Functional clustering and linear regression for peak load forecasting

Goia, Aldo; May, Caterina; Fusai, Gianluca

doi:10.1016/j.ijforecast.2009.05.015

“…Linear regression model proves to be very useful for predicting short-term peak loads [10,29] and annual loads [43]. The later research uses support vector regression and various functional linear regression models for the actual prediction.…”

Section: Regression Modelsmentioning

confidence: 99%

A pattern recognition approach for peak prediction of electrical consumption

Goodwin

¹

,

Yazidi

²

2016

ICA

View full text Add to dashboard Cite

Abstract.Demand peaks in electrical power consumptions pose serious challenges for energy companies as these are typically unforeseen and require the net to support abnormally high consumption levels. Such peaks can be regulated in smart energy grids with the introduction of relatively simple techniques such as load balancing and smart pricing strategies. This is, however, difficult in practice because it requires prediction of peaks prior to their actual occurrence.While most studies formulate the problem as an estimation problem, we take a radically different approach and formulate it as a classical pattern recognition problem. Further, the paper applies classification methods to solve the problem and applies these with real-life data from a Norwegian smart grid pilot project. Some of the key findings are that the algorithms can accurately detect 80% of energy consumption peaks up to one week ahead of time. Furthermore, we introduce a novel Learning Automata based approaches for selecting the optimal prediction model from a pool of models in an online fashion.

show abstract

“…The first approach is a straightforward solution introduced for comparison purposes. This is continued with standard classification techniques in-spired by similar work on load forecasting [10,23,19,6]. Lastly, we present hierarchical classification solutions inspired by techniques presented in the literature on rare item classification [17,35,45].…”

Section: Problem Setting and Modellingmentioning

confidence: 99%

A Pattern Recognition Approach for Peak Prediction of Electrical Consumption

Goodwin

¹

,

Yazidi

²

2014

Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications

View full text Add to dashboard Cite

Abstract.Demand peaks in electrical power consumptions pose serious challenges for energy companies as these are typically unforeseen and require the net to support abnormally high consumption levels. Such peaks can be regulated in smart energy grids with the introduction of relatively simple techniques such as load balancing and smart pricing strategies. This is, however, difficult in practice because it requires prediction of peaks prior to their actual occurrence.While most studies formulate the problem as an estimation problem, we take a radically different approach and formulate it as a classical pattern recognition problem. Further, the paper applies classification methods to solve the problem and applies these with real-life data from a Norwegian smart grid pilot project. Some of the key findings are that the algorithms can accurately detect 80% of energy consumption peaks up to one week ahead of time. Furthermore, we introduce a novel Learning Automata based approaches for selecting the optimal prediction model from a pool of models in an online fashion.

show abstract

“…The literature supports the use of clustering prior to the use of a clustering algorithm. The work in [13] clusters load curves using K-means clustering for forecasting short-term daily peak loads in a heat system. They state that the goal of the clustering is to find characteristic patterns that determine changes in demand peaks.…”

Section: Links To Forecastingmentioning

confidence: 99%

Analysis of clustering techniques on load profiles for electrical distribution

Akperi

¹

,

Matthews

²

2014

2014 International Conference on Power System Technology

10

0

View full text Add to dashboard Cite

any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. Additional information:Use policyThe full-text may be used and/or reproduced, and given to third parties in any format or medium, without prior permission or charge, for personal research or study, educational, or not-for-prot purposes provided that:• a full bibliographic reference is made to the original source • a link is made to the metadata record in DRO • the full-text is not changed in any way The full-text must not be sold in any format or medium without the formal permission of the copyright holders.Please consult the full DRO policy for further details. Abstract-The classification of electrical load profiles has become increasingly important as a driver for distribution companies in understanding substation data. The daily load profile can often give great insight into the types of customers connected to the substation and can assist with developing a long-term forecast. The literature in this area often uses data mining and clustering techniques to determine a load diagram representative for a subset of customers or substations. The type of technique used can often lead to representative load diagrams of unique shapes with differing numbers of customers belonging to each group. This paper analyses clustering techniques on representative load diagrams for primary substations at the distribution level. In particular, this paper will analyse clustering techniques in terms of their performance and effect on load profile groupings. The results show that K-means clustering showed the best performance in generating unique, well-populated cluster groups. This gives a greater understanding of the divisions between substations which can be used for future forecasting.

show abstract

Functional clustering and linear regression for peak load forecasting

Cited by 170 publications

References 14 publications

A pattern recognition approach for peak prediction of electrical consumption

A pattern recognition approach for peak prediction of electrical consumption

A Pattern Recognition Approach for Peak Prediction of Electrical Consumption

Analysis of clustering techniques on load profiles for electrical distribution

Contact Info

Product

Resources

About