To realize real-time prediction and intelligent monitoring of membrane fouling in Membrane Bioreactor (MBR) systems, a membrane fouling prediction model based on an improved Whale Optimization Algorithm (the Whale Optimization Algorithm integrated with a global search strategy, referred to as the Gravitational Search Whale Optimization Algorithm，GS-WOA) and a Bidirectional Long Short-Term Memory (BiLSTM) neural network was developed. First, monitoring data samples were standardized, and the BiLSTM neural network was adopted as the basic prediction framework to fully utilize its bidirectional temporal feature extraction capability for capturing the dynamic variation of the fouling process. Then, the Whale Optimization Algorithm (WOA) was improved by introducing a gravitational search mechanism and adaptive inertia weight to globally optimize BiLSTM hyperparameters such as learning rate, number of hidden neurons, and time step, thereby balancing global exploration and local exploitation. Finally, the optimized model was trained and validated using actual operational data. The results show that the GS-WOA-BiLSTM model achieves a prediction accuracy of R²=0983 7, improving by approximately 6.6% compared with the LSTM model, while the mean absolute error and root mean square error are reduced by 28.1% and 19.9%, respectively. The predicted values exhibited excellent agreement with measured data. This method enables high-precision prediction and trend forecasting of membrane flux and transmembrane pressure, providing reliable technical support for intelligent monitoring and optimized operation of MBR systems.