Ash Fusion Characteristies And Fouling Monitoring Of Low Temperature Heating Surface Based On Support Vector Machine

Fouling has an important impact on the safe operation and economic benefits in power plants.Coal ash fusibility plays a decisive role in the contamination of heating surfaces.Therefore,coal ash fusibility and fouling monitoring of low-temperature heating surface were studied.Support vector machine(SVM)and gray wolf algorithm(GWO)were used to predict the deformation temperature of coal ash and the clean heat absorption of low-temperature heating surface,and the accuracy was high.Compared with traditional neural network,SVM was fit for trainning small sample data to realize intelligent prediction.Compared with genetic algorithm(GA),GWO has the advantages of faster speed and more accurate results in SVM parameter optimization.To explore the effect of sulfate on the ash-fusion temperature(AFT)of coal ash,different contents of sulfate were added to coal ash,and AFT and X-ray diffractometry(XRD)spectral analyses were carried out.The results show that CaSO4 can reduce the AFT of experimental coal ash,but when the content of CaSO4 exceeds 15%.the AFT increases with the increase of CaSO4.However,with the increase of Na2SO4,the AFT of experimental coal decreases as a whole.The deformation temperature of coal ash was predicted according to the oxides content and composition parameters.The SVM model optimized by GWO was used to predict the deformation temperature of different coal ash.and the prediction accuracy was accurate.And SO3 was compared as an independent variable for training prediction.The results show that the relative error of the deformation-temperature prediction is small and the prediction result with the SO3 input is more accurate.To monitor pollution on heating surface of economizer and low temperature superheater in power plant boilers,SVM was used to predict the clean heat absorption of heating surface,and GWO and GA were used to optimize the parameters of SVM.According to the predicted clean heat absorption,the cleaning factor is calculated-and the fouling state of the heating surface was judged by the change of the cleaning factor.Taking a 660 MW unit as an example,the collected clean data samples were trained and validated.The results show that the GWO-GWO has higher prediction accuracy and shorter training time than GA.Finally,the trained models were used to predict the clean heat absorption of the economizer before long blowing and the low temperature superheater betore short blowing respectively,then the clean curve was drawn.The fouling in economizer and low temperature superheater heating surface can be performed well.Thus,a basis for fouling on-line monitoring is offered.