A novel NSSA-LSTM-based forecasting model for ship delays in the Beijing-Hangzhou Grand Canal

The Beijing-Hangzhou Grand Canal serves as a crucial corridor for cargo transportation, connecting the north and south parts of China. With the continuous growth of inland waterway transportation demand, the vessel traffic flow of the Beijing-Hangzhou Grand Canal has become more intensive. The ship delays at busy locks which are featured as nonlinear and unsteady, have significantly affected the canal's navigation efficiency and exerted more pressure on the maritime safety administration. To optimize the lock operational strategies and ensure navigation efficiency, a novel ship delay forecasting model based on the optimized Sparrow Search Algorithm (NSSA) and the Long Short-Term Memory (LSTM) network has been proposed. The exploration and utilization capability of the traditional Sparrow Search Algorithm has been improved by introducing the niching techniques. The optimized Sparrow Search Algorithm is subsequently integrated with the LSTM network, leveraging the algorithm to achieve intelligent selection of hyperparameters as well as improving the model accuracy. Finally, a case study has been conducted at Shiqiao Lock in the northern Jiangsu section of the Grand Canal. Ship delay data ranging from 2017 to 2020 were collected for model validation and analysis. The present study demonstrates that the proposed NSSA-LSTM model exhibits better performance and stability in the forecasting of ship waiting time, with an average reduction of 4.1% in Mean Absolute Error (MAE), 52.8% in Mean Squared Error (MSE), 34% in Mean Relative Error (MRE), 29.6% in Root Mean Square Error (RMSE), and an improved goodness of fit (2.8%). The present study would provide a guidance for the intelligent management of ship locks and maritime safety administration in canals.