Accurate wind energy forecasting is crucial for Bangladesh as it seeks to diversify its energy mix and reduce dependence on fossil fuels. Moreover, precise predictions can optimize wind farm operations, enhancing grid stability and economic viability in the country's emerging renewable energy sector. This study explores wind energy potentiality for Inani Beach, Bangladesh, by forecasting wind speed using various deep learning models. To evaluate their performance based on important criteria, models including Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN)-LSTM, and CNN-BiLSTM (Bidirectional LSTM) were used. Input parameters included time, barometric pressure, height, and humidity for 2014–2015. The CNN-BiLSTM model was specifically proposed for wind speed potential forecasting and compared against other DL models. Results indicated the superior performance of CNN-BiLSTM compared to other algorithms, with CNN-LSTM showing the weakest performance. CNN-BiLSTM achieved the best results, boasting Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) values of 0.36 and 0.52, respectively. In contrast, CNN-LSTM underperformed with MAE and RMSE values of 0.42 and 0.6, respectively.