Student Research Project / Master ThesisMachine Learning & Artificial Intelligence applied to Short-Term Photovoltaic Power Forecasting
Contact: Martín Herrerías < herrerias@hlrs.de >
Background:
Accurate short-term and very-short-term renewable energy production forecasts are critical instruments for the energy transition, as they support grid operators, plant owners, and energy traders in coping with the natural variability of the renewable resources.In the framework of the HyForPV project, HLRS is combining hybrid irradiance forecasts (satellite + numerical weather models + sky-imagers) with very detailed Photovoltaic (PV) Plant physical models to improve the accuracy of short-term regional PV production forecasts. A natural benchmark/extension to this approach is to use purely data-driven methods or hybrid statistical-physical methods; combining existing irradiance forecasts, site measurements, and physical model features, to provide fast and accurate estimates of PV power production.
Scope: Review of the extensive literature on ML & AI methods applied to short-term PV forecasting. Identification of adequate performance criteria. Selection, implementation, and cross comparison of the most promising methods.Detailed documentation (code, methods, and results).
Profile: Fluency in the programming language of your choice.Experience with ML & AI methods (ANN’s, SVM’s, autoregressive models,…).Fluent English in speaking and writing.Basic understanding of PV systems (favourable).Experience with MATLAB and Linux environments (favourable).