PhD Program System Integration of Renewable Energy

SEE Colloquium

07.11.2016 Jürgen Knies (Jade Hochschule), Title: Shaping Energetic Neighbourhood: A Dynamic Approach For A Future Proof Urban Energy Planning (15:00, Room: A5-2-229)

The proclaimed heat transition, with a huge impact on the reduction of the greenhouse gases, addresses especially cities. The simple question “Where should be implemented which technological solution of sustainable energy supply” often leads to a confusion of local stakeholders caused by the wide range of potential solutions. Besides technological and social challenges, facing the related question of the best location or area will be a key factor for a successful local implementation. The approach enables stakeholders to analyze cities spatially from the perspective of heat demand across planning borders like districts and real estates. The result shows areas of suitability of heat supply systems like LowEx - heating grids, typical district heating systems, or local heat generation, e.g. on block or building level. The shapes of this energy driven neighbourhoods are fuzzy. Due to the scenarios the dynamic spatial change of the suitability patterns can be used as a forecast system for urban planning activities.

18.09.2015 Hans-Georg Beyer (University of Agder, Norway), Title: Probabilistic forecasting for wind farms and PV power stations  (14:00, Room: V02 1-114) 

For operational purposes forecasts of the lumped output of a group of wind farms spread over a larger geographic area will often be of interest. It is well documented that a better choice than to simply sum up all sites within an area is to use a model that also is able to take advantage of spatial smoothing effects. Here, the technique of Bayesian model averaging BMA is used to improve the weighting of the single site forecasts for the estimation of the expected lumped output. In addition BMA deliver information for a probabilistic forecast. This scheme is applied here to a set of synthetic (modelled power output data based on wind speed measurements and forecasts) for 43 stations in Norway, separated into 7 groups. The BMA scheme was performed for each group.

01.07.2015 Pierre Pinson (DTU Denmark), Title: Towards future electricity markets with large penetration of renewable generation (14:45, Room: A5 0-055) 

The increased penetration of renewable energy generation in power systems and electricity has induced a number of challenges, from downward pressure on energy prices to worries about investment in necessary capacity for instance. More generally, the current design of electricity and ancillary service markets do not recognize and embrace the very nature of renewable energy generation from e.g. wind and solar power installations, with its variability and limited predictability. The needs to revisit such market design, as well as its potential implications, will be discussed. Examples will include concepts related to probabilistic offers (therefore revealing and accepting generation uncertainty), wind offering ancillary services, time and space coordination of markets, etc. Besides, we will insist on the necessity to bridge the gap between research on toy models and real-world large and complex systems, then introducing ongoing initiatives for relevant parties to work on a common EU dataset for renewables in electricity markets.

21.05.2015 Markus Bortolamedi, Title: Accounting for hidden energy dependency: The impact of energy embodied in traded goods on cross-country energy security assessments (17:00, Room: A1 0-008) 

Energy security ranks high on the policy agenda of many countries. Accordingly, to provide policy guidance, a large and growing body of literature has proposed metrics to measure security of (primary) energy supply which are then applied in cross-country energy security assessments. In general, the data used in these assessments are based on production-oriented energy accounting frameworks. In doing this, these studies neglect additional indirect foreign energy consumption - i.e. consumption of energy embodied in traded goods. This paper highlights this issue. It provides and applies a methodology that allows including indirect foreign energy consumption into commonly used energy security indicators. In particular, it shows that the inclusion of foreign primary energy embodied in traded goods does have an impact on the values of energy security indicators as well as on the results of cross-country energy security assessments - i.e. the indicator-speci?c country rankings.

05.05.2015 Hermann De Meer, Title: Renewable Power Supply: Smart or Risky? (16:00 - Offis F02)

The transformation of the energy system poses challenges and opportunities in various forms. In a decentralized setting, volatile supply and demand need to be matched, increasingly based on local conditions and policies. Voltage and reactive power management within the distribution grid are also demanding for more attention. Many of the means to overcome the challenges are based on flexibilities of loads and equipments as well as on ICT-based automated sensing and control techniques. An exploitation of application process flexibilities and an extended incorporation of ICT
in the distribution power grid, however, may also bare some additional risks such as by application process failures, by ICT failures, or by tempering through adversaries. In this seminar, we therefore focus on the exploitation of load flexibilities in smart environments on one side and on challenges of proactive risk management of critical infrastructures on the other side.

16.04.2015 Carsten Helm, Title: On the economics of electricity markets with renewable energies (17:00; Room: A5 1-159)

Intermittency of renewable energies like wind and solar leads to the following effect: when their supply is high, the price of energy is low, or even zero. Conversely, when prices are high, the supply of renewables is low. This constellations reduces the ability of renewables to earn money, and one might expect that this leads to too little investments in renewable capacities. Our paper shows that this intuition is wrong. In the absence of market restrictions - such as price ceilings - markets will lead to efficient choices of renewable and fossil capacities. We also investigate some properties of electricity markets with renewables. For example, we find that a reduction in the capacity costs initially leads to an exponential build-up of renewable capacities. However, the build-up rate falls dramatically once the market penetration is such that renewables are able to satisfy the whole demand at times of high availability. Thus, the most difficult stages of the energy transition are still to come. 

05.02.2015 Annika Schomburg (DWD), Title: Data assimilation developments within the DWD numerical weather prediction system aiming at the improvement of weather forecasts for the renewable energy sector

The renewable energy production in Germany is strongly growing. The availability of wind and photovoltaic power, however, is fluctuating in time due to its weather dependency. To ensure a stable and cost effective grid management, reliable weather predictions for 100m wind and solar radiation become more and more important. Two years ago, at DWD (Deutscher Wetterdienst) a project called EWeLiNE (Erstellung innovativer Wetter- und Leistungsprognosen für die Netzintegration wetterabhängiger Energieträger) has been launched, aiming at the improvement of the numerical weather prediction systems for the renewable energies wind and photvoltaic power. This is a collaboration with the Fraunhofer IWES institute, they work on the advancement of their power prediction systems using the DWD weather forecasts as input. In this talk an overview on the EWeLiNE project and the DWD numerical weather prediction systems will be given, with the main focus on data assimilation. The question to be answered will be „Why do we need data assimilation, what exactly is data assimilation?“ In this context the DWD data assimilation system for the COSMO model will be presented and some aspects of ongoing research towards the assimilation of additional data within the EWeLiNE project will be highlighted.

22.01.2015 Stephan Späth, Title: Statistische Nachbearbeitung von regionalen Ensemble-Windleistungsvorhersagen

Der Vortrag zeigt die statistische Nachbearbeitung von regionalen Ensemble-Windleistungsvorhersagen, die mit Hilfe des COSMO-DE EPS erstellt wurden. Systematische Fehler des Ensemblemittels und der Ensemblestreuung werden durch nichthomogene Gaußsche Regression korrigiert. Zeitlich konsistente Trajektorien werden mit Ensemble Copula Coupling erstellt und multivariat ausgewertet. Die Untersuchung wird durchgeführt für die Übertragungsnetze von Tennet, 50 Hertz, Amprion sowie für Gesamtdeutschland. Die statistische Nachbearbeitung zeigt große Verbesserungen insbesondere in der Ensemblestreuung. Die zeitlichen Korrelationen der regionalen Windleistung können nahezu korrekt vorhergesagt werden.

08.01.2015 Michael Richter, Title: Photovoltaics - Physics of a Thin-Film Solar Cell

There are manifold reasons why renewable energies are getting more and more important for our energy supply. To name but a few, climate change, limited fossil resources and not least the costs. The photovoltaics are proven to be the most promising type of renewable energy technology regarding the cost efficiency. All non-organic solar cells realize more or less the same physics, however, thin-film solar cells based on Cu(In,Ga)Se2 might be a prominent alternative to crystalline silicone solar cells in near future. The talk gives an overview about the main physical processes including loss mechanisms inside the solar cells.

31.07.2014 Gunther Wittstock, Title: Combinatorial Approaches in Electrochemistry

The lecture introduces the idea of high troughput experimentations and its application to electrochemical material science. The idea follows the concept of natural evolution in which a large number of property combinations are generated and tested to arrive at optimized functionalities. Such approaches have been very successful in live science where building block chemistry can be exploited for library generation and photonic technologies for readout. In electrochemical material science, this concept is more difficult to realize due to the special properties of interfaces between a solid and a liquid that has much more structural features than the liquid and the solid alone. Nevertheless, the large number of combination of suitable materials and the slow speed of discoveries based on deductive approaches has stimulated a growing interest in combinatorial approaches in the most developed research centers for electrochemistry world-wide. Existing methodical approaches for generation of libraries (typically in form of gradient or arrays) and high throughput characterization are explained with example from the fields of fuel cell catalysis, batteries and photoelectrochemical water splitting.

19.06.2014 Anna Mehrens, Title: Introduction to my PhD Topic: Mesoscale Wind Fluctuations

The save integration of wind energy into the power system requires high quality wind power predictions of wind fluctuations. To develop new prediction models, the atmospheric processes and conditions, which lead to strong wind fluctuations have to be known. Moreover a powerful and robust measure to characterize a heterogeneous wind field is required. This talk gives you an overview of different meteorological scales in the time and frequency domain. The focus is on mesoscale fluctuations with a time scale of minutes to hours. Therefore wind field measurements of a Lidar and WRF simulations of an offshore site with a size of several kilometers were used. An offshore site was chosen because convective cells which lead to strong mesoscale fluctuations occur more often above the open sea.

17.04.2014 Cornelius Steinbrink, Title: Smart Grid Co-Simulation across Heterogeneous Domains

The power grid of the future will be enhanced with information- and communication technology in order to guarantee stable power supply with a variety of more or less readily controllable components. This type of grid is typically called “smart grid”. For the research on these complex systems simulation tools are needed, e.g. the platform “mosaik” that allows steady state simulation of large scale scenarios with pre-existing grid component models. However, appropriate simulation environments for distribution grids also need means to analyze short term dynamic behavior like electromagnetic transients. Therefore we have coupled mosaik with a high performance real-time simulator. Challenges and objectives of the development of an improved interface between these two simulation domains are presented here.

20.03.2014 Heinz Welsch, Title: Using Subjective Well-Being Data for Energy Policy Analysis

Survey data on the life satisfaction of about 140,000 individuals in 25 European Countries, 2002-2011, were used to estimate the relationship between subjective well-being (SWB) and production shares of various electricity generation technologies. It is found that individuals’ SWB varies systematically and significantly with differences in the electricity mix across countries and across time. In particular, a greater share of solar and wind power relative to nuclear power is associated with greater SWB at all levels of income. The strength of this relationship has risen drastically after the Fukushima nuclear accident. These findings are taken to indicate a considerable preference for a clean and safe electricity supply in spite of higher costs.

06.03.2014 Justin Heinermann, Title: Efficient Nearest Neighbor Queries using GPUs

In this talk, we present a high-performance implementation of the nearest neighbors classification algorithm. The original field of research was data mining in huge astronomical catalogues, but the work can be easily adapted for wind power prediction. Recent studies have demonstrated the effectiveness of prototype-based learning schemes such as simple nearest neighbor models. However, although being among the most computationally efficient methods for such settings, applying these models on large data sets can easily take hours or even days. In this work, we investigate the practical effectiveness of GPU-based approaches to accelerate such nearest neighbor queries in this context. Our experiments indicate that carefully tuned implementations of spatial search structures for such multi-core devices can significantly reduce the practical runtime.

20.02.2013 Mehrnaz Anvari, Title: Wind Farm Modelling

The renewable sources and their share in electricity production have been increased constantly in recent years. The renewable sources, such as wind, are influenced by atmospheric turbulence, which cause frequent extreme events in power output of wind turbines. Such events can influence the stability of power grids and especially can cause voltage instability in very short term scales. Therefore, understanding and characterization of stochastic behavior of power output has great scientific and practical importance. Our aim is to show that such extreme events are intelligible from interactions between wind turbines in different time intervals. For this purpose, we investigate the power synchronization of different turbines and show that synchronization may explain extreme events in this complex system.

23.01.2014 Björn Wolff, Title: Statistical Learning for Short-Term Photovoltaic Power Predictions

At the end of 2013 the installed capacity of photovoltaics (PV) in Germany reached almost 36 GW, a fourth of the worldwide installed capacity. Due to the high volatile nature of PV power, it becomes increasingly important for the grid integration to develop high quality forecast models. In this work, we employ statistical learning methods (i.e., k-nearest neighbors and support vector regression) for short-term PV power predictions. The main idea is to train models by learning from historical data to predict PV power output up to six hours. In the first part of this presentation, we focus on tuning model parameters and selecting the most important features for this task. After evaluating the performance of both models, the combination of the prediction models results in a even more robust and accurate hybrid model. The second part of the presentation shows an outlook on how statistical learning methods can be used to improve PV power predictions derived from satellite-based irradiance forecasting.

14.11.2013 Stephan Späth, Title: Uncertainty of Short-Term Wind Power Predictions

With increasing shares of installed wind power in Germany, reliable forecasts of wind power become more important. These forecasts suffer from different error sources. We show the evolution of typical wind power forecast, from Numerical Weather Prediction to final wind power time series for whole grid zones. Different sources of error in the several steps of transforming weather forecasts to forecasts of wind power are discussed. We use forecasts from the COSMO-DE EPS, a high-resolution numerical weather ensemble prediction system with 20 realisations to investigate Storm Christian,which hit Northern Germany in October 2013. It is shown that uncertainty information and rapid forecast updates every three hours were helpful for the integration of the largest amount of wind power produced in Germany to date. The ensemble Prediction System was able to predict a huge cutoff event in the Tennet Grid Zone several hours in advance.

31.10.2013 Benjamin Wahl, Title: Granger Causality

The basic principle of Granger causality is as follows: If the prediction of a time series X from its past is improved by adressing the past of another time series Y, then Y causes X. The additional assumption that X and Y describe the whole universe is to exclude the possibility that bot X and Y are driven by some other time series Z. The mathemacal formulation bases on linear stochastic processes. A nonparametric approach will be presented. It has been introduced by Dhamala et al. and bypasses an explicit estimation of the underliing autoregressive process. The method is intended to yield results on wind data. Nonlinear extensions of Granger causality exist but are not considered in this presentation. 

17.10.2013 Anja Ohsenbrügge, Title: Dynamische Regelleistungsbemessung in Future Smart Grids

The nuclear phase-out and the increased share of renewable distributed generation implicate new challenges for transforming the electric power system into an environmentally sustainable, reliable and cost-efficient system. An important component in transforming the electric power system is the reliable substitution of large controllable power plants by small generation units (photovoltaic systems, wind energy converters and combined heat and power plants (CHP)) without impairing the overall safety and quality of the energy supply. In particular, the process of control reserve and balancing power to cover power plant or prognosis faults has to be adapted to today´s complex decentralized structure. This complexity both includes the factors of influence on the reserve dimensioning and the certifiable level of reliability for its provision. The aim of this paper is to develop a dynamic strategy for reserve and reliability that factors in these altered circumstances, in particular the provision of ancillary services by decentralized self-organized coalitions of small active units.

10.10.2013 Wiebke Schulte, Title: Herstellung und Charakterisierung von Katalysatorbibliotheken

Der Vortrag gibt eine kurze Einführung in die üblichen Charakteriesierungsmethoden von Katalysatoren. Dabei handelt es sich um metallische Katalysatoren, die in alkalischen Brennstoffzellen zur Reduktion von Sauerstoff genutzt werden. Neben der Möglichkeit, jeweils einen Katalysator mit Hilfe der rotierenden Ring-Scheiben Elektrode zu untersuchen, können auch sogenannnte Katalysatorbibliotheken untersucht werden. Bei einer Katalysatorbibliothek handelt es sich um ein Substrat, auf das mehrere Katalysatoren unterschiedlicher Zusammensetzung aufgebracht sind. Erste Erfahrungen bei der Herstellung solcher Bibliohteken werden dargestellt. Bei der Charakterisierung stellt die elektrochemische Rastersondenspektroskopie eine vielfältig einsetzbare Methode dar. Durch Variation der Pulsprogramme, können unterschiedliche Parameter der Katalysatoren bestimmt werden.

12.09.2013 Thomas Luhmann, Title: Photogrammetrische Messtechnik für Anwendungen in der Windenergie

Der Vortrag gibt zunächst eine Einführung in die Grundprinzipien der Photogrammetrie, d.h. der Erfassung dreidimensionaler Informationen aus Bildern eines beliebigen Objektes. Dabei werden die Hauptaufgaben der Kamerakalibrierung, Bildorientierung, Korrespondenzanalyse und 3D-Berechnung erläutert. Besonderes Augenmerk wird auf die erreichbare Messgenauigkeit von 3D-Koordinaten gelegt, die in der Praxis im Bereich zwischen 1:20000 und 1:100000 der Objektgröße liegt. Es folgen praktische Beispiele aus dem Bereich der Windenergieanlagen, z.B. die photogrammetrische Vermessung von Turmfundamenten, Biegeversuchen an Rotorblättern oder Deformationsmessungen. Dabei können sowohl statische als auch dynamische Problemstellungen gelöst werden. Abschließend folgt eine Zusammenfassung des derzeit an der Jade Hochschule laufenden Forschungsprojektes WindScan, bei dem es um die Kombination photogrammetrischer Verfahren mit terrestrischem Laserscanning zur Erfassung von Rotorblättern im Betrieb geht.

22.08.2013 Nils André Treiber, Title: Wind Power Prediction with Machine Learning

Die Einspeisung von erneuerbaren Energien ist durch die hohen Fluktuationen bei Wind- und Solarenergie schwierig. Nur mit kurzfristigen Leistungsvorhersagen ist möglich, möglichst viel Regelenergie von schnell anlaufenden, konservativen Kraftwerken einzusparen. In dem Vortrag steht die Windleistungsvorhersage im Mittelpunkt. Nachdem die hier behandelten statistischen Modelle von den physikalischen abgegrenzt sind, wird das datengetriebene raumzeitliche Vorhersagemodell der Computational Intelligence Gruppe vorgestellt. Nach einer kurzen Einführung der dabei implementierten Regressionstechniken (kNN und SVR), werden die erhaltenen Vorhersagen für einzelne Windturbinen diskutiert. Im Anschluss wird der Frage nachgegangen, wie Vorhersagen für ganze Windparks unternommen werden können, indem unterschiedliche Kombinationen von Eingabemustern untersucht werden.

15.08.2013 Markus Bortolamedi, Title: Energiesicherheit - Gerechtfertigte Begründung einer Politikintervention in Energiemärkte

Politische Eingriffe in Energiemärkte werden heutzutage zunehmenden mit dem Argument der Sicherstellung der Energieversorgung gerechtfertigt. Energiesicherheit wird dabei meisthin als das Niveau der Importabhängigkeit von Primärenergie interpretiert. Von einem ökonomischen Standpunkt aus ist dieses Argument jedoch problematisch, da politische Eingriffe in den internationalen Handel, die darauf abzielen die Importabhängigkeit zu verringern, meist mit Wohlfahrtsverlusten verbunden sind. Eine alternative und aus ökonomischer Sicht bessere Begründung wäre ein Marktversagen in der privaten Risikovorsorge gegen Versorgungsstörungen. Marktversagen äußert sich dabei in einem gesellschaftlich ineffizienten Niveau der privaten Risikovorsorge. Ursache dieses Marktversagens können private Haftungsbeschränkungen, abweichende Risikoeinstellungen oder weitere äußere Einflussfaktoren sein, da diese den privaten Vorsorgeanreiz verzerren. [talk]

04.07.2013 Constantin Junk, Title: Short- and Medium-term Wind and Wind Power Prediction

The economic operation of wind farms and the safe integration of wind into the transmission grid requires high-quality wind power predictions for short and medium term forecast horizons. For the named forecast horizons, wind forecasts from numerical weather prediction models are an essential input for wind power predictions models, while it turned out that pure statistical approaches are not sufficient. Until some years ago, weather services provided end-users with deterministic forecasts which are single integrations of numerical weather prediction models. Since the initial conditions and the model physics and dynamics of the numerical weather prediction model introduce flow-dependent forecast uncertainties, a deterministic forecast is only one possible realisation of the future state of the atmosphere. Ensemble Prediction Systems from the European Center for Medium Range Weather Forecasts consider the initial and model uncertainties and allow the quantification of forecast uncertainties which can be of great value for wind energy applications e.g. for wind farm maintenance planning or for the design of trading strategies in future liberalized electricity markets. However, Ensemble Prediction Systems are not able to represent the entire range of forecast uncertainties and have systematic, location and flow dependent errors. These errors require a calibration of the probabilistic ensemble forecasts. The calibration models are applied to the uncalibrated ensemble forecasts of wind which are then transformed to wind power with an appropriate power curve model. Several univariate (bivariate) methods exist for the calibration of wind speed (wind vector) ensemble forecasts for short- and medium-term forecast horizons. We have compared state-of-the-art univariate and bivariate calibration techniques. A bivariate adaptive and recursive calibration method for wind vector ensembles has turned out to yield probabilistic forecasts with highest deterministic and probabilistic forecast skill.

20.06.2013 Peter Michalowski, Title: Aufbau und Charakterisierung von Festkörper-Dünnschicht-Batterien

Im ersten Teil des Vortrags wurden der prinzipielle Aufbau und die Funktionsweise von Akkumulatoren, im wissenschaftlichen Sprachgebrauch oft schlicht als Batterien bezeichnet, dargestellt. Dies beinhaltete die Einführung physikalischer Kenngrößen wie bspw. der gravimetrischen Ladungsdichte, der Zellspannung und der gravimetrischen Energiedichte. Hiervon ausgehend wurde die Verwendung der Li-Ionen-Technologie begründet und hinsichtlich ihrer Parameter mit anderen  Akkumulatoren (Blei, Ni-MH) verglichen. Über die Ansprache bestehender Probleme aktueller Li-Ionen-Batterien wurden die Besonderheiten und das Potenzial von Festkörper-Dünnschicht-Batterien hervorgehoben. Der zweite Teil des Vortrags umfasste die Skizzierung des eigenen Forschungsvorhabens mit Konzentration auf Kathode und Elektrolyt. Dabei wurden entsprechende Materialsysteme vorgestellt und verglichen. Die vorläufige Auswahl von LiMn2O4 als Kathodenmaterial wurde begründet; LiFePO4 als weitere Möglichkeit genannt. Die Präparation erster eigener Proben wurde beschrieben sowie dabei aufgetretene Probleme angesprochen. Es wurden Lichtmikroskop- und REM-Aufnahmen der hergestellten Schichten gezeigt. Ein Ausblick auf das geplante Vorgehen in naher Zukunft schloss den Vortrag ab.

13.06.2013 Oliver Kramer, Title: Wind Power Ramp Event Prediction with Support Vector Machines

Wind energy is playing an important part for ecologically friendly power supply. Important aspects for the integration of wind power into the grid are sudden and large changes known as wind power ramp events. In this work, we treat the wind power ramp event detection problem as classification problem, which we solve with support vector machines. Wind power features from neighbored turbines are used features in a spatio-temporal classification approach. Recursive feature selection is employed to illustrate how this approach is affected by the number of neighbored turbines. The problem of imbalanced training and test sets w.r.t. the number of no-ramp events is analyzed experimentally and the implications on practical ramp detection scenarios are discussed.

30.05.2013 Stefan Weitemeyer, Title: Energiesystem-Modellierung

Aktuelle Studien belegen, dass eine nahezu vollständig auf Erneuerbare Energien (EE) aufbauende Stromversorgung bis zum Jahr 2050 in Deutschland und Europa realisierbar ist. Bei dem geplanten massiven Ausbau erneuerbarer Energien bis zum Jahr 2050 ist aufgrund der natürlichen Fluktuationen der potentialstärksten Formen erneuerbarer Energieflüsse (Solar- und Windenergie) jedoch die Bereitstellung großer Stromspeicherkapazitäten eine der erforderlichen Maßnahmen zur Sicherung eines stabilen Netzbetriebs. Da dies mit hohen Kosten verbunden ist sollte der Bedarf an stationären elektrischen Speichern weitestgehend durch Ausschöpfung alternativer und kostengünstigerer Maßnahmen gering gehalten werden. Im Vortrag wurden hierzu verschiedene Ergebnisse aus dem Forschungsfeld der Energiesystem-Modellierung präsentiert, die den Einfluss von Maßnahmen wie Netzausbau, Aufbau von Überkapazitäten und des Optimalen Mixes zwischen Solar- und Windenergie untersuchen. Da diese Arbeiten eine Reihe entscheidender Zusammenhänge und Fragen hinsichtlich eines robusten Zielsystems der europäischen Stromversorgung nicht ausreichend abdecken, wurde zudem erste Ergebnisse sowie der Ausblick auf ein neues Modellwerkzeug präsentiert, welches im Rahmen einer SEE-Promotionsarbeit entwickelt wird.

15.08.2013 Christian Hinrichs, Title: Dezentrale Optimerung über interdependenten Suchräumen

Viele praxisrelevante Optimierungsprobleme lassen sich effizient zentral lösen, indem die dem Suchraum der gültigen Lösungen inhärenten Beschränkungen und Nebenbedingungen strukturell ausgenutzt werden. Auf diese Weise können ungültige oder ungünstige Lösungen schnell erkannt und von der weiteren Betrachtung ausgeschlossen werden, um somit die Suche nach optimalen Lösungen zu beschleunigen. Solche Ansätze basieren jedoch auf globalem Wissen und lassen sich nicht unmittelbar auf dezentrale Systeme übertragen, in denen der Suchraum in Form von disjunkt verteilten Komponenten vorliegt. Ein möglicher Ansatz wäre hier, die lokal verfügbare Information zu kommunizieren und zentral zu sammeln und auszuwerten. Dies ist jedoch nicht immer möglich oder wünschenswert. Eine weitere Einschränkung besteht darin, dass die verteilten Suchräume oftmals nicht unabhängig voneinander sind, sondern Interdependenzen besitzen und somit in Verbindung zueinander ausgewertet werden müssen, was eine kommunikationsarme parallele Suche nach optimalen Lösungen erschwert. Zu diesem Zweck wird ein schwarmbasiertes Verfahren entwickelt, welches sich Selbstorganisationstechniken aus der Natur zu Nutze macht. Im Vergleich zu existierenden populationsbasierten Heuristiken repräsentieren die Individuen des Schwarms hier jedoch keine vollständigen Lösungskandidaten für das Optimierungsproblem. Stattdessen wird jede Anlage, mit ihren Fahrplanalternativen als lokalem Suchraum, als autonomes Individuum betrachtet. Dieses exploriert seinen eigenen Suchraum hinsichtlich derjenigen Teillösung, die gemäß seines individuellen Wissens zu einer optimalen Gesamtlösung beiträgt. Die Individuen interagieren dabei über Nachbarschaftsbeziehungen und teilen ihre lokalen Informationen miteinander, um so ihre Entscheidungen iterativ zu optimieren. Mittels einer Ausbreitungsstrategie wird lokales Wissen zudem sukzessive erweitert. Auf diese Weise sucht das System zunächst parallel nach vielversprechenden Teillösungen, welche dann iterativ zu Gesamtlösungen kombiniert werden. Priorisierungsmechanismen sorgen dafür, dass die Individuen schließlich die beste gefundene Konfiguration auswählen und das System somit ohne