D6.2–Load and generation forecasting methods and prototypes: Energy forecasting, Renewable energy, Consumption, Neural Networks.

Per Printz Madsen, Lara Pérez Dueñas, Carlos Castaño Moraga

Publikation: Bog/antologi/afhandling/rapportRapportForskningpeer review

Resumé

The objective of the forecasting module in ENCOURAGE Platform is to provide the system with data of what the consumption and the production will be in the next hours, so that the energy management modules can decide the future strategies in order to optimise energy flows and decrease overall consumption.
Energy consumption forecasting and renewable energy generation forecasting are two completely different problems that have been addressed separately, although they require similar inputs and a similar architecture.
The modelling and forecasting of the energy consumed by a building usually leads to implementation of daily profile models. Each building, as shown by the 3 demonstrators’ energy baseline analysis, has a specific energy consumption pattern that is composed by daily, weekly, and seasonal cycles. Individual models are run to predict future demand for the interval of interest, typically one day ahead.
After assessment of the different available methods, the modelling has been achieved through neural networks, and compared to a persistence model that has served as baseline for comparison, obtaining good results.
Forecasting of renewable power generation is a significantly more difficult task than energy load forecasting due to immediate impact of weather conditions – wind or cloudiness – on power output. Renewable generation depends on environmental conditions (wind speed, solar irradiation), which are subject to large fluctuations and their reliable forecasts are not always available.
Most existing suppliers use anyway some kind of statistical approach to make the energy prediction. In the market there are few but strong providers of such services, and it has been preferred to use an external provider rather than developing ENCOURAGE’s own energy production algorithm.
The external service chosen belongs to one of the partners of the consortium (Gnarum), so tests have been carried on to adapt the forecasting methods to the distributed small-scale generation case, with satisfactory results.
OriginalsprogEngelsk
ForlagAalborg Universitet
Antal sider112
StatusUdgivet - 2014

Emneord

  • Energy forecasting, renewable energy
  • Neural Network

Citer dette

@book{a872a96f9bd7462a86a9eb3ac8cf0def,
title = "D6.2–Load and generation forecasting methods and prototypes: Energy forecasting, Renewable energy, Consumption, Neural Networks.",
abstract = "The objective of the forecasting module in ENCOURAGE Platform is to provide the system with data of what the consumption and the production will be in the next hours, so that the energy management modules can decide the future strategies in order to optimise energy flows and decrease overall consumption.Energy consumption forecasting and renewable energy generation forecasting are two completely different problems that have been addressed separately, although they require similar inputs and a similar architecture.The modelling and forecasting of the energy consumed by a building usually leads to implementation of daily profile models. Each building, as shown by the 3 demonstrators’ energy baseline analysis, has a specific energy consumption pattern that is composed by daily, weekly, and seasonal cycles. Individual models are run to predict future demand for the interval of interest, typically one day ahead.After assessment of the different available methods, the modelling has been achieved through neural networks, and compared to a persistence model that has served as baseline for comparison, obtaining good results.Forecasting of renewable power generation is a significantly more difficult task than energy load forecasting due to immediate impact of weather conditions – wind or cloudiness – on power output. Renewable generation depends on environmental conditions (wind speed, solar irradiation), which are subject to large fluctuations and their reliable forecasts are not always available.Most existing suppliers use anyway some kind of statistical approach to make the energy prediction. In the market there are few but strong providers of such services, and it has been preferred to use an external provider rather than developing ENCOURAGE’s own energy production algorithm.The external service chosen belongs to one of the partners of the consortium (Gnarum), so tests have been carried on to adapt the forecasting methods to the distributed small-scale generation case, with satisfactory results.",
keywords = "Energy forecasting, renewable energy, Neural Network",
author = "Madsen, {Per Printz} and Due{\~n}as, {Lara P{\'e}rez} and Moraga, {Carlos Casta{\~n}o}",
year = "2014",
language = "English",
publisher = "Aalborg Universitet",

}

D6.2–Load and generation forecasting methods and prototypes : Energy forecasting, Renewable energy, Consumption, Neural Networks. / Madsen, Per Printz; Dueñas, Lara Pérez ; Moraga, Carlos Castaño .

Aalborg Universitet, 2014. 112 s.

Publikation: Bog/antologi/afhandling/rapportRapportForskningpeer review

TY - RPRT

T1 - D6.2–Load and generation forecasting methods and prototypes

T2 - Energy forecasting, Renewable energy, Consumption, Neural Networks.

AU - Madsen, Per Printz

AU - Dueñas, Lara Pérez

AU - Moraga, Carlos Castaño

PY - 2014

Y1 - 2014

N2 - The objective of the forecasting module in ENCOURAGE Platform is to provide the system with data of what the consumption and the production will be in the next hours, so that the energy management modules can decide the future strategies in order to optimise energy flows and decrease overall consumption.Energy consumption forecasting and renewable energy generation forecasting are two completely different problems that have been addressed separately, although they require similar inputs and a similar architecture.The modelling and forecasting of the energy consumed by a building usually leads to implementation of daily profile models. Each building, as shown by the 3 demonstrators’ energy baseline analysis, has a specific energy consumption pattern that is composed by daily, weekly, and seasonal cycles. Individual models are run to predict future demand for the interval of interest, typically one day ahead.After assessment of the different available methods, the modelling has been achieved through neural networks, and compared to a persistence model that has served as baseline for comparison, obtaining good results.Forecasting of renewable power generation is a significantly more difficult task than energy load forecasting due to immediate impact of weather conditions – wind or cloudiness – on power output. Renewable generation depends on environmental conditions (wind speed, solar irradiation), which are subject to large fluctuations and their reliable forecasts are not always available.Most existing suppliers use anyway some kind of statistical approach to make the energy prediction. In the market there are few but strong providers of such services, and it has been preferred to use an external provider rather than developing ENCOURAGE’s own energy production algorithm.The external service chosen belongs to one of the partners of the consortium (Gnarum), so tests have been carried on to adapt the forecasting methods to the distributed small-scale generation case, with satisfactory results.

AB - The objective of the forecasting module in ENCOURAGE Platform is to provide the system with data of what the consumption and the production will be in the next hours, so that the energy management modules can decide the future strategies in order to optimise energy flows and decrease overall consumption.Energy consumption forecasting and renewable energy generation forecasting are two completely different problems that have been addressed separately, although they require similar inputs and a similar architecture.The modelling and forecasting of the energy consumed by a building usually leads to implementation of daily profile models. Each building, as shown by the 3 demonstrators’ energy baseline analysis, has a specific energy consumption pattern that is composed by daily, weekly, and seasonal cycles. Individual models are run to predict future demand for the interval of interest, typically one day ahead.After assessment of the different available methods, the modelling has been achieved through neural networks, and compared to a persistence model that has served as baseline for comparison, obtaining good results.Forecasting of renewable power generation is a significantly more difficult task than energy load forecasting due to immediate impact of weather conditions – wind or cloudiness – on power output. Renewable generation depends on environmental conditions (wind speed, solar irradiation), which are subject to large fluctuations and their reliable forecasts are not always available.Most existing suppliers use anyway some kind of statistical approach to make the energy prediction. In the market there are few but strong providers of such services, and it has been preferred to use an external provider rather than developing ENCOURAGE’s own energy production algorithm.The external service chosen belongs to one of the partners of the consortium (Gnarum), so tests have been carried on to adapt the forecasting methods to the distributed small-scale generation case, with satisfactory results.

KW - Energy forecasting, renewable energy

KW - Neural Network

M3 - Report

BT - D6.2–Load and generation forecasting methods and prototypes

PB - Aalborg Universitet

ER -