Master Thesis Data-Driven Identification and Forecasting of Thermal Flexibility in Residential Heat Pump Systems

About this role

Residential heat pumps can provide electrical flexibility by leveraging the thermal storage capacity available within the entire energy system, including the building thermal mass, domestic hot water (DHW) storage and, where available, buffer storage tanks. Bosch already has concepts for thermal characterization, flexibility estimation, as well as energy consumption forecasting. Your thesis should therefore focus on making these concepts operational by identifying system-specific thermal characteristics from monitoring data and using them to forecast energy consumption, as well as available flexibility.

The question is: how can the thermal characteristics and usable storage capacities of residential heat pump energy systems, including building thermal mass, DHW storage, as well as buffer storage, be automatically identified from monitoring data, and how can this information be integrated into forecasting concepts to estimate electrical energy consumption, as well as available thermal flexibility? You will review existing Bosch modelling concepts for thermal characterization and flexibility estimation, as well as analyzing empirical field monitoring data to define data pipelines, input/output requirements and relevant data quality criteria.

In addition, you will automatically identify building- and storage-specific thermal parameters (e.g., thermal resistance, capacitance). You will also have the option of benchmarking data-driven identification approaches against Neural Ordinary Differential Equations (Neural ODEs) in terms of parameter plausibility, computational effort, as well as model accuracy. Furthermore, you will integrate the parameterized models into forecasting pipelines in order to estimate electrical load profiles and quantify time-dependent flexibility margins.

You will compare forecasting, as well as flexibility metrics at the level of individual buildings and aggregated fleets (including statistical analysis of the effects of aggregation). As part of your thesis, you will develop an automated end-to-end prototype workflow covering data pre-processing, model parameter identification, instantiation and forecast generation. You will validate the prototype against measurement data, documenting your findings, as well as the system's limitations.