Genetic algorithm optimized LSTM modeling for dynamic water level regulation in smart garden rainwater recycling systems

This study addresses the dynamic regulation needs of smart garden rainwater recycling systems by proposing a water level predicion model based on a Genetic Algorithm-optimized Long Short-Term Memory network (GA-LSTM). Taking a garden in a southern Chinese city as an example, a bivariate time-series dataset was constructed by integrating precipitation data from meteorological stations and water level sensor data. The genetic algorithm was used to optimize hyperparameters of the LSTM, such as time step and the number of hidden layers. Experimental results show that the GA-LSTM model achieved a root mean square error (RMSE) of 0.143 and a coefficient of determination (R²) of 0.946, demonstrating significant optimization compared to other models. Through the synergistic effect of the rainwater recycling system and the water level prediction model, this study realized dynamic water storage and replenishment regulation for urban garden landscapes.