Navigation

Colloquium lecture: 17. November 2020, Philipp Winklmann

Bild der Präsentationsfläche

Datenanalyse von realen Ladevorgängen von Elektrofahrzeugen /
Vorhersage des Nutzerverhaltens unter Verwendung von ML-Verfahren

It is important to understand and predict the charging behavior of electric vehicle owners in order to use EVs as potentials. Otherwise, the ever-increasing number of approved electric vehicles can quickly become a problem for the power grid. The most important aspects of this charging behavior are the duration until the user will pick up his vehicle and the amount of energy the vehicle can absorb. These data must be estimated in order to plan the electric vehicle and thus avoid peak demand.
In this thesis a comparatively new data set is investigated and in the second part used as a basis for the evaluation of different prediction techniques. We compare the results of several regressors and cluster the loading sessions according to different aspects in order to make the most ideal predictions possible. In addition to achieving the best possible results, another goal of this work is to investigate how much effort is required to achieve good results.
In the course of the investigations it is shown that the results are improved by checking different features, clusters and other factors. However, the greatest influence is the removal of data that deviates from normal behavior. When clustering by UserIDs, it is noticeable that by setting high demands on a cluster, 75% of the data falls into a residual cluster that is difficult to estimate, but the remaining 25% have less than 25% of the initial error.
Since the high demands are loading sessions that a user must have performed, it can be assumed that correspondingly regularly visited parking lots achieve very good predictive values and can thus optimally distribute the loading of cars.

 

 

Zeit: 10:15 Uhr

Zoom-Meeting

————————————

Zoom-Meeting beitreten
https://fau.zoom.us/j/91355939802?pwd=V3AxUll3K01lU0ErNTBQYUV5YVFjQT09

Meeting-ID: 913 5593 9802
Kenncode: 871417
Schnelleinwahl mobil
+12532158782,,91355939802#,,,,,,0#,,871417# Vereinigte Staaten von Amerika (Tacoma)
+13017158592,,91355939802#,,,,,,0#,,871417# Vereinigte Staaten von Amerika (Germantown)

Einwahl nach aktuellem Standort
+1 253 215 8782 Vereinigte Staaten von Amerika (Tacoma)
+1 301 715 8592 Vereinigte Staaten von Amerika (Germantown)
+1 312 626 6799 Vereinigte Staaten von Amerika (Chicago)
+1 346 248 7799 Vereinigte Staaten von Amerika (Houston)
+1 669 900 6833 Vereinigte Staaten von Amerika (San Jose)
+1 929 205 6099 Vereinigte Staaten von Amerika (New York)
Meeting-ID: 913 5593 9802
Kenncode: 871417
Ortseinwahl suchen: https://fau.zoom.us/u/aD4Tu0ds2

Über SIP beitreten
91355939802@zoomcrc.com

Über H.323 beitreten
162.255.37.11 (USA
Westen)
162.255.36.11 (USA
Osten)
213.19.144.110 (Amsterdam
Niederlande)
213.244.140.110 (Deutschland)
Kenncode: 871417
Meeting-ID: 913 5593 9802