BA/PA - Improved Parallel Computing and Feasibility Determination for Mixed-Integer Linear Programming based Dispatch Optimization

Verbesserte Parallele Berechnung und Lösbarkeitsanalyse für eine Gemischt Ganzzahlig Lineare Betriebsoptimierung

Improved Parallel Computing and Feasibility Determination for Mixed-Integer Linear Programming based Dispatch Optimization

Background

Shipping is one of the most efficient transportation methods and at the same time one of the largest emitters of greenhouse gases. According to conservative forecasts, greenhouse gas emissions in shipping will increase by at least 150% by 2050 without additional measures. The strategic goals of the International Maritime Organization (IMO) counteract this trend, aiming to reduce emissions by 70% by 2040 compared to 2008 levels. Achieving these goals can only succeed through comprehensive measures for decarbonization and energy efficiency improvement. In addition to direct efficiency-improvement measures, such as optimizing hull and propeller geometries, improving combustion engines, or using alternative fuels, improving the architecture as well as distribution and supply strategies of a ship's various energy systems is promising. These improvement measures include, among others:

• Diesel-electric propulsion concepts that allow for a more energy-efficient operation by providing greater flexibility in engine selection and operation strategy. - The integration of alternative forms of power generation, such as fuel cells, photovoltaics, or wind-assisted propulsion, which reduce fossil fuel consumption.

• Sector coupling/combined heat and power generation between electrical and thermal systems, for example, through waste heat utilization using heat recovery steam generators on main engines or high-temperature fuel cells, to reduce the demand for additional thermal power generation.

• The integration of energy storage systems for electric or thermal energy, which enables decoupling of power generation and usage. Particularly during short-term peak loads, such as maneuvering, storage systems can stabilize the generator load and thus ensure more efficient operation.

The integration of these measures significantly increases the potential complexity of energy systems compared to conventional configurations with direct mechanical propulsion, shaft generators, and steam boilers. Although energy system design, even for conventional configurations, is a highly complex task, it generally occurs manually and iteratively, mostly based on the shipyard's experience. This approach can easily cause unconventional but potentially more efficient concepts to not be pursued further, as they are perceived as unproven and therefore risky. The great complexity and hesitant adoption of unconventional solutions make energy system design an ideal candidate for holistic optimization to increase energy efficiency. Such optimization, particularly considering the configuration, is however highly challenging due to the size and complexity of the systems and configuration options.

Task Description

At the Institute of Mechatronics in Mechanical Engineering, a methodology was developed to tackle the challenge of a holistic system optimization by using genetic and mixed-integer algorithms to optimize coupled ship energy systems - the Maritime Energy System Optimizer (MESO).

In the current implementation of MESO, the mixed-integer linear programming based dispatch optimization necessitates by far the largest amount of computational resources. As the dispatch optimization is part of the fitness calculation (= determining the quality of an individual in the genetic algorithm), it can be done largely in parallel for an entire generation of individuals. The effectiveness of this parallel computing strategy is however heavily reduced by the fact that determining INfeasibility of a dispatch optimization problem can often require significantly longer than actually solving a feasible problem.

Strategies to improve both, the parallel computing concept, and the infeasibility determination speed, are to be developed in this thesis. This development may include the following steps:

• Familiarization with the existing implementation and analysis of the challenges.

• Literature research on potential solvers and dynamic solver parametrization.

• Development of simple performance metrics for the optimization.

• Development of an improvement strategy.

• Implementation of the proposed strategy.

• Test of the proposed strategy on the TUHH HPC-Cluster and comparison to the existing implementation based on the performance metrics.

Contact

If you have any questions/suggestions/ideas, feel free to contact me: send mail to Mattis Molinski

Institute Website

Zeitplanung:

Checklist

• Introduction / tour in M4

• Urheberrechtsvereinbarung signed: https://www.tuhh.de/t3resources/sls/pdf/ZPA/Formulare_oeffentlich/Rechte_an_Abschlussarbeiten.pdf

• if applicable: signed confidential agreement

• official registration

Helpful links:

• Thesis Work / Studentische Arbeiten 

• Criteria for the evaluation of Bachelor's and Master's theses, handout version

• Laborordnung

• Spotlights / Kurzvorträge

• Templates (Latex, ppt etc.)

Document Upload Final Thesis / Dokumentenabgabe Abschlussdokument

File of final presentation / Dokumentenabgabe Abschlusspräsentation

Link for further files / Link für weitere Dokumente