Research Article
Hourly Forecasting of Solar Photovoltaic Power in Pakistan Using Recurrent Neural Networks
Table 5
Results of LSTM, Bi-LSTM, and SARIMA.
| Weather | Model | RMSE | square | CV (RMSE) % |
| Sunny | LSTM (2 layers) | 0.06 | 0.99 | 0.15 | Bi-LSTM | 0.06 | 0.99 | 0.15 | SARIMA | 1.43 | 0.26 | 4.57 | Cloudy | LSTM (2 layers) | 0.058 | 0.99 | 0.21 | Bi-LSTM | 0.0025 | 0.99 | 0.0095 | SARIMA | 1.76 | 0.40 | 7.03 | Rainy | LSTM (2 layers) | 0.157 | 0.91 | 0.60 | Bi-LSTM | 0.12 | 0.95 | 0.54 | SARIMA | 25.45 | 0.48 | 10.79 | Partial cloudy | LSTM (2 layers) | 0.18 | 0.81 | 0.51 | Bi-LSTM | 0.06 | 0.99 | 0.17 | SARIMA | 16.54 | 0.229 | 6.55 | Dusty | LSTM (2 layers) | 0.18 | 0.80 | 0.50 | Bi-LSTM | 0.08 | 0.99 | 0.22 | SARIMA | 2.53 | 0.274 | 7.75 | Fog | LSTM (2 layers) | 0.17 | 0.85 | 0.84 | Bi-LSTM | 0.072 | 0.98 | 0.33 | SARIMA | 3.81 | 0.19 | 25.4 |
|
|