Abstract:

The bypass evaporation system for desulfurization wastewater can reduce the efficiency of the boiler and increase coal consumption by extracting part of the hot flue gas at the inlet of the air preheater. To achieve accurate prediction of the energy consumption caused by the extracted hot flue gas, a hybrid model prediction method combining a mechanistic model and an artificial neural network was proposed. Operational data from a 660MW power plant in Guangdong Province were collected as samples. Six parameters, including the wind temperature at the inlet of the air preheater, the flow rate of air, the extracted flue gas volume and temperature, the boiler load, and the coal feed rate, were used as inputs to establish a backpropagation neural network (BPNN) model for predicting the air heat transfer through the air preheater. The optimal structure of the network was determined by simulating and analyzing different hidden layer structures, and it was found to be 6-9-1, with a coefficient of determination (R²) of 0.99478 and a relative error of about 1% for the prediction model. The overall prediction effect of the model was good. Based on this, the qualitative law of energy consumption for the bypass evaporation system was obtained by combining the mechanistic model and typical operating conditions of fluctuating unit loads and extracted flue gas volumes.

Key words: desulfurization wastewater, hot flue gas, bypass evaporation system, energy consumption prediction, hybrid model

摘要：

脱硫废水旁路蒸发系统抽取部分空预器入口热烟气会使锅炉效率降低、煤耗增加。为了实现对抽取热烟气造成能耗增加的精准预测，提出了一种机理模型与人工神经网络有机结合的混合模型预测方法。采集广东某660MW电厂的运行数据作为样本，以空气预热器进口的风温、空气的流量、抽取烟气量、抽取烟气温度、锅炉负荷以及给煤量6个参数作为输入，建立了用于预测经过空气预热器空气换热量的BP神经网络模型，对不同隐含层结构进行模拟计算，分析比较确定了网络的最优结构是6-9-1，最终模型的决定系数R²是0.99478，预测模型的相对误差在1%附近波动，整体预测效果较好。在此基础上结合机理模型，针对机组负荷波动及抽取烟气量波动的典型工况进行了能耗预测，获得了旁路蒸发系统能耗的定性规律。

关键词: 脱硫废水, 热烟气, 旁路蒸发系统, 能耗预测, 混合模型