基于试验和BP神经网络的CFB锅炉脱硫效率研究

Research on the Desulfurization of CFB Boiler Based on Experiment and BP Neural Network

在一台300 MW以洗中煤和煤矸石按6∶4比例混合后制得的混矸煤为燃料的工业CFB锅炉上,研究了床温、Ca/S摩尔比、一/二次风配比主要因素对脱硫效率的影响。试验结果表明:脱硫效率随着Ca/S摩尔比增大而提高,并且提高的幅度越来越小,最后趋于平稳;脱硫效率随着床温的升高而逐渐降低,并且减小幅度越来越小;脱硫效率随二次风率的增加先缓慢增大,后迅速下降,存在一个最佳的二次风率。以试验数据为基础,建立了3种算法的BP神经网络模型对脱硫效率进行预测,通过对比分析发现:基于附加动量法的BP神经网络预测脱硫效率时,在单隐含层神经元数为15时,其平均偏离度为3.76%,最大相对误差为8.97%,这能够较好地预测CFB锅炉的脱硫效率。

An industrial experiment was conducted on a 300 MW circulating fluidized bed （CFB） boiler burning mixed waste coal which consists of middlings and coal gangue according to the proportion of 6： 4. The effect of sev- eral main factors was researched, concerning Ca/S molar ratio, dense-phase zone bed temperature, ratio of the primary/secondary wind. The experiment shows that： the desulfurization efficiency increases with the rise of Ca/S molar ratio, and it first increases fast, then slowly, the last stable ; the desulfurization efficiency decreases with the rise of dense - phase zone bed temperature under the same Ca/S molar ratio, and the rate of reduction tends to small; the desulfurization efficiency increases slowly in the early stage, and then decreases rapidly in the late stage, with the rise of secondary air ratio. Based on the experimental data, three BP neural network models were established to predict desulfurization efficiency. By comparison, it shows that the BP neural network model based on momentum back propagation and 15 of single hidden layer neurons number can predict desulfurization efficiency with mean diversion extent of 3.76% and maximum error of 8.97%, the model can better predict particle circulating flow rate.