Optimized Machine Learning Method for PV Power Prediction

Heck, Horst; Muntwyler, Urs; Schüpbach, Eva (6 September 2021). Optimized Machine Learning Method for PV Power Prediction Proceedings of the EU PVSEC 2021 (online), pp. 1341-1345. 10.4229/EUPVSEC20212021-6CO.11.2

Full text not available from this repository. (Request a copy)

Prediction of PV power is useful to estimate and plan power production, net stability, and own consumption. Input data for the predictions are physical parameters like solar irradiation (horizontal or inclined), temperature (of air and PV module), etc. To identify such input parameters, several methods have been proposed in the open literature. Physical models, statistical models, or a machine learning approach can be used to predict PV power. Here, we developed our own machine learning (ML) algorithm and trained it with AC-power data from our own PV monitoring network in Switzerland. Results are presented on how to optimize our algorithm in view of obtaining a precise prediction for PV power production. Such information is important for owners of PV plants to steer their own production/consumption. Especially own consumption of solar electricity in winter needs to be maximised, as PV will be enforced to successfully implement the Swiss Energy Strategy 2050.

Item Type:

Conference or Workshop Item (Paper)


School of Engineering and Computer Science > Institut für Optimierung und Datenanalyse IODA
School of Engineering and Computer Science > Institute for Energy and Mobility Research IEM > IEM / Photovoltaic systems


Heck, Horst;
Muntwyler, Urs and
Schüpbach, Eva


Q Science > QA Mathematics
T Technology > TK Electrical engineering. Electronics Nuclear engineering






Horst Heck

Date Deposited:

16 Nov 2021 10:09

Last Modified:

16 Nov 2021 10:09

Publisher DOI:




Actions (login required)

View Item View Item
Provide Feedback