Neural Network Based on Simulated Annealing Algorithm and Back Propagation and Application in the Groundwater Level Forecast of Irrigation District
Bo Zhou · Journal of North China Institute of Water Conservancy and Hydroelectric Power · 2007
In order to make BP network escape local minimum without descending training speed,a mixed algorithm is adopted,which combines simulated annealing with back propagation to train the forward feedback neural networks.That is to say,overall train with simulated annealing and accurate train with BP algorithm could accelerate network convergence velocity and escape local minimum.So overall minimum point can be found in the condition of the ideal network study velocity.At last this method is applied to groundwater level forecast in irrigation district,and the feasible and satisfactory results are obtained.