From: A scoping review of deep neural networks for electric load forecasting
Reference | Model | RMSE / RMSPE | MAPE / MAE | Horizon | Data Context |
---|---|---|---|---|---|
(Katsatos & Moustris, 2019) | MLP | 20.5 kWh | 16.4% | 24 h | Data collected from public building in Athens, Greece |
(Koprinska et al., 2018) | CNN | 476.9 kW / 2392.88 kW / 642.52 kW | 340.64 kW / 1884.86 kW / 497.62 kW | 24 h | Publicly available data from Australia, Portugal, and Spain |
(Ferlito et al., 2015) | NARNN | 15.7% – 17.97% | – | 3mo – 12mo | Data collected from public building in Eboli, Italy |
(Timur et al., 2020) | RBF | 0.33 MWh | 11.83% | 1mo | Data collected from hospital building in Adana, Turkey |
(Selvi & Mishra, 2018) | Unspecified ANN | – | 2.9% | 1 h | Data collected from DSO in Delhi, India. |
(Torabi & Hashemi, 2012) | Unspecified ANN | – | 1.96% | 1 h | Data from Bandar Abbas, Iran |
(Eseye et al., 2019) | Unspecified ANN | – | 1.96% | 24 h | Data collected from buildings in Espoo, Finland. |
(Aurangzeb et al., 2021) | CNN | – | 39% | 24 h | Data collected from households in Australia for Smart Grid Project |
(Jarábek et al., 2018) | LSTM | – | 15% | 24 h | Data collected from enterprises in Slovakia. |
(Khan & Jayaweera, 2018) | Unspecified NN | – | 5.85% – 16.25% / 7.92 kWh – 22.93 kWh | 24 h | Data collected from smart meters in Ireland. |
(Barzola-Monteses et al., 2020) | LSTM | 5.085 kW | 3.714 kW | 24 h | Data collected from public building in Guayaquil, Ecuador |
(Vinagre et al., 2015) | MLP | – | 13.6% | 5 min | Data collected from office SCADA system |
(Rosato et al., 2019) | LSTM | 2.252 kW – 7.061 kW | – | 24 h | Data collected from power plant in Denver, CO |
(Qi et al., 2020) | CNN-LSTM | – | 1% | 24 h | Data collected from industrial area in China. |
(Pramono et al., 2019) | CNN-LSTM | 203.23 kW | 2.02% / 142.23 kW | 1 h | Data collected from public datasets (New England, USA & Switzerland) |
(Kim & Cho, 2019a) | CNN-LSTM | 0.6114 kW | 34.84% / 0.3493 kW | 1 h | Data collected from UCI ML repository on households |
(Al Khafaf et al., 2019) | LSTM | – | 3.15% | 3 day | Data collected from households in Victoria, Australia |
(Ves et al., 2019) | Ensemble | 6.19 kWh | 1.59% / 5.60 kWh | 4-24 h | Publicly available data UK-DALE |
(Bot et al., 2020) | Ensemble | 0.8 kW – 1.6 kW | – | 12 h | Data collected from Smart Home in Davis, USA |
(Ai et al., 2020) | Ensemble | – | 0.0199% | 24 h | Data collected from households in Norway. |
(Chenglei et al., 2015) | PSO-ANN | – | 0.0169% - 0.0606% | 24 h | Data from buildings in America |