基于SVM的高效煤粉锅炉爆燃保护策略研究

为了提高高效煤粉锅炉的燃烧稳定性,提出了一种基于支持向量机(SVM)的爆燃保护控制策略。提取锅炉关键参数构建特征向量,采用SVM对系统历史数据进行离线训练,应用径向基函数、网格搜索算法生成系统状态分类器,并引入氧含量因子校正训练模型。锅炉运行时,分类器通过在线数据预测系统预爆燃状态并控制PLC模块执行保护程序。测试结果表明,氧含量因子取0.4时,分类器的最高交叉验证匹配率大于97%,最高预测准确率大于95%,失配率小于10%。保护策略能够有效地识别锅炉预爆燃状态,同时在锅炉正常工作状态下保持低误判率,增加了系统运行的鲁棒性。

Research on furnace explosion protection strategy of efficient pulverized coal fired boiler based on SVM

In order to improve the combustion stability of pulverized coal fired boiler,a furnace explosion protection strategy based on Support Vector Machine (SVM) was proposed in the paper.The state vector was constructed with key parameters of the boiler;afterwards a system state classifier was generated by training off-line history data of boiler with SVM,applying radial basis function and grid search algorithm.Besides,the oxygen content factor was introduced to regulate the training model.During running time of boiler,the classifier predicted pre-furnace state from on-line data then executed protection program through PLC (Programmable Logic Controller) module.Results show that when the oxygen content factor takes value of O.4,the maximum cross validation matching rate of the classifier is over 97％,the maximum predicting accuracy is over 95％,and mismatching rate is less than 10％.The protection strategy is able to identify the pre-furnace explosion state of boiler effectively as well as keeping low error judging rate under normal working state,and enhance the robustness of the system