摘要
为提高煤/气混燃机组的锅炉效率,以某300 MW煤/气混燃锅炉为对象,对热量混燃比、分级配风、排烟温度、飞灰含碳量及锅炉效率进行了优化。基于优化试验数据,采用支持向量机(SVM)、反向传播(BP)神经网络及遗传算法优化BP(GABP)神经网络等算法,建立了锅炉效率智能算法预测模型。采用均方根误差和平均绝对百分比误差,分别评价了模型预测精度和预测值相对于试验值的平均偏离情况。试验与预测结果表明:随着高炉煤气热量混燃比的增加,锅炉效率由92.87%降到了90.82%,高炉煤气与焦炉煤气的最佳配比应小于1.3;锅炉效率随着分离燃尽风开度的增加而降低,最优开度为4层,均开至100%。通过对比3种预测模型的性能发现,遗传算法优化BP神经网络算法具有较强的逼近能力和泛化能力。该模型对锅炉效率的预测能力优于其他2种模型。
To improve the boiler efficiency of coal/gas co-fired power unit,with certain 300 MW co-fired boiler as object,the optimal experimental research on heat co-firing ratio,air staging,flue gas temperature,unburned carbon in fly ash and boiler efficiency are conducted. The boiler efficiency intelligent algorithm prediction models of support vector machines（ SVM）,back propagation（ BP） neural network and genetic algorithm optimizing BP（ GABP） neural network are established based on the experimental data. The root mean square error and mean absolute percentage error are adopted to evaluate the prediction accuracy and the average deviation of predicted value to experimental value respectively. The experiment and prediction results show that the boiler efficiency decreases from 92. 87% to 90. 82% with the increase of heat co-firing ratio of blast furnace gas. The optimum ratio of blast furnace gas and coke oven gas should be less than 1. 3. The boiler efficiency decreases with increase of damper opening of separated over-firing air,and the optimal damper opening is 100% for the four-layer of separated overfiring air. By comparing the performance of three prediction models,it is found that the GABP algorithm has stronger approximation and generalization capabilities,and the predictive ability of this model on boiler efficiency is better than the other two models.
出处
《自动化仪表》
CAS
2018年第1期20-24,共5页
Process Automation Instrumentation
基金
河北省自然科学基金资助项目(E2016502058)
关键词
热量混燃比
分级配风
燃烧优化
智能算法
预测模型
Heat co - firing ratio
Air staging
Combustion optimization
Intelligent algorithm
Prediction model
