Study On Applications Of Neural Network To Coal Ash Fusion Characteristics Prediction

China is the world’s major economic powers which depend on coal in most areas. Andfor a long period of time, the dominant position of coal in our countries’s primary energystructure will be difficult to change. Therefore, it is particularly important to the rational useof coal. Coal gasification technology is an important way to solve the low utilizationefficiency of coal, as well as the large-scale coal gasification technology choosing liquid slagash discharge entrained bed gasification.Due to high gasification temperature, gasifierdischarge most of the coal ash into liquid slag in molten state under gravity and air drag inquench chamber. When temperature is too high cooling wall hanging slag thickness of orrefractory lining life fell sharply. Therefore, for coal ash melt temperature prediction isparticularly important.This article first analyzes the influencing factors of coal ash melting characteristics, andsums up that the main8oxides composition has the periorous impact of coal ash melting. Dueto the between complex relationship coal ash melting and its various oxides composition it’sdifficult to certained function model to be describe by certained function models. And theneural network has a great advantage to solve the problem of complicated nonlinearself-organization, self-learning, distributed storage, the characteristics of nonlinear, good faulttolerance and so on. Therefore, this paper adopts BP、generalized regression (GRNN) andimproved GA-BP, three kinds of neural network to forecast the coal ash melting sex. theGRNN network built by selecting the optimal smoothing factor，To determine the number ofhidden layer, vector selection, selection of initial weights, based on the analysis of theprediction error, choice of hidden layer nodes, selecting and training function, analysis of theconstruction of BP network， BP network’s initial weights are optimized by the geneticalgorithms to Create the GA-BP network. Fitting effect of three kinds of networks and errorsare compared. The result implies that the neural network has certain feasibility andsuperiority in coal ash melting characteristics prediction.Compared with traditional BP network GRNN is more suitable for small sample,but the predictive results of the improvedGA-BP network is superior to the traditional BP network and GRNN network.