[ ] SIENA

Stimulation Initiative for European Neural Applications

Esprit Project 9811

Case Studies of Successful Applications

Current Prediction for Shipping Guidance in IJmuiden

Company background

Rijkswaterstaat is the department of Public Works and Water Management of the Ministry of Transport, Public Works and Water Management of the Netherlands. The North Sea Directorate of Rijkswaterstaat provides, among other things, forecasts of hydrological parameters to Dutch port authorities.

The problem

The IJ-channel, that leads to the port of IJmuiden, 25 km west of Amsterdam, is accessible to ships with drafts up to 16.46 m. The relatively strong cross-channel current in front of the harbour moles has an important bearing on navigational safety of deep draft ships. To ensure a safe passage of the harbour moles a current criterion has been established. This criterion is exceeded in almost every tidal cycle. Previously a reliable current forecast was not available. In making the sailing plan the pilot had only a rough current estimate based on current curves for mean spring, mean and neap tide conditions. The sometimes considerable residual wind driven currents were estimated by the pilot on the basis of experience, backed up by the actual and forecast wind conditions. If the current conditions can only roughly be estimated the pilot will build in extra safety margins. Without reliable current forecast, when weather conditions are unfavourable the pilot will more often decide to postpone a planned channel trip for at least one tidal cycle (12.5 hrs). So a reliable forecast is economically advantageous because of the reduction in waiting time that can be achieved for deep draft ships.

Neural network application

Rijkswaterstaat in collaboration with All Fours Neural Network Applications trained a neural network for current prediction. The neural network has been trained with nine months of current measurements and simultaneous wind and water level data from several locations in the North Sea. The neural network is implemented on a PC in the Hydro Meteo Centre Rijnmond (HMR) at Hook of Holland. A 24 hour current forecast on the basis of on-line measurements and forecasts of wind and water level is provided four times a day. The current forecast is relayed to the port authority of IJmuiden. Since April 1994, the current prediction system is in use helping IJmuiden pilots to make their sailing plans.

Benefits

The neural network is capable to provide reliable current forecasts. This makes a more efficient use of the IJ-channel possible. The development time for the system was very short (three months), against relatively low costs. The neural network solution requires only limited computing time; the calculation of a 24 hour 10 minute time step forecast, including the data preprocessing, takes less than 2 minutes on a PC. A good physical model for current prediction would require too much computing time for operational forecast.

Generalization

Many technical prediction problems which are too complex for efficient physical modelling are handled by human specialists in a qualitative way on the basis of their experience. If a large number of example-data is available, neural network techniques often can provide a means to increase prediction reliability of these problems at relatively low costs.

Contact person

Ir. J.C. Wüst, Rijkswaterstaat, North Sea Directorate, P.O. Box 5807, 2280 HV Rijswijk, The Netherlands. Tel +31 70 3366704 .

References for further reading

H. Wüst et al. Snelle en nauwkeurige stroomverwachting: Scheepvaartbegeleiding met neurale netwerken (in Dutch), Land + Water, Amsterdam, 12/08/94, pp 24-27.

H. Wüst et al. Neural Network Current Prediction for Shipping Guidance. in Proc. OCEANS '94 (Brest, France), pp I-58 -- I-63, IEEE

J.C. Wüst. Current Prediction for Shipping Guidance. In: Neural Networks: Artificial Intelligence and Industrial Applications. Proceedings of the 3rd Annual SNN Symposium on Neural Networks, Nijmegen, The Netherlands, 14-15 September 1995, Nijmegen. B. Kappen and S. Gielen (eds). Springer-Verlag, London.