| In recent years,the frequent natural disasters have brought great threat to people’ livesand property, so it’s impending to establish a rapid and efficient emergency rescue system.The timely and effective emergency distribution is a priority for the rescue, and the accuratelyforecast of emergency materials is a prerequisite to the distribution. Revolved aroundemergency relief work, this paper launched a study of demand forecasting and distribution foremergency materials.Firstly, predict the demand for disaster area in an indirect way. The back-propagationnetwork is a focus methods to predict disaster casualties. In view of the shortcomings, such aslow learning efficiency, easy to fall into local optimum and so on, this paper improves thestandard BP algorithm by additional momentum and adaptive adjustment of learning rate;then set up a predictive model for disaster casualties, and the forecast of casualties number in15large earthquake verifies the rationality of prediction model and the validity of algorithm;after that, estimate the tents demand in Wenchuan earthquake based on the demandforecasting model. Secondly, based on the characteristics of emergency supplies, construct adistribution model that take the satisfaction of disaster areas as a precondition, and theshortest delivery as a goal. Finally, design an improved genetic algorithm for the distributionproblem. Put forward an improved proportional selection operator to guarantee not only thediversity of population, but also approach to the optimal solution; blending the simulatedannealing algorithm with strong local search ability in the simple genetic algorithm(SGA) toovercome the “premature” convergence of simple genetic algorithm, as well as avoiding theSGA falling into local optimum; at last, demonstrate the high efficiency the improved geneticalgorithm by TSP, at the same time, verify the rationality of distribution model and theavailability of the improved genetic algorithm by a emergency supplies distribution instance. |
PDF Full Text Request