Fig 1.
Environmental variables and their impact on WT power output.
Fig 2.
Research objectives for enhancing WT power output prediction using ML and DL models.
Fig 3.
Density distributions of environmental parameters affecting WT energy output: (a) Temperature at 2m, (b) RH at 2m, (c) DWPT at 2m, (d) Wind speed at 10m, (e) Wind speed at 100m, (f) Wind direction at 10m, (g) Wind direction at 100m, (h) Wind gusts at 10m.
Table 1.
Descriptive statistics of environmental variables in the dataset used for WT power prediction.
Fig 4.
Heatmap of correlation matrix among environmental variables and power output.
Fig 5.
Overview of ML model selection and training for WT power output prediction.
Table 2.
Summary of ML models: Key parameters, scaling requirements, baseline usage, and training data needs.
Fig 6.
Flowchart of ML methodology for predicting WT power output.
Table 3.
Comparison of DL models used in the study for predicting WT power output.
Fig 7.
DL workflow flowchart for predicting WT power output from environmental variables.
Table 4.
Performance comparison of ML models for predicting WT power output.
Fig 8.
Comparative Actual vs. Predicted Power output by ML model for the first 1000 observations: (a) AdaBoost, (b) LR, (c) SVR, (d) LightGBM, (e) XGBoost, (f) CatBoost, (g) RF, (h) ET.
Fig 9.
Aggregated actual vs. predicted power output across all ML models.
Fig 10.
R2 values for ML models in predicting WT power output.
Fig 11.
MAE values for ML models in WT power prediction.
Fig 12.
RMSE values for ML models in WT power prediction.
Table 5.
Performance comparison of DL models for predicting WT power output.
Fig 13.
Training and validation R2 performance of DL models (a) CNN, (b) LSTM, (c) RNN, (d) ANN.
Fig 14.
Training and validation MAE performance of DL models (a) CNN, (b) LSTM, (c) RNN, (d) ANN.
Fig 15.
Training and validation loss performance of DL models (a) CNN, (b) LSTM, (c) RNN, (d) ANN.
Fig 16.
Predicted vs. Actual Power output for DL models on a sample of 1000 data points (a) CNN, (b) LSTM, (c) RNN, (d) ANN.
Fig 17.
Aggregated actual vs. predicted power output across all DL models.
Table 6.
Summary of comparative analysis between ML and DL models for predicting WT power output.