switching sequences is a regular language and considered to be the language accepted by a non-deterministic finite state automaton (NDFA). The proposed method is expected to be useful for abstraction based control synthesis methods for hybrid systems.
System Identification of an Aggregation of Domestic Refrigerators
During the talk I will present the results of my internship and learn you some more about the Netherlands.
This is the abstract of the report:
To be able to store electrical energy in smart grids, the flexibility in power consumption of domestic refrigerators is being investigated. This report presents a first order model of a domestic refrigerator and validates it using measurements of a real refrigerator. The influence of food in the refrigerator and the ambient temperature on the power usage of the refrigerator are investigated. An aggregation of 10.000 refrigerators has been identified, using a first order model based on an energy balance. This model is successfully used to design a one day ahead tractable power reference for the aggregation.
Jose de Jesus Barradas Berglind
Fatigue Load Control for Wind Turbines
On this talk I will present some results based on online fatigue damage estimation, that we later use for wind turbine control. This will build up on the previous talk I gave, but I will try to make this presentation self-contained.
Per Printz Madsen
ENCOURAGE: What have we achieved
The EU/Artemis project ENCOURAGE is in the last phase of the project period. The main achievements, seen from our side, are partly the distributed architecture based on RabbitMQ and a "cloud" server, and partly MPC controllers, dedicated control language and runtime interpreter build on top of this architecture. Besides that there is a Dashboard for visualization and an energy system for trading energy inside a local community. I will present these components in more or less details.
Seyed Ehsan Shafiei
Input Signal Design for Identification of MIMO Processes
Using subspace identification method for predictive control design for refrigeration systems, I ran into the problem of input signal design. It turned out that this is really a less-noticed part in many control system design relying on system identification methods. Amongst the few papers I read in this regard, I found some methods interesting for sharing with you in my talk. The problem of input signal design is specifically significant when dealing with multivariate systems which are ill-conditioned and/or contrained. Subsequently, a new and simple approach we proposed for subspace identification of refrigeration systems is presented. However, the design of input signal is still the minor contribution of the paper we are submitting, but got noticed.
Peter Fogh Odgaard
Industrial Feasibility Study of a Model Predictive Controller for a Wind Turbine
In my presentation I will present recent results from an industrial feasibility study of a model predictive control of a wind turbine. The controller outline, model, objectives and constraints will be presented. The controller is evaluated on a Vestas simulator and compared with a Vestas controller for the same wind turbine. The model predictive controller is evaluated with respect to power generation, structural fatigue loads. Notice that this presentation will be confidential.
Decreasing skips between paths when projecting 2D planning into 3D terrain for more efficient field operations
Today research within agricultural technology focuses on productivity, operating costs and the efficient use of resources. Autonomous machines have the potential to significantly contribute to this through the real-time application of artificial intelligence and optimization algorithms to conduct agricultural operations in a manner beyond human capabilities and in fully autonomous mode. 2D coverage planning assumes that the field is flat and ignores elevation changes. While this assumption is valid in many flat agricultural fields, there is a significant portion of agricultural production areas where topography has impacts on operational patterns. When projecting 2D planning result to 3D terrain, the actual distance between paths on the topographic surface increases, and there will be skips between adjacent paths on the slopes which might result in economic impacts. In this regard, an approach is developed to evaluate the expected economic impact of using conventional field coverage approaches and optimize it in a manner to reduce unused areas between paths. A new 3D coverage approach able to minimize skips is presented. Developed approaches are tested on simulated and real fields of different surface topography.
Tero Ahonen, Lappeenranta University of Technology, Finland
Smart Variable-Speed-Driven Pump and Fan Systems - Research in Lappeenranta University of Technology, Finland
Pump and fan systems benefit from their variable-speed operation.
Especially industrial pump and fan systems are composed of separate devices, where the frequency converter may be the only "smart" device in the system allowing new control and diagnostic possibilities. Tero Ahonen from Lappeenranta University of Technology (LUT) tells about these possibilities and also about his home university in Finland.
Aggregation and Control of Supermarkets in a Smart Grid
In this talk I will present some of the results obtained throughout my master's project. We consider control strategies for aggregation of a portfolio of supermarkets towards the electricity balancing market. The supermarkets are able to shift the power consumption in time by pre-cooling the contained foodstuff. It is shown how the flexibility of an individual supermarket can be modeled and how this model can be used by an aggregator to manage the portfolio to deliver upward and downward regulation. Two different interfaces are presented; first a simple ON/OFF interface is shown and secondly a more complex interface, based on the supermarket following a power reference and feeding back it’s flexibility in form of a bucket model, is considered. Furthermore, VPP control strategies for both interfaces will be presented. Lastly, I would like to introduce my current work in the SmartC2Net project.
Barrier Certificates - Converse Results
In this talk, I will discuss the task of determining whether a dynamical system is safe. Successful safety verification certifies that no admissible trajectories of a system reach an unsafe subset of the state space. The talk will discuss the problem of finding a barrier certificate - the barrier certificate method identifies a manifold separating the initial states from the unsafe states. The separating manifold is given as a sublevel set of a function. I will show that under a mild conditions, if a system is safe there is a barrier certificate.
The project DANTEQ - fish hoovers, spinning water molecules and partial differential equations
The project DANTEQ (short for "Development and assessment of novel technologies improving the fishing operation and on board processing with respect to environmental impact and fish quality") aims to provide new skills and develop new methods to optimize the handling of fish on board fishing vessels with regard to environmental impact and raw material quality.
It consists of three packages: Handling Systems (taking and storing fish on board, processing fish on board), Cooling and Freezing Systems (modeling freezing processes with partial differential equations, cells alive system) and Energy Systems (optimization with regard to energy consumption and environmental impact).
Supermarket Consumption Control in Danish Smart Grid
In order to partially buffer intermittent behavior of wind and solar sources, Danish future smart grid increases the need for controlling the consumption. Solution to encourage the change in the behavior among the population of consumers is to introduce variable electricity prices. Supermarket is an example of a multi agent system, where each of its subsystems (HVAC, Solar Panel, Refrigeration Unit) is controlled autonomously by its local controller. We are examining the supermarket subsystems' possibility of changing their nominal parameters in order to decrease the total cost of operation. In order to do that we introduce a supervisory controller that according to the information received from environment and all subsystems decides what is the best behavior for the whole system.
Jesus Barradas B
Fatigue Estimation for Wind Turbine Control using Hysteresis Operators
In general fatigue can be understood as the weakening or breakdown of a material subject to stress, especially a repeated series of stresses; this is particularly relevant for wind generation settings, due to the harsh conditions that they are exposed. Some of the different approaches to estimate the damage caused by fatigue will be discussed and compared, with particular emphasis in a hysteresis operator method as a control convenient measure of fatigue. The intuition behind the hysteresis operator approach is that the underlying process is rate independent, meaning that it does not matter how fast the stress occurs but its amplitude.
Model Reduction by Moment Matching for Linear Switched Systems
A moment-matching method for the model reduction of linear switched systems (LSSs) is developed. The method is based upon a partial realization theory of LSSs and it is similar to the Krylov subspace methods used for moment matching for linear systems. The results are illustrated by numerical examples. During the talk, the emphasize will be given on the generalization of some concepts from linear system theory -like convolution representation, transfer function, Markov parameters and moment matching- to the LSSs.
Certificates of Positivity in the Bernstein Basis
In this talk, I will explain how to exploit a Bernstein Basis to formulate an optimization problem with polynomial inequality constraints as a linear programming problem.
A certificate of positivity is an easily checkable condition that determines if a function is positive. As an example, linear matrix inequalities (LMIs) are used extensively in control to certify the positiveness of quadratic forms. Certificates of positivity exist for general polynomials. We express polynomials in Bernstein form, and conclude that a polynomial is positive if all its Bernstein coefficients are positive. This enables the analysis and design of polynomial control systems.
Jan Dimon Bendtsen, Miroslav Krstic (UCSD)
Control of thermodynamical system with input-dependent state delays
We consider control of a cooling system with several consumers that require cooling from a common source. The flow feeding coolant to the consumers can be controlled, but due to significant physical distances between the common source and the consumers, the coolant flow takes a non-negligible amount of time to travel to the consumers, giving rise to input-dependent state delays. We first present a simple bilinear model of the system, followed by a state feedback control design that is able to stabilize the system at a chosen equilibrium in spite of the delays. We also present a heuristic, performance-oriented improvement to the design. The strategy is illustrated with some simulation examples.
Kristian Edlund, DONG
Workshop on Optimization and Smart Grids - Slides by Stephen Boyd
Everyone is talking smart grid as the solution that will save the world! And while I agree with that sentiment, I have seen a lot of research which builds on assumptions that are unrealistic to obtain within the near future either due to cost or regulatory issues.
In this talk, I will present Power Hub, which is the DONG Energy implementation of a smart grid solution. We take a more pragmatic approach and target solutions where we can optimize within the current regulatory regime in order to exploit the flexibility in the energy system. I will present a sample case in details and show you the potential gain by using better planning
Low-Voltage Consumption Coordination for Loss Minimization and Voltage Control
In this talk I present a strategy for minimizing active power losses in the low-voltage distribution grid, by coordinating the consumption from electric vehicles and power generation from solar panels. I will show that minimizing losses, also reduces voltage variations, and illustrate how this may be employed for increasing the number of electric vehicles and photovoltaic systems in the grid, without violating grid constraints.
Wind farm simulation and control
In this FLL, I will give you an introduction to the OFFWIND (www.offwind.eu) project and to wind farm simulation in general. I will present the online OFFWIND wind farm simulation tool (tools.offwind.eu), as well as other wind farm simulation tools like SimWIndFarm and SOWFA.
The primary objective of the OFFWIND project is to develop tools for advanced operation assessment and forecasting for offshore wind farms. So the majority of the work focus on accurately simulating the wind flow through a wind farm, using CFD tools. Here at AAU, our small contribution to the project is to control the turbines and to distribute the power reference to each turbine, such that more accurate flow simulations can be made. In the end it is the goal to provide an interactive online tool, that is capable of simulating and predicting power output from any given location and configuration of a wind farm.
Tom S. Pedersen
Control and modeling of heat pumps in private homes to help balancing the grid
In the READY! project power consumption by heat pumps has been used to counteract fluctuations in the production caused in particular by a large number of wind turbines. In the project it is possible to read data from a number of houses and for a few it is possible to ON/OFF control the heat pump. The control algorithms are design to optimize power purchase on the day-ahead market and to make an intra-day control. Our role is to design control algorithm’s for control of a large number of houses and to make models used in the control
The experienced refrigeration system
In this FLL, I will talk about the possible use of learning/repetitive control to improve the existing control in refrigeration systems. The reason for doing this is that the typically used PI(D) controllers often does not work optimally in all operating points and are not always tuned well. Furthermore, there is an incentive in the industry to create alternatives to e.g. advanced model based control, as these methods can be cumbersome and hard to implement in large scale as each refrigeration system is different. I will cover topics such as reference modifying repetitive control, 3R2C modeling, EnergyPlus, repeatable-to-nonrepeatable ratio, supermarket refrigeration system modeling, precooling, and thermal storage. The idea is to make a control which requires as little system knowledge as possible, learns from previous mistakes, and thus create a more intelligent refrigeration system and hence the title of the talk "the experienced refrigeration system".
Simple Model for Describing and Estimating Wind Turbine Dynamic Inflow
Wind turbines operate with sudden change in pitch angle, rotor or wind speed. In such cases the wake behind the turbine, achieve steady state conditions only after a certain delay. This phenomenon is commonly called dynamic inflow. There are many models for dynamic inflow. The most accurate use a method that can be characterised as the blade element momentum method plus a dynamic equation for the induction factor. This method then needs calculations along the blade for a number of sections including numerical solution of equations. This is numerical demanding. The simplest models amounts to placing a lead-lag filter after rotor torque and thrust calculated from static tables of the power and thrust coefficients. The filter constants will then vary with average wind speed. The filtered versions of torque and thrust are then an approximate modelling of the dynamic inflow. The dynamic inflow model suggested here places itself in between the most complex and the most simple both in accuracy, numerical demands and physical appeal. The suggested models behavior is demonstrated by simulation and the usefulness for extended Kalman filtering is assessed both via simulated data and real full scale turbine data.
Integration of Heterogeneous Industrial Consumers to Provide Regulating Power to the Smart Grid
Integration of consumers to the smart grid can be implemented with different policies. In the literature, two main control policies have been studied which are entitled direct control and indirect control. In this work, we apply the direct and indirect control to utilize the flexibility of a heterogeneous portfolio of industrial consumers. The problem is formulated as a mixed integer linear program which we solved it with CVX. Different scenarios have been simulated to compare these approaches.
Jesper Abildgaard Larsen
On Monday February 25th, the 3rd student satellite from Aalborg University, AAUSAT3, was launched. During my talk I'll present you with an overview of the satellite, along with the first results from the satellite after its first month in orbit.
Daniel Uhrenholt/Anders La Cour
Introduction to SolidWorks
We will show how to use the SolidWorks program to generate drawing of objects in 3D. This will include the basic operations for constructing a variety of different objects that later can be constructed in real life by Daniel or external parters.
Wave Disturbance Reduction of a Floating Wind Turbine Using a Reference Model–based Predictive Control
Floating wind turbines are considered as a new and promising solution for reaching higher wind resources beyond the water depth restriction of monopile wind turbines. But on a floating structure, the wave–induced loads significantly increase the oscillations of the structure. Furthermore, using a controller designed for an onshore wind turbine yields instability in the fore–aft rotation. In this paper, we propose a general framework, where a reference model models the desired closed–loop behavior of the system. Model predictive control combined with a state estimator finds the optimal rotor blade pitch such that the state trajectories of the controlled system tracks the reference trajectories. The framework is demonstrated with a reference model of the desired closed–loop system undisturbed by the incident waves. This allows the wave-induced motion of the platform to be damped significantly compared to a baseline floating wind turbine controller at the cost of more pitch action.
Tom Nørgaard Jensen
Distributed control of a water supply system
I will be presenting some results and open problems from my postdoc project which is in collaboration with Grundfos. In the project we consider the problem of regulating the pressure at designated notes in a water supply system to desired reference values using distributed control actions.
Path planning and trajectory control of unmanned ground vehicle (UGV) for weed control
Adaptive Surveying and Early treatment of crops with a Team of Autonomous Vehicles (ASETA) aims to use autonomous helicopters to photograph a sugar beet field in order to guide autonomous ground robots for weed control. The main task is to receive the coordinates of the regions which have to be visited for further investigation and/or treatment using a single UGV or a team of UGVs and generate a path to enable each UGV to visit a set of in-field regions in an optimized order. A path planning algorithm has been developed to enable a UGV to visit a set of in-field regions under the constraint of only driving over crop rows in order to minimize crop damage. A control trajectory is then generated with the ability to smooth sharp turnings to avoid wheel slippage and soil compaction. The developed path planning can be expanded to 3D in order to have more accurate navigation in fields of rolling terrains.
Smart Grid Dispatch Strategy for ON/OFF Demand-Side Devices
We consider an aggregator managing a portfolio of runtime and downtime constrained ON/OFF demand-side devices. The devices are able to shift consumption in time within certain energy limitations. We show how an agile control strategy can manage the portfolio of devices to collectively follow a power reference requiring only little information of the devices in the portfolio. Further, we illustrate that the performance of this agile control strategy is comparable with the performance of more advanced mixed-integer based control strategies that have full knowledge of all devices in the portfolio.
Vision for Smart Grid Controls: 2030 and Beyond
As part of a major Smart Grid Vision Project, in 2011, IEEE Standards Association commissioned a number of vision reports on the future electricity system, including IEEE Smart Grid Visions for Communication, Power, IT, Control Systems, and, Vehicular Technologies.
The SG Vision Project was chartered with creation of a rich set of forward looking use cases, applications scenarios for SG, and corresponding enabling technologies for SG of the future snap shots of years 2015, 2020, 2030, and beyond that in order to characterize the SG paradigm and be the world's beacon for SG paradigm and relevant to it technologies. It was targeted to produce publishable products (long term vision, reference architecture model(s), and a technology roadmap)
The framing of the SG vision was to be a pie in the sky globally optimal scenario for the SG with minimal to no boundary conditions. In other words the vision should not be bounded by existing technological, regulatory, business paradigms, standards projects, or product plans.
Within the IEEE Control Systems Society, a large number of smart grid control experts across the world worked during the past year on the report 'Vision for Smart Grid Controls: 2030 and Beyond', which is currently reaching its final form with the publisher. I had the honor to be one of the contributors. In this presentation, I will present some of the highlights from this report.