Research On Intelligent Fault Diagnosis Of Coalfired Power Plant Boiler Based On D-S Multi Information Fusion Theory

With the continuous development of China’s economy,the social demand for power industry is also increasing.Thermal power generation plays a very important role in China’s power structure.Large thermal power units have complex structure and numerous systems.Once some system fails,it’s easy to expand to a larger fault if it’s not found and handled in time,which causes unplanned shutdown of the unit,not only affect the stability of thermal power generation,but also increase the operating cost and energy consumption.Therefore,this paper studies the boiler combustion system of the existing coal-fired power plant,and uses the D-S multi information fusion theory to carry out the fault diagnosis of the boiler combustion system of the coal-fired power plant.so as to improve accuracy of the fault diagnosis of the thermal power unit and ensure the safe and stable operation of the thermal power unit,reduce operation cost and energy consumption.Firstly,the paper introduces the technological process of the boiler system of coal-fired power plant,and briefly describes the common faults of combustion system.According to the collection of measuring points and expert knowledge in real-time database,60 main measuring points of combustion system are selected.In addition,The Chauvenet Criterion and Savitzky-Golay method were used to remove outliers and reduce noise from the data of60 measurement points to provide high-quality samples for the subsequent diagnosis and analysis of the combustion system.Then,a combustion system fault diagnosis method based on improved Murphy rule is proposed.Based on the preprocessed data,the Relief algorithm is used to extract the eigenvalue to obtain 11 main parameters affecting the coking fault of the boiler.Taking 11 main parameters as input variable to train four intelligent learning algorithms,SVM,LVQ,PNN and BP by using normal tag data and fault tag data to obtain four fault diagnosis classification models.By using the confusion matrix,ROC curve and AUC area,the performance of four fault diagnosis models is evaluated in order to get the m function needed for information fusion.Finally,the improved Murphy rule in D-S theory is used to fuse the results of four fault diagnosis models,and obtain the final fault diagnosis results.Secondly,a combustion system fault diagnosis method based on FEKNN is proposed.The main feature of this method is that it doesn’t need fault tag data,and only need normal data for fault diagnosis of thermal power units.MDS dimension reduction is carried out for a large number of normal operation historical data after data preprocessing.Then,the mean shift clustering method is used to cluster the normal historical data to form the normal operation state database of combustion system.Finally,FEKNN fault diagnosis method is used to get the running state curve of combustion system for fault diagnosis of combustion system.Finally,the paper describes the construction of intelligent diagnosis and monitoring system in thermal power group.It main introduces the overall architecture,real-time database,distributed cloud computing system and communication architecture of the system,and graphically displays the operation modules of the system.