Regional groundwater depth forecast by BP model of post-water-saving reconstruction in the Hetao Irrigation District of Inner Mongolia

This paper applies a three and four-layer Artificial Neural Network model (BP) based on long term regional groundwater, hydrological and climate data, for simulating dynamic movement of annual and monthly groundwater depth of the Hetao Irrigation District and forecasting the trend of change after reconstructing a water-saving engineering project. The average depth will decrease 0.51 m compared with present average depth in 2010. At the same time, it discusses the influence of different BP hidden layers, hidden units number and learning rates on the fitting errors, training efficiency, etc. The range of learning rates was presented as lr=0.01～0.1. The results showed that the BP network could be applied for regional groundwater forecast, and has a higher calculating precision. This will provide an important foundation in the planning, design, and management of the water saving reconstruction project in the large-scale irrigation district, and for the application of BP model in groundwater hydrological prediction.