Research On Fuel System Operation Optimization And Fault Warning Method For Thermal Power Units

In recent years,renewable energy generation has been developing rapidly,but thermal power generation still occupies the main position in China’s power structure,and its fuel cost accounts for more than 70%of the cost of power generation.Strengthening fuel wisdom management and improving coal management efficiency have become one of the effective means for thermal power enterprises to improve their competitiveness in the power market.The contradiction between power generation and coal supply is more prominent in thermal power sector,and the combustion coal quality is unstable and often deviates from the design value,the operating conditions of auxiliary engines in the fuel system are more complicated,and the probability of failure is also greatly increased,and as an important auxiliary equipment in the fuel system,effective condition monitoring and fault warning of the coal mill is helpful to improve the safety and economy of the whole unit.In this paper,the intelligent stacking of incoming coal in a 2*350MW coal-fired power station in the north and the fault warning of medium-speed coal mill are the research objectives,and intelligent algorithms are used to carry out the corresponding research work.In response to the problem of complex incoming coal types and confusing stacking in the coal yard under study,we extracted the historical incoming coal information of 2021 batch 860 from this coal yard in the fuel system database,selected the key coal quality information of low-level calorific value,total sulfur and volatile fraction in the coal quality testing information,and used K-means algorithm and DBSCAN algorithm to cluster the coal quality information data set and compare the results respectively,and finally selected K-means,which has better clustering effect,is selected as the method of coal quality division.It divides the selected historical incoming coal quality information into 4 categories,and the coal quality components within each category are close to each other,and the frequency and weight ratio of incoming coal of different categories are counted,and the strip coal yard is divided into corresponding proportions,and the incoming coal of the same category is stacked in each partition.Meanwhile,in order to solve the fault warning problem of medium-speed coal mill,a method based on multivariate state estimation and vector similarity is proposed in this paper,and the monitoring parameters closely related to the coal blockage fault of coal mill are selected by using Pearson correlation coefficient,and the historical data containing the information of coal blockage fault are collected at an interval of 5 s.The equally spaced sampling method is used to establish a healthy process memory matrix for the normal working training set data,define a similarity function between the estimated vector of the model and the actual observed vector,and set an early warning threshold using the sliding window method,and the model issues an alarm when the similarity is lower than the early warning threshold for three consecutive time series points.The K-means algorithm was used to classify incoming coal for stacking,and the automatic stacking process of incoming coal in the coal yard was designed to improve the intelligence level of incoming coal stacking and the efficiency of coal yard management.When the data of the normal operation test set was input into the model,the average relative error of the model predicted each parameter was less than 1%,which proved that the model could predict the normal operation of the coal mill well.The model can give an alarm 375 s before a trip due to coal blockage,thus realizing a fast and accurate intelligent warning of coal mill failure.