A Genetic Neural Network for Tide Forecasting

The tide forecasting is an important task in the daily operation and optimal scheduling in tidal power station. Tidal level forecasting has its difficulty because the level is always effected by non-cyclical factors which are required for the prediction of tidal level. Traditional artificial neural network(ANN) model has been widely used in tide forecasting, however, the ANN model has some shortages, such as slow training, low precision or easy to fall into local optimum. Based on the ANN model, this paper presents a genetic neural network (GNN) model has been improved for forecasting the tidal level, which weights and thresholds are optimized by genetic algorithm. This GNN model can easily decide the nonlinear input-output relation and has a higher precision. Three models which adopt different algorithms will be used to test the performance of the proposed model. The result indicates the hourly tidal levels can be accurately predicted by the GNN model using a short-term hourly tidal records.