Deep learning and wavelet transform integrated approach for short-term solar PV power prediction

•A solar forecasting model based on Wavelet Transform and LSTM-dropout network has been proposed.•The model is based on previous output PV power data andseveral meteorological data.•Output performance is compared with several other contemporary ML approaches.•The comparative indices include MAPE, RMSE, MAE and R2 score. A novel short-term solar power prediction model is presented in this work, by utilizing the learning ability of Long-Shot-Term-Memory network (LSTM) based deep learning (DL) technique and the concept of wavelet transform (WT). In this proposed WT-LSTM model, the WT is used to decompose the recorded solar energy time-series data into different frequency series followed by the statistical feature extraction process. The LSTM with dropout based DL model is proposed to predict the futuristic value of solar energy generation in different time-horizon (hourly and day basis), where the statistical WT based features combined with several other meteorological factors such as temperature, wind speed, pressure, cloudy-index, humidity and altimeter index are modelled as input to the LSTM model. The efficiency of the suggested WT-LSTM model has been proved by comparing statistical performance measures in terms of RMSE, MAPE, MAE and R2 score, with other contemporary machine learning and deep-learning based models.