Neural Network-based Load Forecasting in Distribution Grids for Predictive Energy Management Systems

Conference: International ETG Congress 2017 - International ETG Congress 2017
11/28/2017 - 11/29/2017 at Bonn, Deutschland

Proceedings: ETG-Fb. 155: International ETG Congress 2017

Pages: 6Language: englishTyp: PDF

Personal VDE Members are entitled to a 10% discount on this title

Authors:
Sauter, Patrick; Karg, Philipp; Pfeifer, Martin; Kluwe, Mathias; Hohmann, Soeren (Institute of Control Systems, Karlsruhe Institute of Technology, Karlsruhe, Germany)
Zimmerlin, Martin; Leibfried, Thomas (Institute of Electric Energy Systems and High-Voltage Technology, Karlsruhe Institute of Technology, Karlsruhe, Germany)

Abstract:
In this paper we present a new approach for load forecasting in distribution grids based on neural networks. The application focus of the method are predictive energy management systems with a model predictive control (MPC) approach. These control algorithms need predictions of load profiles from 15 minutes up to several days. Due to the moving horizon principle of MPC, the short-term prediction values are of higher importance than the long-term prediction values. Hence, our prediction method focuses in particular on the short-term prediction by taking instantaneous measurement values into account. With this approach, the method yields significantly better results than state of the art forecasting methods. This is shown by means of a case study with one year data from a German distribution grid, where the root-mean-squared error of the prediction can be reduced by 40-80 %.