Research On Forecasting Method Of Rotary Kiln Calcination Temperature

Cement rotary kiln is the core equipment during the raw material calcination process on NSP Cement production line, the function status of the cement rotary kiln is directly related to the clinker quality, clinker production, fuel consumption and business costs. Effective cement rotary kiln calcination temperature forecasting can provide reference for regulating the rotary kiln thermal parameters in advance. Therefore, the research on the forecasting method of cement rotary kiln calcination temperature is of great theoretical and practical significance.Aiming at the production process of the NSP rotary kiln has the characteristics of large inertia, multivariate, pure hysteresis and nonlinearity, the thesis studies the NSP cement production process, and proposes a forecasting method of rotary kiln calcination temperature based on fruit fly optimized extreme learning machine. First of all, the thesis analysis the NSP rotary kiln and its sintering mechanism, then selects the amount of feeding coal, the speed of second fan and the rotary speed of the kiln as the input affecting variants and the calcination temperature as the output target, which are the basis of the forecasting model. Secondly, for extreme learning machine, the network structure is most important to its performance, this thesis presents a fruit fly optimized extreme learning machine algorithm(FOA-ELM), the new method uses fruit fly algorithm which has a characteristic of global optimization to optimize the network structure of extreme learning machine. Experimental results show that the improved algorithm has good stability and spends less training time comparing with the conventional neural networks. Thirdly, combined with the modeling variable determined in the first step, a forecasting method of cement rotary kiln calcination temperature based on FOA-ELM is proposed, this method is expected to improve prediction efficiency comparing to the generally neural network based models.Finally, the simulation experiments utilizing test data collected in a working rotary kiln. The experimental results show that the presented cement rotary kiln calcination temperature forecasting method based on fruit fly optimized Extreme Learning Machine in this thesis can effectively predict the calcination temperature and provide production guidance for the site operators. In addition, comparing to the traditional neural networks based prediction methods, the new method spends less time during the model learning process, so the new method can acquire higher efficiency.