Forecasting of Solar Power Potential in Pakistan Using Deep Learning

In power systems, renewable energy type, solar energy is significant in its present utilization and penetration; thus, there is a need for deterministic models for its forecasting. This work aims to develop an effective deep-learning model for predicting solar power potential in Pakistan, especially in the following cities: Lahore, Karachi, Peshawar, Quetta, and Islamabad. Three deep learning models, Artificial Neural Networks (ANN), Long Short-Term Memory (LSTM), and Convolutional Neural Networks (CNN) were used to forecast the solar irradiance by using performance measures of root mean square error (RMSE), mean absolute error (MAE), and correlation coefficient (R) was employed. According to the results from different models, it is revealed that CNN outperformed all the models in most of the cities. In the case of Lahore city, CNN achieved RMSE 0.638, but ANN and LSTM achieved 0.0648 and 0.0638, respectively. The results indicated that CNN had a high level of performance and stability in all the studied cities; thus, it was the most suitable model for forecasting solar power in Pakistan.