Optimization of Coal Blending with Backpropagation Neural Networks (BPNN) and Genetic Algorithms (GA) in Tangential In-Furnace Blending Boilers

One of Phase 1 Fast Track Program (FTP-1) is a coal-fired power plant with a capacity of 3 × 315 MW and the main fuel is coal. Coal has a very important role in determining combustion characteristics. Coal with good quality will improve the quality of combustion and operation of a power plant. But in reality, often a power plant does not get coal according to specifications, so it is necessary to find a solution to this problem. Coal blending is the process of mixing good quality coal with low-quality coal to obtain medium-quality coal. One method of coal blending is mixing in the furnace whereby only placing one type of coal in each coal burner that is known as furnace blending. The coal blending is done by mixing medium rank coal (MRC) and low-rank coal (LRC) with a composition of 50%:50% which is fed into the boiler through four burners with different elevations. In this research, the optimal search for blending MRC and LRC coal also the composition of the feed on the burner layer is carried out with the backpropagation neural network (BPNN) and genetic algorithm (GA) model in Matlab software. Based on the results obtained in this optimization system, it was found that the coal blending of MRC 1 (BA company), LRC 3 (PLNBB company) and the layer burner composition 1 (composition of MRC in the lower burner layer and LRC in the upper burner layer) produce optimal output (value −0.39402) which is predicted to produce a load of 280 MW, boiler efficiency of 84.15%, flue gas temperature 151.92 ℃, NO_x 21.35 mg/Nm³, SO_x 400.19 mg/Nm³, unburned carbon in fly and bottom ash 4.38 and 3.83%wt.