|
Ref. | Method | Application | Results obtained | Metrics used | Data units |
|
[50] | ANN + fuzzy logic | Energy (electricity demand) | The method was lowering the average. | RMSE = 42% | Hourly |
[51] | TCNN, LSTM, | Energy (solar power) | In terms of accuracy and capability to maintain a longer effective history, TCNN beats other models. | MAE | Half-hourly |
[52] | TCNN, TCAN | Energy (solar power) | In terms of accuracy, TCAN outperforms some state-of-the-art deep learning prediction models, such as TCNN. | MAE | Half-hourly |
[53] | LSTM + TA | Solar generation (energy) | Employing partial autocorrelation to calculate the input lag, the TA method improves performance over regular LSTM. | RMSE = 0.25 MAE = 0.12 | Daily |
[54] | ETS + LSTM | Energy (electricity demand) | Superior performance and competitiveness with both classical models | MAPE 5% | Monthly |
[55] | LSTM | Energy (electricity consumption) | Prediction results were obtained with the least degree of error. | RMSE | Every minute |
[56] | LSTM | Energy | LSTM outperformed ARIMA and SARIMA | RMSLE, RMSE, MASE, and MAPE | Daily |
[57] | NARX-ANN-PSO | Energy | The model is capable of determining the appropriate input and output lag terms. | MSE, RMSE, MAPE | Hourly |
[58] | Combined ANN | Energy (electricity price) | The model’s performance may be utilized as a baseline for making EV charging decision | MAPE | Hourly |
[59] | LSTM + fuzzy | Energy (electricity demand and wind speed) | The nonstationary and irregular characteristics are effectively addressed by the CFML model. Under the MAPE metric, the model perfection wind speed and electrical power load by an average of 49% and 70%, respectively. | MAPE | - |
[60] | ISO-TS-RBF-RFNN | Power load forecasting | The model performs the best in terms of long-term load forecasting accuracy. | MAPE | Monthly, daily, weakly, hourly |
|