Leakage Forecasting of Karst Reservoir Using Back-Propagation Artificial Neural Network: A Case of Shuibuya Hydroproject

This is a preliminary attempt towards an extending use of artificial neural network (ANN) in leakage forecasting of valuable water resources in karst reservoirs, aiming to sustainable reservoir management. We presented an ANN system to evaluate the leakage behaviors of karst reservoir in response to their controls. Data from southwestern China were used for the training subset and 4 cases for the testing subset. After systematic training-testing efforts, the back-propagation (BP) ANN model was optimally estimated. A case sample of Shuibuya karst reservoir in Qingjiang River of central China was conducted to elaborate the effectiveness of our presented approach. The results showed that BP-ANN system was available and efficient in predicting the leakage nature of karst reservoirs with reasonable accuracy.