Thesis - General comparison of ML methods for Li-Ion battery voltage prediction - Graz, Österreich - AVL List GmbH

Beschreibung

In the automotive industry, SOC (State of Charge) and SOH (State of Health) estimates play an essential role for electric vehicles. These estimates increase safety and improve charging efficiency. We use machine learning algorithms to improve these estimations and thus promote progress in battery technology for a more efficient and sustainable automotive future.

• Modeling of an electrical equivalent circuit of a battery (ECC)
  • State Space Model (SS-Model) on time invariant model parameters (R & C constant)
• Implementation and comparison of different ML algorithms
  • like NN, LSTM, Decision Trees, Gaussian Processes, Feature Engineering
  • for system identification to predict the output voltage of a Li-Ion battery
  • the ML algorithms are to be trained and validated on a synthetic data set generated by an ECC
• Comparison of at least two different ML algorithms (NN mandatory) to
  • R & C parameter estimation
• Time variant model parameters (R & C) (optional)
• How can prior physical knowledge be used to improve the prediction (opt.)
• Ongoing studies in the fields of Electrical Engineering, Computer Science/Data Science or Digital Engineering
• Good knowledge of English
• Programming skills in Python
• Knowledge of optimization methods and machine learning