PERFORMANCE TEST OF NEW PV-FORECASTING MODELS IN REALISTIC ENVIRONMENTS
L. Gaisberger, R. Höller, P. Rechberger, O. Voronych, W. Traunmüller, S. Moser - PERFORMANCE TEST OF NEW PV-FORECASTING MODELS IN REALISTIC ENVIRONMENTS - EU PVSEC 2019, Marseille, Frankreich, 2019, pp. 1432-1436
In this paper new methods of PV-forecasting are evaluated in different simulation studies to calculate
their performance in terms of economic value. The focus lies on energy management in residential PV-systems with
battery storage and micro-grids powered by different sources.
The micro-grid simulation was built up without physical models and consists of two diesel generators, (gensets) a PVsystem
and a fixed load profile. The results show that the necessary spinning reserve can be reduced by the use of PVpower
forecasting. This leads to better utilization of the PV-System as well as cost reductions in comparison to systems
without forecast. The total energy costs can be lowered by around 3.3 % to 3.4 % depending on the PV generation
profile and the forecast quality. These results apply only for the chosen system setup, whereas for larger PV arrays or
higher number of gensets the economic benefit is expected to be higher.
The second application investigated in this paper is a residential PV-system with an energy storage. This simulation
model was built up with a physical inverter and battery model. An energy management controller optimizing for
financial benefits using forecasts resulted in annual electricity costs lowered by 2.1 % to 2.5 % compared to a standard
storage battery controller. We assumed that the load profile was assumed to be predicted perfectly. This moderate
results arise out of the tariff gap between electricity consumption (duties and taxes are added) and feed-in, leaving little
scope for optimization.