飞灰含碳量影响因素权重计算

Weight Coefficient Calculation of Factors that Influence Carbon Content in Fly Ash

以影响飞灰含碳量指标的参数作为输入参数,飞灰含碳量作为输出参数建立神经网络模型。首先,针对网络训练参数选择的不确定性,通过正交试验法对建模过程中的训练参数进行选择,结果表明能明显提高模型的精度。其次,针对影响飞灰含碳量的因素复杂及耦合性强的特点,提出了基于连接权、基准值的敏感性与偏差分析方法,计算得到输入参数的敏感性系数及偏差系数,以两个系数的乘积代表运行参数对指标的影响程度,即权重。最后,根据权重系数大小排序,确定出影响飞灰含碳量指标的主要影响参数。参数对指标影响程度的具体量化,能指导运行人员在日常优化运行时重点监控、调整关键参数,对进一步避免优化工作盲目性具有一定意义。

For establishing neural network model, carbon content in fly ash is often selected as the output parame- ter and factors influencing carbon content in fly ash are selected as the input parameters. First, for the problem that the uncertainty of selection of network training parameters, the train parameters are chosen by the orthogonal test, then the neural network model of high precision is obtained. Second, for the problem that factors influencing carbon content in fly ash are complex and have strong coupling, sensitivity analysis and deviation analysis based on con- nection weight are introduced to calculate the sensitivity coefficients and deviation coefficients, which reflecting the effects of input parameters on output indicators. The weight of operational parameters on indicator is represented by the product of two coefficients called weight coefficient. Finally, according to the orders of weight coefficients, the main factors influencing the indicators are determined. The determination of main factors is very significant for our operating personnel to strengthen the focus on monitoring and adjustment of the key parameters as well as avoid blindness in tuning.