Search

Short-term load forecasting (STLF) is crucial for the daily operation of power grids. However, the non-linearity, non-stationarity, and randomness characterizing electricity demand time series renders STLF a challenging task. Various forecasting approaches have been proposed for improving STLF, including neural network (NN) models which are trained using data from multiple electricity demand series that may not necessarily include the target series. In the present study, we investigate the performance of a special case of STLF, namely transfer learning (TL), by considering a set of 27 time series that represent the national day-ahead electricity demand of indicative European countries. We employ a popular and easy-to-implement feed-forward NN model and perform a clustering analysis to identify similar patterns among the load series and enhance TL. In this context, two different TL approaches, with and without the clustering step, are compiled and compared against each other as well as a typical NN training setup. Our results demonstrate that TL can outperform the conventional approach, especially when clustering techniques are considered. [ABSTRACT FROM AUTHOR]

Short-term load forecasting (STLF) is crucial for the daily operation of power grids. However, the non-linearity, non-stationarity, and randomness characterizing electricity demand time series renders STLF a challenging task. Various forecasting approaches have been proposed for improving STLF, including neural network (NN) models which are trained using data from multiple electricity demand series that may not necessarily include the target series. In the present study, we investigate the performance of a special case of STLF, namely transfer learning (TL), by considering a set of 27 time series that represent the national day-ahead electricity demand of indicative European countries. We employ a popular and easy-to-implement feed-forward NN model and perform a clustering analysis to identify similar patterns among the load series and enhance TL. In this context, two different TL approaches, with and without the clustering step, are compiled and compared against each other as well as a typical NN training setup. Our results demonstrate that TL can outperform the conventional approach, especially when clustering techniques are considered.

• Households change their heating patterns when they switch to a heat pump. • Typically, a setback temperature 2–5 °C lower than the desired thermostat setting. • We calculate the increase in heat demand with varied parameters. • The increase in heat demand is determined by heat loss and thermal mass. • This would be a useful metric for 'heat pump readiness'. Heating with air source heat pumps is an important step in the transition to Net Zero Carbon for many households. When predicting the resulting energy consumption it is frequently assumed that heating demand is unchanged from before the switch from gas heating. However, heating patterns usually change, with a net increase in heating demand to achieve the same level of thermal comfort. This paper discusses the change in pattern and uses modelling to determine the resulting increase in energy demand. It shows that the increase can be predicted with information about dwelling heat loss and thermal mass as well as weather and the pre and post switch heating pattern. The increase can be 20% or more for homes with high heat loss. It proposes that the metric for heating demand increase would be a useful measure of heat pump readiness, and that the parameters required to assess this should be provided on energy performance certificates. [ABSTRACT FROM AUTHOR]

[Display omitted] Heating homes using gas boilers is incompatible with the UK's target of net-zero greenhouse gas emissions by 2050. One solution is to shift to heat pumps (HPs) supplied from decarbonised power plant, but this could place an unmanageable burden on the electricity supply network. National heat demand profiles depend on the heating patterns adopted by households which, in turn, depend on the type of heating system and its control. The largest data sets available, from around 6600 gas-heated homes and 600 homes with HPs, are used to create an empirical model of Great Britain's (GB) half-hourly domestic heat demand. The model is used to estimate the annual half-hourly heat demand of the GB housing stock for both current and future weather conditions. The demand profile when using HPs is compared to the current profile for gas heating. In a cold year, the calculated total annual heat demand of a typical mix of ground source and air-source HPs was 422TWh, 8% greater for than for gas-heated homes. However, the peak heat demand of 157GW was 8% lower than for gas heating, and the maximum heat ramp rate of 21GW/h, 67% lower. These results are due to the different ways that households use gas boilers and HPs. The accurate modelling of heating patterns is necessary to achieve reliable predictions of national heat demand. Policy initiatives, financial incentives or other interventions that influence the daily pattern of HP usage could also have a marked and positive influence on the GB heat demand profile. [ABSTRACT FROM AUTHOR]