Aquaculture plays a predominant role in meeting the global demand for food, providing substantial employment, and contributing significantly to various economic sectors. Water quality in aquaculture ensures sustainable and profitable aquaculture. Aquaculture water quality (A-WQ) parameters are the influencing factors on the health, survival, productivity of aquatic species. In this research, deep learning (DL) techniques Gated Recurrent Unit (GRU) and Bidirectional Gated Recurrent Unit (Bi-GRU) have been proposed to predict the aquaculture water quality parameters for a real-world MAC dataset. In addition, a comprehensive analysis of the impact of hyperparameters \((h_p)\) on the models’ performance has been studied. As a result, an optimal set of hyperparameters have been identified that provides improved prediction accuracy in lesser computation time. It is observed from the comparative analysis of both models, Bi-GRU has the capability to achieve same prediction accuracy of GRU with lesser number of epochs \((E_p)\) . This shows the capability of Bi-GRU to achieve with good prediction accuracy with lesser computation time.