What is the Problem?
Renewable energies such as PV and wind are unpredictable by nature to a large extent. Despite all the new forecasting methods, they still conservatively bid in the market and pay penalty to the ancillary services market based on causer pay.
Why is it Interesting?
If there is a way that the predictability of renewable generation can be improved, it helps plant operators as well as bring larger economic benefits to the plant owners. It would be more interesting if the proposed approach can be prediction-method-agnostic.
What is the Approach?
An optimal battery sizing methodology is proposed to improve renewable generation predictability using “Seasonal-Trend decomposition based on LOESS1 (STL)” decomposition technique, self-similarity estimation, and enhancing it through filtering.
What is New?
The idea of improving predictability and using the self-similarity concept in this context was offered for the first time. In addition, the STL decomposition technique application in this study was new in the power system engineering community. Moreover, the battery degradation over its useful lifetime was added to the optimisation problem.
How was it Tested?
The optimisation problem is solved using Gurobi in MATLAB. The battery sizes are examined in terms of power magnitude and battery SOC limits using the test dataset. Then, predictability improvement was tested by developing four prediction methods for three different horizons. Finally, the cost-effectiveness of the battery application was examined by standard LCOE calculation against average AS market prices.