Improving the performance of machine learning models by integrating partly physical control response models in short-term forecasting of aggregated power system loads

Pekka Koponen, Harri Niska, Reino Huusko

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

Abstract

Combining the strengths of different modelling approaches and various information sources is studied in short-term forecasting of aggregated electrical loads that are controllable and include e.g. thermal storage capacity. Measurement data driven models tend to fail in forecasting power during rare situations such as dynamic control actions and extreme weather conditions. The thermal dynamics of the loads, large outdoor temperature variations, and changes in the technologies contribute to this challenge. Here we study a model integration approach using field trial data covering about 7000 houses and 27 months. Control responses and load saturation are forecast using a physically based structure. The residual is forecast with a machine learning model designed and tuned to learn also system dynamics. The load forecast is the sum of these component forecasts. The forecasting accuracy of this hybrid method is compared with using the machine learning alone. The results show improvement in the accuracy
Original languageEnglish
Title of host publicationProceedings ITISE 2017. Granada, 18-20, September, 2017.
EditorsOlga Valenzuela, Ignacio Rojas
PublisherUniversity of Granada
Pages795-806
Volume2
ISBN (Print)978-84-17293-01-7
Publication statusPublished - 2017
MoE publication typeA4 Article in a conference publication
EventInternational Work-Conference on TIme SEries Analysis, ITISE 2017 - Granada, Spain
Duration: 18 Sep 201720 Sep 2017

Conference

ConferenceInternational Work-Conference on TIme SEries Analysis, ITISE 2017
Abbreviated titleITISE 2017
CountrySpain
CityGranada
Period18/09/1720/09/17

Fingerprint

Learning systems
Dynamical systems
Temperature
Hot Temperature

Keywords

  • forecasting
  • machine learning
  • physically based models
  • smart grid

Cite this

Koponen, P., Niska, H., & Huusko, R. (2017). Improving the performance of machine learning models by integrating partly physical control response models in short-term forecasting of aggregated power system loads. In O. Valenzuela, & I. Rojas (Eds.), Proceedings ITISE 2017. Granada, 18-20, September, 2017. (Vol. 2, pp. 795-806). University of Granada.
Koponen, Pekka ; Niska, Harri ; Huusko, Reino. / Improving the performance of machine learning models by integrating partly physical control response models in short-term forecasting of aggregated power system loads. Proceedings ITISE 2017. Granada, 18-20, September, 2017.. editor / Olga Valenzuela ; Ignacio Rojas. Vol. 2 University of Granada, 2017. pp. 795-806
@inproceedings{20ee4dd5525b4280a252dfd01226550f,
title = "Improving the performance of machine learning models by integrating partly physical control response models in short-term forecasting of aggregated power system loads",
abstract = "Combining the strengths of different modelling approaches and various information sources is studied in short-term forecasting of aggregated electrical loads that are controllable and include e.g. thermal storage capacity. Measurement data driven models tend to fail in forecasting power during rare situations such as dynamic control actions and extreme weather conditions. The thermal dynamics of the loads, large outdoor temperature variations, and changes in the technologies contribute to this challenge. Here we study a model integration approach using field trial data covering about 7000 houses and 27 months. Control responses and load saturation are forecast using a physically based structure. The residual is forecast with a machine learning model designed and tuned to learn also system dynamics. The load forecast is the sum of these component forecasts. The forecasting accuracy of this hybrid method is compared with using the machine learning alone. The results show improvement in the accuracy",
keywords = "forecasting, machine learning, physically based models, smart grid",
author = "Pekka Koponen and Harri Niska and Reino Huusko",
note = "Project code: 101968",
year = "2017",
language = "English",
isbn = "978-84-17293-01-7",
volume = "2",
pages = "795--806",
editor = "Olga Valenzuela and Ignacio Rojas",
booktitle = "Proceedings ITISE 2017. Granada, 18-20, September, 2017.",
publisher = "University of Granada",
address = "Spain",

}

Koponen, P, Niska, H & Huusko, R 2017, Improving the performance of machine learning models by integrating partly physical control response models in short-term forecasting of aggregated power system loads. in O Valenzuela & I Rojas (eds), Proceedings ITISE 2017. Granada, 18-20, September, 2017.. vol. 2, University of Granada, pp. 795-806, International Work-Conference on TIme SEries Analysis, ITISE 2017, Granada, Spain, 18/09/17.

Improving the performance of machine learning models by integrating partly physical control response models in short-term forecasting of aggregated power system loads. / Koponen, Pekka; Niska, Harri; Huusko, Reino.

Proceedings ITISE 2017. Granada, 18-20, September, 2017.. ed. / Olga Valenzuela; Ignacio Rojas. Vol. 2 University of Granada, 2017. p. 795-806.

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

TY - GEN

T1 - Improving the performance of machine learning models by integrating partly physical control response models in short-term forecasting of aggregated power system loads

AU - Koponen, Pekka

AU - Niska, Harri

AU - Huusko, Reino

N1 - Project code: 101968

PY - 2017

Y1 - 2017

N2 - Combining the strengths of different modelling approaches and various information sources is studied in short-term forecasting of aggregated electrical loads that are controllable and include e.g. thermal storage capacity. Measurement data driven models tend to fail in forecasting power during rare situations such as dynamic control actions and extreme weather conditions. The thermal dynamics of the loads, large outdoor temperature variations, and changes in the technologies contribute to this challenge. Here we study a model integration approach using field trial data covering about 7000 houses and 27 months. Control responses and load saturation are forecast using a physically based structure. The residual is forecast with a machine learning model designed and tuned to learn also system dynamics. The load forecast is the sum of these component forecasts. The forecasting accuracy of this hybrid method is compared with using the machine learning alone. The results show improvement in the accuracy

AB - Combining the strengths of different modelling approaches and various information sources is studied in short-term forecasting of aggregated electrical loads that are controllable and include e.g. thermal storage capacity. Measurement data driven models tend to fail in forecasting power during rare situations such as dynamic control actions and extreme weather conditions. The thermal dynamics of the loads, large outdoor temperature variations, and changes in the technologies contribute to this challenge. Here we study a model integration approach using field trial data covering about 7000 houses and 27 months. Control responses and load saturation are forecast using a physically based structure. The residual is forecast with a machine learning model designed and tuned to learn also system dynamics. The load forecast is the sum of these component forecasts. The forecasting accuracy of this hybrid method is compared with using the machine learning alone. The results show improvement in the accuracy

KW - forecasting

KW - machine learning

KW - physically based models

KW - smart grid

M3 - Conference article in proceedings

SN - 978-84-17293-01-7

VL - 2

SP - 795

EP - 806

BT - Proceedings ITISE 2017. Granada, 18-20, September, 2017.

A2 - Valenzuela, Olga

A2 - Rojas, Ignacio

PB - University of Granada

ER -

Koponen P, Niska H, Huusko R. Improving the performance of machine learning models by integrating partly physical control response models in short-term forecasting of aggregated power system loads. In Valenzuela O, Rojas I, editors, Proceedings ITISE 2017. Granada, 18-20, September, 2017.. Vol. 2. University of Granada. 2017. p. 795-806