Coal-based Boiler Production Prediction And Optimization Based On Improved BP And GA

With the economy development and environment protection, the competition among industries which take the coal as energy has increased as well. However, most of the boilers combustion adjustment is based on the production experience which has great randomness and less scientific guidance, and will lead to problems such as wasting fuel, reducing the boiler operation efficiency and so on. Therefore it is an urgent for industries to solve the problems to reduce the production cost and improve the combustion efficiency to enhance competitiveness. At the same time, Coal-based boiler has also brought serious environment pollution. As one of the major pollutants produced by the coal-based boiler, NOx(Nitrogen Oxide) will lead to poisonous pollutant which will harm human health such as acid rain, chemical smoke and so on. Therefore it attracts much attention to reduce the NOx emissions effectively. Above all, it is very significant to compute and analyze the coal-fired boiler efficiency and the NOx emissions, and provide the prediction and guidance of the parameters which needs to be adjusted reasonably.To meet the requirement of high efficiency and low pollutant emissions, based on the optimized BP neural network and genetic algorithm, the study develops the operation model by the historical data and optimizes those data to meet the high efficiency and keep the pollution emissions in a certain range. Firstly, according to the coal-based boiler operation, we select parameters which could affect the efficiency and NOx emissions as the inputs of the BP neural network. Secondly, we code the weights and thresholds of BP neural network by real number coding, and optimize them by genetic algorithm to get a better data prediction model. After that, we take an experiment to compare the prediction results between optimized BP neural network and ordinary BP neural network. Finally, we provide the optimization objective function according to the high efficiency and low emissions problems, and take the genetic algorithm to optimize the adjustable combustion parameters of a particular condition which can guide operators to adjust parameters such as fuel quantity and supply air rate and so on to achieve the production optimization. The experiment shows that the combustion model established by the optimized BP neural network is better than that of ordinary BP neural network, and combines it with genetic algorithm can guide the production better, and improve the production efficiency and reduce the pollutant emissions.Finally, in order to improve the practicability of the research, we use C# to develop a coal-based boiler combustion prediction and optimization system to meet the requirement of a thermal power plant. The software system takes real-time weather into consideration to build the BP neural network model with the history production data, and predicts the real-time production data. The comparison between real production data and prediction data will guide the operators to control the operation change. At the same time, the operation parameters optimization provides the guidance of real time operation adjustment.