Prediction Of Ash Accumulation On Heating Surface Based On Time Series Neural Network

As an important foundation for the realization of intelligent coal-fired power plant,the prediction of ash heating area can provide accurate ash accumulation data for the soot blowing optimization system of coal-fired power plant in the future,which is of great significance for the research of energy industry,the economy,safety and intelligent automatic control of coal-fired system.However,in the process of coal-fired boiler operation in thermal power plants,ash deposition on the heating surface is an important reason to reduce the efficiency and safety of boiler operation.Therefore,many researchers at home and abroad make soot blowing strategies based on the establishment of a fixed prediction model.However,in the whole process of making soot blowing strategies,they not only think that the soot deposition process is fixed,but also think that Soot blowing is a work of timely soot blowing operation,while neglecting that soot blowing requires certain preparation time.In view of the above problems,the main research work of this paper includes the following contents:?1?Because there are many factors that affect the ash of heating area,the research of ash of heating area is required to be higher.In this paper,the coal-fired boiler of a coal-fired power station is taken as the research object.By analyzing the ash principle of the heating area,the cleaning factor is selected to represent the health state of the heating surface.At the same time,we deal with all kinds of problems of the collected original data,which establishes a solid foundation for the subsequent prediction model of ash deposition.?2?By analyzing the time series characteristics of the data collected by DCS system,the neural network with time series characteristics is selected for the corresponding prediction research,and the established prediction model is applied to the case simulation of ash prediction of heating area.In this paper,the neural feedback Elman neural network and NAR nonlinear autoregressive neural network are used to establish the prediction model.Before establishing the model,in order to construct a reasonable network model,the algorithm principle of autoregressive model?AR?is adopted,and the minimum information criterion?AIC?and trial and error method are used to determine the number of inputs and hidden layer nodes,so as to realize a complete network structure.?3?In order to compare the network prediction model more effectively,this paper improves the structure of Elman neural network,which has the characteristics of output feedback.Through the establishment of the above three neural network prediction models based on the cleaning factor,and after modeling,in order to prove the influence of the prediction starting point on the prediction results,three data points₁T=250,₂T=300,₃T=350 are used to simulate and compare the prediction results.Finally,the corresponding time of ₃T data point is selected as the appropriate prediction starting point.?4?The weight and threshold of the network are randomly generated in the network operation,so the prediction results will also have certain differences.In order to solve this problem,the accuracy of the prediction model at the ₃T data point is verified by the normal probability density curve?PDF?,and the prediction results are evaluated by various error indexes.The results show that the predicted results of NAR network model are in good agreement with the actual monitoring data,which shows that the network model is effective for the prediction of ash in heating area.