The Research On Predicting Model Of Coals Quality And Optimizing Models Of Power Coal Blending In Power Plant

The contradiction of imbalance between regional distribution of coal resources and location of major consumer was very serious in China. The power plants of Hunan consume a great amount of coals, but Hunan is lack of coal, especially high-quality coal. Coal blending was the best way to overcome the contradiction, because they can raise combustion efficiency of coal, reduce pollutant emission and make full use of poor quality coal.In resent years, it has become one of hot points to solve the problem of blended coals through artificial intelligence technology. In this paper, Elman neural network was used to predict characteristic parameter of blended coals. The Simulated Annealing method was adopted to optimize the components of blended coals. This main works of this dissertation are as following.Based on element analysis and proximate analysis of blended coals and single coal, the non-linear relationship between qualities of the blended coals and its components was proved by t distribution method.The Elman neural network prediction model was established. The moisture content, ash content, volatile content, ignition temperature and calorific value of the blended coals were predicted by the characteristic prediction model. At the same time, different elements of blended coals were also predicted by different Elman neural network element prediction models.Relationship analysis method was used to testify the relationship between predictive values and actual values. It showed they had good dependencies, the correlation coefficients of all the parameters were more than 0.95. High reliability and confidence of the predicted result were verified using confidence interval method.Power coal blending model was established, whose constraints were performance of the blended coals allowed by boiler, and the objective function was the price of the blended coals. Simulated annealing (SA) method was used to solve the power coal blending model to search the lowest cost of coals. Elman neural network was used to predict qualities of blended coals. The result solved by SA was compared with the optimal result solved by exhaustive method. It showed that they had a small error, but algorithm efficiency of SA is eight times higher than exhaustive method.