Day-Ahead Forecasting of Solar Irradiance: KNN-Based Ensembles
| Autor: | Kamarianakis, Y., Pantazis, Y., Kalligiannaki, E., Kotsovos, K., Gereige, I., Abdullah, M., Tzavaras, A., Katsaounis, T. |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2022 |
| Předmět: | |
| DOI: | 10.4229/wcpec-82022-4bv.4.1 |
| Popis: | 8th World Conference on Photovoltaic Energy Conversion; 1248-1252 This work presents a novel, outlier-robust forecasting scheme, which is derived by combining two specifications. Specifically, a recently proposed time series KNN (TSKNN) ensemble procedure is combined with splinebased daily profiles (SDP), estimated via median regression. An experiment that mimics real-world implementation of an online system is performed, to evaluate day-ahead solar irradiance forecasts of the proposed method, dubbed TSKKNR0, versus TSKNN, SDP and the persistence model. Alternative specifications are evaluated with a new relative error metric which is outlier-robust. The analyzed measurements are collected at KAUST, Thuwal, Saudi Arabia for a full year, with 10-minute granularity. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|