Evaluation of Regression Algorithms in Residential Energy Consumption Prediction

Since electrical energy is generated and consumed simultaneously, predicting the energy consumption helps network operators to optimize their planning assuring stable supply of energy. Residential consumers, in particular, are subject to investigation as they share a large proportion of the total co...

Full description

Saved in:
Bibliographic Details
Published in:2019 3rd European Conference on Electrical Engineering and Computer Science (EECS) pp. 22 - 25
Main Authors: Schirmer, Pascal A., Mporas, Iosif, Potamitis, Ilyas
Format: Conference Proceeding
Language:English
Published: IEEE 01.12.2019
Subjects:
Energy consumption
Energy Consumption Prediction
Linear regression
Load Forecasting
Load modeling
Prediction algorithms
Predictive models
Regression Models
Task analysis
Training
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Since electrical energy is generated and consumed simultaneously, predicting the energy consumption helps network operators to optimize their planning assuring stable supply of energy. Residential consumers, in particular, are subject to investigation as they share a large proportion of the total consumed energy and have irregular components in their energy consumption signal. In this article we evaluate the performance of different regression methods in the energy prediction task. Specifically Linear Regression, Decision Trees, Deep Neural Networks, Recurrent Neural Networks, Gated Recurrent Units and Long Short Time Memory are evaluated in terms of their performance in predicting residential energy consumption. For the evaluation of the regression algorithms a large scale dataset monitoring London households for a period of several years was used. It was shown that LSTM regression model outperforms all other regression algorithms improving the accuracy of the predictor and lowering the mean absolute error up to 26.7% when compared to the baseline Linear Regression predictor.
DOI:10.1109/EECS49779.2019.00018