Project
Development of a Deterministic PV Power Forecasting Workflow for Household Load Management
Idea and context
REgy was developed as an energy-coaching app concept intended to encourage household electricity consumption in line with renewable-energy availability.
To that end, a workflow for PV feed-in forecasting was first developed. The app remained at concept and prototype stage.
Forecasting workflow
A multi-day photovoltaic power-forecasting model was implemented in Python using pvlib and German Weather Service numerical weather prediction data. The prototype was extended and evaluated in a supervised bachelor’s thesis at OTH Regensburg through comparison with measured PV output of three PV plants and the commercial forecasting service Solcast.
Results
The supervised bachelor’s thesis enabled a comprehensive analysis of the predictive model’s performance. The table below provides a summary. The results show that the workflow was able to compete with commercial software.
nMAE
| Location | REgy | Solcast |
|---|---|---|
| A | 5,28 % | 5,52 % |
| B | 3,75 % | 3,63 % |
| C | 2,49 % |
nRMSE
| Location | REgy | Solcast |
|---|---|---|
| A | 9,43 % | 8,65 % |
| B | 7,05 % | 7,26 % |
| C | 5,19 % |