Description

  • Participants of the course are allocated individual research topics to work over the seminar course. Group works are possible.
  • A kickoff session will be conducted to introduce the course and topics.
  • Attendance is only mandatory on the proposal and final presentation dates.
  • Regular follow-up meetings with the participants and supervisor are held to aid the participants with the seminar tasks.

The seminar tries to mimic the event flow of a conference publication. Therefore, the deliverables are the following:

  • a scientific paper (submitted as the seminar report).
  • a conference-style final presentation.

More details on ECTS, learning outcomes, etc is offered at the kickoff session and later on moodle.

Topic Allocation

To apply for topics, the top 3 topics (please provide the acronyms) and the current transcript of records must be sent to a.k.schneider@tum.de. Applications will only be accepted from 27.09.24. Information on deadlines can be found in the next section.

Process and Time Line (Winter semester 2024/25)

Topics are chosen in two blocks. Students who want to be sure of getting a fixed place in the seminar before the start of the semester must apply by 04.10. (12 noon). These students will be informed on 07.10. which topic they will be working on.

Students who want to apply later can do so until 18.10. (12 noon) and will then be informed on 21.10. whether they could still get a place in the seminar.

The most important dates for the seminar are summarized below:

Registration period: 27.09.24 - 18.10.24.

First deadline for submitting the top 3 topic requests: 04.10.24, 12 noon.

Assignment of topics block 1: 07.10.24

Kick-off (introduction to the seminar, presentation of the topics still available): 16.10.24, 3pm-4:30pm, room N3815.

2. deadline for submission of the top 3 topic requests: 18.10.24, 12 noon.

Assignment of topics block 2: 21.10.24

Proposal Presentations: November 06 & 13, 2024, 3pm-4:30pm, N3815.

Final Presentations: January 29 & February 05, 2025, 3pm-4:30pm, N3815.

NOTE: THE FIRST APPLICATION ROUND IS CLOSED. ACRONYMS OF TOPICS WHICH ARE TAKEN ARE MARKED IN RED, STILL AVAILABLE TOPICS FOR APPLICATION ROUND 2 ARE MARKED IN GREEN.

AcronymTopic descriptionRequired skillsContactTime Topic Added

DEEP_NILM

Deep Learning for Energy Disaggregation

Research method: Prototyping

Research question:

  • How can Deep Learning be applied to Energy Disaggregation / NILM?
  • How does Deep Learning perform compared to other methods?

Possible approach

  • Understand state-of-the-art based on literature
  • Develop efficient Python code to train DL models on public NILM datasets
  • Evaluate performance by benchmarking with methods described in literature


  • Python programming skills

Christoph Goebel

christoph.goebel@tum.de

09/2023

DEEP_HEAT

Deep Reinforcement Learning for HVAC Control

Research method: Prototyping

Research question:

  • How can Deep Reinforcement Learning be applied to HVAC Control problems?
  • How does Deep Reinforcement Learning perform compared to other methods?

Possible approach

  • Understand state-of-the-art based on literature
  • Develop efficient Python code to train DRL models using HVAC model
  • Evaluate performance by benchmarking with methods described in literature


  • Python programming skills

Christoph Goebel

christoph.goebel@tum.de

09/2023

ES_EVAL

Techno-Economic Evaluation of Combined Energy Storage Technologies in Neighborhoods

Research method: Data Analysis & Modeling

Research question:

  • How can different types of energy storage be combined in a neighborhood setting?
  • How can different scenarios be compared technically and economically?

Possible approach:

  • Understand state-of-the-art based on literature
  • Develop models to understand technical limitations and economics of energy storage
  • Define scenarios and investigate potential of different types of energy storage and their combinations


  • Python programming skills
  • Background in energy system modelling

Christoph Goebel

christoph.goebel@tum.de

04/2024

P2P_ENERGY_TRADING

Decentralized P2P Energy Trading Under Network Constraints

Research method: Prototyping

Research question:

  • How can peer to peer energy trading in distribution grids be realized while respecting physical constraints?
  • Which approaches exist?
  • How can these approaches be benchmarked using realistic distribution grid models?

Possible approach

  • Understand state-of-the-art based on literature
  • Develop efficient Python or Julia code implementing existing methods
  • Evaluate performance of methods using realistic models of distribution grids


  • Python/ Julia programming skills

Christoph Goebel

christoph.goebel@tum.de

09/2024

ELECROTHERMAL_DEMAND

Containerized Implementation of Integrated Electrothermal Building Energy Demand Model

Research method: Prototyping

Research question:

  • How can electricity and heat demand be concurrently simulated?
  • How can the simulation be packaged into a Docker container and controlled via APIs?

Possible approach

  • Understand state-of-the-art models based on literature
  • Develop efficient Python or Julia code implementing one of the methods
  • Package code in a Docker container and evaluate the implementation via EMS use cases


  • Python/ Julia programming skills

Christoph Goebel

christoph.goebel@tum.de

09/2024

MULTI_PERIOD_OPF

Computing Multi-Period Optimal Power Flow in Distribution Grids

Research method: Prototyping

Research question:

  • Which methods can be used to calculate multi-period optimal power flow in distribution grids?
  • How do these methods scale for different distribution grid sizes?

Possible approach

  • Understand state-of-the-art models based on literature
  • Formulate mathematical optimization problem
  • Develop efficient Python or Julia code using state-of-the-art methods to solve the problem
  • Evaluate solution method using realistic distribution grids with solar, load, and storage


  • Python/ Julia programming skills

Christoph Goebel

christoph.goebel@tum.de

09/2024

PINN_LOADFLOW

Benchmarking Physics-Informed Neural Networks for Distribution Grid Load Flow Calculation

Research method: prototyping

Research question:

  • How can the method of physics-informed NNs be used to calculate load flows in power distribution systems?
  • How do these methods perform compared to standard numerical optimization methods?

Possible approach:

  • Understand state-of-the-art models based on literature
  • Develop efficient Python or Julia code implementing one of the methods
  • Compare PINN-based approach to numerical optimization based on quality and solving speed


  • Python/ Julia programming skills
  • First experience with deep learning

Christoph Goebel

christoph.goebel@tum.de

09/2024

BAT_AGE_PRED

Machine learning-based prediction of battery aging

Research method: prototyping

Research question:

  • How to predict lithium-ion battery aging using ML models?
  • How performant are ML models compared to other methods?

Possible approach:

  • Understand state-of-the-art based on literature
  • Develop an efficient Python program that fits various ML models on battery aging data
  • Benchmarking of the developed models
  • Python/ Julia programming skills
  • First experience with deep learning

Christoph Goebel

christoph.goebel@tum.de

09/2023

RC_MODELS

R-C Models for Heat Transfer in Buildings

Research method: Prototyping

Research question:

  • How to design an easy-to-use Python library for fitting R-C-models of the heat transfer in buildings?
  • How to use fitted models to simulate buildings in control environments?

Possible approach

  • Understand state-of-the-art based on literature
  • Develop efficient Python code to fit R-C models
  • Evaluate implementation, e.g., by benchmarking results and evaluating computational complexity
  • Python/ Julia programming skills

Christoph Goebel

christoph.goebel@tum.de

09/2023

RC_MODEL_FITTING

Using EnergyPlus to fit R-C Models

Research method: Prototyping

Research question

  • How can EnergyPlus models be approximated by R-C models?
  • How effective and efficient is the approach?   

Possible approach

  • Learn to work with EnergyPlus
  • Develop efficient Python code fit R-C models to output data
  • Benchmark method
  • Python/ Julia programming skills

Christoph Goebel

christoph.goebel@tum.de

09/2023

PINN_STATE_ESTIMATION

Benchmarking Physics-Informed Neural Networks for Distribution Grid State Estimation 

Research method: Prototyping

Research question

  • How can the method of physics-informed NNs be used to predict the state of power distribution systems?
  • How do these methods perform compared to standard numerical optimization methods?

Possible approach:

  • Understand state-of-the-art models based on literature
  • Develop efficient Python or Julia code implementing one of the methods
  • Compare PINN-based approach to numerical optimization based on quality and solving speed


  • Python/ Julia programming skills
  • First experience with deep learning

Christoph Goebel

christoph.goebel@tum.de


09/2024

WPF_USECASES

Identifying Use Cases for Ultra-Short-Term Wind Power Forecasting

Research method: Literature Research

Research question:

  • Who are ultra-short-term (<1h) wind power forecasting stakeholders? 
  • In what applications are ultra-short-term wind power forecasts required?
  • How can the benefit of improving wind power forecasts in these applications be quantified?

Possible approach

  • Thorough literature research on, e.g., energy markets, trading wind energy, operating wind power plants, etc.
  • Identification and characterization of practical use cases for ultra-short-term wind power forecasting.
  • Propose of a quantification method for the value of wind power forecasts in two to three of your identified use cases.
  • Curiosity in wind energy and energy markets
  • Structured way of working

Annika Schneider

a.k.schneider@tum.de 

09/2024

WPF_BENCH

Benchmarking Wind Power Forecasting Methods

Research Method: Literature Review (& Prototyping)

Research Question:

  • How are wind power forecasting models benchmarked in state-of-the-art studies?
  • How can we set up a fair, easy-to-use wind power forecasting benchmarking framework?

Possible Approach:

  • Identifying state-of-the-art wind power forecasting models
  • Identifying today's approaches on how to benchmark new wind power forecasting models
  • Proposal (and implementation) of a wind power forecasting benchmarking framework
  • Curiosity in wind energy
  • High interest in deep learning time series forecasting methods
  • Structured way of working
  • Optional: basic Python skills

Annika Schneider

a.k.schneider@tum.de 

09/2024

COMP_ENERGY

Evaluating the Energy Impact of Computation for Machine Learning

Research method: Literature Review / Prototyping

Research question:

  • How is energy estimation of computation done in research and how do the methods compare?

  • For ML applications, does faster training always mean more energy efficient, or are there other model or pipeline design considerations?
  • What are the popular libraries for measuring energy impact of computation for python and what do they provide?
  • (Stretch) Using one of these libraries, can we benchmark and compare the energy impact of one or more ML models in the EMT context (ex. solar or wind power prediction)? 

Possible approach:

  • Read and review papers on the environmental impacts of Machine Learning

  • Provide summaries of methods and develop guidelines on the types of ML model design, configurations, parameters, or algorithms that lead to lower energy consumption

  • Look for popular power estimation libraries from the literature and on GitHub
  • (Stretch) Use one of these libraries to benchmark state of the art ML techniques and document the results

Interest in Computational Energy

Interest in ML and computer architectures 

Basic Python skills

Ehimare Okoyomon
e.okoyomon@tum.de

04/2024

IMPL_HP

Implementation of a Paper with Heat Pump Model and Thermal Storage

Research method: Prototyping

Research question:

  • How can MILP be used in practice to optimize the design and operation of heat pumps?
  • Can the results of the paper be reproduced with public data?

Possible approach:

  • Programming skills in Python
  • Knowledge of linear optimization helpful

Elgin Kollnig
elgin.kollnig@tum.de


Simon Zollner
simon.zollner@tum.de


04/2024

IMPL_AGING

Implementation of various Battery Ageing Models in an Energy Management System

Research method: Prototyping

Research question:

  • How can different battery ageing models be compared fairly?
  • How can approaches from the literature be implemented and compared with each other in a prototype-based and reproducible way?

Possible approach:

 

  • Programming skills in Python
  • Knowledge of linear optimization
  • Interest in batteries

Elgin Kollnig
elgin.kollnig@tum.de


Simon Zollner
simon.zollner@tum.de


04/2024

GRID_TARIFF

Comparison of the influence of different variable grid tariffs

Research method: Literature review

Research question:

  • What options are there for designing variable grid tariffs? 
  • What effects do the different variable grid tariffs have on grid utilization and profitability for end consumers?

Possible approach:

  • Identification of different variable grid tariffs
  • Identification of the advantages and disadvantages of the different grid tariffs
  • (Development of new options for variable grid tariffs)

Elgin Kollnig
elgin.kollnig@tum.de

09/2024

FED_HOU

Federated machine learning for household load forecasting

Research method: Prototyping

Research question:

  • Can we forecast household load without seeing a household's sensitive data?
  • Which forecasting models are suitable for a federated setting?

Possible approach:

  • Literature review on federated forecasting
  • Find & preprocess suitable datasets
  • Build federated forecasting pipeline
  • Compare results vs. non-federated forecasting methods
  • Programming skills in Python
  • Basic statistics/ML knowledge

Jan Marco Ruiz de Vargas
janmarco.ruiz@tum.de

04/2024

COR_ENCR

Fast computation of correlation measures with homomorphic encryption

Research method: modeling

Research question:

  • Can two parties owning a data vector compute a correlation measure together, without revealing their private data to each other?
  • Can we devise a (partial/ fully) homomorphic encryption scheme to calculate various correlation measures?
  • How fast can this scheme be?

Possible approach:

  • Understand homomorphic encryption
  • Understand correlation measures
  • Model communication & computation protocol

 


Jan Marco Ruiz de Vargas
janmarco.ruiz@tum.de

09/2024

SINDY_BAT

SINDy (Sparse Identification of Nonlinear Dynamical Systems)-based Battery Aging Model

Research method: Prototyping

Research question:

  • How to implement  SINDy to formulate battery aging problem?
  • How to design the  a feature library theta for SINDy model, and how different could the results be?
  • How does the SINDy model perform comparing to other battery aging model (e.g. NNs / other symbolic regression model etc..)?

Possible approach:

  • Literature review on SINDy and battery aging prediction
  • Implement SINDy model and train with existing dataset
  • Compare the results with the existing battery aging model.
  • Programming skills in Python
  • Basic ML and/or Battery aging knowledge

Sheng Yin
sheng.yin@tum.de

09/2024
MARL_EMS

Multi Agent Reinforcement Learning (MARL) for Power Split Control

Research method: Prototyping

Research question:

  • How can MARL solve power split control problem?
  • Can MARL out-perform centralized RL?

Possible approach:

  • Literature review on MARL and RL in power split control
  • Build RL and MARL environment for Power split Control
  • Implement Basic RL and MARL algorithm to compare the results
  • Programming skills in Python
  • Basic ML and/or RL knowledge
Sheng Yin
sheng.yin@tum.de
09/2024


  • No labels