Research Of Mid-speed Mill For Multiobjective Optimization Based On Limited Coal

In the coal-fired units, direct blowing pulverizing system with medium speed mill is applied more frequently for several merits such as the simple system, low pulverizing unit cost, low steel consumption, less occupied space and low-risk of explosion and so on. Coal mill is an important component of coal pulverizing system, which provides proper coal powder for the combustion. What's more, the ventilation quantity must match the quantity of primary air required by the combustion system. The Primary air is the main motive force of boiler's fuel feed system and is used to dry and carry coal power. It relates to the factual combustion conditions of the furnace chamber directly. The variation of coal characteristics generally exerts great impact on the combustion in the boiler. As a result, coal fineness and air velocity delivering the coal power is reasonably adjusted according to coal characteristics in order to reduce the energy consumption and to improve the boiler thermal efficiency. Currently, the optimization of air-coal ratio with artificial intelligence and global optimization for combustion have been put in use in electric systems, and many problems are being researched and explored. Various references show that some achievements are obtained about forecasting boiler thermal efficiency, NOx emissions and unburned carbon content in fly ash, however, comprehensive optimization study for the primary air system of pulverized mill has rarely been reported.This paper builds the neural network model of boiler thermal efficiency and pulverizing unit cost by analyzing the determining factors and its influence on direct blowing pulverizing system with medium speed mill, and combines the BP neural network theory with Genetic Algorithms Optimization Extreme for parameter optimization on practical performances. Matlab programs are explored and corresponding conclusions are as follows:1) A deep analysis of determining factors and influencing levels is made, and some parameters are considered such as determining factors on primary air and damper opening during operation and controlling of the mill. The neutral network theory is combined with physical model. The BP neutral network using for non-linear model construction reaches to the ideal precision.2) The forecasting results by the completed model are conducted as fitness of the genetic algorithm. The global optimum values of the function and the corresponding input values are searched through selection, cross and mutation. 3) The programs for building models and searching for optimum values are conducted by Matlab. According to the programs, the best values of input factors such as primary air quantity and damper opening are acquired. After getting the trend of parameters, it can provide guidance for the inexperienced operators through tiny adjustment.By case verification, the model overcomes traditional fuzzy evaluation that only depends on the feelings one can judge the trend of primary air and damper opening of separator mill and then modulate them. Obviously, it not only ensures safety during operation, but also improves the economic benefits and so that is meaningful for saving energy and reducing consumption.